Loading...
The URL can be used to link to this page
Your browser does not support the video tag.
Home
My WebLink
About
Final Workshop Document (combined) 6.18.10
FY 2010-2011 Board of Trustees Budget Retreat June 18, 2010 1 FY 2010-11 Budget and Strategic Planning Calendar 2 Strategic Plan Desired Outcomes 3 General Financial Information 4 School Census and Enrollment Information 5 Cost Comparison Information 6 Budget Scenarios 7 Recommended Service Level Adjustments 8 Summary Points and Recommendations 9 Studies on School Size and Class Size 10 The Excellence Factor 11 Block Schedule Information 12 2009 WA Parents Survey Summary Report 1 MEMORANDUM Date: June 16, 2010 TO: Honorable Mayor/President and Board of Trustees FROM: Tom Brymer, Town Manager/Superintendent Westlake Academy SUBJECT: WAFY 2010-11 Budget Retreat Information In preparation for the Board’s budget retreat to discuss and give direction regarding the FY 2010-11 Academic Services Budget, please find for your review and information the following information: Page Section 1: WA FY 2010-11 Budget and Strategic Planning Numbers Calendar…….…………………………………………………. Section 1-1 Section 2: WA Strategic Plan Desired Outcomes…………………… Section 2-1 Section 3: General Financial Information • FY 09-10 financial snapshot including Municipal Support…… …. Section 3-1 • FY 09-10 financial snapshot for Academic Services Budget only…………………………………………………………. Section 3-2 • Revenues FY 09-10……………………………………………… ….. Section 3-3 • Cost per student per FY 09-10 Academic Services Only………… Section 3-5 • Cost per student per FY 09-10 Including Municipal Support……. Section 3-6 • Enrollment and Academic Overview……………………………….. Section 3-7 • Position Summary……………………………………………………. Section 3-8 • Approved WA Salary Scale and Actual ISD Scales for FY 2009-2010 School Year …………………………………………… Section 3-9 • Municipal Direct and Indirect Cost Information……………………. Section 3-10 • Debt service information related to the WA campus……………… Section 3-12 • Program information………………………………………………….. Section 3-13 Section 4: School Census/Enrollment Information • Census of Students by School District Residency……………….... Section 4-1 • Census of Students by School District Residency and by Grade.... Section 4-2 • Census of Students by School Staff, Town Staff, and Remainder.. Section 4-3 • Enrollment Levels……………………………………………………… Section 4-4 2 Page Numbers Section 5: WA Cost Comparison Information to Texas Charter School Average Costs • Excerpt from Report entitled; Paying for the Vision: Charter School Revenues and Expenditures- Texas………………………… Section 5-1 Section 6: FY 10-11 Budget Scenarios • Option A – Block Schedule (No new students or classes) • Scenario 1- No SLA’s, hold enrollment and costs flat………. Section 6-1 • Scenario 2- No SLA’s/10% reduction…………………………. Section 6-2 • Scenario 3- No SLA’s/20% reduction…………………………. Section 6-3 • Scenario 4- includes SLA’s/2% salary scale adjustment…… Section 6-4 • Option B – Block Schedule (New students and classes) • Scenario 1- No SLA’s, hold enrollment and costs flat………. Section 6-5 • Scenario 2- No SLA’s/10% reduction…………………………. Section 6-6 • Scenario 3- No SLA’s/20% reduction…………………………. Section 6-7 • Scenario 4- includes SLA’s/2% salary scale adjustment…… Section 6-8 • Option C – Block Schedule (New classes, no new students) • Scenario 1- No SLA’s, hold enrollment and costs flat………. Section 6-9 • Scenario 2- No SLA’s/10% reduction…………………………. Section 6-10 • Scenario 3- No SLA’s/20% reduction…………………………. Section 6-11 • Scenario 4- includes SLA’s/2% salary scale adjustment…… Section 6-12 Section 7: FY 10-11 Recommended Service Level Adjustments……. Section 7-1 . Section 8: Summary Points and Recommendations………………..…. Section 8-1 Section 9: Studies on School Size and Class Size • High School Size: Which Works Best and for Whom? ................... Section 9-1 • Teachers, Schools, and Academic Achievement………………..…. Section 9-27 • The Effects of Class Size on Student Achievement……………..…. Section 9-69 • Class Size and Teacher Quality…………………………………..….. Section 9-117 • Information from NEA on Class Size……………………………..….. Section 9-133 3 Page Numbers Section 10: The Excellence Factor Presentation on the Blacksmith Campaign……………………………………………………………………..… Section 10-1 Section 11: Block Schedule Information Block Scheduling in Large, Urban High Schools: Effects on Academic Achievement, Student Behavior and Staff Perceptions……..…. Section 11-1 (Information to be provided at meeting) Section 12: 2009 WA Parents’ Survey Summary Report………………. Section 12-1 3 Village Circle, Suite 202 ~Westlake, Texas 76262 Metro: 817-430-0941 ~ Fax: 817-430-1812 ~ www.westlake-tx.org MEMORANDUM Date: December 7, 2009 TO: Municipal Leadership Team Westlake Academy Leadership Team FROM: Tom Brymer, Town Manager/CEO Westlake Academy SUBJECT: Council Meeting Schedule and the Strategic Planning/Budget Calendar for Both Westlake Academy and the Town of Westlake The 2010 Board/Council meeting calendar has been compiled. We have incorporated, as much as possible, key dates for meetings focused on strategic planning and/or budgeting for both Westlake Academy and the Town. Here is some additional detail presented in a different format than what is on the proposed calendar: Westlake Academy Academic Services Budget and Strategic Plan • Dec. 2009- WALT meeting to discuss FY 10-11 Budget • Jan. 2010- meet with WA affiliates for joint planning for FY 2010 • Jan. 4, 2010 present draft WA Strategic Plan to BOT • Feb. 1, 2010- discussion with BOT of broad overview of WA’s 2010-2011 (early input, including review of updated WA financial forecast) • March/April 2010- parents’ survey • June 18, 2010 WA FY 10-11 Proposed Academic Services Budget and Strategic Planning Retreat (note: this is a Friday and would be from 9:00am to 1:00pm) • August 16, 2010- Board adoption of WA FY10-11 Academic Services Budget Town of Westlake Municipal Budget • Jan. 25, 2010- Review of Strategic Plan progress to date/discuss Citizens’ Survey • Feb. 22, 2010- Council ranking of municipal programs/services • May 12, 2010- New Council orientation • May 24, 2010- After election strategic planning retreat to preview the municipal budget (early input), Town financial forecast, and discuss progress and updates to the strategic plan • Aug. 13, 2010- Presentation of FY 10-11 proposed municipal budget in retreat (note: this is a Friday and would be from 9:00am-1:00pm) • Sept. 27, 2010- Council adoption of FY 10-11 municipal budget Section 1-1 Governance Structure MISSION / VISION Westlake Academy is governed by a six (6) member Board of Trustees comprised of the Board President and five (5) Trustees. These positions serve a dual role — as both the Academy’s governing board and the Town Council for the Town of Westlake. The Trustees serve two (2) year staggered terms of office and are responsible for the governance of the school, which includes adopting policies related to educational services and programs, approving the annual budget and appointment of members to the Westlake Academy Foundation. The Board also appoints the Chief Executive Officer of the school, who is responsible for the Academy’s daily operations, staff appointment/ management, plant maintenance, and financial administration. In addition, the CEO functions as the Town Manager for the Town of Westlake. VALUES (GUIDING PRINCIPLES) “Westlake Academy is a nurturing, community-owned International Baccalaureate charter school whose mission is to achieve academic excellence and to develop life-long learners who become responsible global citizens.” Academic success through the IB curriculum Student and family oriented environment Being a municipally owned and operated charter school Engaged students with international awareness Community involvement and support Diverse learning experience Section 2-1 The strategic plan is the result of many hours of review and discussion with the Board of Trustees, and Westlake Academy Staff, which spanned a year-long process where we gathered information, discussed our long-range goals and developed comprehensive guidelines to ensure student success. The Board met, during a planning retreat, to develop the mission / vision statement and identify the values that were important to the Academy. The plan was then expanded to include five (5) Desired Outcomes which the Staff used to develop Specific Outcomes, Indicators and Activities to accomplish these directives. Strategic Plan Overview Vision / Mission Values (Guiding Principles) Desired Outcomes Specific Outcomes Specific Outcomes are designed to support the five (5) main Desired Outcomes identified as important to the success of the students. Assigned completion dates for each Indicator Activities are actions Staff will complete to achieve the goal set out in the corresponding Indicator Team Leaders are identified to monitor our progress toward the stated goals Indicators are measurable statements/ activities designed to show that Specific Outcomes have been achieved. Specific Outcome = 1.0 Students will be well prepared to matriculate and succeed in college. Short-Range and On-going Goals Team Leader(s): Heads of Section Indicator = 1.1 100% of students are passing classroom exams Estimated Completion Date: On-going Goal Activity = 1.1.1 Continued integration of IB curricula into all daily educational programs Indicators Specific Outcomes are assigned a deadline relative to the time anticipated it will take to complete the task. Short Range = 1 - 2 years Mid-Range = 2 - 3 years Long Range = 3 - 5 years and On-going Section 2-2 Description of Desired Outcomes ~ High Student Achievement Develop inquiring, knowledgeable, caring and disciplined young people who use their unique talents to create a better and more peaceful world through intercultural understanding and respect ~ Strong Parent & Community Connections To involve all stakeholders in building a better Westlake Academy community ~Financial Stewardship & Sustainability To ensure sufficient, well-managed resources to support and ad- vance the mission of Westlake Academy ~ Student Engagement – Extracurricular activities To ensure that all athletics, community service and extra-curricular activities are held to the same standard of excellence as the curricula programs to promote well-balanced students ~ Effective Educators & Staff Recruit, develop and retain a core faculty and staff with the personal qualities, skills and expertise to work effectively with the IB inquiry-based, student-centered curricula Westlake Academy is the only school in the state of Texas to provide their students with all three IB programs — Primary Years, Middle Years and the Diploma Program; and one of five schools to offer all three in the United States. Provides numerous extra-curricular opportunities, in addition to the rigorous curriculum, such as athletics, drum line, music, yearbook, Black Cow journalism and newspaper publishing, historian and garden club, student council, science and history competition, Model United Nations, etc. Our first graduating class, 24 students, have received over $1 million dollars in scholarship offers! Rated as “Exemplary” by the Texas Education Agency for three (3) of the last seven (7) years Highlights of Westlake Academy PSAT Statistics for 2008/2009 Critical Reading 9th 10th 11th National Mean 41.6 41.6 46.7 Texas Mean 40.3 40.3 44 Westlake Students 48.9 50.6 56.5 Math 9th 10th 11th National Mean 44 44 48.9 Texas Mean 43.5 43.5 47 Westlake Students 47.2 51.3 59.2 Writing 9th 10th 11th National Mean 41 41 45.8 Texas Mean 40.4 40.4 44.4 Westlake Students 46.5 48.5 54.7 Section 2-3 WESTLAKE ACADEMY Budget Snapshot – FY 2009-2010 Expenditures Including Municipal Support Fiscal Year 09-10 Adopted Academic Services Budget $ 4,057,713 Municipal Funded Direct Operational Costs 248,638 Subtotal Direct Operation Expenditures 4,306,351 Municipal Funded Indirect Operational Costs (Support Services) 339,469 Subtotal All Operating Costs 4,645,820 Annual Debt Service Payment 1,499,751 Grand Total $ 6,145,571 In an effort to be as transparent as possible about the Town’s finances, including Westlake Academy, this “Budget Snapshot” shows all costs for FY 09-10 for Westlake Academy. These costs totaling $6.1 million include all academic services expenditures, debt service, facility maintenance costs, and support services costs. 66%4% 6% 24% Academic Services Budget Municipal Funded Direct Operational Costs Municipal Funded Indirect Operational Costs Annual Debt Service Payment 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Academy Expenditures 66% Municipal Indirect & Direct Expenditures 34% Section 3-1 WESTLAKE ACADEMY Academic Services Budget Only - FY 2009-2010 Includes All Funds This “Budget Snapshot” shows the breakdown of the Academy’s $4.0 Academic Services Budget and includes all teaching, counseling, athletic, administration, library, extra-curricular costs, insurance and utilities to fund a K-12 IB chartered school operation. 10% 3% 3% 84% Blacksmith Revenue $420,000 Federal Revenue $119,976 Local Revenue $135,464 State Revenue $3,637,109 76% 12% 6% 5% Payroll & Related $3,090,575 Contracted Services $506,290 Supplies & Materials $234,205 Other Operating $184,643 Debt Service/Lease $42,000 Expenditures Revenues Section 3-2 WESTLAKE ACADEMY – REVENUES (Excerpt from FY 09‐10 Academic Services Budget) $0 $100,000 $200,000 $300,000 $400,000 $500,000 FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09 $239,446 $335,727 $388,377 $411,505 $418,313 Blacksmith Program Contributions Revenue Sources Westlake Academy’s revenues to fund the Academic Services Budget come from two major sources. First, public education funding is provided by the State of Texas which accounts for 83% of this budget revenues. The funding is provided on a per student basis and is projected to be $7,279 per student for FY 2009-2010. The following chart shows the State’s per student funding level (Foundation state funds only) for the Academy since inception as well as the projected FY 2009-2010 and FY 2010-2011: Donations provided by the Westlake Academy Foundation’s Blacksmith Program constitute the second largest funding source for the Academic Services Budget which totals 10%. Absent other revenue sources, if it were not for this private donation support from the parents that participate in the Foundation’s Blacksmith campaign, the Academic Services Budget for Westlake Academy would not be fully funded and could not provide educational services at their current level. The Academic Services Budget receives no ad valorem revenues as the Town of Westlake does not levy an ad valorem tax. Since the inception of the Foundation’s Blacksmith program, annual amounts raised from the Academy parent body are as follows: $0$1,000$2,000$3,000$4,000$5,000$6,000$7,000$8,000 FY 03/04 FY 04/05 FY 05/06 FY 06/07 FY 07/08 FY 08/09 FY 09/10 $4,809 $5,729 $5,624 $6,530 $6,623 $7,057 $7,279 State Funding Per Student Section 3-3 WESTLAKE ACADEMY – REVENUES (Excerpt from FY 09‐10 Academic Services Budget) Academic Services Revenue Sources Estimated Adopted Actual Actual Budget Budget FY 06/07 FY 07/08 FY 08/09 FY 09/10 Local & Intermediate Sources Blacksmith Apprentice Program $ 398,780 $ 394,571 $ 420,000 $ 420,000 Investment Earnings 23,644 22,335 5,500 3,000 Lunchroom Revenues 67,165 23,572 5,750 8,600 Other Local Sources 22,746 78,583 105,661 75,858 Athletic Activities Income 44,444 48,006 Total Local & Intermediate Sources 512,335 519,061 581,355 555,464 State Revenues Foundation School Program 2,259,642 2,500,817 2,935,315 3,443,120 TRS - On-Behalf Payments 107,553 149,090 154,638 174,285 Total State Revenues 2,367,195 2,649,907 3,089,953 3,617,405 TOTAL REVENUES $ 2,879,530 $ 3,168,968 $ 3,671,308 $ 4,172,869 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% FY 06/07FY 07/08FY 08/09FY 09/10 Revenues by Percentage Other Local Sources Lunchroom Revenue Investment Earnings Blacksmith Apprentice Program State Revenues $0 $1,000,000 $2,000,000 $3,000,000 $4,000,000 $5,000,000 FY 06/07FY 07/08FY 08/09FY 09/10 2, 3 6 7 , 1 9 5 2, 6 4 9 , 9 0 7 3, 0 8 9 , 9 5 3 3, 6 1 7 , 4 0 5 Revenues by Dollars Other Local Sources Lunchroom Revenue Investment Earnings Blacksmith Apprentice Program State Revenues Section 3-4 COST PER STUDENT ACADEMIC SERVICES ONLY (Excerpt from FY 09‐10 Academic Services Budget) $‐ $1,000 $2,000 $3,000 $4,000 $5,000 $6,000 $7,000 $8,000 $9,000 FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 $7,016 $7,111 $6,944 $7,725 $8,721 $8,955 $8,579 Expenditures per Student (Academic Services Only) Section 3-5 COST PER STUDENT INCLUDING MUNICIPAL SUPPORT (Excerpt from FY 09‐10 Academic Services Budget) $‐ $2,500 $5,000 $7,500 $10,000 $12,500 $15,000 FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 $10,628 $11,930 $12,014 $12,305 $13,012 $13,865 $12,993 Expenditures per Student (Including Municipal Support) Section 3-6 FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 Grade Levels Taught1‐6K‐7K‐8K‐9K‐10K‐11K‐12 Total Students 195264324346379417473 Percentage of Increase 35%23%7%10%10%13% FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 Teachers 16.023.828.131.634.036.040.6 Admin/Support 4.55.56.56.59.012.213.1 Total 20.529.334.638.143.048.253.6 Percentage of Increase 43%18%10%13%12%11% FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 Primary‐Teachers 12.014.315.515.015.014.014.9 Primary‐Specialty Teachers 4.04.95.05.05.65.97.1 Total Primary 16.019.220.520.020.619.922.0 Secondary‐Teachers 4.37.611.613.415.118.6 Total Secondary 0.04.37.611.613.415.118.6 Primary‐Admin/Support 4.55.56.56.59.012.211.3 Secondary‐Admin/Support 0.00.00.00.00.01.01.8 Total Admin/Support 4.55.56.56.59.013.213.1 TOTAL 20.529.034.638.143.048.253.6 Employee Count by Section (Full Time Equivalents) Description Description Description Enrollment Employee Count (Full Time Equivalents) Enrollment and Academic Overview FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 Student/Teacher Ratio (1)12.1911.0911.5310.9511.1511.5811.66 Student/Teacher Ratio (2)16.2516.5018.0017.3017.2317.3818.19 9.519.109.369.088.818.658.82 FY 03/04FY 04/05FY 05/06FY 06/07FY 07/08FY 08/09FY 09/10 98%98%99%98%99%100% Writing97%97%99%91%93%100% Social StudiesN/AN/A99%95%98%100% Mathematics97%91%95%91%93%98% Science93%85%78%88%96%99% TEA School Rating ExemplaryRecognizedRecognizedRecognizedExemplaryExemplary Excerpt from FY 09‐10 Academic Services Budget Read/Eng Language Arts Ratios Description Student/Total Staff Ratio (2) Based on ALL teachers (including Specialty Teachers) (1) Based on home room class size Description Texas Assessment of Knowledge and Skills (TAKS) Enrollment and Academic Overview Section 3-7 Fiscal Year FY 03/04 Actual FY 04/05 Actual FY 05/06 Actual FY 06/07 Actual FY 07/08 Actual FY 08/09 Actual FY 09/10 Proposed Classes Served1-6K-7K-8K-9K-10K-11K-12 Head of School1.00 1.00 1.00 1.00 1.00 1.00 - Head of Primary - - - - 1.00 1.00 1.00 Head of Secondary - - - - - 0.50 1.00 Admin Coordinator - - - - - 1.00 1.00 PYP Coordinator - 0.30 0.50 0.50 1.00 0.50 0.60 MYP Coordinator - - 0.40 0.40 0.40 0.50 0.50 DP Coordinator - - - - - 0.50 0.50 Primary - K - 2.00 2.00 2.00 2.00 2.00 2.00 Primary 1 2.00 2.00 2.00 2.00 2.00 2.00 1.70 Primary 2 2.00 2.00 2.00 2.00 2.00 2.00 2.00 Primary 3 2.00 2.00 3.00 2.00 2.60 2.00 2.00 Primary 4 2.00 1.70 2.50 2.00 2.00 2.00 2.00 Primary 5 2.00 2.00 2.00 2.00 2.00 2.00 2.00 Primary 6 2.00 2.00 2.00 2.00 2.00 2.00 1.80 Primary - PE 1.00 1.00 1.00 1.00 1.00 1.00 1.00 Primary - Traveling - - - - - - 0.40 Primary - Art 0.80 0.80 0.60 0.80 0.80 0.80 0.90 Primary - Music 0.80 0.40 0.80 1.00 1.00 1.00 1.00 Primary - Spanish 1.00 1.00 1.00 1.00 1.00 1.00 1.20 Special Education - - 1.00 2.00 2.00 2.00 2.00 Secondary - Art - - 0.60 0.60 1.00 0.80 1.00 SdElih 100 200 200 200 300 Westlake AcademyPosition Summary Secondary English - - 1.00 2.00 2.00 2.00 3.00 Secondary - Humanities - 1.00 0.60 0.60 1.60 2.00 3.10 Secondary - IT - - - 0.30 0.30 0.30 0.30 Secondary - Math - 0.40 1.00 1.50 2.00 1.75 2.00 Secondary - Science - 0.50 1.00 1.50 2.00 2.50 3.25 Secondary - Spanish - 0.40 0.50 1.00 1.00 1.50 2.00 Secondary - PE - 0.25 0.40 1.00 1.00 1.25 0.65 Secondary - CAS - - - - 0.60 - 0.25 Strings Staff - - - - - - 0.25 Librarian - 0.40 1.00 1.00 1.00 1.00 1.00 Counselors - 0.50 1.00 1.00 1.50 2.00 2.00 Athletic Director - - - - - - 0.75 Teaching Aides - 0.60 - - - 2.50 1.75 Nurse 1.00 0.50 0.50 0.50 0.50 1.00 1.00 IT - - - 0.70 0.70 0.70 1.70 Asst. to Head of School 1.00 1.00 1.00 1.00 1.00 - - Registrar - 1.00 1.00 1.00 1.00 1.00 1.00 Office Aide - - 0.60 0.60 1.60 2.60 3.00 Lunchroom Personnel 0.25 0.40 0.40 0.40 0.40 0.40 0.40 Day Porter - - - - - - 0.60 Total Positions 18.85 25.15 32.40 36.40 43.00 48.10 53.60 Excerpt from FY 2009-2010 Academic Services Budget Position Summary Section 3-8 Years W e s t l a k e A c a d e m y B i r d v i l l e I S D C a r r o l l I S D G r a p e v i l l e - C o l l e y v i l l e I S D K e l l e r I S D N o r t h w e s t I S D A v e r a g e M e d i a n W A t o A v e r a g e W A t o M e d i a n 0 44,598 48,000 45,900 46,100 46,143 47,500 46,729 46,143 -4.6%-3.3% 1 45,263 48,200 46,247 46,247 46,693 47,750 47,027 46,693 -3.8%-3.1% 2 45,570 49,000 46,597 46,391 47,298 48,000 47,457 47,298 -4.0%-3.7% 3 46,163 49,380 47,105 46,564 47,908 48,250 47,841 47,908 -3.5%-3.6% 4 46,356 49,580 47,426 46,704 48,303 48,500 48,103 48,303 -3.6%-4.0% 5 47,130 49,780 47,737 46,871 48,819 48,750 48,391 48,750 -2.6%-3.3% 6 47,532 49,980 48,058 47,300 49,363 49,100 48,760 49,100 -2.5%-3.2% 7 47,884 50,180 48,380 47,491 49,663 49,450 49,033 49,450 -2.3%-3.2% 8 48,180 50,380 48,701 47,662 49,963 49,800 49,301 49,800 -2.3%-3.3% 9 48,339 50,580 49,022 47,890 50,263 50,150 49,581 50,150 -2.5%-3.6% 10 48,708 50,780 49,344 48,014 50,563 50,500 49,840 50,500 -2.3%-3.5% 11 49,111 50,980 49,665 48,344 50,863 50,900 50,150 50,863 -2.1%-3.4% 12 49,363 51,180 49,986 48,468 51,213 51,300 50,429 51,180 -2.1%-3.6% 13 49,542 51,380 50,308 48,669 51,527 51,700 50,717 51,380 -2.3%-3.6% 14 49,767 51,580 50,629 48,857 51,827 52,100 50,999 51,580 -2.4%-3.5% 15 50,466 51,780 50,951 49,150 52,613 52,500 51,399 51,780 -1.8%-2.5% 16 50,629 51,980 51,272 50,112 52,913 53,000 51,855 51,980 -2.4%-2.6% 17 50,903 52,180 51,599 50,456 53,513 53,750 52,300 52,180 -2.7%-2.4% 18 51,852 52,815 52,697 51,453 53,863 54,500 53,066 52,815 -2.3%-1.8% 19 52,823 53,930 53,454 52,584 54,388 55,250 53,921 53,930 -2.0%-2.1% 20 54,010 54,990 53,775 53,582 55,263 56,000 54,722 54,990 -1.3%-1.8% 21 54,902 55,865 54,756 54,632 56,138 56,750 55,628 55,865 -1.3%-1.7% 22 55,573 56,665 56,153 55,542 56,463 57,500 56,465 56,463 -1.6%-1.6% 23 55,995 58,768 57,306 56,343 57,163 58,250 57,566 57,306 -2.7%-2.3% 24 56,484 60,872 58,200 57,361 58,107 59,000 58,708 58,200 -3.8%-2.9% 25 58,226 62,975 58,675 58,267 58,777 59,750 59,689 58,777 -2.5%-0.9% 26 59,096 58,996 59,178 59,095 60,500 59,442 59,137 -0.6%-0.1% 27 59,712 59,318 60,141 59,764 61,250 60,118 59,953 -0.7%-0.4% 28 60,343 59,639 61,052 60,444 62,000 60,784 60,748 -0.7%-0.7% 29 60,443 59,960 61,962 61,107 62,750 61,445 61,535 -1.6%-1.8% 30 61,187 60,282 62,871 61,689 63,500 62,086 62,280 -1.4%-1.8% -2.3%-2.6% ISD SCALES FOR FY 20092010 SCHOOL YEAR APPROVED WESTLAKE ACADEMY SALARY SCALE AND ACTUAL Section 3-9 MUNICIPAL DIRECT AND INDIRECT COST INFORMATION (Excerpt from FY 09-10 Academic Services Budget) VII. General Academy Financial Structure and Facilities Overview The land and buildings that comprise the Westlake Academy campus are owned by the Town of Westlake, a municipality incorporated under State law as a Type A general law city. Westlake Academy opened in 2003 with Grades 1 through 6 and has added a grade each subsequent year. With the commencement of the 2009-2010 school-year, the final component will be added with the 12 Campus Facilities th grade inaugural graduating class. The campus is located on twenty three (23) acres adjacent to J.T. Ottinger Road, which is near the intersection of State Highways 114 and 170. The existing campus includes three (3) stand alone academic buildings with a total of twenty nine (29) classrooms, administrative offices, restrooms, a library, breakout area, a performance hall, dining and kitchen area, locker rooms, and a gym. These buildings total 61,000 square feet in size. The Academy interior design includes wood, vinyl covering, carpet and tile flooring. The hallways are lined with wooden lockers and carpet flooring. The Sam and Margaret Lee Arts & Sciences Center is scheduled to open in August of 2009. This new facility will add another 8,400 square feet of building space to the campus and will include, one art room, two science labs and five offices, a workroom, conference room, restrooms and breakout space. The Town of Westlake has issued certificates of obligation and general obligation bonds to fund the construction of the campus. Presently there is $33.7M (principal and interest) of outstanding bonded indebtedness to retire these bonds. Included in this total, is the 2008 issuance of $2.5 million in G.O. bond debt to fund a portion of the new $5.1 million Sam and Margaret Lee Arts & Sciences Building. Capital Costs/Debt Service Annual debt service payments are expensed to the Town’s municipal budget and the revenue stream presently utilized to make the annual debt service payment is comprised of municipal sales tax. However, the legal pledge securing these bonds, should this present sales tax revenue source not be adequate to retire this debt, is an ad valorem tax which the Town would be required to implement on all property within its corporate limits. The Town of Westlake levies no ad valorem tax and there is no direct limit on debt for the Town. The Constitution of the State of Texas provides that the ad valorem tax levied by the Issuer for general purposes and for the purpose of paying debt service requirements of the Issuer’s general obligation debt shall not exceed $1.50 for each $100 of assessed valuation of taxable property. While there are future facility needs, there are currently no buildings or large projects contained in the Town’s CIP (capital improvement plan) due to the unavailability of funding. In addition to owning the Westlake Academy campus facilities, the Town of Westlake is responsible for the daily operations of the campus. Direct Operating Costs Direct costs to operate the campus are Section 3-10 expensed in two separate budgets. One budget is the Academic Services Budget, the subject of this transmittal letter, which is on a September 1st to August 31st fiscal year. These costs include all teaching, counseling, athletic, administration, library, extra-curricular costs, insurance and utilities to fund a K-12 IB chartered school operation. The second budget contains operational expenses related to maintaining the campus facility that are expensed in the Town’s municipal budget which is on an October 1st to September 30th fiscal year. Further, the Town of Westlake provides various administrative support services for the Academy to avoid duplication of costs. This fact was used as part of the charter application process with the TEA to help bolster the Town’s case for having a community school without duplicating cost structures and required resources. These indirect operating costs are contained in the Town’s municipal budget and are estimated to be approximately $330K in FY 09-10. These support services for the Academy that generate indirect costs paid by the Town of Westlake are: Indirect Operating Costs • Human resources • Information technology • Risk management • Finance and accounting services • Facility maintenance services • General administrative services (CEO) • Board support services including policy advisement and strategic planning (CEO) • Official Board records maintenance and election administration (Town Secretary) As other indices in this letter show, such as the growing Academy’s enrollment and staffing levels, this has created an increased demand for the level of support services that the Town provides the Academy. In summary, the Financial Structure Summary full cost structure for Westlake Academy for FY09-10, including the adopted FY 09-10 Academic Services Budget, is: Adopted FY 09-10 Adopted Academic Services Budget $ 4,057,713 Estimated Town Funded Direct Operational Costs 248,638 Subtotal Direct Operation Expenditures 4,306,351 Estimated Town Funded Indirect Operational Costs (Support Services) 339,469 Subtotal All Operating Costs 4,645,820 Annual Debt Service Payment 1,499,751 Grand Total $ 6,145,571 Section 3-11 Pymt #Fiscal Year Principal Interest Total 1 2010 525,000 973,250 1,498,250 2 2011 555,000 946,778 1,501,778 3 2012 580,000 918,450 1,498,450 4 2013 610,000 888,528 1,498,528 5 2014 640,000 856,906 1,496,906 6 2015 675,000 823,442 1,498,442 7 2016 710,000 787,864 1,497,864 8 2017 740,000 760,197 1,500,197 9 2018 775,000 730,089 1,505,089 10 2019 810,000 691,257 1,501,257 11 2020 850,000 649,870 1,499,870 12 2021 895,000 606,461 1,501,461 13 2022 940,000 560,744 1,500,744 14 2023 985,000 512,719 1,497,719 15 2024 1,040,000 462,385 1,502,385 16 2025 1,090,000 408,530 1,498,530 17 2026 1,130,000 362,154 1,492,154 18 2027 1,185,000 314,082 1,499,082 19 2028 1240000 263668 1503668 Debt Service Fund Long-Term Debt Summary All Bonds 19 2028 1,240,000 263,668 1,503,668 2020291,110,000 209,945 1,319,945 2120301,155,000 161,095 1,316,095 2220311,210,000 110,060 1,320,060 2320321,260,000 56,160 1,316,160 20,710,000$ 13,054,629$ 33,764,629$ Excerpt from FY 09-10 Town of Westlake Budget TOTAL ‐ 200,000 400,000 600,000 800,000 1,000,000 1,200,000 1,400,000 1,600,000 20 1 0 20 1 2 20 1 4 20 1 6 20 1 8 20 2 0 20 2 2 20 2 4 20 2 6 20 2 8 20 3 0 20 3 2 Annual Funding Requirements Interest Principal Section 3-12 Westlake Academy Activity Budget Percent PYP Program 1 General Instruction 832,204 20.13% 1 PE 116,859 2.83% 2 Supplies & Operations 556,171 13.45% 3 Administration 150,181 3.63% 4 Support Services 128,366 3.10% 4 Spanish 69,263 1.68% 5 Music 49,876 1.21% 6 Art 70,683 1.71% 7 Library 37,917 0.92% 8 Strings 7,143 0.17% $ 2,018,662 48.82% MYP Program 1 Art 22,767 0.55% 1 Humanities 72,655 1.76% 1 Information Technology 23,417 0.57% 1 Language Arts 131,948 3.19% 1 Math 81,647 1.97% 1 Science 105,549 2.55% 1 Spanish 111,816 2.70% 1 PE/Athletics 66,776 1.61% 2 Library 21,726 0.53% 3 Supplies & Operations 374,995 9.07% 4 Administration 136,289 3.30% 5 Support Services 73,553 1.78% $ 1,223,140 29.58% Tom Brymer 817-490-5720 Program Department Director Department Phone Section 3-13 DP Program 1 Humanities 36,327 0.88% 1 Information Technology 11,709 0.28% 1 Language Arts 65,974 1.60% 1 Math 40,823 0.99% 1 Science 52,775 1.28% 1 Spanish 55,908 1.35% 1 Supplies & Operations 169,584 4.10% 2 Administration 68,043 1.65% 3 Support Services 33,263 0.80% 4 Library 9,812 0.24% 5 Art 11,384 0.28% $ 555,601 13.44% Special Education 1 Occupational Therapy (Social Skills) 18,000 0.44% 1 Speech Therapy (Social Skills) 25,000 0.60% 1 Inclusion (Mainstream/Social Skills) 81,620 1.97% 1 Resource 13,518 0.33% 1 Compliance 56,064 1.36% 1 Student Assessment 31,600 0.76% 1 Supplies 500 0.01% $ 226,302 5.47% Information Technology 1 IT Support 22,217 0.54% 1 Windows/Mac Support 33,325 0.81% 1 Enterprise Support 11,108 0.27% 1 Data Center 11,108 0.27% 1 Website Services 11,108 0.27% 1 LAN/WAN Operations 11,108 0.27% 1 Physical & Electronic Network Security 11,108 0.27% $ 111,083 2.69% 4,134,788$ 100% Section 3-14 Ce n s u s of St u d e n t s Ce n s u s of St u d e n t s 9% 17 % 9% 32% 55 % 68 % 19 % 68 % Ke l l e r IS D Ca r r o l l IS D No r t h w e s t IS D Ot h e r IS D ' s We s t l a k e Re s i d e n t s St u d e n t s Re s i d i n g Ou t s i d e of Westlake 02 / 0 1 / 2 0 1 0 BO T Me e t i n g Se c t i o n 4- 1 Total#%#%#%#%#%#%#%#% GradeStudents K301033%00%1343%310%413%2067%27%13% 130723%27%1240%310%620%2377%517%13% 2372054%00%514%1027%25%1746%00%13% 3361542%13%822%822%411%2158%13%13% 4371746%25%1027%719%13%2054%25%00% 5361233%26%1131%822%38%2467%26%00% 6361336%514%1131%719%00%2364%26%00% 7391333%25%1333%923%25%2667%25%13% 836617%617%1131%1233%13%3083%26%00% 93539%720%1131%926%514%3291%13%00% 1036925%411%1644%411%38%2775%26%26% 1129310%27%1345%621%517%2690%00%13% 417128 31%33 8%134 32%86 21%36 9%28969%215%82% *Westlake resident students not included in this list of Studentsof Students WestlakeCarrollKellerNorthwestOther SECONDARY BOUNDARIESPRIMARY of Studentsof Studentsof Studentsof Studentsof Studentsof Students Sub-TotalDue to SchoolDue to Town SecondaryEmployeeEmployeeDistrictsISD* ISD* ISD* ResidentsFY 2008-2009 Included in appropriate District WESTLAKE ACADEMY STUDENT CENSUS -May 2009 Excerpt from FY 09-10 Academic Services Budget WESTLAKE ACADEMY STUDENT CENSUS -May 2009 Carroll, 8% Keller, 32% Northwest, 21% Other, 9% Westlake, 31% Section 4-2 Ce n s u s of St u d e n t s Ce n s u s of St u d e n t s 2% 36 24 G1 1 G1 2 8% 39393536 G8G9 G1 0 G1 1 40404139 G5G6G7G8 90 % 373944 G2G3G4 To w n St a f f Ch i l d r e n Sc h o o l St a f f Ch i l d r e n Op e n En r o l l m e n t Ch i l d r e n 3439 KG1 02 0 4 0 6 0 02 / 0 1 / 2 0 1 0 BO T Me e t i n g Op e n En r o l l m e n t Ch i l d r e n Se c t i o n 4- 3 En r o l l m e n t Le v e l s En r o l l m e n t Le v e l s To t a l St u d e n t s 40 0 50 0 34 6 37 9 41 7 47 3 (Actual –487) 30 0 19 5 26 4 32 4 34 6 10 0 20 0 0 20 0 3 ‐20 0 4 1‐6 20 0 4 ‐20 0 5 K‐7 20 0 5 ‐20 0 6 K‐8 20 0 6 ‐20 0 7 K‐9 20 0 7 ‐20 0 8 K‐10 20 0 8 ‐20 0 9 K‐11 2009‐2010 K‐12 02 / 0 1 / 2 0 1 0 BO T Me e t i n g Se c t i o n 4- 4 Section 5-1 American Federation of Teachers Educational Foundation 555 New Jersey Avenue, N.W. Washington, DC 20001-2079 Authors of the Report F. Howard Nelson Edward Muir Rachel Drown Project Officer Duc -Le To, Institute of Education Sciences U.S. Department of Education Washington, DC 20202 May 2003 This report was prepared for the Office of Educational Research and Improvement, U.S. Department of Education under Contract Number ED98-CO-0029. The views expressed herein are those of the contractor. No official endorsement by the U.S. Department of Education is intended or should be inferred. Permission is hereby granted to reproduce and distribute copies of this work for nonprofit educational purposes, provided that copies are distributed at or below cost, and that the author, source and copyright notice are included on each copy. The full text of this publication is available at www.aft.org. For copies of this report (limited number of copies available) call 202/879-4428. Section 5-2 Texas 209 TEXAS In addition to base funding, a student weighting system provides extra funds for special education, low- income students, limited-English proficiency and other high -cost programs. No dedicated facilities funding exists for charter schools. In 1997-98, the typical charter school enrolled 203 students, of which 36 percent were low-income students compared to 62 percent in host school districts. Limited-English proficient students made up only 5 percent of charter school enrollment, one-fourth the percentage of host school districts. Special education students comprised 4.4 percent of enrollment in the typical charter school compared to 11.3 percent in the school districts. Charter schools averaged $99 per pupil (in membership, not just special education) in spending to support special education, while host school districts spent $596 per pupil. Charter schools spent $900 per pupil on administration, twice the $485 spent by school districts. Charter school instruction costs averaging $2,716 per pupil, however, fell significantly below the $3,346 per pupil spent on instruction by host school districts. The pupil-to -teacher ratio averaged 22.7, six students more per teacher than in host school districts. Charter school federal revenue fell short of that for host school districts by about $100 per pupil. Total host school district revenue, averaging $5,908 per pupil, topped total charter school revenue of $5,121 per pupil. Differences in spending for special education, transportation and food service explain the entire revenue differential. The average fund balance totaled 12.6 percent of revenue. See end of chapter for a more extensive summary. Texas enacted charter school legislation in 1996 and has amended it several times since. Between the 1997-98 and 1999-2000 school years, the number of charter schools increased six- fold to a total of 168. The Texas accountability and data collection system provides detailed data for both school districts and charter schools that allows researchers to address finance issues more thoroughly than in any other state in our study. All Texas charter schools typically receive five-year charters and must participate the comprehensive school reporting system known as the Public Education Information Management System (PEIMS). Texas provides one model of how to match funding to the specific educational needs of students; the state also demonstrates how financial inequalities among school districts result in financial inequalities among charter schools. Several pre-existing nonprofit community organizations not only sponsored charter schools, but also provided management and financial services to many of the charter schools. This chapter profiles the revenue, expe nditures and related student and staffing data for the 19 open-enrollment charter schools in Texas that were in operation during 1997-98. Texas provides for both “open enrollment” and “campus” charter schools. Authorized by school districts, campus charter schools obtain funding through the normal budget allocation process of the Section 5-3 Paying for the Vision: Charter School Revenue and Expenditures 210 school district.264 Open-enrollment charter schools resemble the autonomous charter schools in many other states and accounted for all but one of the Texas charter schools in 1997-1998. The state board of education approves open-enrollment charter schools, and nearly any entity except for-profit organizations can apply. As open-enrollment charter schools are fiscally and legally autonomous from districts, they therefore report to and are monitored by Texas Education Agency (TEA) staff. TABLE 13.1 CEI and Foundation Allowance in Host School Districts in Texas, 1997 -98 Cost Foundation Index (CEI) Method 1 Method 2 Austin 1.0355 $ 4,647 $ 3,734 Corpus Christi 1.0390 2,272 3,866 Dallas 1.0568 4,053 3,716 Flour Bluff 1.0443 2,203 3,794 Houston 1.0603 3,291 3,605 Irving 1.0497 3,662 3,851 Mission 1.0674 860 3,883 North East 1.0390 3,833 3,882 San Antonio 1.0497 1,757 3,875 Spring Branch 1.0568 4,663 3,908 Waco 1.0390 2,067 3,868 Note: Method 1 is highly correlated with school district wealth. The student weighting system does not apply to school districts and charter schools electing to use method 1. Funding for charter schools is equa l to the revenue that each student would have generated at his or her regular public school district of residence under the state’s school finance system. Geographic location and student characteristics are important determinants of charter school funding in Texas. Revenue for charter schools is based on the foundation allowance generated in the school districts in which students reside, and charter schools and school districts can choose between either a foundation allowance based mostly on property wealth (referred to, in Texas, as method 1) or a more equalized foundation allowance to which a student weighting system applies (method 2). Wealthy school districts, and charter schools drawing students from those districts, tend to benefit from method 1. The foundation allowance, averaging about $4,300 in 1997-98, incorporates a number of funding variables unique to each school district, including property wealth, tax effort and adjustments for district size and cost of education differentials.265 In Texas, 80 percent of school districts have a foundation allowance between $3,900 and $5,400. 264 Campus charter schools must negotiate the details of their charters with their districts, including fiscal autonomy. Fundin g, which is determined by negotiations with the district, flows through the district to the school. Teachers in campus charter schools remain employees of the district, and the schools do not have legal autonomy from their granting district. Although districts can grant an unlimited number of campus charters, only four districts have done so. All except two of the 24 campus charter schools (as of 1999-2000) are located in Dallas and Houston. In general, the Texas Education Agency (TEA) monitors or provides ongoing assistance to these schools in the same way as any other public school rather than as charter schools. 265 See Nelson, Muir and Drown (2000) for a more complete description of the financial incentives in the Texas charter school funding system. Section 5-4 Texas 211 As shown in Table 13.1, the foundation allowance for school districts in which the 1997-98 charter schools were located varied from $3,605 in Houston (method 2) to $4,663 in Spring Branch (method 1). Houston is a relatively poor school district (though five other school districts in Table 13.1 had less wealth per pupil), while Austin is wealthy. Very wealthy school districts that do not happen to qualify for state aid do not lo se their foundation allowance to charter schools. Austin joined this category in 1999-2000. In effect, wealthy school districts lose no revenue to charter schools. The base funding formula for school districts accounts for geographic variations in resour ce costs using a Cost of Education Index (CEI). The CEI is incorporated into the foundation allowance. The CEI for host school districts ranged from 1.0355 (3.55 percent more) in Austin (American Institute for Learning and Texas Academy of Excellence) to 1.0674 (6.74 percent more) in Mission (One -Stop Charter School). In 1997-98 all charter schools were located in school districts that did not benefit from district size adjustments.266 Under method 2, the formula has weights for special education, limited-English proficiency and various at-risk factors based on characteristics of students enrolled in charter schools.267 Student weighting adds approximately 17 percent to funding in the typical school district, which increased foundation support to an average of approximately $5,100 per pupil in 1997-98. Although a charter school receives all of its funding from the state, the state indirectly recovers an amount equivalent to the weighted foundation allowance from the sending school district because the student no longer counts as a student in that district. As noted above, the state does not recover charter school costs from the very wealthy school districts that do not qualify for state foundation aid. Study Design and Data Texas has a comprehensive system of school district reporting for student and staffing information, finances and student achievement. Open enrollment charter schools participate in this system on the same basis as school districts. Staffing and demographic data are also derived from the fede ral Common Core Data (CCD) collection.268 Charter schools are compared to host school districts—those districts in which the charter school building is physically located. Since charter school students come from many different school districts, the compariso n to a host school district is not financially exact. The charter school is the unit of analysis in our study, so small schools get the same weight as big schools. 266 The small district (fewer than 1,600 students) adjustment factor depends on enrollment, grade taught and whether the school district area exceeds 300 square miles. A mid -size district (between 1,600 and 5,000 students) adjustment factor is based on enrollment alone. Funds generated by district size adjustments are passed on to charter schools. Flour Bluff (Seashore Learning Academy), which enrolled slightly more than 5,000 students, was the smallest school district with a charter school in 1997-98. 267 Texas supplies an online tool for calculating charter school funding in specific school districts based on the educational and demographic characteristics of student enrollment. The Web site is: www.tea.state.tx.us/schoolfinance/funding/charter.html. 268 See http://nces.ed.gov/ccd/. Section 5-5 Paying for the Vision: Charter School Revenue and Expenditures 212 Despite the uniform financial reporting requirements, some charter schools are omitted fro m some of the analyses due to poor data.269 Charter schools are also required to submit independent financial audits. One year after the close of the 1998 fiscal year, four of 19 schools in our study had not submitted their audits.270 Two other schools were excluded from our study because the charter school audits were too intertwined with their nonprofit sponsoring organizations.271 Subsequent to 1997-98, Texas developed a more stringent system of financial oversight for its open-enrollment charter schools.272 In 1999-2000, TEA assigned three employees to its charter school audit department, which deals solely with charter schools. A school having documented fiscal problems is assigned a “financial monitor,” who closely regulates its accounting and reporting.273 In addition to these mechanisms, the TEA’s child nutrition department conducts audits to ensure schools are accurately counting and reporting students who are eligible for free and reduced-price lunch. In other parts of the National Charter School Finance Study, Policy Studies Associates, Inc. conducted case study research based on interviews with key staff and visits to 18 charter schools and host school districts in six states (including Texas). Interviews were also conducted with state agency staff, charter school authorizers and charter school associations. Fox River Learning, Inc. produced in-depth analyses of charter school and school district expenditures based on general ledgers. Though not the focus of our report, the PSA and FRL projects informed the conceptualization and analysis in this chapter. Student and Staffing Profiles Like school district funding, Texas charter school funding provides extra revenue for high attendance rates, special education, limited-English proficiency and at-risk factors based on the characteristics of students attending charter schools. This section begins by comparing the student demographics of charter schools and host school districts. (See Table 13.2.) Enrollment and attendance. Unlike most states, Texas funds charter schools based on average daily attendance (ADA), the same method used to fund school districts, so schools with low attendance rates receive less funding. It is usually assumed that low-income and at-risk students have the highest absentee rates, although many schools enrolling a large proportion of low- income students have very good attendance rates. ADA funding discourages the establishment of charter schools that would serve at-risk children and, in some cases, leads to funding problems for charter scho ols that choose to enroll these students. Attendance averaged 90.1 percent across the 19 charter schools compared to 94.7 percent in host school districts. So based on lower attendance, the average charter school received approximately 4.6 percent less funding— 269 West Houston Charter School classified 97 percent of its expenditures as instructional (compared to an average of 50 percent in other charter schools) and none as administrative (compared to 17 percent in other charter schools). One-Stop Charter School did not report revenue. Building Alternatives Charter School reported revenue data for half a year. 270 Renaissance, Sanchez, Texas Academy of Excellence and Waco. 271 Building Alternatives had no revenue or fund balances that can be separated from its sponsor. The University of Houston Charter School of Technology submitted unaudited fund transactions, although the school is technically part of the state government’s audit of the University of Houston. 272 Campus charter schools are monitored by their sponsoring district, rather than by TEA. 273 As of March 2000, there were two open-enrollment charter schools with financial monitors. Section 5-6 Texas 213 averaging $215 per pupil—than host school districts. Most of the attendance differential, however, occurred in only a few charter high schools for at-risk students.274 TABLE 13.2 School and Student Characteristics in Texas, 1997-98 Charter Host Schools Districts Enrollment 203 105,338 Average daily attendance (ADA) 90.1% 94.7% Host district enrollment > 10,000 95.0% na Special education students Emotionally impaired 0.4% 1.0% Learning disabled 2.1% 6.0% Speech 0.6% 2.0% Severe/other 1.0% 2.4% Total 4.4% 11.3% LEP students 5.4% 20.4% Economically disadvantaged 35.5% 62.3% Career & technology education 23.4% 15.8% FTE teachers Pre-K 1.5% 2.1% Kindergarten 8.5% 6.5% Elementary 28.3% 41.6% Secondary 47.8% 31.7% Ungraded 14.0% 18.5% Total 100.0 100.0 Pupil-to -teacher ratio 22.7 16.6 FTE staff per 100 students Teachers 5.6 6.1 Instructional aides 0.6 1.0 Instructional supervisors 0.1 0.0 Guidance counselors 0.4 0.2 Library/media 0.0 0.2 Administrators 0.9 0.3 Admin. support staff 0.6 0.5 Student support staff 0.3 0.8 All other support staff 1.6 3.4 Note: Data from federal CCD and the Texas Education Agency internet site. A pupil weighting system provides more funding for a variety of special needs, including speech, resource rooms, and self-contained settings. § Speech and similar low-intensity categories are weighted an additional 0.16 students (totaling 1.16 students). § Learning disabilities and similar categories receive an additional 85 percent weighting (totaling 1.85 students). 274 American Institute for Learning, Blessed Sacrament Academy, Building Alternatives and One-Stop had attendance rates of less than 80 percent. Section 5-7 Paying for the Vision: Charter School Revenue and Expenditures 214 § Auditory, visual, orthopedic, autistic, mental retardation and similar high-intensity categories are weighted an additional 1.20 students (totaling 2.20 students). § Mainstreamed special education students are weighted as an additional 1.1 students (totaling 2.1 students). From a legal perspective, charter schools must provide necessary special education services as required by state and federal special education rules. In practice, charter schools have limited resources (human and financial) and often seek to use inclusion only, rather than a range of services. The higher levels of funding for more costly special education helps encourage charter schools to provide a broader array of special education services. On the other hand, special education funding seldom covers the total costs for either charter schools or schools districts. Special education. Even with the funding weights, charter schools in 1997-98 were much less likely to serve special education students than host school districts. Special education students made up 4.4 percent of charter school enrollment compared to 11.3 percent in host school districts. Eight charter schools reported having no special education students. The gap between charter schools and school districts existed in low-cost special needs, like learning disabilities, as well as high-cost areas. When West Houston Charter School is omitted from the averages, a school that serves one -third of its students in spec ial education, the remaining schools average a special education population totaling 2.6 percent of enrollment.275 High schools typically serve fewer students in special education. The relatively large number of charter high schools in Texas explains some of the divergence between charter schools and school districts during 1997-98. Costs associated with special education students account for about $500 of the expenditure disparity between charter schools and host school districts. (See Table 13.7.) Charter schools spent an average of about $100 per pupil (pupils enrolled, not just special education students) compared to almost $600 per pupil in host school districts. Stated differently, charter schools spent about $2,200 per special education student and school districts spent about $4,800 per special education student. However, only four of 19 charter schools reported any spending on special education.276 Limited-English proficiency. Limited-English proficient (LEP) students in Texas secure 10 percent extra funding through the weighting system in the funding formula. Charter schools averaged a total LEP enrollment of about 5 percent compared to 20 percent in host school districts. Only four of 19 charter schools reported having LEP students, with SER-Niños Charter School reporting that four out of five students were LEP. Charter school spending on bilingual education averaged $47 per pupil (pupils enrolled, not per LEP student) compared to $325 per pupil in host school districts. (See Table 13.7.) Low-income students. Compensatory education students in Texas receive 20 percent extra funding through the weighting system. Eligibility is based on participation in the free and reduced-price lunch program. The average charter school reported 35.5 percent of enrollment as 275 The “severe/other” average drops from 1.0 to 0.2 percent. 276 Boys and Girls Prep, Building Alternatives, SER-Niños, and West Houston. Section 5-8 Texas 215 economically disadvantaged compared to 62.3 percent in host school districts. Six charter schools reported an economically disadvantaged student enrollment rate of less than 5 percent.277 Charter schools spent an average of $250 per student (pupils enrolled, not just disadvantaged students in programs) in programs designed to accelerate learning of underachieving students, while school districts spent $541 per student in these programs. (See Table 13.7.) Vocational education. Students served in career and technology programs obtain 67 percent more funding than regular students through the Texas financing system. Concentrated in high schools--and since many Texas charter schools in 1997-98 were high schools with vocational, career-based or alternative education programs --it is not surprising that charter schools averaged 23.4 percent of enrollment in career and technology programs compared to the host school district average of 15.8 percent. Seven charter high schools had career and technology programs, with Dallas Can! and Renaissance enrolling their entire student body in these programs. Grade level. Neither school districts nor charter schools generate funding based on grade level. While it is generally believed that high school costs more than other grade levels, Texas does not collect data in a way that enables measuring the cost differential. Despite the financial disincentive, charter high schools and middle schools were more common than elementary schools in 1997-98. Fully 48 percent of Texas charter school teachers taught in middle or high schools compared to 32 percent in host district schools. The career and technology funding weight indirectly boosts funding for charter high schools and other public high schools in Texas perhaps serving as an individual grade level funding adjustment. In 1997-98, the charter schools in Texas numbered 19. In 1998-99, 89 charter schools operated for the entire school year, 45 of which were designated to serve at-risk students (not necessarily low-income students). The TEA-sponsored evaluation provides some idea of how charter school student demographics changed (TEA, 2000). In 1999-2000, special education students grew to 8.5 percent of enrollment (from 4.4 percent shown in Table 13.2), but LEP students represented only 3.4 percent of students. At least two -thirds of charter schools offered high school grades (many in K-12 configurations), but there probably was a shift to less high school students and more elementary students as the number of charter schools great ly expanded during those years. Pupil-to-teacher staffing ratios. Even though charter schools have a reputation for smaller class size, the average pupil-to-teacher ratio (an indirect measure of class size) in 1997-98 was 22.7 in charter schools compared to 16.6 students per teacher in host school districts. With 0.6 instructional teacher aides for every 100 students, charter schools were less likely than host school districts (1.0 aides per 100 students) to use teacher aides. One reason for the lower pup il- to-teacher ratio and greater use of teacher aides in host school districts may be their greater levels of special education services. Reflecting the large proportion of charter high schools in Texas, charter schools were twice as likely to employ guidance counselors (0.4 counselors per 100 students) as host school districts. Due to size and the resulting diseconomies of small scale, charter schools employed nearly one administrator per 100 students—three times the level of host school districts, which employed only one administrator per 300 students. Administrative 277 North Hills, Pegasus, Renaissance, Seashore, University of Houston and West Houston. Section 5-9 Paying for the Vision: Charter School Revenue and Expenditures 216 support staff, however, numbered about the same in both charter schools (0.6 Full-Time Equivalent per 100 students) and host school districts (0.5 FTE per 100 students). On the other hand, hos t school districts employed 3.4 FTE support staff for every 100 students, compared to just 1.6 FTE support staff per 100 students in charter schools. The large discrepancy is a product of many factors, one being that charter schools may be more likely to contract for transportation, food service and janitorial services. Some of the difference may also be explained by the use of volunteer workers such as parents. Perhaps more important, charter schools are less likely to provide school breakfast and lunch programs and student transportation. Charter schools spent $106 per student on food service compared to $307 per student in host school districts. (See Table 13.6.) Transportation costs averaged $61 per student (enrolled, not transported) in charter schools compared to $134 per student in school districts. TABLE 13.3 Teacher Characteristics in Texas, 1997 -98 Charter Host Schools Districts Teachers by highest degree held No degree 4% 1% Bachelor’s 75% 65% Master’s 16% 32% Doctorate 5% 1% Teachers by years of experience Beginning teachers 20% 7% 1-5 years 51% 26% 6-10 years 9% 17% 11-20 years 16% 27% Over 20 years 3% 23% Years of experience 4.9 12.3 Average actual salaries Teachers $ 25,646 $ 34,851 Professional support $ 27,836 $ 42,949 Campus administration $ 44,228 $ 54,734 Central administration $ 47,996 $ 73,370 Turnover rate of teachers 50% 13% Teachers by program Regular education 90% 61% Special education 0% 11% Compensatory education 2% 5% Career & technology education 3% 4% Bilingual/ESL education 3% 13% Gifted & talented education 1% 3% Other 0% 2% Note: Data from federal CCD and the Texas Education Agency internet site. Teacher characteristics. The Texas data collection system provides information on teachers that relates both directly and indirectly to personnel costs. (See Table 13.3.) Among charter school teachers, 21 percent have at least a master’s degree compared to 33 percent of teachers in host school districts. Similarly, one in five charter school teachers had no teaching experience prior to 1997-98 compared to 7 percent of teachers in host school districts; and approximately three of four teachers in a typical charter school had no more than five years of teaching experience, Section 5-10 Texas 217 while the same is true for only one in three teachers in a typical host school district. As might be expected, turnover was higher in the younger, less experienced charter school teaching force with nearly half leaving at the end of the year compared to an average of just 13 percent of teachers in host school districts who left the building they had been teaching in at year end. An additional reason for a high rate of charter school teacher turnover may be low salaries. Legislation exempts charter schools from the state’s minimum salary schedule. Teacher salaries in the typical charter school averaged $25,600, $9,000 less than in host school districts. The salary differential is evident with other charter school personnel as well, the gap being about $15,000 for professional support, $10,000 for campus administration and $30,000 for central administration. Nine of 10 charter school teachers were regular education teachers compared to six in 10 teachers in host school districts. Statewide, none of the charter school teachers were classified as special education teachers; however, an average of 11 percent in host school districts were so classified. Similarly, only 3 percent of charter school teachers were bilingual teachers compared to an average of 13 percent of teachers in host school districts. In part, charter school teacher characteristics in 1997-98 are a reflection of the fact that these were fairly new schools. Over time, however, the gap between charter school and school d istrict teachers characteristics regarding levels of education, experience, pay and turnover may close. Some staffing differences may have to do with the mission of many charter high schools whereby they try to provide something different and appealing for students who have been failing in regular public schools. But the growth in the number of charter schools from 19 to 89 between 1997-98 and 1998-99 resulted in few changes in teacher characteristics. In fact, according to the TEA-sponsored evaluation (TEA, 2000), the number of charter school teachers without degrees increased from 4 percent to 11 percent; and even though average experience increased from 4.9 to 5.3 years, average teacher pay of $26,044 increased just slightly from $25,646 a year earlier. Financial Position Texas charter schools are required to submit independent financial audits. The following analysis of fund balances depends on these audits. Table 13.4 shows charter schools revenue, expenditures and the change in fund balances for all funds combined (restricted and unrestricted funds, including all spending for facilities and debt). Revenue in 1997-98 averaged $5,169 per pupil and exceeded expenditures by $369 per pupil (7.1 percent of revenue). Excess revenue varied considerably among the charter schools.278 Revenue fell short of expenditures by $864 per pupil at the Medical Center Charter School in Houston and by $498 per pupil at Boys and Girls Preparatory Academy.279 The average fund balance increased to 15.8 percent of expenditures. These two schools, along with Blessed Sacrament Academy, were also the only three schools 278 Revenue exceeded expenditure by more than $1,000 per pupil in North Hills, SER-Niños, and Seashore charter schools. 279 Due to a large transfer to the parent organization, the ending fund balance decreased to $463 per pupil at Academy of Transitional Studies but still stood at 62 percent of expenditures. Section 5-11 Paying for the Vision: Charter School Revenue and Expenditures 218 with a negative fund balance exceeding $50 per pupil. Five others had fund balances exceeding 40 percent of 1997-98 expenditures.280 TABLE 13.4 Revenue, Expenditures and Fund Balances (Per-Pupil Amounts) in Texas, 1997 -98 Average Min Max 1996-97 fund balance $ 274 $ -1,617 $ 3,032 Revenue (and other sources) 5,169 4,060 7,407 Expenditures (and transfers) 4,800 3,709 6,101 Excess of reve nue over expenditures 369 -864 1,826 1997-98 fund balance 649 -2,250 3,457 As percent of spending 15.8% -37.8% 61.9% Note: Minimums and maximums do not add up to totals. Data from audited financial statements of 13 charter schools. Audits not available for Renaissance, Sanchez, Texas Academy of Excellence, University of Houston and Waco. Also excludes Building Alternatives. Revenue Despite an equalization aid formula that has gradually reduced spending inequalities in recent years, Texas still has substantial spending inequities based on property wealth and tax effort. Charter school funding reflects school district spending inequalities. (See Table 13.1.) The revenue data shown in Table 13.5, which includes both unrestricted and restricted funds including support for debt and facilities, is derived from the Texas uniform finance reporting systems. The data indicate that average charter schools received $5,121 per pupil, about $785 less than host school districts. Charter school state funding totaled $4,414 per pupil (86 percent of revenue), which includes the foundation allowance (See Table 13.1) of revenue which is increased by student weightings. Federal grants add $475 per pupil (9.3 percent) to charter school funding compared to $568 per pupil (9.6 percent of revenue) in school districts. “Other local” funding, composed of donations, contributions, private foundation grants and income from enterprises averaged $232 per pupil in charter schools, or 4.6 percent of revenue, and accounted for $321 per pupil of host school district revenue. 280 One-Stop (40 percent), North Hills (46 percent), SER-Niños (48 percent), Academy of Transitional Studies (57 percent) and Seashore Learning Academy (62 percent). Section 5-12 Texas 219 TABLE 13.5 Revenue Per Pupil in Texas, 1997-98 Percent of Total Charter Schools Host Districts Charter Host Average Min Max Average Min Max Schools Districts Local tax $ - $ - $ 7 $ 3,133$ 1,736$ 4,533 0.0%48.4% Other local & intermediate 232 -764 321 235 518 4.5%5.0% State 4,414 3,437 7,463 1,883 639 3,993 86.2%29.1% Federal Title I 127 -623 206 70 350 2.5%3.2% School lunch (240) 85 -555 175 -323 1.7%2.7% Federal start-up 180 -1,328 ---3.5%0.0% Special education ---41 15 71 0.0%0.6% Other federal 83 na na 146 na na 1.6%2.5% To tal federal 475 -1,430 568 320 941 9.3%9.6% Total revenue $ 5,121 $ 4,043 $ 7,958 $ 5,908 $ 5,504 $ 7,002 100.0%100.0% Note: N=16. Minimums and maximums do not add up to totals. Data provided by Texas Education Agency. Excludes Building Alternatives, One-Stop and North Hills (including host school districts). The averages disguise considerable revenue variation among charter schools due to school district funding inequalities, attendance rate variations and funding associated with students in high-cost programs. Total revenue ranged from just over $4,000 per pupil (Pegasus Charter School in Dallas) to almost $8,000 per pupil (Renaissance Charter School in Irving). State funding varied from about $3,437 per pupil (University of Houston Charter School) to about $7,463 per pupil (Renaissance Charter School). Federal and state grants ranged from no federal funding to as high as $1,430 per pupil (Dallas Can! Academy). Private funding probably led to the large variation in “Other local” funding, which ranged from $0 to $764 per pupil. On average, revenue for school districts exceeded charter school revenue by about $785 per pupil. The difference was due to: § Local property taxes raised to pay for debt on facilities (about $380 per pupil in host school districts). (See Table 13.6.) § Higher average daily attendance in school districts (94.7 percent) compared to charter schools (90.1 percent). (See Table 13.2.) § Funding for transportation, food service, the education of high-cost handicapped children and several other adjustments incorporated into the student weighting system.281 Averaging 4.6 percent, the ADA differential accounts for about $150 of the difference between charter schools and host school districts. Food service revenue widens the funding gap because school districts are more likely to provide food service than charter schools, which may elect not 281 In the average Texas district, special educatio n weights add 8.7 percent to the foundation allowance. See Nelson, Muir and Drown (2000) for a more complete description of the Texas charter school funding system. Section 5-13 Paying for the Vision: Charter School Revenue and Expenditures 220 to provide breakfast or lunch programs. The federal food service funding gap is $90 per pupil.282 Fueled in part by fe deral charter school start-up grants, federal funding for charter schools ($475 per pupil) is only a little less than for school districts ($568 per pupil). State funding includes some assistance for transportation if provided by charter schools; however, since no Texas public school district is required to provide transportation, charter schools are also exempt. The state also provides additional funding for special education transportation. Wealthy school districts get little or no state transportation funding because transportation is funded on an equalized basis. While Table 13.5 contains no specific amount of transportation aid, charter schools spent an average of $61 per pupil compared to $134 per pupil in host school districts. (See Table 13.6.) Charter schools can apply for all state categorical programs with the exception of funds for facilities assistance.283 A “technology allotment” is the only program routinely available to all school districts and all charter schools. Charter schools averaged $24 per pupil in state technology allotments and school districts averaged $32 per pupil. Several charter schools did not report a technology allotment; the eight that did averaged $54 per student. Examples of programs in the “other federal” category (and a verage amounts received by host school districts) include: § Immigrant education—$5 per pupil. § Drug-free schools—$10 per pupil. § Goals 2000—$4 per pupil. § Eisenhower Professional Development—$6 per pupil. § Title VI innovative education—$6 per pupil. The charter school average for some forms of aid is less than the school district average because many charter schools do not apply for or qualify for some forms of federal assistance. Title I averages $254 per pupil (enrolled in school, not just Title I eligible) in charter schools actually getting Title I assistance, which is a little higher than the school district average. School lunch aid averages $255 per pupil for the six charter schools providing qualifying programs. Expenditures The expenditure data in Tab les 6 and 7 combines restricted, unrestricted, operating and non- operating funds. In total, the average charter school spent $5,166 per pupil, about $650 per pupil less than the host school district average of $5,819 per pupil. Expenditures by function. Charter schools devoted an average of 49.6 percent of expenditures to instruction, almost as much as the 52.7 percent spent by school districts, and also devoted fewer funds to instructional resources and media, curriculum development, staff development and 282 Meal charges are included in local revenue and the amount is probably higher in school districts since food service spending is $307 per pupil and only $106 per pupil in charter schools. (See Table 6.) 283 State assistance applies to the debt service tax rate. Since charter schools cannot tax property, they are not eligible for assistance. A few other exceptions exist for programs funded outside the foundation program. Section 5-14 Texas 221 instructional leadership. But reflecting the numerous charter high schools serving at-risk students, the average charter school devoted a larger percentage of spending to guidance counseling and social work services than did the school districts. TABLE 13.6 Expenditures Per Pupil in Texas, 1997 -98 Percent of Total Charter Schools Host Districts Charter Host Average Min Max Average Min Max Schools Districts Total operating expenditures Instruction $ 2,560 $ 1,018 $ 4,203 $ 3,069 $ 2,697 $ 3,539 49.6% 52.7% Instructional media 24 - 120 80 65 110 0.5% 1.4% Curriculum/staff development 70 - 278 107 20 237 1.4% 1.8% Instructional leadership 62 - 707 90 52 148 1.2% 1.5% School leadership 467 - 2,457 328 248 364 9.0% 5.6% Guidance/counseling 272 - 1,321 185 155 220 5.3% 3.2% Social work services 22 - 194 19 6 31 0.4% 0.3% Health services 31 - 190 52 27 76 0.6% 0.9% Transportation 61 - 478 134 17 195 1.2% 2.3% Food 106 - 541 307 244 382 2.1% 5.3% Co-curricular 45 - 252 71 36 181 0.9% 1.2% General administration 434 2 1,181 157 141 212 8.4% 2.7% Maintenance and operation 388 87 1,467 575 490 609 7.5% 9.9% Charter school lease costs 564 - 2,541 na na na 10.9% na Security/monitoring 24 - 166 40 11 78 0.5% 0.7% Data processing services 47 - 327 56 31 85 0.9% 1.0% Intergovernmental charge - - - 6 - 49 0.0% 0.1% Total operating 5,136 3,651 8,331 5,278 4,903 6,026 99.4% 90.7% Community services 22 - 216 38 3 102 0.4% 0.7% Debt service 7 - 47 380 168 700 0.1% 6.5% General fund capital outlay1 - - - 123 82 189 0.0% 2.1% Total expenditures $ 5,166 $ 3,851 $ 8,331 $ 5,819 $ 5,218 $ 6,808 100.0% 100.0% Note: N=16. na indicates not applicable. Minimums and maximums do not add up to totals. Data provided by Texas Education Agency. Excludes Building Alternatives, One-Stop and West Houston (including host school districts). 1 Excludes an average of $470 per pupil financed by bond proceeds and other non-revenue sources. Demons trating diseconomies of small scale, charter schools spent about $900 per pupil on school leadership and general administration (17.4 percent of total spending) compared to $485 per pupil in host school districts (8.3 percent of total spending). The spending data on administration is consistent with findings about administrative staffing in Table 13.2. Facilities and the costs associated with operating and maintaining them were similar for both charter schools and host school districts. Rental and lease costs for charter school facilities comprised 10.9 percent of spending. Plant operations and maintenance (including janitorial service and utilities) consumed another 7.5 percent, for a total of 18.4 percent of expenditures. In Section 5-15 Paying for the Vision: Charter School Revenue and Expenditures 222 school districts, 6.5 percent of spending went to debt service, 2.1 percent was devoted to capital outlay284 and 9.9 percent went to operations and maintenance, for a total of 16.5 percent. Transportation and food service spending accounted for about $275 of the $650 per pupil spending differential between charter schools and host school districts. Approximately 2.3 percent of host school district spending ($134 per pupil) went to pupil transportation. Charter school transportation costs averaged $61 per pupil, comprising 1.2 percent of spending. Averaging $106 per pupil, charter school expenditures on school lunch programs fell $200 per pupil short of school district spending, which averaged $307 per pupil. Expenditures by object. The first panel of Table 13.7 divides total spending into payroll (salaries and benefits), charter school lease costs, other operating expenditures, debt service and capital outlay. Payroll made up 75 percent of spending in host school districts compared to just 60 percent in charter schools. The University of Houston Charter School reported a payroll higher than any school district, at $6,245 per pupil. The Medical Center Charter School, however, reported payroll costs of just $1,306 per pupil. The “other operating” expenditures category— which excludes charter school lease costs, debt service and school district capital outlay—was almost 30 percent of spending in charter schools compared to only 15 percent in school districts. Although the exact amount cannot be determined, some of this differential is due to contracting out for professional and nonprofessional services, depreciation of assets and management fees for nonprofit sponsors. Expenditures by program. The bottom panel of Table 13.7 displays expenditures by “program intent” code. Charter schools allocated an average of $3,822 per pupil (74 percent) of total expenditures to programs comparable to the $3,998 per-pupil allocation (69 percent of total expenditures) in host school districts. But 65 percent of charter school spending was devoted to regular instruction compared to just 39 percent in host school districts. One of the big differences between charter schools and host school districts is spending on special education. Host school districts put 10 percent (approximately $600 per pupil) of spending to special education. While charter schools devoted only 2 percent ($100 per pupil) to special education.285 Similarly, 5.6 percent ($325 per pupil) of school district spending went to bilingual education compared to 1 percent (approximately $47 per pupil) in charter schools. Career and technology education made up 2.3 percent of school district spending but barely any of charter school spending. Accelerated education is “challenging and meaningful instructional programs to close the achievement gap for disadvantaged children in at-risk situations.” Host school districts spent 9 percent of their resources on accelerated education compared to under 5 percent of charter school resources, with two charter high schools for at-risk students, Blessed Sacrament and American Institute for Learning, accounting for a majority of that amount. In addition, only two charter schools reported any spending on gifted and talented education, while host school districts classified 1.4 percent of expenditures as resources devoted to gifted and talented education. These program- 284 Excludes an average of $470 per pupil financed by bond proceeds and other non-revenue sources. These expenditures are primarily financed by selling tax-exempt securities that are subsequently repaid through the debt service fund. 285 Stated differently, charter schools spent about $2,200 per special education student and school districts spent about $4,800 per special education student. Section 5-16 Texas 223 area spending patterns are consistent with the staffing data in Table 13.3, which shows 90 percent of charter school teachers in regular education compared to 61 percent in host school districts. TABLE 13.7 Expenditures Per Pupil by Object and Program in Texas, 1997-98 Percent of Total Charter Schools Host Districts Charter Host Average Min Max Average Min Max Schools Districts Total expenditures by object Payroll $ 3,081 $ 1,306 $ 6,245 $ 4,426 $ 3,712 $ 5,248 59.6% 75.6% Charter school lease costs 564 - 2,541 na na na 10.9% na Other operating 1,514 750 2,534 919 729 1,293 28.3% 15.7% Debt service 7 - 47 387 168 700 0.1% 6.6% General fund capital outlay1 - - - 121 82 189 na 2.1% Total 5,165 3,666 8,337 5,855 5,218 6,808 100.0% 100.0% Operating expenditures by program Regular 3,388 739 8,196 2,273 1,988 2,700 65.6% 39.1% Gifted & talented 4 - 58 84 18 130 0.1% 1.4% Career & technology 34 - 294 133 89 194 0.7% 2.3% Students with disabilities 99 - 1,363 596 328 951 1.9% 10.2% Accelerated e ducation 249 - 1,824 541 246 848 4.8% 9.3% Bilingual 47 - 796 325 6 541 0.9% 5.6% Athletics/related Activities 2 - 31 46 25 95 0.0% 0.8% Total 3,822 1,817 8,196 3,998 3,530 4,744 74.0% 68.7% Non-program expenditures 1,314 141 2,583 1,280 1,037 1,373 25.4% 22.0% Total operating expenditures 5,136 3,651 8,331 5,278 4,903 6,026 99.4% 90.7% Community services 22 - 216 38 3 102 0.4% 0.7 % Debt service 7 - 47 380 168 700 0.1% 6.5% General fund capital outlay1 - - - 123 82 189 0.0% 2.1% Total $ 5,166 $ 3,851 $ 8,331 $ 5,819 $ 5,218 $ 6,808 100.0% 100.0% Note: N=16. na indicates not applicable. Minimum s and maximums do not add up to totals. Data provided by Texas Education Agency. Excludes Academy of Transitional Studies, Building Alternatives, One-Stop and West Houston (including host school districts). 1Excludes an average of $470 per pupil financed by bond proceeds and other non -revenue sources. Summary Texas provides school districts and charter schools additional funding for special education, limited-English proficiency and at-risk students based on the actual student characteristics of charter schools. Nevertheless, the typical Texas charter school enrolls fewer special education, bilingual and economically disadvantaged students than host school districts. Several charter schools serving at-risk high school youth suffered significant funding red uctions because Texas bases funding on average daily attendance rather than membership. Even with the extra funding for students in high-cost programs, special education students made up only 4.4 percent of charter school enrollment compared to 11.3 percent in host school districts. Costs associated with special education students account for about $500 of the expenditure disparity between charter schools and host school districts. Charter schools averaged an LEP enrollment of about 5 percent of total enrollment compared to 20 percent in host school districts. Compensatory education students in Texas received 20 percent extra funding through the weighting system. The average Section 5-17 Paying for the Vision: Charter School Revenue and Expenditures 224 charter school reported 35.5 percent of enrollment as economically disadvantaged compared to 62.3 percent in host school districts. Even though they have a reputation for smaller class size, the pupil-to-teacher ratio is 22.7 in the average charter school compared to 16.6 students per teacher in host school districts. Perhaps reflecting the diseconomies of small scale, charter schools employed nearly one administrator per 100 students—three times the level of host school districts. On the other hand, host school districts employed 3.4 FTE support staff for every 100 students compared to j ust 1.6 FTE support staff per 100 students in charter schools. Charter schools were not as well funded as school districts, but much of the difference is explained by differences in how the state funds high-cost programs. Charter school revenue in 1997-98 averaged $5,169 per pupil and exceeded expenditures by $369 per pupil (7.1 percent of revenue). On average, state and local revenue for school districts exceeded charter school revenue by about $775 per pupil. The difference is due to: facilities funding that charter schools do not receive (about $380 per pupil), higher average daily attendance in school districts (94.7 percent) compared to charter schools (90.1 percent), funding for transportation, extra state funding for compensatory education and the ed ucation of high-cost handicapped children, and several other adjustments incorporated into the student weighting system. Charter schools focus on regular classroom instruction. The average charter school devoted 65 percent of spending to regular instruct ion programs compared to just 39 percent in host school districts. One of the big differences is spending on special education. Host school districts put 10 percent of spending toward special education; charter schools expended just 2 percent on special education. Similarly, school districts devoted 5.6 percent of school district spending to bilingual education compared to the charter school average of 1.0 percent. Charter schools also devoted fewer funds to instructional resources and media, curriculum development, staff development and instructional leadership. Demonstrating diseconomies of small scale, charter schools spent about $900 per pupil on school leadership and general administration (17.4 percent of total spending) compared to $485 per pupil in host school districts (8.3 percent of total spending). Facilities and costs for operating and maintaining facilities were similar for both charter schools and host school districts. Transportation and food service spending accounts for about $275 of the spe nding differential between charter schools and host school districts. Section 5-18 FY 10/11 - No SLA's (Hold Costs Flat) Option A - 1 FY 08-09 Audited FY 09/10 Adopted FY 09/10 Estimated FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 53.6 53.6 53.6 53.6 53.6 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted Budget FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,444,082$ 2,444,082$ 2,444,082$ 2,444,082$ 2,444,082$ 21 Resources & Media 62,513 91,303 81,303 81,303 81,303 81,303 81,303 81,303 22 Staff Development 36,973 67,405 57,405 57,405 57,405 57,405 57,405 57,405 23 Instructional Leadership 11,660 92,034 92,034 92,034 92,034 92,034 92,034 92,034 24 School Leadership 179,554 210,441 210,441 210,441 210,441 210,441 210,441 210,441 25 Guidance & Counseling 163,575 153,340 148,340 148,340 148,340 148,340 148,340 148,340 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 58,813 58,813 58,813 58,813 58,813 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 9,018 9,018 9,018 9,018 9,018 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 108,220 108,220 108,220 108,220 108,220 31 Administrative 527,000 272,714 287,714 287,714 287,714 287,714 287,714 287,714 32 Maintenance & Operations 279,485 300,156 290,156 290,156 290,156 290,156 290,156 290,156 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 140,404 140,404 140,404 140,404 140,404 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 3,927,930$ 3,927,930$ 3,927,930$ 3,927,930$ 3,927,930$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 3,992,788$ 3,992,788$ 3,992,788$ 3,992,788$ 3,992,788$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (111,001)$ (111,001)$ (111,001)$ (111,001)$ (111,001)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (111,001)$ (111,001)$ (111,001)$ (111,001)$ (111,001)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 422,428$ 311,427$ 200,426$ 89,425$ (21,576)$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (699,108)$ (699,108)$ (699,108)$ (699,108)$ (699,108)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,197,358)$ (2,200,886)$ (2,197,558)$ (2,197,636)$ (2,196,014)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-1 FY 10/11 - No SLA's (Hold Costs Flat) Option A - 2 FY 08-09 Audited FY 09/10 Adopted FY 09/10 Estimated FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 53.6 53.6 53.6 53.6 53.6 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted Budget FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - 9 Fund100 (WAF)- - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,199,674$ 2,199,674$ 2,199,674$ 2,199,674$ 2,199,674$ 21 Resources & Media 62,513 91,303 81,303 73,173 73,173 73,173 73,173 73,173 22 Staff Development 36,973 67,405 57,405 51,665 51,665 51,665 51,665 51,665 23 Instructional Leadership 11,660 92,034 92,034 82,831 82,831 82,831 82,831 82,831 24 School Leadership 179,554 210,441 210,441 189,397 189,397 189,397 189,397 189,397 25 Guidance & Counseling 163,575 153,340 148,340 133,506 133,506 133,506 133,506 133,506 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 52,932 52,932 52,932 52,932 52,932 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 8,116 8,116 8,116 8,116 8,116 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 97,398 97,398 97,398 97,398 97,398 31 Administrative 527,000 272,714 287,714 258,943 258,943 258,943 258,943 258,943 32 Maintenance & Operations 279,485 300,156 290,156 261,140 261,140 261,140 261,140 261,140 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 126,364 126,364 126,364 126,364 126,364 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 3,535,137$ 3,535,137$ 3,535,137$ 3,535,137$ 3,535,137$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 3,599,995$ 3,599,995$ 3,599,995$ 3,599,995$ 3,599,995$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ 281,792$ 281,792$ 281,792$ 281,792$ 281,792$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ 281,792$ 281,792$ 281,792$ 281,792$ 281,792$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 815,221$ 1,097,013$ 1,378,805$ 1,660,597$ 1,942,389$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (306,315)$ (306,315)$ (306,315)$ (306,315)$ (306,315)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (1,804,565)$ (1,808,093)$ (1,804,765)$ (1,804,843)$ (1,803,221)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-2 FY 10/11 - No SLA's (Hold Costs Flat) Option A - 3 FY 08-09 Audited FY 09/10 Adopted FY 09/10 FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 53.6 53.6 53.6 53.6 53.6 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted Budget FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - 9 Fund100 (WAF)- - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 1,955,266$ 1,955,266$ 1,955,266$ 1,955,266$ 1,955,266$ 21 Resources & Media 62,513 91,303 81,303 65,042 65,042 65,042 65,042 65,042 22 Staff Development 36,973 67,405 57,405 45,924 45,924 45,924 45,924 45,924 23 Instructional Leadership 11,660 92,034 92,034 73,627 73,627 73,627 73,627 73,627 24 School Leadership 179,554 210,441 210,441 168,353 168,353 168,353 168,353 168,353 25 Guidance & Counseling 163,575 153,340 148,340 118,672 118,672 118,672 118,672 118,672 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 47,050 47,050 47,050 47,050 47,050 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 7,214 7,214 7,214 7,214 7,214 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 86,576 86,576 86,576 86,576 86,576 31 Administrative 527,000 272,714 287,714 230,171 230,171 230,171 230,171 230,171 32 Maintenance & Operations 279,485 300,156 290,156 232,125 232,125 232,125 232,125 232,125 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 112,323 112,323 112,323 112,323 112,323 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 3,142,344$ 3,142,344$ 3,142,344$ 3,142,344$ 3,142,344$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 3,207,202$ 3,207,202$ 3,207,202$ 3,207,202$ 3,207,202$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ 674,585$ 674,585$ 674,585$ 674,585$ 674,585$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ 674,585$ 674,585$ 674,585$ 674,585$ 674,585$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 1,208,014$ 1,882,599$ 2,557,184$ 3,231,769$ 3,906,354$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ 86,478$ 86,478$ 86,478$ 86,478$ 86,478$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (1,411,772)$ (1,415,300)$ (1,411,972)$ (1,412,050)$ (1,410,428)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-3 FY 10/11 - No SLA's (Hold Costs Flat) Option A - 4 FY 08-09 Audited FY 09/10 Adopted FY 09/10 FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 55.6 55.6 55.6 55.6 55.6 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,641,664$ 2,694,497$ 2,748,387$ 2,803,355$ 2,859,422$ 21 Resources & Media 62,513 91,303 81,303 82,929 84,588 86,279 88,005 89,765 22 Staff Development 36,973 67,405 57,405 58,553 59,724 60,919 62,137 63,380 23 Instructional Leadership 11,660 92,034 92,034 93,875 95,752 97,667 99,621 101,613 24 School Leadership 179,554 210,441 210,441 214,650 218,943 223,322 227,788 232,344 25 Guidance & Counseling 163,575 153,340 148,340 151,307 154,333 157,420 160,568 163,779 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 59,989 61,189 62,413 63,661 64,934 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 9,198 9,382 9,570 9,761 9,957 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 110,384 112,592 114,844 117,141 119,484 31 Administrative 527,000 272,714 287,714 293,468 299,338 305,324 311,431 317,660 32 Maintenance & Operations 279,485 300,156 290,156 295,959 301,878 307,916 314,074 320,356 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 143,212 146,076 148,998 151,978 155,017 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 4,155,189$ 4,238,292$ 4,323,058$ 4,409,519$ 4,497,710$ 38 Community Services 66,822 64,858 64,858 66,155 67,478 68,828 70,204 71,608 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 4,221,344$ 4,305,771$ 4,391,886$ 4,479,724$ 4,569,318$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (339,557)$ (423,984)$ (510,099)$ (597,937)$ (687,531)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (339,557)$ (423,984)$ (510,099)$ (597,937)$ (687,531)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 193,872$ (230,111)$ (740,210)$ (1,338,147)$ (2,025,678)$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (927,664)$ (1,012,091)$ (1,098,206)$ (1,186,044)$ (1,275,638)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,425,914)$ (2,513,869)$ (2,596,656)$ (2,684,572)$ (2,772,544)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-4 FY 10/11 - No SLA's (Hold Costs Flat) Option B - 1 FY 08-09 Audited FY 09/10 Adopted FY 09/10 Estimated FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 550 568 586 604 622 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted Budget FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,792,800 3,916,928 4,041,056 4,165,184 4,289,312 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 4,106,765 4,230,893 4,355,021 4,479,149 4,603,277 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 4,240,379$ 4,364,507$ 4,488,635$ 4,612,763$ 4,736,891$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,762,682$ 2,762,682$ 2,762,682$ 2,762,682$ 2,762,682$ 21 Resources & Media 62,513 91,303 81,303 81,303 81,303 81,303 81,303 81,303 22 Staff Development 36,973 67,405 57,405 57,405 57,405 57,405 57,405 57,405 23 Instructional Leadership 11,660 92,034 92,034 92,034 92,034 92,034 92,034 92,034 24 School Leadership 179,554 210,441 210,441 210,441 210,441 210,441 210,441 210,441 25 Guidance & Counseling 163,575 153,340 148,340 148,340 148,340 148,340 148,340 148,340 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 58,813 58,813 58,813 58,813 58,813 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 9,018 9,018 9,018 9,018 9,018 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 108,220 108,220 108,220 108,220 108,220 31 Administrative 527,000 272,714 287,714 287,714 287,714 287,714 287,714 287,714 32 Maintenance & Operations 279,485 300,156 290,156 290,156 290,156 290,156 290,156 290,156 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 140,404 140,404 140,404 140,404 140,404 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 4,246,530$ 4,246,530$ 4,246,530$ 4,246,530$ 4,246,530$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 4,311,388$ 4,311,388$ 4,311,388$ 4,311,388$ 4,311,388$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (71,009)$ 53,119$ 177,247$ 301,375$ 425,503$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (71,009)$ 53,119$ 177,247$ 301,375$ 425,503$ 49 Fund Balance 504,653$ 504,653$ 533,429$ 462,420$ 515,539$ 692,786$ 994,161$ 1,419,664$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (659,116)$ (534,988)$ (410,860)$ (286,732)$ (162,604)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,157,366)$ (2,036,766)$ (1,909,310)$ (1,785,260)$ (1,659,510)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-5Section 6-5 FY 10/11 - No SLA's (Hold Costs Flat) Option B - 2 FY 08-09 Audited FY 09/10 Adopted FY 09/10 Estimated FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 550 568 586 604 622 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted Budget FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - 9 Fund100 (WAF)- - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,792,800 3,916,928 4,041,056 4,165,184 4,289,312 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 4,106,765 4,230,893 4,355,021 4,479,149 4,603,277 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 4,240,379$ 4,364,507$ 4,488,635$ 4,612,763$ 4,736,891$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,486,414$ 2,486,414$ 2,486,414$ 2,486,414$ 2,486,414$ 21 Resources & Media 62,513 91,303 81,303 73,173 73,173 73,173 73,173 73,173 22 Staff Development 36,973 67,405 57,405 51,665 51,665 51,665 51,665 51,665 23 Instructional Leadership 11,660 92,034 92,034 82,831 82,831 82,831 82,831 82,831 24 School Leadership 179,554 210,441 210,441 189,397 189,397 189,397 189,397 189,397 25 Guidance & Counseling 163,575 153,340 148,340 133,506 133,506 133,506 133,506 133,506 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 52,932 52,932 52,932 52,932 52,932 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 8,116 8,116 8,116 8,116 8,116 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 97,398 97,398 97,398 97,398 97,398 31 Administrative 527,000 272,714 287,714 258,943 258,943 258,943 258,943 258,943 32 Maintenance & Operations 279,485 300,156 290,156 261,140 261,140 261,140 261,140 261,140 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 126,364 126,364 126,364 126,364 126,364 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 3,821,877$ 3,821,877$ 3,821,877$ 3,821,877$ 3,821,877$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 3,886,735$ 3,886,735$ 3,886,735$ 3,886,735$ 3,886,735$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ 353,644$ 477,772$ 601,900$ 726,028$ 850,156$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ 46 Other (Uses)(200,000) (100,000) (100,000) - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ 353,644$ 477,772$ 601,900$ 726,028$ 850,156$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 887,073$ 1,364,845$ 1,966,745$ 2,692,773$ 3,542,929$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (234,463)$ (110,335)$ 13,793$ 137,921$ 262,049$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (1,732,713)$ (1,612,113)$ (1,484,657)$ (1,360,607)$ (1,234,857)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-6Section 6-6 FY 10/11 - No SLA's (Hold Costs Flat) Option B - 3 FY 08-09 Audited FY 09/10 Adopted FY 09/10 FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 550 568 586 604 622 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted Budget FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,792,800 3,916,928 4,041,056 4,165,184 4,289,312 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 4,106,765 4,230,893 4,355,021 4,479,149 4,603,277 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 4,240,379$ 4,364,507$ 4,488,635$ 4,612,763$ 4,736,891$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,210,146$ 2,210,146$ 2,210,146$ 2,210,146$ 2,210,146$ 21 Resources & Media 62,513 91,303 81,303 65,042 65,042 65,042 65,042 65,042 22 Staff Development 36,973 67,405 57,405 45,924 45,924 45,924 45,924 45,924 23 Instructional Leadership 11,660 92,034 92,034 73,627 73,627 73,627 73,627 73,627 24 School Leadership 179,554 210,441 210,441 168,353 168,353 168,353 168,353 168,353 25 Guidance & Counseling 163,575 153,340 148,340 118,672 118,672 118,672 118,672 118,672 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 47,050 47,050 47,050 47,050 47,050 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 7,214 7,214 7,214 7,214 7,214 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 86,576 86,576 86,576 86,576 86,576 31 Administrative 527,000 272,714 287,714 230,171 230,171 230,171 230,171 230,171 32 Maintenance & Operations 279,485 300,156 290,156 232,125 232,125 232,125 232,125 232,125 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 112,323 112,323 112,323 112,323 112,323 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 3,397,224$ 3,397,224$ 3,397,224$ 3,397,224$ 3,397,224$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 3,462,082$ 3,462,082$ 3,462,082$ 3,462,082$ 3,462,082$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ 778,297$ 902,425$ 1,026,553$ 1,150,681$ 1,274,809$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ 778,297$ 902,425$ 1,026,553$ 1,150,681$ 1,274,809$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 1,311,726$ 2,214,151$ 3,240,704$ 4,391,385$ 5,666,194$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ 190,190$ 314,318$ 438,446$ 562,574$ 686,702$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (1,308,060)$ (1,187,460)$ (1,060,004)$ (935,954)$ (810,204)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-7Section 6-7 FY 10/11 - No SLA's (Hold Costs Flat) Option B - 4 FY 08-09 Audited FY 09/10 Adopted FY 09/10 FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 550 568 586 604 622 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,792,800 3,916,928 4,041,056 4,165,184 4,289,312 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 4,106,765 4,230,893 4,355,021 4,479,149 4,603,277 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 4,240,379$ 4,364,507$ 4,488,635$ 4,612,763$ 4,736,891$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,811,564$ 2,958,145$ 3,017,308$ 3,077,654$ 3,139,207$ 21 Resources & Media 62,513 91,303 81,303 82,929 84,588 86,279 88,005 89,765 22 Staff Development 36,973 67,405 57,405 58,553 59,724 60,919 62,137 63,380 23 Instructional Leadership 11,660 92,034 92,034 93,875 95,752 97,667 99,621 101,613 24 School Leadership 179,554 210,441 210,441 214,650 218,943 223,322 227,788 232,344 25 Guidance & Counseling 163,575 153,340 148,340 151,307 154,333 157,420 160,568 163,779 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 59,989 61,189 62,413 63,661 64,934 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 9,198 9,382 9,570 9,761 9,957 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 110,384 112,592 114,844 117,141 119,484 31 Administrative 527,000 272,714 287,714 293,468 299,338 305,324 311,431 317,660 32 Maintenance & Operations 279,485 300,156 290,156 295,959 301,878 307,916 314,074 320,356 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 143,212 146,076 148,998 151,978 155,017 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 4,325,089$ 4,501,940$ 4,591,979$ 4,683,819$ 4,777,495$ 38 Community Services 66,822 64,858 64,858 66,155 67,478 68,828 70,204 71,608 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 4,391,244$ 4,569,419$ 4,660,807$ 4,754,023$ 4,849,104$ 42 Excess (Deficiency of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (150,865)$ (204,912)$ (172,172)$ (141,260)$ (112,213)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (150,865)$ (204,912)$ (172,172)$ (141,260)$ (112,213)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 382,564$ 177,653$ 5,481$ (135,779)$ (247,992)$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (738,972)$ (793,019)$ (760,279)$ (729,367)$ (700,320)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,237,222)$ (2,294,797)$ (2,258,729)$ (2,227,895)$ (2,197,226)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-8Section 6-8 FY 10/11 - No SLA's (Hold Costs Flat) Option C - 1 FY 08-09 Audited FY 09/10 Adopted FY 09/10 Estimated FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,762,682$ 2,762,682$ 2,762,682$ 2,762,682$ 2,762,682$ 21 Resources & Media 62,513 91,303 81,303 81,303 81,303 81,303 81,303 81,303 22 Staff Development 36,973 67,405 57,405 57,405 57,405 57,405 57,405 57,405 23 Instructional Leadership 11,660 92,034 92,034 92,034 92,034 92,034 92,034 92,034 24 School Leadership 179,554 210,441 210,441 210,441 210,441 210,441 210,441 210,441 25 Guidance & Counseling 163,575 153,340 148,340 148,340 148,340 148,340 148,340 148,340 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 58,813 58,813 58,813 58,813 58,813 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 9,018 9,018 9,018 9,018 9,018 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 108,220 108,220 108,220 108,220 108,220 31 Administrative 527,000 272,714 287,714 287,714 287,714 287,714 287,714 287,714 32 Maintenance & Operations 279,485 300,156 290,156 290,156 290,156 290,156 290,156 290,156 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 140,404 140,404 140,404 140,404 140,404 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 4,246,530$ 4,246,530$ 4,246,530$ 4,246,530$ 4,246,530$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 4,311,388$ 4,311,388$ 4,311,388$ 4,311,388$ 4,311,388$ 42 Excess (Deficiency) of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (429,601)$ (429,601)$ (429,601)$ (429,601)$ (429,601)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (429,601)$ (429,601)$ (429,601)$ (429,601)$ (429,601)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 103,828$ (325,773)$ (755,374)$ (1,184,975)$ (1,614,576)$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (1,017,708)$ (1,017,708)$ (1,017,708)$ (1,017,708)$ (1,017,708)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,515,958)$ (2,519,486)$ (2,516,158)$ (2,516,236)$ (2,514,614)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-9Section 6-9 FY 10/11 - No SLA's (Hold Costs Flat) Option C - 2 FY 08-09 Audited FY 09/10 Adopted FY 09/10 Estimated FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - 9 Fund100 (WAF)- - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,730,822$ 2,730,822$ 2,730,822$ 2,730,822$ 2,730,822$ 21 Resources & Media 62,513 91,303 81,303 73,173 73,173 73,173 73,173 73,173 22 Staff Development 36,973 67,405 57,405 51,665 51,665 51,665 51,665 51,665 23 Instructional Leadership 11,660 92,034 92,034 82,831 82,831 82,831 82,831 82,831 24 School Leadership 179,554 210,441 210,441 189,397 189,397 189,397 189,397 189,397 25 Guidance & Counseling 163,575 153,340 148,340 133,506 133,506 133,506 133,506 133,506 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 52,932 52,932 52,932 52,932 52,932 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 8,116 8,116 8,116 8,116 8,116 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 97,398 97,398 97,398 97,398 97,398 31 Administrative 527,000 272,714 287,714 258,943 258,943 258,943 258,943 258,943 32 Maintenance & Operations 279,485 300,156 290,156 261,140 261,140 261,140 261,140 261,140 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 126,364 126,364 126,364 126,364 126,364 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 4,066,285$ 4,066,285$ 4,066,285$ 4,066,285$ 4,066,285$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 4,131,143$ 4,131,143$ 4,131,143$ 4,131,143$ 4,131,143$ 42 Excess (Deficiency) of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (249,356)$ (249,356)$ (249,356)$ (249,356)$ (249,356)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ 46 Other (Uses)(200,000) (100,000) (100,000) - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (249,356)$ (249,356)$ (249,356)$ (249,356)$ (249,356)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 284,073$ 34,717$ (214,639)$ (463,996)$ (713,352)$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (837,463)$ (837,463)$ (837,463)$ (837,463)$ (837,463)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,335,713)$ (2,339,241)$ (2,335,913)$ (2,335,991)$ (2,334,369)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-10Section 6-10 FY 10/11 - No SLA's (Hold Costs Flat) Option C - 3 FY 08-09 Audited FY 09/10 Adopted FY 09/10 FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - 9 Fund100 (WAF)- - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,698,962$ 2,698,962$ 2,698,962$ 2,698,962$ 2,698,962$ 21 Resources & Media 62,513 91,303 81,303 65,042 65,042 65,042 65,042 65,042 22 Staff Development 36,973 67,405 57,405 45,924 45,924 45,924 45,924 45,924 23 Instructional Leadership 11,660 92,034 92,034 73,627 73,627 73,627 73,627 73,627 24 School Leadership 179,554 210,441 210,441 168,353 168,353 168,353 168,353 168,353 25 Guidance & Counseling 163,575 153,340 148,340 118,672 118,672 118,672 118,672 118,672 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 47,050 47,050 47,050 47,050 47,050 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 7,214 7,214 7,214 7,214 7,214 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 86,576 86,576 86,576 86,576 86,576 31 Administrative 527,000 272,714 287,714 230,171 230,171 230,171 230,171 230,171 32 Maintenance & Operations 279,485 300,156 290,156 232,125 232,125 232,125 232,125 232,125 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 112,323 112,323 112,323 112,323 112,323 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 3,886,040$ 3,886,040$ 3,886,040$ 3,886,040$ 3,886,040$ 38 Community Services 66,822 64,858 64,858 64,858 64,858 64,858 64,858 64,858 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 3,950,898$ 3,950,898$ 3,950,898$ 3,950,898$ 3,950,898$ 42 Excess (Deficiency) of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (69,111)$ (69,111)$ (69,111)$ (69,111)$ (69,111)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (69,111)$ (69,111)$ (69,111)$ (69,111)$ (69,111)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ 464,318$ 395,206$ 326,095$ 256,984$ 187,872$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (657,218)$ (657,218)$ (657,218)$ (657,218)$ (657,218)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,155,468)$ (2,158,996)$ (2,155,668)$ (2,155,746)$ (2,154,124)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-11Section 6-11 FY 10/11 - No SLA's (Hold Costs Flat) Option C - 4 FY 08-09 Audited FY 09/10 Adopted FY 09/10 FY 10/11 FY 11/12 FT 12/13 FY 13/14 FY 14/15 1 # of Students 490 498 498 498 498 498 2 # of Teachers 53.6 59 59 59 59 59 Assumptions 3 Blacksmith (Average Donation/Student)888$ -$ -$ -$ -$ -$ 4 State FSP ($/Student)6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 6,896$ 5 Teacher Average Compensation (Salary & Benefits)59,000$ 59,000$ 59,000$ 59,000$ 59,000$ 59,000$ FY 08/09 Audited FY 09/10 Adopted FY 09/10 Estimated Budget FY 10/11 Proposed Budget FY 11/12 Proposed Budget FT 12/13 Proposed Budget FY 13/14 Proposed Budget FY 14/15 Proposed Budget 6 Local & Intermediate Sources 7 Blacksmith Apprentice Program (WAF)394,129$ 420,000$ 435,095$ -$ -$ -$ -$ -$ 8 Gifts and Donations 20,000 - - - - - - - 9 Fund100 (WAF)- - - - - - - - 10 Investment Earnings 4,617 3,000 3,000 6,750 6,750 6,750 6,750 6,750 11 Lunchroom Revenues 6,339 8,600 8,600 3,000 3,000 3,000 3,000 3,000 12 Other Local Sources 81,163 75,858 75,858 75,858 75,858 75,858 75,858 75,858 13 Athletic Activities Income 32,743 48,006 48,006 48,006 48,006 48,006 48,006 48,006 14 Total Local & Intermediate Sources 538,992 555,464 570,559 133,614 133,614 133,614 133,614 133,614 State & Federal Revenues 15 Foundation School Program 2,960,589 3,443,120 3,379,040 3,434,208 3,434,208 3,434,208 3,434,208 3,434,208 16 State Revenue 202,539 193,989 193,989 193,989 193,989 193,989 193,989 193,989 17 Federal Revenue 56,134 119,976 119,976 119,976 119,976 119,976 119,976 119,976 18 Total State & Federal Revenues 3,219,263 3,757,085 3,693,005 3,748,173 3,748,173 3,748,173 3,748,173 3,748,173 19 TOTAL REVENUES 3,758,254$ 4,312,549$ 4,263,564$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ 3,881,787$ Expenditures by Function 20 Instructional 2,120,372$ 2,452,007$ 2,444,082$ 2,960,264$ 3,019,469$ 3,079,858$ 3,141,455$ 3,204,285$ 21 Resources & Media 62,513 91,303 81,303 82,929 84,588 86,279 88,005 89,765 22 Staff Development 36,973 67,405 57,405 58,553 59,724 60,919 62,137 63,380 23 Instructional Leadership 11,660 92,034 92,034 93,875 95,752 97,667 99,621 101,613 24 School Leadership 179,554 210,441 210,441 214,650 218,943 223,322 227,788 232,344 25 Guidance & Counseling 163,575 153,340 148,340 151,307 154,333 157,420 160,568 163,779 26 Social Work Services - - - - - - - 27 Health Services 58,845 53,813 58,813 59,989 61,189 62,413 63,661 64,934 28 Transportation - - - - - - - 29 Food Services 15,448 9,018 9,018 9,198 9,382 9,570 9,761 9,957 30 CoCurricular/Extracurricular Activities 111,738 108,220 108,220 110,384 112,592 114,844 117,141 119,484 31 Administrative 527,000 272,714 287,714 293,468 299,338 305,324 311,431 317,660 32 Maintenance & Operations 279,485 300,156 290,156 295,959 301,878 307,916 314,074 320,356 33 Charter School Lease Costs - - - - - - 34 Security Monitoring - - - - - - 35 Data Processing 88,720 140,404 140,404 143,212 146,076 148,998 151,978 155,017 36 Intergovernmental Charges - - - - - - 37 Total Operating 3,655,883$ 3,950,855$ 3,927,930$ 4,473,789$ 4,563,264$ 4,654,530$ 4,747,620$ 4,842,573$ 38 Community Services 66,822 64,858 64,858 66,155 67,478 68,828 70,204 71,608 39 Debt Service 82,991 42,000 42,000 - - - - - 40 General Fund Capital Outlay - - - - - - 41 Total Academic Services Expenditures 3,805,696$ 4,057,713$ 4,034,788$ 4,539,944$ 4,630,743$ 4,723,357$ 4,817,825$ 4,914,181$ 42 Excess (Deficiency) of Revenues Over (under) Expenditures (47,442)$ 254,836$ 228,776$ (658,157)$ (748,956)$ (841,570)$ (936,038)$ (1,032,394)$ 43 Technology/FF&E -$ 127,822$ 100,000$ -$ -$ -$ -$ -$ 44 Other Financing Sources (Uses) 45 Capital Leases 117,640$ -$ -$ -$ -$ -$ -$ -$ 46 Other (Uses)(200,000) (100,000) (100,000) - - - - - 47 Total Other Financing Sources (Uses (82,360)$ (100,000)$ (100,000)$ -$ -$ -$ -$ -$ 48 Net Change to Fund Balance (129,802)$ 154,836$ 28,776$ (658,157)$ (748,956)$ (841,570)$ (936,038)$ (1,032,394)$ 49 Fund Balance 504,653$ 659,489$ 533,429$ (124,728)$ (873,683)$ (1,715,254)$ (2,651,291)$ (3,683,686)$ 50 Municipal Expenditures 51 Direct 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 248,638$ 52 In-direct 339,469 239,469 239,469 339,469 339,469 339,469 339,469 339,469 53 Sub-Total Direct & Indirect Expenditures 588,107 488,107 488,107 588,107 588,107 588,107 588,107 588,107 54 Excess (Deficiency of Revenues Over (under) Including Direct & In-direct Expenditures (717,909)$ (333,271)$ (459,331)$ (1,246,264)$ (1,337,063)$ (1,429,677)$ (1,524,145)$ (1,620,501)$ 55 Debt Service 1,500,457 1,499,751 1,499,751 1,498,250 1,501,778 1,498,450 1,498,528 1,496,906 56 Total Expenditures Booked to Town of Westlake 2,088,564$ 1,987,858$ 1,987,858$ 2,086,357$ 2,089,885$ 2,086,557$ 2,086,635$ 2,085,013$ 57 Excess (Deficiency of Revenues Over (under) Including Direct, In-direct & Debt Service Expenditures (2,218,366)$ (1,833,022)$ (1,959,082)$ (2,744,514)$ (2,838,841)$ (2,928,127)$ (3,022,673)$ (3,117,407)$ Budget Scenarios Assumes no growth in revenue or expenses, and no Blacksmith Donations Section 6-12Section 6-12 WESTLAKE ACADEMY SERVICE LEVEL ADJUSTMENT FISCAL YEAR 2010/2011 Service Adjustment New or Expanded Service PYP Spanish Teacher Section: Primary Increased: $58,350.00 Fund: 199 Function Code: 11 Object Code: 6119 Brief Program Description: The number of students in the PYP spanish program requires additional support in order to meet our goals as outlined in the strategic plan. Proposed Position Title Hourly Rate Annual Hours Salary Description FY 10/11 FY 11/12 FY 12/13 FY 13/14 FY 14/15 IMPACT ON OPERATING BUDGET Payroll and Related Base Annual Salary $49,033.00 $0.00 $0.00 $0.00 $0.00 Related Expenses $8,317.00 $0.00 $0.00 $0.00 $0.00 Total Payroll & Related $57,350.00 $0.00 $0.00 $0.00 $0.00 Operating Expenditures new amount $0.00 $0.00 $0.00 $0.00 $0.00 Base budget $0.00 $0.00 $0.00 $0.00 $0.00 Total Operating Expense $ 0.00 $0.00 $0.00 $0.00 $0.00 Capital Expenditures Classroom Start-up $1,000.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Capital Expense $1,000.00 $0.00 $0.00 $0.00 $0.00 Total Additional Expense 58,350.00 0.00 0.00 0.00 0.00 ESTIMATED COST SAVINGS ON OPERATING BUDGET Expenditure Reduction explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Expenditure Reduction $ 0.00 $0.00 $0.00 $0.00 $0.00 Revenue Enhancement explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Revenues $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Cost Savings 0.00 0.00 0.00 0.00 0.00 Budget Impact 0.00 0.00 0.00 0.00 0.00 FUNDING 199 General Fund $58,350.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Other (Explain) $0.00 $0.00 $0.00 $0.00 $0.00 TOTAL 58,350.00 0.00 0.00 0.00 0.00 Section 7-1 WESTLAKE ACADEMY SERVICE LEVEL ADJUSTMENT FISCAL YEAR 2010/2011 Service Adjustment New or Expanded Service Physics/Math Teacher Section: Secondary Increased: $58,350.00 Fund: 199 Function Code: 11 Object Code: 6119 Brief Program Description: In order to maintain compliance with the state science requirements for high school, we need to add a physics teacher. In addition, this teacher will reduce the number of preps for the existing teaching staff in the secondary math department. Proposed Position Title Hourly Rate Annual Hours Salary Physics/Math Teacher $49,033 Description FY 10/11 FY 11/12 FY 12/13 FY 13/14 FY 14/15 IMPACT ON OPERATING BUDGET Payroll and Related Base Annual Salary $49,033.00 $0.00 $0.00 $0.00 $0.00 Related Expenses $8,317.00 $0.00 $0.00 $0.00 $0.00 Total Payroll & Related $57,350.00 $0.00 $0.00 $0.00 $0.00 Operating Expenditures new amount $0.00 $0.00 $0.00 $0.00 $0.00 Base budget $0.00 $0.00 $0.00 $0.00 $0.00 Total Operating Expense $ 0.00 $0.00 $0.00 $0.00 $0.00 Capital Expenditures Classroom Start-up $1,000.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Capital Expense $1,000.00 $0.00 $0.00 $0.00 $0.00 Total Additional Expense 58,350.00 0.00 0.00 0.00 0.00 ESTIMATED COST SAVINGS ON OPERATING BUDGET Expenditure Reduction explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Expenditure Reduction $ 0.00 $0.00 $0.00 $0.00 $0.00 Revenue Enhancement explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Revenues $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Cost Savings 0.00 0.00 0.00 0.00 0.00 Budget Impact 0.00 0.00 0.00 0.00 0.00 FUNDING 199 General Fund $58,350.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Other (Explain) $0.00 $0.00 $0.00 $0.00 $0.00 TOTAL 58,350.00 0.00 0.00 0.00 0.00 Section 7-2 WESTLAKE ACADEMY SERVICE LEVEL ADJUSTMENT FISCAL YEAR 2010/2011 Service Adjustment New or Expanded Service Operating expenses for school buses Section: Academy Increased: $20,000.00 Fund: 199 Function Code: 11 Object Code: 6311 Brief Program Description: WAF donated two MFSAB buses to be used for Academy-related field trips and events. This expense is intended to cover routine maintenance and cleaning as well as the cost of fuel consumption. Proposed Position Title Hourly Rate Annual Hours Salary Description FY 10/11 FY 11/12 FY 12/13 FY 13/14 FY 14/15 IMPACT ON OPERATING BUDGET Payroll and Related Base Annual Salary $0.00 $0.00 $0.00 $0.00 $0.00 Related Expenses $0.00 $0.00 $0.00 $0.00 $0.00 Total Payroll & Related $ 0.00 $0.00 $0.00 $0.00 $0.00 Operating Expenditures new amount $20,000.00 $0.00 $0.00 $0.00 $0.00 Base budget $0.00 $0.00 $0.00 $0.00 $0.00 Total Operating Expense $20,000.00 $0.00 $0.00 $0.00 $0.00 Capital Expenditures explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Capital Expense $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Additional Expense 20,000.00 0.00 0.00 0.00 0.00 ESTIMATED COST SAVINGS ON OPERATING BUDGET Expenditure Reduction explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Expenditure Reduction $ 0.00 $0.00 $0.00 $0.00 $0.00 Revenue Enhancement explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Revenues $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Cost Savings 0.00 0.00 0.00 0.00 0.00 Budget Impact 0.00 0.00 0.00 0.00 0.00 FUNDING 199 General Fund $20,000.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Other (Explain) $0.00 $0.00 $0.00 $0.00 $0.00 TOTAL 20,000.00 0.00 0.00 0.00 0.00 Section 7-3 WESTLAKE ACADEMY SERVICE LEVEL ADJUSTMENT FISCAL YEAR 2010/2011 Service Adjustment New or Expanded Service IB Annual Program Fees (PYP, MYP, DP) Section: Academy Increased: $2,000.00 Fund: 199 Function Code: 11 Object Code: 6499 Brief Program Description: The 5% increase this year reflects a cost increase in dues. These dues allow the Academy to maintain its status as an IB school. Proposed Position Title Hourly Rate Annual Hours Salary Description FY 10/11 FY 11/12 FY 12/13 FY 13/14 FY 14/15 IMPACT ON OPERATING BUDGET Payroll and Related Base Annual Salary $0.00 $0.00 $0.00 $0.00 $0.00 Related Expenses $0.00 $0.00 $0.00 $0.00 $0.00 Total Payroll & Related $ 0.00 $0.00 $0.00 $0.00 $0.00 Operating Expenditures new amount $27,000.00 $0.00 $0.00 $0.00 $0.00 Base budget $25,000.00 $0.00 $0.00 $0.00 $0.00 Total Operating Expense $2,000.00 $0.00 $0.00 $0.00 $0.00 Capital Expenditures explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Capital Expense $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Additional Expense 2,000.00 0.00 0.00 0.00 0.00 ESTIMATED COST SAVINGS ON OPERATING BUDGET Expenditure Reduction explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Expenditure Reduction $ 0.00 $0.00 $0.00 $0.00 $0.00 Revenue Enhancement explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Revenues $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Cost Savings 0.00 0.00 0.00 0.00 0.00 Budget Impact 0.00 0.00 0.00 0.00 0.00 FUNDING 199 General Fund $2,000.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Other (Explain) $0.00 $0.00 $0.00 $0.00 $0.00 TOTAL 2,000.00 0.00 0.00 0.00 0.00 Section 7-4 ` WESTLAKE ACADEMY SERVICE LEVEL ADJUSTMENT FISCAL YEAR 2010/2011 Service Adjustment New or Expanded Service Graduation Expenses Section: Secondary Increased: $10,000.00 Fund: 199 Function Code: 11 Object Code: 6499 Brief Program Description: This item was not budgeted previously and needs to be added as a recurring item to cover the cost of programs, flowers, banners etc. associated with Westlake Academy’s graduation ceremony. Proposed Position Title Hourly Rate Annual Hours Salary Description FY 10/11 FY 11/12 FY 12/13 FY 13/14 FY 14/15 IMPACT ON OPERATING BUDGET Payroll and Related Base Annual Salary $0.00 $0.00 $0.00 $0.00 $0.00 Related Expenses $0.00 $0.00 $0.00 $0.00 $0.00 Total Payroll & Related $ 0.00 $0.00 $0.00 $0.00 $0.00 Operating Expenditures new amount $10,000.00 $0.00 $0.00 $0.00 $0.00 Base budget $0.00 $0.00 $0.00 $0.00 $0.00 Total Operating Expense $10,000.00 $0.00 $0.00 $0.00 $0.00 Capital Expenditures explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Capital Expense $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Additional Expense 10,000.00 0.00 0.00 0.00 0.00 ESTIMATED COST SAVINGS ON OPERATING BUDGET Expenditure Reduction explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Expenditure Reduction $ 0.00 $0.00 $0.00 $0.00 $0.00 Revenue Enhancement explain $0.00 $0.00 $0.00 $0.00 $0.00 explain $0.00 $0.00 $0.00 $0.00 $0.00 Total Revenues $ 0.00 $0.00 $0.00 $0.00 $0.00 Total Cost Savings 0.00 0.00 0.00 0.00 0.00 Budget Impact 0.00 0.00 0.00 0.00 0.00 FUNDING 199 General Fund $10,000.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Special Revenue Fund $0.00 $0.00 $0.00 $0.00 $0.00 x Other (Explain) $0.00 $0.00 $0.00 $0.00 $0.00 TOTAL 10,000.00 0.00 0.00 0.00 0.00 Section 7-5 SUMMARY POINTS AND RECOMMENDATIONS Past decisions regarding salary structure (ISD Mkt vs. Charter Mkt) have had a major impact on our cost structure. Past decisions regarding class sizes have created a cost structure that remains under-funded. Inability to add students renders a revenue stream that is fairly static and reduces the ability to cover full costs. Removing Black Smith funding = forced cost reductions and No SLAs. To achieve cost reductions, must reduce instruction which leads to deleterious effects on the strategic plan’s desired outcomes. Section 8-1 Educational Evaluation and Policy Analysis Fall 1997, Vol. 19, No. 3, pp. 205-227 High School Size: Which Works Best and for Whom? Valerie E. Lee University of Michigan Julia B. Smith Western Michigan University The study described in this article investigates the relationship between high school size and student learning. We used three waves of data from NELS:88 and hierarchical linear modeling (HLM) methods to examine how students' achievement growth in two subjects (reading and mathematics) over the high school years is influenced by the size of the high school they attend. Three research questions guided the study: (a) Which size high school is most effective for students'learning?, (b) In which size high school is learning most equitably distributed?, and (c) Are size effects consistent across high schools defined by their social compositions? Results suggest that the ideal high school, defined in terms of effectiveness (i.e., learning), enrolls between 600 and 900 students. In schools smaller than this, students learn less; those in large high schools (especially over 2,100) learn considerably less. Learning is more equitable in very small schools, with equity defined by the relationship between learning and student socioeconomic status (SES). An important finding from the study is that the influence of school size on learning is different in schools that enroll students of varying SES and in schools with difering proportions of minorities. En- rollment size has a stronger effect on learning in schools with lower-SES students and also in schools with high concentrations of minority students. Implications for educational policy are discussed. Why study high school size? This study extends existing empirical and synthesis work on school structure and organization. Although one strand of that work investigates the effects of school reforms (particularly school restructuring) on learning, re- sults from studies in that strand also provide strong evidence that students learn more in smaller high schools and that learning is also more equitable in smaller school settings (Bryk, Lee, & Holland, 1993; Lee, Bryk, & Smith, 1993; Lee & Smith, 1993, 1995, 1996; Lee, Smith, & Croninger, 1997). But exactly how small should high schools be? There would seem to be a point of diminishing re- turns, where reducing size could constrain the courses that are offered and the subject matter ex- pertise among teachers to the point where learning is diminished. Findings about school size from stud- ies of school restructuring have relevance to edu- cational policy in that they show that most existing high schools are too large to maximize their stu- dents' educational progress. Unfortunately, they lack the specificity of a more practical question: "Exactly what size works best?" Beyond "ideal size," two other questions moti- vate this study. Both target the issue of equity in student learning. A second question asks, "Does an ideal school size, defined in terms of maximal learning, also support an equitable social distribu- tion of learning within the same school?"' A third question focuses on the social composition of schools: "Do size effects vary by the types of stu- dents enrolled in the school?" We explore these questions with data from three waves of the Na- tional Educational Longitudinal Study of 1988 (NELS:88). With nested data and questions that focus on school effects, we employ the methodol- ogy most appropriate to these circumstances: hier- archical linear models (HLMs). 205 Section 9-1 Lee and Smith An enduring issue for educational policy is the optimal size of a school. "Optimal" typically has been defined using two potentially conflicting cri- teria: (1) how organizational size affects group members (a sociological criterion) and (2) the best school size for optimum economic efficiency (an economic criterion). Although these goals are cer- tainly related, some researchers have suggested that maximizing performance can lead away from effi- cient functioning and vice versa (Goss, 1994; Morrison, 1993). At least since the end of World War II, this topic has been hotly debated in policy circles. These discussions, often motivated by a need to consider school consolidation, have focused more on the economic than the sociological crite- ria and have a decidedly bureaucratic bent. Although elementary schools are often small, based on an interest in providing intimate relations and a sup- portive environment for young children, high schools are seen as needing to be much larger in order to accomplish their purpose. The number of students in a school can either facilitate or constrain contact among members (teachers and students), affecting important relationships in both academic and social domains. Our direct interest and familiarity with this issue emerged from a series of studies sponsored by the Center on the Organization and Restructuring of Schools (CORS) at the University of Wisconsin. In these studies, in which we used the NELS data- base to evaluate the effects of restructuring on stu- dent outcomes, we focused on outcomes of two types: (1) learning (defined as gains in achievement over the high school years) and (2) the social dis- tribution of learning (defined by how learning is associated with students from families with vary- ing social class backgrounds). These outcomes are, of course, related. Although those studies focused on elements of school restructuring defined by CORS' mission (e.g., the organization of the cur- riculum, the character of instruction, the profes- sional lives of teachers), they also took account of other structural and compositional features of schools that might provide alternative explanations for the results of restructuring (e.g., average SES, minority concentration, sector, and size). At the outset, we included school size in ana- lytic models for the purpose of statistical control. However, the consistency of the residual effects of school size on student outcomes was striking. Al- though the analyses also included variables known to be related to school size (such as sector, minor- ity concentration, and several characteristics of school social organization), direct size effects per- sisted. Over the course of the five-year life of CORS, we began to see these findings about school size as important in their own right: Both effectiveness and equity were shown to be higher in small high schools. As size was not an explicit item on the CORS agenda, however, we did not pursue that is- sue further in that venue. Within the format of those studies, we investi- gated size as a linear effect. Because the size vari- able is positively skewed (with a large number of small schools), we subjected it to a natural loga- rithmic transformation for inclusion in analyses that made stringent assumptions of normality for dis- tributions for continuous variables. Consistently significant negative coefficients for "log size" on student learning led to an interpretation that "smaller is better." However, we recognized that the true re- lationship probably is not linear. It seemed reason- able to explore whether there may be an "ideal" high school size where both effectiveness and eq- uity are maximized for a given outcome. It adds an additional layer of complexity to consider whether the effects of size are consistent across different students and on learning in more than one subject. These are the questions we pursue here. Research Background Arguments Underlying Research on School Size Two research strands. Research on size, a stan- dard ecological feature of educational organiza- tions, falls into two categories. Most studies have targeted high schools. The first research stream, that reflects an economy-of-scale argument, focuses on the potential for increased savings through reduced redundancy and increased resource strength as schools get bigger. The second strand directs at- tention toward how size influences other organiza- tional properties of schools. As schools grow, it is natural that they become more formal and bureau- cratic. Certain consequences flow from such changes, including a typically more specialized instructional program. Conclusions from the two streams go in opposite directions: The efficiency argument suggests benefits from increased size, whereas the organizational argument favors smaller schools. Economy ofscale. When considering efficiency in a service-production organization, increasing the numbers of persons served can generate greater 206 Section 9-2 High School Size efficiency under two criteria (Buzacott, 1982). First, increasing numbers of recipients maximize the ef- ficient delivery of a given service. For example, if one goal of a high school is to provide a curricu- lum tailored to particular aptitude levels (i.e., ad- vanced, average, or basic), then more students would help maximize the delivery of this instruc- tion by increasing the numbers of students of simi- lar ability. Other examples apply to curriculum goals targeted to different student interests, special needs, or other selection criteria. Depending on the im- portance of these goals, meeting them efficiently means that the school must have enough students to sustain separate programs or classes. As the num- ber of students with common needs increases, schools can create more specialized programs. The second criterion relates to physical resources. Supplies and materials needed to deliver services are more economically obtained through larger purchases (Buzacott, 1982). If the cost of supplies (such as writing paper) is reduced when purchased in greater amounts or if other costs (such as light- ing or heat) can be sustained at a relatively flat level regardless of the numbers served, then spreading the relatively lower per-person cost over a larger base reduces overall spending on core costs. The economy-of-scale argument may be applied to the cost of "producing" a given level of achieve- ment in students. The argument would lead to con- clusions favoring school consolidation and larger size (Kenny, 1982). The logic is that savings should accrue as core costs are spread over a larger pupil base. The savings can then be applied toward strengthening (or expanding) the school's academic offerings in response to individual differences in interest and ability among students. This shift should result in either a general increase in resource strength, greater program specialization, or both. Program specialization is seen as an advantage within this research. Although this argument as- sumes that greater size results in an economically more efficient operation (Guthrie, 1979; Michelson, 1972), savings projected by proponents of school consolidation have not materialized (Chambers, 1981; Fox, 1981). Large schools usually expand their support and administrative staffs to handle the greater bureaucratic demands. In rural areas (where consolidation continues to be a big issue), higher costs for distributing materials and transporting stu- dents offset any savings (Chambers, 1981). Evidence that size and academic outcomes are positively related is weak, although Bidwell and Kasarda (1975) offer evidence of an indirect rela- tionship. They showed that the availability of re- sources is positively but indirectly related to achievement, with the effect mediated through hir- ing better-trained teachers and more staff to sup- port students' special needs. School and district size are often confused, particularly for high schools (as many districts operate a single high school). The relationship between school district size and re- source availability is inconsistent across commu- nities, contingent on the socioeconomic status of the community (Friedkin & Necochea, 1988). Al- though larger districts in low-income areas typi- cally have access to more resources than small dis- tricts, the higher incidence of "exceptional prob- lems" in such populations introduces constraints in such schools that contribute to lower achieve- ment. Academic and social organization. Recent re- search documents a relationship between organi- zational size and program specialization. In prin- ciple, larger schools have more students with simi- lar needs and thus are better able to create special- ized programs to address those needs. In contrast, small schools must focus resources on core pro- grams, with marginal students (at either end of a distribution of ability or interest) either excluded from programs or absorbed into programs that may not meet their needs as well (Monk, 1987; Monk & Haller, 1993). Research on tracking suggests, however, that extensive differentiation in curricu- lar offerings and students' academic experiences has debilitating consequences (Gamoran, 1989; Oakes, 1985). Increasing size promotes curriculum specialization, resulting in differentiation of stu- dents' academic experiences and social stratifica- tion of student outcomes (Lee & Bryk, 1989). Is increased specialization good or bad? Although specialization fits the ideal of the comprehensive high school, where a major goal of schools is to cater to individual differences among students, an alternate perspective, one that focuses on the more communal aspects of learning, would see special- ization differently. This perspective has motivated some recent empirical work on curriculum effects that links differences in students' academic experi- ences to stratification in academic outcomes (Garet & DeLaney, 1988; Lee & Bryk, 1988, 1989; Lee & Smith, 1993). Private and public schools alter course offerings differently with a change in size. Catholic schools add academic courses as they grow bigger, while public schools typically add courses 207 Section 9-3 Lee and Smith in personal development and other nonacademic areas (Bryk et al., 1993). Basic sociological theory suggests that as an or- ganization grows, human interactions and ties be- come more formal (Weber, 1947). Organizational growth typically generates new bureaucratic struc- tures as connections between individuals becomes less personal. These structures, in turn, can inhibit communal organization (Bryk & Driscoll, 1988). This hypothesis has been substantiated in the re- search studies that identify the organizational char- acteristics of effective schools. In much of the lit- erature on school climate, for example, size oper- ates as an "ecological" feature of a school's social structure, part of the physical or material environ- ment that influences the nature of social interac- tions (Barker & Gump, 1964; Bryk & Driscoll, 1988; Garbarino, 1980; Morocco, 1978). The organizational research strand concludes that smaller school size is beneficial for students in sev- eral ways. The more constrained curriculum in small high schools is typically composed of aca- demic courses, resulting in virtually all students following the same course of study, regardless of their interests, abilities, or social background. This results in both higher average achievement and achievement that is more equitably distributed (Lee & Bryk, 1988, 1989). Social relations are also more positive in smaller schools. The preponderance of sociological evidence about high schools suggests that "smaller is better" (Lee, Bryk, & Smith, 1993). Study objectives. Within any high school, there are clear tensions relating to the number and types of students it serves. Obviously, high schools need to provide some variety in curriculum options based on the interests, competencies, and future plans of their students. An expanded student base would increase a school's ability to provide those options in that more students translate into more resources (e.g., the ability to hire teachers with expertise in different subjects, numbers of students to fill op- tional courses at both ends of the academic cur- riculum). On the other hand, more students also translate into more problems, as well as more bod- ies to monitor. From this concern, constraining unit size may help to promote the human dimensions of schooling. We intend our work to build on empirical and theoretical work that has touched or embraced the issue of high school size. Although most studies have couched the issue within a "bigger-versus- smaller" mode, our objective is to expand on that approach by estimating an appropriate balance point. We identify the balance point by estimating how much students learn as a function of school size, although we recognize that size might influ- ence other outcomes differently. Our first objec- tive is to identify an ideal high school size, defined in terms of student learning. A second objective is to define the optimal size in terms of the equitable distribution of learning within schools. A third ob- jective is to identify whether the ideal size is con- stant across different types of high schools, defined in terms of the social background of the students they serve. Method Data Sample. We use the first three panels of NELS:88, collected by the National Center for Education Sta- tistics, on the same students as 8th-, 10th-, and 12th- graders. Besides survey data from students, their parents, their teachers, and their schools, NELS includes cognitive test scores equated to the same scales from the same students at each wave. NELS:88 represents very appropriate data to pur- sue this study, as size effects may be estimated on achievement growth between the beginning and end of high school for large and nationally representa- tive samples of students and schools. As we focus on students during their high school years, our sample includes those with data at the three waves who stay in the same high school until graduation: 9,812 students in the 789 public, Catholic, and elite private high schools with sufficient data for the analysis methods we use. Details about filters for selecting this sample are described more fully by Lee, Smith, and Croninger (1997). At the base year, NELS selected eighth-graders in middle-grade schools. The design called for fol- lowing these students into high school, but high schools were not sampled directly. The two-stage sampling plan oversampled certain types of schools (private schools and those with high enrollments of Asians and Latinos); thus, analyses required the use of design weights to generalize to the popula- tion of U.S. high schools and students. NCES sup- plied student- and school-level design weights at the base year, but for the follow-ups, they supplied only weights for students. Because our analyses focus on school effects, it is important that we use school weights. A major advantage of NELS, on which we wished to capitalize, is its representa- tiveness. We solved the problem by constructing 208 Section 9-4 High School Size our own design weights for schools.2 Measures. The outcomes in this study measure learning over the course of high school in two sub- jects: mathematics and reading. We chose these subjects because (a) they are important to students' future success, (b) they are very different from one another, and (c) they may be differentially ame- nable to school effects. We limited our investiga- tion to two subjects to simplify the study; NELS also includes test data in science and history. Learn- ing is measured as achievement growth (or gain) in those subjects between the beginning (8th grade) and the end of high school (12th grade). The NELS test scores were scaled using item response theory (IRT) procedures to capture students' pattern of performance in terms of the number correct on an estimated continuum of items scaled by difficulty level and equated across grades and forms. Thus, simple difference scores are not subject to the re- gression to the mean that traditionally has plagued the use of simple gain scores. We follow John Willet's advice here, who suggested that "the dif- ferences score is not the outcast that many critics have claimed" (1994, p. 673). Because additional information improved the scaling procedures over time, the reliability of the estimated scores actually increased from the base year to the second follow-up (from 0.80 to 0.85 in reading, from 0.89 to 0.94 in mathematics-NCES, 1995, p. 67). In addition, the construct validity of the IRT score estimates was investigated by com- paring the pattern of correlations between students' IRT scores and other relevant measures of demo- graphics and process. In general, the scaling pro- cedure resulted in consistently reliable information about students' mathematics and reading levels across the different points of comparison (NCES, 1995). The independent variable of special focus is, of course, school enrollment size.3 Although our ear- lier work used this measure in a logarithmically transformed metric, here we considered school enrollment size in its raw metric. In a preliminary sensitivity analysis, we used the continuous mea- sure, which is highly skewed in a positive direc- tion. In most analyses, we divided the continuous variable into eight categories: 300 students or less, 301-600, 601-900, 901-1,200, 1,201-1,500, 1,501-1,800, 1,801-2,100, and over 2,100 students. These categories were arrived at based on sensitiv- ity analyses (described below). One analysis used two piecewise continuous measures (for smaller and larger schools). Other variables in the models, in- cluded as statistical controls, are also described below, where we lay out our analytic models. De- tails of construction of all variables used in the study are provided in Appendix A. Analytic Models HLM models estimating size effects on learning. The nested structure of the research questions- estimating the effects of school size on student learning--coupled with the NELS data structure suggests the need for a hierarchical linear model (HLM) approach (Bryk & Raudenbush, 1992). We use a two-level HLM structure (students nested in schools). In Level 1 (within schools) we model stu- dents' achievement growth over the four years of high school as a function of the characteristics of students. Outcomes at this level include both learn- ing (gains in achievement in reading and mathemat- ics) and its equitable distribution (the relationship between students' family social class, or SES, and achievement gains). HLM reliability estimates for these gain-score outcomes (lambda) are respect- able: 0.724 (mathematics gain) and 0.556 (reading gain). For all research questions, we use the same Level 1 model, which includes controls for student de- mographics (SES, race/ethnicity, gender) and abil- ity (separate composite measures of achievement at eighth grade in subjects other than the outcome). Student SES is of special interest, as the SES/learn- ing slope is our indicator of social equity within schools.4 In Level 2 models (between schools), the outcomes are average learning in these two sub- jects (Question 1) and the SES/learning slopes in each school (Question 2). The Level 2 HLM mod- els include statistical adjustment for school demo- graphics and sector on both learning and its equi- table distribution. Besides size, Level 2 models control for school SES, minority concentration, and sector (public, Catholic, elite private). Descriptive information on group means for all measures in- cluded in our models is provided in Appendix A. Sensitivity analysis. How, exactly, to model school size effects on learning motivated a set of sensitivity analyses. Decisions about cutoff points for the eight size categories were guided by these analyses. The multivariate sensitivity analyses used a multilevel residual technique. We saved the re- siduals from a two-level HLM model similar to that described above, but without school size, on the four learning and equity outcomes (average learn- 209 Section 9-5 Lee and Smith *.* 2 * * ** *= .. 4-4 -< * * * * ** ** *** ** * o** * *70 **12*5 1*202 * * * . * * S** * *** *- ** * * .*. * * -2 * *tA ** ** * * * S *4 *C * * * ~** * * -3 ** * * * . ** ~* * * * -4 , * * * I I I I I I I I I I 250 500 750 1000 1250 1500 1750 2000 2250 2500 High School Enrollment Key * Mathematics gain residual for one school Figure 1. Distribution of residuals from mathematics achievement gains compared to high school enrollment. ing in math and reading and their social distribu- tions). These residuals (adjusted learning) were plotted against the continuous form of the school size variable. Figure 1 displays the scatterplot for residualized mathematics learning against high school enrollment.5 The scatterplot in Figure 1 indicates that residual mathematics learning (i.e., differences in school average gains after adjusting for student- and school-level covariates) varies by school size and that the relationship is curvilinear. The peak in this graph suggests an optimal size range. Schools whose enrollments range between 500 and 1,000 students appear to be favored in terms of math learn- ing. In schools smaller than that, learning drops. More dramatically, learning is lowest in the largest schools. As the analyses suggested roughly equiva- lent optimal school size ranges across the two sub- jects, we omitted the plot for residualized learning in reading. In general, smaller schools appeared to be more effective in terms of student learning. But it was also evident that the relationship is not linear. To accommodate this in our HLM models, we con- verted the continuous size measure into categories of 300 students/group and dummy-coded them. To estimate the effects of a set of dummy variables, it is necessary to designate an excluded comparison group. We selected the size category of 1,201-1,500 students. Although the comparison group choice is arbitrary, we chose this category because it repre- sented the mean high school size attended by stu- dents in this sample. Differential size effects. Does high school size influence learning differently in schools defined by the types of students who attend them? It seems unlikely that a single optimal size would be appro- priate for all types of schools and students. Ques- tion 3 explores how school size influences learn- ing in schools with different social compositions. Focusing on school average SES and minority con- centration, we pursued an interaction analysis strat- egy to explore the possibility of differential size effects. For average school SES, we created a set of ef- fect-coded interaction terms with the size catego- ries and entered them into the full HLM model along with the set of size dummies. Because the variable measuring the proportion of minority stu- dents in high schools also is not normally distrib- uted, we created a dichotomous variable whereby schools that enroll 40% minority students or more were coded "1" and schools with less than 40% were coded "0." Because minority concentration was thus a dummy-coded variable, it would result in a large number of effects-coded categorical in- teraction terms. When included in models together with the main-effect dummies for size, the HLM models became quite unstable. We solved this prob- lem by creating two piecewise linear terms for 210 Section 9-6 High School Size school size and computed product terms of each with minority concentration.6 Presentation of results. Descriptive results are presented as subgroup means on all variables used in the study, separated by school size groupings. The multivariate models exploring our research questions included multiple quantitative analyses and a large number of quantitative estimates of ef- fects within each analysis. We wanted to display our multivariate results in a form that minimized the amount of numbers and focused on school size effects. Thus, we chose to present the HLM results in graphic form. All graphs that represent school size effects are constructed from the full HLM models described above. These models include sta- tistical adjustment for a large set of control mea- sures describing both students and schools. School size effects are presented in two different metrics. Those that answer Questions 1 and 2 are presented in a between-school effect size (standard deviation or SD) metric that compares learning in schools of various sizes to schools enrolling 1,200- 1,500 students. For Question 3, we present results as adjusted group means in average gain-score points on the NELS reading and mathematics tests for each school size category. We selected the graphical mode of presentation because it tells "the size story" in a form understandable to a nontech- nical audience. For readers interested in technical details of analyses and in the effects of control vari- ables, we provide numerical results of the full HLM analyses of our research questions in Appendix B. Results Descriptive Analyses Characteristics ofstudents and schools by school size. The distribution of size for the high schools in our NELS sample is positively skewed with a me- dian size of about 1,200. Although there are quite a few small schools in the sample and even more in the population, of course, more students in the popu- lation attend large schools. Table 1 displays TABLE 1 Means of Variables Describing Students and Schools for Several Categories of High School Size (n = 9,812 Students in 789 Schools) A. Means of variables describing students School size Below 301- 601- 901- 1,201- 1,501- 1,801- Over 300 600 900 1,200 1,500 1,800 2,100 2,100 Student sample 912 830 1,667 1,645 1,319 1,205 1,263 971 1. Outcomes Mathematics gain 8.91 12.13 15.69 13.44 12.20 11.61 10.18 7.84 Reading gain 4.54 6.28 7.61 6.46 5.05 4.60 4.34 3.45 2. Control variables Ability, matha 0.03 0.17 0.17 0.18 0.12 0.18 0.05 0.11 Ability, readinga 0.05 0.21 0.14 0.19 0.13 0.21 0.07 0.15 % female 52.8 51.5 47.9 49.9 52.7 52.4 52.9 50.4 % minority 14.5 24.3 14.3 18.0 16.6 15.6 23.5 21.5 Social classb -0.12 0.07 0.11 0.05 0.03 0.08 -0.04 -0.06 B. Means of variables describing schools School size Below 301- 601- 901- 1,201- 1,501- 1,801- Over 300 600 900 1,200 1,500 1,800 2,100 2,100 School sample size 75 67 148 139 83 70 101 106 Average SESb -0.21 0.09 0.18 0.08 0.09 0.18 -0.15 -0.32 % high minorityc 20.3 26.9 16.3 21.2 15.8 14.5 26.1 33.3 % public 95.0 92.5 75.5 81.2 90.8 89.4 92.8 95.9 % Catholic 2.5 4.5 10.9 12.2 6.6 6.6 0.9 3.1 % independent 2.5 3.0 13.6 6.6 2.6 4.0 6.3 1.0 aStudents' average achievement at 8th grade in the three other subjects used as a proxy measure of ability, mean (M) = 0, SD = 1. bVariables are z-scored at M = 0, SD = 1 on this sample. "cSchools with more than 40% minority students (Black or Latino) coded 1, others coded 0, due to non-normal distribution. 211 Section 9-7 Lee and Smith 2.0 1.8 - Mathematics 1.6 - Reading 1.4 1.2 1.0 Effects 0.8 on 0.6 - Gains in 0.4 Achievement 0.2 <300 1201-1500* 1501-1800 1801-2100 >2100 (8th-12th 0.0 grade) -0.2 inSD301-600 601-901 901-1200 Units -0.6 -0.8 -1.0 -1.2 High School Enrollment -1.4 -1.6 -1.8 -2.0 * 1201-1500 students was used as the comparison group; thus by definition effect sizes are zero. FIGURE 2. Effects of high school size on achievement gains in mathematics and reading. unweighted sample sizes and weighted means of the variables included in this study by the eight size groupings. Variables are grouped by whether they describe students (panel A) or schools (panel B). Because a close-to-fixed number of students was sampled in each NELS school as part of the origi- nal sampling design, both student and school sample sizes are reasonably well distributed across schools of different sizes. There are, however, some- what more students and schools in the moderate size categories.7 In general, panel A of Table 1 shows that learn- ing gains are largest in moderate-sized to small schools, although not in the smallest ones. How- ever, such schools also enroll somewhat more able and higher-SES students. School factors are more varied by size categories (panel B). Average SES is somewhat higher in the moderate-sized schools (600-900 and 1,500-1,800 students), close to av- erage in the middle categories, and lowest in very small and very large high schools (group differ- ences under 0.2 SD). Not surprisingly, minority concentration is highest in larger schools. The large majority of schools in all categories are public; pri- vate schools are more common (but still under 25%) in the 600-1,200 range, which may explain the higher SES and ability means for these groups.8 Differences in both student and school characteris- tics by size, while not large, suggest the importance of taking such differences into account in estimat- ing size effects on learning. Multivariate Models: Which Size High School Works Best? Effects on learning in mathematics and reading. Results of analyses assessing the effects of high school size on achievement gains in mathematics and reading over the course of high school are dis- played in Figure 2. Effects, presented in a between- school effect-size (SD) metric,9 were estimated in a two-level HLM model that includes adjustment for all the characteristics of students and schools listed in Table 1. We interpret effect sizes (ES) as large if 0.5 SD or more, moderate if 0.3-0.5 SD, small in the 0.1-0.3 SD range, and trivial if less than 0. 1 SD (Rosenthal & Rosnow, 1984). Because the contrasts are somewhat arbitrary, we do not fo- cus on statistical significance. Our discussion fo- cuses, rather, on the relative magnitudes of school size effects displayed in Figure 2. Size effects are larger for learning in mathemat- ics (dark bars) than in reading (light bars), though the pattern is similar for both subjects. All effects 212 Section 9-8 High School Size are compared to the 1,201-1,500 size category that, by definition, has no effect. These results indicate that students who attend high schools that enroll between 600 and 900 students have optimal learn- ing. Gains are less in smaller schools (particularly those with less than 300 students); learning is also considerably less in large schools (with more than 2,100 students). Effect sizes are very large for math- ematics learning (over 1 SD in two cases); moder- ate effects accrue for gains in reading comprehen- sion. Though the 600-900 category contains some- what more private schools (about one fourth of schools in this category are private), school sector is taken into account in computing these effects.'1 Effects on the equitable distribution of learning. The effects of school size and the equity param- eters-within-school slopes of SES on learning in mathematics and reading-were estimated simul- taneously in the same HLM models as the effects on learning shown in Figure 2. We display these effects separately (Figure 3), however, because the findings here are somewhat different. In virtually all schools, the relationship between SES and achievement growth is positive; higher-SES stu- dents learn more. Thus, by definition, negative size effects are more equitable, as they reflect a de- creased relationship between SES and learning. In general, size effects are larger on equity than on learning (mainly because the SD of the SES/gain slopes used to compute these effects is smaller, as described in note 9). Although size effects on learn- ing are larger for mathematics than reading, size effects on equity are generally more substantial for reading. The pattern here is also clear, but somewhat dif- ferent: Learning is distributed more equitably in smaller schools. The pattern of school size effects on the SES/gain slopes is generally linear, rather than exhibiting a special advantage for high schools of moderate size. In reading, the equity advantage is largest in schools with 300-600 students (ES of about -3 SD); in mathematics, the largest equity advantage occurs in the smallest high schools (ES around -1 SD). As with size effects on achieve- ment gains, learning is distributed more inequita- bly in the largest schools (especially in reading). Readers interested in effects of each variable in the full HLM models (and in nominal significance lev- els) may consult Appendix B-2 for the numerical results for the full Level 2 HLM models on math- ematics and reading gains and slopes shown in Fig- ures 2 and 3." 4.0 3.6 3.2 2.8 High School Enrollment 2.4 2.0 Effects 1.6 on 1.2 SES/ 0.8 Gain <300 301-600 601-901 901-1200 0.4 Slope 0.0 (8th-I2th -0.4 --1201-1500* 1501-1800 1801-2100 >2100 grade) -0.8 in SD -1.2 Units -1.6 -2.0 -2.4 SES/Math Gain -2.8 -3.2 [ SES/Reading Gain -3.6 -4.0 *1201-1500 students was used as the comparison group; thus by definition effect sizes are zero. FIGURE 3. Effects of high school size on the relationship between SES and achievement gains in mathematics and reading. 213 Section 9-9 Lee and Smith 16.0 15.0 - * Low-SES Schools ] High-SES Schools 14.0 Gains in 14.0 Mathematics 13.0 Achievement 12.0 (8th-12th grade) 11.0 10.0 - 9.0- 8.0 7.0 6.0 5.0 0.0 <300 300-600 600-900 900-1200 1200-1500 1500-1800 1800-2100 >2100 High School Enrollment FIGURE 4. Average gains in mathematics achievement by high school size in low-SES and high-SES high schools. Multivariate Models: Which Size High School Works Best for Whom? Size effects in low- and high-SES schools. With HLM, we also estimated whether the effect of school size is constant across schools with differ- ent social compositions. Our first analyses investi- gate how school size effects on learning vary by the social class composition of a high school. We approached this task by creating an interaction term for each school size category with school average SES and included these in the HLM analyses on learning in each subject. Rather than presenting the results here in effect-size units, we display adjusted gains for schools in each size and school SES cat- egory. We designated low-SES schools as those whose average school SES is one SD below the sample average for school SES. Similarly, high- SES schools are those with an average school SES one SD above the sample mean.12 We display the results of these analyses for students' gains in math- ematics achievement in Figure 4. Although we con- ducted an identical analysis for achievement gains in reading, the interaction effects were not statisti- cally significant for that outcome. Thus, we focus our discussion on patterns identified for learning in mathematics. Three findings are evident from Figure 4. The first is unsurprising, although noteworthy and trou- bling. Students learn considerably more mathemat- ics in higher-SES schools (light bars) than in lower- SES schools. Recall that these differences in learn- ing are computed with statistical adjustment for sev- eral other social characteristics of students and schools taken into account. The second finding is more surprising. The optimal school size is quite similar in both low- and high-SES schools. That is, schools in the 600-900 enrollment range have the highest achievement gains in both groups. Students who attend schools that are larger or smaller than this optimal size don't learn as much mathematics. The third finding is the most striking and the most important. School size appears to matter more in schools that enroll less-advantaged students. Al- though learning differences are notable for low- and high-SES schools of 600-900 students (about 2 points of gain on a 40-point test), in schools with less than 300 students, this difference is larger (about 3.5 points). In the largest schools, the dif- ferences in learning are striking (about 5 points). We also know from Table 1 that the average school SES in very large and very small high schools is low (-0.32 SD in the largest schools, -0.21 SD in the smallest schools). Our findings suggest that 214 Section 9-10 High School Size 16.0 Low Minority Enrollment (<40%) * High Minority Enrollment (>40%) 15.0 14.0 Gains in 13.0 Mathematics 12.0 Achievement (8th- 11.0 12th 10.0 - grade) 9.0 8.0 7.0 6.0 5.0 4.0 3.0 0.o - <300 300-600 600-900 900-1200 1200-1500 1500-1800 1800-2100 >2100 High School Enrollment FIGURE 5. Average gains in mathematics achievement by school size in high schools with low- and high-minority enrollments. large numbers of socially disadvantaged students attend high schools of a size where, in fact, stu- dents like them appear to learn the least. Size effects in schools with high and low minor- ity enrollments. We also explored whether school size effects were constant between schools with student bodies of high and low minority concen- tration. The distribution of the variable measuring the proportion of minority (Black and Latino) stu- dents in U.S. high schools is bimodal: Large pro- portions of high schools enroll very few minority students; smaller proportions of high schools en- roll mostly minority students. There are relatively fewer schools in the middle of the distribution (mixed-race or integrated schools). This distribu- tion, which reflects a substantively important pat- tern of school segregation that has persisted for many decades, suggested creating a dummy-coded variable to tap minority concentration (see Appen- dix A). The size-by-minority concentration interaction terms are somewhat different. We used a two-piece linear model to capture large- and small-sized schools (see Appendix A for exact codings and Bryk & Raudenbush, 1992, for another application for these codings). We computed product terms of these variables with an effect-coded form of the minor- ity concentration dummy variable. Using the size codings in the two piecewise terms, we calculated average gains in each size category for schools with low and high minority concentrations. The results of the HLM analysis for gains in mathematics are displayed in Figure 5; reading gains are in Figure 6. In both subject areas, the interaction terms be- tween school minority concentration and school size were statistically significant. Mathematics. The differentiation of learning gains in mathematics in schools with low minority enrollment (light bars) and high minority enroll- ment (black bars) is less striking than the contrasts by school SES in Figure 4. Again unsurprising (but troubling) is the finding that students learn less mathematics in schools with more minority stu- dents. As we saw for school SES, the optimal size for schools with differing minority concentrations is the same, although with these analyses, the peak is for schools in the 900-1,200 range.13 Very small schools with high minority enrollments seem to show slightly higher gains. We know from Table 1 that very small schools enroll fewer minority stu- 215 Section 9-11 Lee and Smith 10.0 - Low Minority Enrollment (<40%) * High Minority Enrollment (>_40%) 9.0 Gains in 8.0 Reading Achievement 7.0 (8th-6.0 6.0 - 12th grade) 5.0 4.0 3.0 2.0 1.0 0.0 -9 0-1 - <300 300-600 600-900 900-1200 1200-1500 1500-1800 1800-2100 >2100 High School Enrollment FIGURE 6. Average gains in reading achievement by school size in high schools with low- and high-minority enroll- ments. dents. As we saw in Figures 3 and 4, the most so- cially differentiating environments are large. Very large schools with high minority enrollments have quite low learning gains, and differences are great- est in such schools. Clearly, large schools are quite problematic environments for learning, especially those that enroll high proportions of minority and low-SES students. Reading comprehension. Although our analy- ses discovered no interactions for school size and average school SES on learning in reading, such interactions were statistically significant with school minority concentration (see Figure 6). In general, the patterns are the same on the two outcomes. For schools with both low and high concentrations of minority students, students in schools in the 600- 1,200 size ranges learn most in reading. There are especially large learning differences in the largest schools. Especially for high-minority schools en- rolling over 1,800 students, on average, students gain little in reading comprehension over the course of high school. In the very largest schools, regard- less of minority concentration, students gain almost nothing. The actual magnitudes of the gains are lower in reading than in mathematics; this is surely an artifact of the relative length of the two tests (21 items on the reading test, 40 items on the math- ematics test). Discussion Effects of School Size on Student Learning We summarize the findings from this study in four general conclusions about the optimal size of high schools. "Optimal" is defined in terms of stu- dents' learning over the course of high school in reading comprehension and mathematics. The dis- cussion is organized as follows. First we present the conclusions that flow directly from the study, drawing on Figures 2-6. This is followed by a sum- mary of recommendations about school size in some important writings about high schools. We close with a discussion of issues underlying the relationship between size and learning. Conclusion 1: High schools should be smaller than many are. The results from the multivariate hierarchical analyses in this study take into account many demographic characteristics of students and structural characteristics of schools other than school size. Results shown in Figure 2 provide clear evidence of a learning advantage for students at- tending moderate-sized high schools, although size effects are not identical for learning in the two sub- 216 Section 9-12 High School Size ject areas considered here. Size effects also differ somewhat for effectiveness (i.e., learning levels) and equity (i.e., the distribution of learning by student SES). Figure 3 suggests that smaller schools are more equitable. We feel confident in concluding that many high schools should be considerably smaller than they are if we wish to maximize learn- ing in the nation's schools. Students learn more in relatively small high schools; learning is more eq- uitable in small places. Conclusion 2: High schools can be too small. One motivation for this study was to investigate current policy claims that smaller high schools are better. As mentioned, it seems logical that high schools could be too small to offer adequate aca- demic programs to students (unless resource bases are very high and their client base quite homoge- neous). The results shown in Figure 2 confirm that. Students learn less in high schools with fewer than 600 students, as well as in very large ones. How- ever, learning is more equitably distributed in very small schools. Obviously, the aims of effectiveness and equity are not completely parallel. In general terms and considering both outcomes, our results lead us to recommend an enrollment size of be- tween 600 and 900 students as "ideal" for a high school. Conclusion 3: Ideal size does not vary by the types ofstudents who attend. Some of our analyses investigated whether our recommendation about an ideal size should be generalized to schools defined by their students' differing social characteristics. This issue is important because there is a tendency for socially disadvantaged students to be educated in very large or very small schools (see Table 1). Our investigations examined whether either smaller or larger high schools would be more advantageous for schools that enroll different types of students. We focused on schools differentiated by their so- cial class and minority concentrations. We were somewhat surprised that the same pattern of results was evident: Schools whose sizes fall in the mod- erate size range (600-900 students) produced greater achievement gains for low- and high-SES schools and for schools with low and high minor- ity concentrations. Thus, our recommendation for the ideal size of a high school (Conclusion 2) holds across schools regardless of their students' social class and ethnic backgrounds. Conclusion 4: Size is more important in some types of schools. Although high school size is an important determinant of learning for all students, it seems to matter for some students more than for others. Our results indicate that size is especially important for the most disadvantaged students. Fig- ures 4-6 indicate that learning for these students falls off sharply as the schools such students attend become larger or smaller than the ideal. These find- ings are important because minority students are most likely to attend large schools, and students of lower social class are likely to be found in either large or very small schools (Table 1 shows this). We argue that this conclusion is especially impor- tant if we wish to increase social equity in educa- tional outcomes in America's secondary schools. Popular Writings About School Size The issue of high school size has received much attention in theoretical and popular writings about education, as well as in reports spelling out ideas for reforming schools. Empirical research on the topic is, however, neither numerous nor strong. Although we reviewed relevant empirical work at the beginning of this article, reflecting on our re- sults led us to consider a broader range of writings. Our conclusions about the ideal size of a high school seem to be in line with recommendations about high school size made by other scholars, although the latter were not drawn directly from empirical analy- ses. One example is James Bryant Conant, acknowl- edged as the father of the comprehensive high school. In his influential 1959 book about the American high school, Conant indicated that a school with a graduating class of 100 should be sufficiently large to implement his recommended curriculum (although he favored schools somewhat larger than this). Quite obviously, contemporary comprehensive high schools are considerably larger than Conant's minimum. John Goodlad wrote about high schools almost three decades after Conant. In A Place Called School, he commented: "The burden of proof, it appears to me, is on large size. Indeed, I would not want to face the challenge of justifying a senior ... high of more than 500 to 600 students (unless I were willing to place arguments for a strong foot- ball team ahead of arguments for a good school, which I am not)" (Goodlad, 1984, p. 310). In an essay about size and adolescent development, Garbarino (1980) argued for the particular impor- tance of school size for marginal students. Echo- ing Barker and Gump's (1964) study, he described a threshold effect whereby advantages from in- creases in high school size over about 500 students 217 Section 9-13 Lee and Smith were minimal. Neither Conant, Goodlad, nor Garbarino provided any evidence for their num- bers. Reflecting the theme of differential effects of size for students of varying backgrounds, Bryk, Lee, and Holland did present evidence that school size has more influence on social equity than on achieve- ment in Catholic and public high schools. Without making a specific size recommendation, they con- cluded: "Quite simply, it is easier to create a more internally differentiated academic structure in a larger school" (Bryk et al., 1993, p. 270). Although the Coalition for Essential Schools has made no specific recommendations about high school size, Theodore Sizer, in Horace's Compromise, included "keep[ing] the structure simple and flexible" among the five "imperatives for better schools" (1984, p. 214). Over the last few years, the Carnegie Founda- tion has sponsored two very influential reports on school reform. Their 1989 report, Turning Points, spelled out policies for changing middle-grade schools. The first recommendation was to "create small communities for learning" (Carnegie Coun- cil on Adolescent Development, 1989, p. 9). The report gave no explicit guidelines for middle-school size, but listed such elements as "schools-within- schools or houses" as key (p. 9). Carnegie's most recent policy statement on school reform is the 1996 report Breaking Ranks. Using the word "personal- ization," terminology identical to a major principle for the Coalition for Essential Schools, the first of the report's six major themes recommended that "High schools must break into units of no more than 600 students so that teachers and students can get to know each other" (NASSP, 1996, p. 5). A recent and popular book described a "radical" school reform effort in Philadelphia, where 90 small charter schools were created within the city's 22 comprehensive high schools (Fine, 1994). Though the small Philadelphia schools were specialized in some sense (as charters typically are), the tenor of the book definitely favors small high schools and emphasizes the communal environment fostered within them. Typical of charter school staff was one teacher's description of the effect of expanding the size of the charters (from 200 to 400 students): "[T]he seams of the charters feel too tightly stretched" (p. 131). The major worry, however, fo- cused on deterioration in social relationships within and between groups of students and teachers. Learn- ing was not a major focus of Fine's book. These writings, most quite recent, draw out a consistent theme: High schools should be smaller than they are. Our first conclusion supports this theme. A major assumption underlying suggestions for reducing high school size is that human rela- tions in smaller schools will be more personalized. We are struck with the consistency of the recom- mendations for an ideal size (600 seems very popu- lar), although our scrutiny of these writings uncov- ered little empirical support for that specific rec- ommendation. For sure, not every policy recom- mendation requires specific evidence to support it (some rest on solid moral ground). But we wonder how these writers arrived at such specific and con- sistent recommendations. We were also surprised that these writings did not seem to recognize that perhaps a high school could be too small. Although our findings do cen- ter somewhere around the same number as an ideal size, they also suggest that very small high schools might not be advantageous for their students' learn- ing. If either personalization or equity were ends in themselves, then "the smaller the better" would probably hold true. However, it is difficult to over- look that the major aim of schools in general, and high schools in particular, is (and we contend that it should be) to increase learning. Thus, we won- der why these writers don't worry about very small size. Importance of Findings for Policy and Practice A causal link? Does a reduced school size really "cause" students to learn more? Although the struc- ture of our analyses of enrollment size on student learning would imply this, we are cautious about drawing a direct causal link between the number of students a high school serves and how much stu- dents learn in school. Rather, we suspect that size acts as a facilitating or debilitating factor for other organizational forms or practices that, in turn, pro- mote student learning. At the outset, we described two conflicting theo- ries about school size, one of which focuses on curriculum. Increasing size makes it easier to offer a specialized high school curriculum, which in turn allows schools to differentiate what their students learn-to better respond to individual differences. We mentioned that smaller schools (especially Catholic schools) are somewhat more likely to of- fer a core curriculum that all (or most) students may follow, regardless of their abilities or aspirations. This type of curriculum responds more to common 218 Section 9-14 High School Size needs than individual differences. As our findings favor smaller schools (but not too small), we sug- gest that there is a balance that might favor enough courses to serve students well, but not too many to foster differentiation. Another theoretical focus of school size writings is social relations. This theory clearly favors small schools; social relations between school members are likely to be more collegial (among teachers or between teachers and administrators) and more personalized (between teachers and students, among all school members). Goodlad (1984) raised another policy issue important among school people and within the communities served by schools: sus- taining winning sports teams. Despite its impor- tance to many constituents of U.S. high schools, we are hesitant to raise this concern to the level of theory. However, the extracurriculum in any high school, and students' participation in it, is an im- portant element in the high school experience. And it is surely influenced by school size. In fact, it is reasonable to hypothesize that school size influ- ences many other outcomes besides cognitive de- velopment, e.g., social relations, students' engage- ment with learning, self-esteem, sense of belong- ing, participation in extracurricular activities, and leadership roles. Moreover, the ideal size is likely to be quite different for other outcomes. In this study, however, we focused our attention on learn- ing in two subjects over the course of high school. Although our analyses support the presence of a direct link between high school size and student learning, logic argues otherwise. More likely, our findings about size represent a proxy explanation for basic features of the organization and process of high schools: the character of the curriculum, relationships among school members, and the extracurriculum. We are pursuing how size influ- ences these structural outcomes in field-based re- search. To understand the school size effects we have shown, we suggest that the effects on learn- ing are probably indirect, mediated by their influ- ence on basic features of the academic and social organization and functioning of schools (variables that were not in our models). Under this explana- tion, size serves as a facilitating or inhibiting factor for fundamental educational processes in schools. On the other hand, policymakers might argue that changing the size of a school is much easier than altering its basic organizational features (particu- larly if such change doesn't involve the cost of build- ing new schools). Empirical results might really influence public policy. High school size, and its effects on students, is one topic of empirical research that the general public can understand. Social policy may be out in front of solid empirical research in this instance. A series of recent front-page articles in the New York Times (Dillon, 1995; Dillon & Berger, 1995; Firestone, 1995; Gonzalez, 1995) presents inter- esting stories about several of the 46 small and ex- perimental high schools that opened in New York City over the previous two years. The major crite- rion defining these schools is smallness (in the 110- 660 range). Reflecting one of the themes we men- tioned-social relations-Joseph A. Fernandez, the former New York City Schools chancellor, decried that "Our high schools were just too large, and there were a lot of problems with kids not feeling people even knew who they are," as he launched the move- ment in 1992 (Dillon & Berger, 1995, p. B11). According to the Times articles, 50 more small schools were on the drawing boards in New York, with support for the movement from the $50 mil- lion educational grant from the Annenberg Foun- dation to New York City. New York and Philadel- phia are just two of many cities on the "small school bandwagon." These developments, where changes are proceed- ing without research that supports them, suggest (a) the timeliness of empirical work on this topic and (b) an unusual receptivity among practitioners to research results that offer post-hoc support for their decisions (the changes often occur before re- search results are known). These policy develop- ments also indicate that a move to small schools may actually result in a number of schools that are probably too small to be effective for their students' learning. This is one issue about which scholars do not have to argue for the importance of research to mobilize school professionals toward reform. In this case, reform efforts are in full gear. How do we change school size? Clearly, the New York experiment (with generous foundation sup- port) represents one way to approach changing the average size of a high school: Create brand-new schools (or smaller schools within the walls of larger existing schools). Our findings suggest that this approach, opening many very small schools, might not be wise. In fact, the New York Times series re- ported several problems in these high schools. Given the present fiscal environment and modest public support for investment (financial or psychic) in social betterment, it seems unlikely that 219 Section 9-15 Lee and Smith America's public school districts will embark on a new building campaign to create many new smaller high schools. This is especially unlikely in our larg- est cities, where financial resources are particularly scarce. A reasonable alternative to building new schools is a movement to create a set of smaller schools- within-schools inside larger high schools. In fact, this movement is now flourishing.14 This policy seems to us a reasonable approach to breaking up large school units, which our study has shown are especially problematic places for learning. How- ever, we suggest a few cautions that policymakers should consider if they wish to adopt the schools- within-schools approach to reducing unit size. First, it is quite important that the actual size of the re- sulting units be considered. Our research suggests that quite small units may be problematic. Consid- eration should be given to our findings about the "ideal" size. Second, we are concerned that deci- sions to create schools-within-schools might be used as a way to create a number of "specialty shops" (Powell, Farrar, & Cohen, 1985)-to dif- ferentiate students and their high school experiences by ability, vocational focus, or other organizational means. We worry that this kind of specialization in smaller units could create further social stratifica- tion in educational opportunities and outcomes, a "side effect" we feel should be avoided. Rather, each small unit should reflect the demographic diver- sity of the school as a whole. School size and student disadvantage. We close with a call to emphasize the special importance of high school size for economically disadvantaged and minority students. U.S. policy and custom about which students attend which schools relates such decisions to the local level. Usually access to schools is based on where students live. Residen- tial segregation in the U.S. is increasing rather than decreasing over time (Farley & Frey, 1994), and de facto school segregation by race and class are now common and seemingly acceptable to the Ameri- can public. Secondary school students of color, and those who come from low-income families, tend to be concentrated in large U.S. public schools with others quite like themselves (at least demographi- cally). Our findings suggest that size is much more important for learning in schools with high con- centrations of disadvantaged students. Thus, schools with many minority or low-income students (often the same schools) should be especially anxious to reduce the size of the units in which their students actually learn. At the very least, the results from this study sug- gest that the size of a high school influences, di- rectly and/or indirectly, how much students learn. Our results favor moderate-sized high schools, nei- ther so small that the curriculum students experi- ence is inadequate nor so large that some students drop through the cracks in some schools to create socially stratified learning experiences. Students most likely to slip through the cracks and to end up at the low end of the curriculum are those with an economic or ethnic disadvantage, and our study indicates that these are exactly the students for whom "ideal size" is most important. APPENDIX A Description of Variable Construction for All Measures Used in the Study of School Size and Learning Dependent Measures ofAchievement Gains Achievement gains Mathematics gain between 8th and 12th grades was constructed as a simple difference in scores between * BY2XMIRR-Mathematics IRT-estimated number right (8th grade). * F22XMIRR-Mathematics IRT-estimated number right (12th grade). Reading gain between 8th and 12th grades was constructed as a simple difference in scores between * BY2XRIRR-Reading IRT-estimated number right (8th grade). * F22XRIRR-Reading IRT-estimated number right (12th grade). School Size School Size * F1C2-Total enrollment as of October 1989. Principal's report of high school size (on NELS restricted school file). * School size categories (300 and below, 301-600, 601-900; 901-1,200, 1,201-1,500, 1,501-1,800, 1,801-2,100, over 2,100) were constructed from F1C2. * Two piecewise size terms were computed as fol- lows. First, enrollment was centered at 900 students (i.e., 900 was subtracted from each school's en- rollment size). The first linear term, representing smaller schools, was continuous up to 0 and coded 0 thereafter. The second term, representing larger schools, was coded 0 for all schools smaller than 900 (the break point) and continuous thereafter. To 220 Section 9-16 High School Size compute the values in Figure 6 (to be comparable to other displays in the article), the means were re- computed to their actual figures. Control Variables Student Background (within-school controls) Socioeconomic Status * F2SES 1-Socioeconomic status z-scored com- posite. Minority Status * F2RACE1--Student race (recoded to: 0 = White or Asian; 1 = Black, Latino, or Native American). Gender * F2SEX-Student gender (recoded to: 0 = male; 1 = female). Academic Controls Analyses included different controls for the two curriculum areas. Controls were constructed as fol- lows: * For math gain: Z-score of sum of BYTXRIRS, BYTXHIRS, BYTXSIRS. * For reading gain: Z-score of sum of BYTXMIRS, BYTXHIRS, BYTXSIRS. School Demographics and Structure (between- school controls) Average Socioeconomic Status * AVSES- SES composite, aggregated to the school level. Minority Concentration * FIRACE-Student race (recoded to: 0 = White or Asian; 1 = Black, Latino, or Native American), aggregated to the school level, and recoded to: 1 = 40% or more, 0 = less than 40% minority. Sector Created from G10CTRL2, the school control measure on the NELS first follow-up restricted school file. Schools that were public, Catholic, or NAIS (members of the National Association of In- dependent Schools) were retained; other private schools were dropped. Created two dummy-coded variables: * CATHOLIC--Coded 1 for Catholic, 0 for pub- lic, NAIS schools. * NAIS--Coded 1 for NAIS, 0 for public, Catho- lic schools. APPENDIX B-1 Average Gains in Reading and Mathematics, Weighted and Unweighted, by School Enrollment Category Gains in mathematics Gains in reading Unweighted Weighted Unweighted Weighted Size category M (SD) M (SD) M (SD) M (SD) 300 or less -0.87 -0.66 -0.34 -0.26 (0.38) (0.28) (0.83) (0.74) 301-600 -0.09 -0.16 0.07 0.05 (0.24) (0.18) (0.80) (0.78) 601-900 1.37 1.38 0.52 0.49 (0.63) (0.56) (0.94) (0.86) 901-1,200 0.61 0.68 0.48 0.44 (0.16) (0.16) (0.88) (0.86) 1,201-1,500 0.07 0.10 0.14 0.19 (0.19) (0.16) (0.99) (0.88) 1,501-1,800 -0.16 -0.09 -0.08 0.06 (0.28) (0.24) (0.96) (0.96) 1,801-2,100 -0.50 -0.58 -0.46 -0.45 (0.22) (0.19) (0.81) (0.76) 2,100 or more -1.57 -1.59 -0.77 -0.89 (0.67) (0.56) (0.92) (0.88) 221 Section 9-17 APPENDIX B-2 HLM Between-School Model for Investigating the Effects of School Size on Gains in Mathematics and Reading (N = 9,812 Students in 789 Schools)a Dependent variables Gain in mathematics Gain in reading achievement, grades 8-12 achievement, grades 8-12 Effects on average between-school achievement gains (intercept) Base estimateb 12.847*** 5.813*** Average SESc 0.408*** 0.262** High-minority enrollment 0.217*** -0.013 Catholic school 0.790 -0.093 NAIS school -0.023 -0.365 School sized 300 or less -0.931*** -0.532* 301-600 -0.089 0.149 601-900 1.512*** 0.539* 901-1,200 0.589*** 0.290 1,501-1,800 -0.152- -0.254 1,801-2,100 -0.415** -0.455* over 2,100 -1.842*** -0.911** Effects on relationship between SES and gains (slope) Base estimateb 1.656*** 1.387*** Average SESc 0.342- -0.720 High-minority enrollment -0.361 -0.043 Catholic school -0.213 -1.092 NAIS school -0.161 -1.382 School sizec 300 or less -1.187- -2.161 301-600 -0.985*** -3.153* 601-900 -0.667- -2.156* 901-1,200 -0.123 -0.487 1,501-1,800 0.984** 2.115* 1,801-2,100 1.481*** 3.795** Over 2,100 1.264** 3.876** HLM-computed SD Intercept 2.276 1.494 SES/gain slope 0.950 0.347 p : .10. *p ? .05. **p 5 .01. ***p ? .001. "aThese HLM effects are estimated using the constructed school-level weight, as described in the text and note 2. bHLM results computed with within-school adjustments for SES, minority status, gender, and 8th-grade ability. "cAll effects (except the average values on the intercept and SES/gain slopes) are presented in a standardized effect-size metric. Effects computed by dividing the HLM gamma coefficient for each outcome by the school-level standard deviation (SD) for that outcome computed from the Level 1 HLM models. These SDs are in the bottom panel of this table. dAll school-size effects are compared to schools that enroll 1,200-1,500 students, which is the excluded category. Section 9-18 APPENDIX B-3 Unweighted HLM Estimates of School Size Effects on Gains in Mathematics and Reading (N = 9,812 Students in 789 Schools)a Dependent variables Gain in mathematics achievement, grades 8-12 Gain in reading achievement, grades 8-12 Effects on average between-school achievement gains (intercept) b School sizec 300 or less -0.292* 0.417- 301-600 -0.469 0.038 601-900 0.473*** 0.630*** 901-1,200 0.347* 0.588*** 1,501-1,800 -0.130 0.012 1,801-2,100 -0.341* -0.320* over 2,100 -0.574* -0.564** Effects on relationship between SES and gains (slope) b School sizec 300 or less -1.220* -1.113 301-600 -0.571 -1.666 601-900 -0.651~ -0.967 901-1,200 -0.602- -1.382 1,501-1,800 0.571- 0.005 1,801-2,100 0.736* 1.937- Over 2,100 0.786* 2.859** HLM-computed SD Intercept 1.494 1.451 SES/gain slope 0.347 0.398 -p 5 .10. *p 5 .05. **p 5 .01. ***p 5 .001. "aThese HLM models include all variables described in the HLM models described elsewhere in this article (Level 1: SES, gender, race/ethnicity, and 8th-grade ability; Level 2: average school SES, minority concentration, Catholic, and NAIS sector). bAs in Appendix B-2, all size effects are presented in a standardized effect-size metric. Effects computed by dividing the HLM gamma coefficient for each outcome by the school-level standard deviation (SD) for that outcome shown in the bottom panel of this table. Note that these are somewhat smaller than those computed in weighted HLM runs (Appendix B-2). cAll school-size effects are compared to schools that enroll 1,200-1,500 students, which is the excluded category. Section 9-19 Lee and Smith APPENDIX B-4 HLM Between-School Model Investigating School Size-by-Average SES Interactions on Gains in Mathematics (N = 9,812 Students in 789 Schools)a Gain in mathematics achievement, grades 8-12 Effects on average between-school mathematics gains (intercept) Base estimateb 10.733*** Average SESc 0.593* High-minority enrollment 0.653** Catholic school -1.793*** NAIS school -0.100 School size main effects (effects-coded)d 300 or less -0.740*** 301-600 -0.075 601-900 0.889*** 901-1,200 0.334** 1,501-1,800 -0.171 1,801-2,100 -0.277- Over 2,100 -1.117 School size-by-average SES interaction termsd < 300 x AVSES -0.089 301-600 x AVSES -0.496** 601-900 x AVSES -0.541 ** 901-1,200 x AVSES -0.446* 1,501-1,800 x AVSES -0.056 1,801-2,100 x AVSES 0.119 > 2,100 x AVSES 0.144 p < .10. *p : .05. **p ? .01. ***p < .001. "aThese HLM effects are estimated using the constructed school-level weight, as described in the text and note 2. Although this analysis also included estimates on the SES/gain slope, and interaction terms on that outcome, those results are not presented here. There were no significant interactions effects on the SES/gain slope. bHLM results were computed with within-school adjustments for SES, minority status, gender, and 8th-grade ability. cAll effects are presented as unadjusted gamma coefficients from HLM, rather than as effect sizes. As described in note 10, in order to consider balanced interaction terms, the school size categories were recorded in an effects-coded metric (1, -1), rather than the dummy coding (1, 0) in the other analyses in this article. The set of interaction terms were created as products between average school SES and each effect-coded school size indicator. The average math gains shown in Figure 4 were computed by summing these main effects and interaction terms for each size category, separately in lower-SES (1 SD below the mean) and higher-SES schools (1 SD above the mean), as explained in the text. dAll school-size effects are compared to schools that enroll 1,200-1,500 students, which is the excluded category. Notes An earlier version of this article was presented at the 1996 annual meeting of the American Educational Re- search Association in New York. This research was sup- ported by a grant from the American Educational Re- search Association, which receives funds for its AERA Grants Program from the National Science Foundation and the National Center for Education Statistics (U.S. Department of Education) under NSF Grant #RED- 9452861. Opinions reflect those of the authors and do not necessarily reflect those of the granting agencies. We appreciate the helpful comments of Richard Shavelson on an earlier draft of this article and help from Anthony Bryk for the sensitivity analyses described in the article. 'The phrase "social distribution of achievement" means, quite literally, the magnitude of difference in achievement related to the characteristics of students' social background-socioeconomic status, race/ ethnicity, or gender. A school in which this relationship is small would be more equitable, in that the difference in achievement in that school would be less strongly dif- ferentiated by students' background characteristics. For more discussion of this issue, see Bryk, Lee, and Hol- land (1993) or Lee and Bryk (1988, 1989). 2More detail on this procedure is available in Appen- dix A of the article by Lee and Smith (1995) or from the authors. The construction method for the school weights included the probabilities of (a) the sector in which stu- dents in each school had spent their 8th-grade year, (b) the total enrollment of each high school, and (c) the ag- gregated student-level weights supplied by NCES. 224 Section 9-20 High School Size 3Because the study relies on an accurate measure of school size available only on the restricted NELS data files, we note that the first author holds a current license from NCES for using NELS restricted data (L- 912050011). The second author holds a separate license through her home university. 4In HLM parlance, SES is set to be "free" and the other within-school control variables are "fixed" (i.e., the variability of the fixed variables is constrained to zero between schools). As such, SES is centered around the group (school) mean, whereas gender, minority status, and ability are all grand-mean centered for this sample. More detail on HLM centering procedures is provided by Bryk and Raudenbush (1992). We control for initial status, or ability, in these analyses by computing a z- scored average of the students' 8th-grade scores in the three NELS tests besides the subject being assessed (e.g., the ability control for gains in mathematics achievement included base-year test scores in reading, science, and social studies). "5Because the patterns for residual learning in reading (as well as science and history, the other subjects tested in NELS) were quite similar, we have not included them here for the purposes of parsimony. 6We created two piecewise size terms (representing small and large high schools). We used a school size of 900 as the cut point for estimating the values for the piecewise terms, as our analyses indicated that 900 was close to optimal. The exact codings for the two piece- wise terms are shown in Appendix A. 7The sample here is almost the same as we used in our studies of school restructuring (Lee & Smith, 1995, 1996; Lee, Smith, & Croninger, 1997). Because HLM was the analysis mode in those studies, the sample at 10th grade was restricted to high schools with at least five NELS students. This selection criterion resulted in dropping many small private schools (particularly those with only one NELS student). The sample for this study is, thus, somewhat biased toward larger schools. However, the number of smaller schools is large enough to support the types of analyses performed here. All students in the sample schools with test scores at 8th and 12th grades were retained. Students dropped from the sample through all these filters were somewhat more advantaged than those retained. Thus, the biases introduced by the sample selection criteria under- rather than overestimate the ef- fects we observe. "8Because the constructed school-level weights in- cluded school size as one component (see note 2), we compared the patterns of achievement gains by school size in weighted and unweighted analyses. Appendix B- 1 displays group-mean comparisons for gains (in a z- score metric). The patterns are generally quite similar except in the smallest schools, where the unweighted gains are somewhat larger. There is a pattern of some- what smaller SDs for weighted than unweighted group means. 9Because our hypotheses focus on school effects in this study, we followed advice from Bryk and Raudenbush (1992, chapter 5) to focus on between- school effect sizes. These were computed by dividing the gamma coefficients for each dummy-coded school size category on achievement gains (or SES slopes on gains) by the between-school SD in the outcome esti- mated in a Level 1 HLM model. The HLM-estimated Level 1 SDs are as follows: math gain, 2.276; reading gain, 1.494; SES/math gain slope, 0.950; SES/reading gain slope, 0.347 (see Appendix B2). The procedure we follow here has been used in many other published stud- ies using HLM, where effects are typically reported in between-school SD (effect-size) units. See Lee and Bryk (1989), Lee and Smith (1993, 1995, 1996), or Raudenbush, Eqmsukkawat, Di-Ibor, Kamali, and Taoklam (1993). Clearly, these effect sizes are larger than a more tradi- tional interpretation of this concept, mainly because the between-school SDs we used to compute them are smaller than the student-level SDs (e.g., within-school SD of math gain is 7.78, of reading gain, 6.80). Some readers may prefer to interpret the magnitudes of the effects of school size in a more traditional effect-size metric. Our findings suggest that school size has impor- tant effects on learning in either metric. We argue, how- ever, that it is only variance between schools that may be explained by school size. Thus, we use between-school effect-size estimates in Figures 2-6, where our results are presented. "MoThe shifts in between-school variance (tau) from the fully unconditional to the within-school model to the size-effects model to the full-school model are provided below (Table 2). This information allows us to estimate the proportion of variance explained at each stage. That proportion, in each case estimated relative to the amount of original variance between schools in the outcome, is provided in parentheses below each variance estimate. "The full HLM models from which the values in Fig- ures 2 and 3 were computed are displayed in Appendix B-2. Unweighted school size effects are displayed in Ap- pendix B-3. We draw three conclusions by comparing results in Appendices B-2 and B-3: (1) Unweighted school-size-effect estimates are somewhat smaller than weighted estimates; (2) the pattern of effects is very simi- lar between weighted and unweighted analyses, even if the magnitudes are somewhat different; and (3) the esti- mated between-school SDs of outcomes from Level 1 analyses are also larger for the unweighted than the weighted HLMs (also reported in note 8). Which results are right? As we argue in the text, school-level case- weighting is necessary because of the NELS sampling design. But we also recognize the inherent difficulties of using school weights here. We have thus included full results so that readers may draw their own conclusions. 12Because each size category was effects-coded (1, - 1), the various interaction terms were computed by mul- 225 Section 9-21 Lee and Smith TABLE 2 Variance Estimates for Successive HLM Models Unconditional Within- Size effects Full HLM model school model model school model variance variance variance variance (explained) (explained) (explained) Outcome Mathematics 7.003 5.221 4.344 2.295 (-) (0.254) (0.380) (0.672) Reading 4.025 2.884 2.737 2.088 (-) (0.283) (0.320) (0.481) tiplying each effect-coded categorical variable by school average SES. Along with the size main effect and the same control variables that were included in analyses for Figures 1 and 2, we included the set of seven interac- tion terms in Level 2 of the HLM analysis. We then com- puted means for each school size category by summing the appropriate terms and substituting either -1 (for low- SES schools) or +1 (for high-SES schools) in these equa- tions. Details of the computations are available from the authors. "3The peak at a slightly different location is probably an artifact of the cut point we used for the two piecewise terms-900 students. Thus, we concentrate more on the general patterns than the actual peak. "4Inf our earlier study (Lee & Smith, 1995), we made linear estimates of school size effects. We suggested in that study that policymakers should consider schools- within-schools (SWS) as a feasible and cost-effective way to reduce high school size. NELS principals were asked to indicate whether they actually had this policy in place in 1990, so we investigated whether the policy related to a school's size (i.e., whether principals were reporting the size of the smaller unit or the larger one). We found that this option was essentially a public school phenomenon (almost no private schools reported it). Among the 672 public schools in our sample, 86 (or 13%) offered SWS. They were larger high schools (average size-1,691) compared to those without the option (av- erage size-1,275). Although the schools with the SWS option enrolled somewhat more minority students (34% versus 24%), selection criteria such as average achieve- ment at high school entry and average SES were very similar in public schools with and without that option. References Barker, R., & Gump, R. (1964). Big school, small school: High school size and student behavior Stanford, CA: Stanford University Press. Bidwell, C., & Kasarda, J. (1975). School district orga- nization and student achievement. American Socio- logical Review, 40(1), 55-70. Bryk, A. S., & Driscoll, M. E. (1988). The school as community: Theoretical foundations, contextual in- fluences, and consequences for students and teach- ers. Madison, WI: Center on Effective Secondary Schools, University of Wisconsin. Bryk, A. S., Lee, V. E., & Holland, P. B. (1993). Catho- lic schools and the common good. Cambridge, MA: Harvard University Press. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: Applications and data analysis meth- ods. Newbury Park, CA: Sage. Buzacott, J. A. (1982). Scale in production systems. New York: Pergamon. Carnegie Council on Adolescent Development. (1989, June). Turning points: Preparing American youth for the 21st century. New York: Carnegie Corporation of New York. Chambers, J. G. (1981). An analysis of school size un- der a voucher system. Educational Evaluation and Policy Analysis, 3, 29-40. Conant, J. B. (1959). The American high school today. New York: McGraw-Hill. Dillon, S. (1995, May 25). Islands of change create fric- tion. New York Times, p. Al, A15. Dillon, S., & Berger, J. (1995, May 22). New schools seeking small miracles. New York Times, p. Al, B 11. Farley, R., & Frey, W. H. (1994). Changes in the segre- gation of Blacks and Whites. American Sociological Review, 59(91), 23-45. Fine, M. (Ed.). (1994). Chartering urban school reform: Reflections on public high schools in the midst of change. New York: Teachers College Press. Firestone, D. (1995, May 24). When teachers unite to run school. New York Times, p. A1, A12. Fox, W. F. (1981). Reviewing economics of size in edu- cation. Journal of Education Finance, 6, 273-296. Friedkin, N. E., & Necochea, J. (1988). School size and performance: A contingency perspective. Educational Evaluation and Policy Analysis, 10(3), 237-249. Gamoran, A. (1989). Measuring curriculum differentia- tion. American Journal of Education, 97, 129-143. Garbarino, J. (1980). Some thoughts on school size and its effects on adolescent development. Journal ofYouth and Adolescence, 9(1), 19-31. Garet, M. S., & Delaney, B. (1988). Students' courses and stratification. Sociology ofEducation, 61, 61-77. 226 Section 9-22 High School Size Gonzales, D. (1995, May 23). A bridge from hope to social action. New York Times, p. Al, A14. Goodlad, J. I. (1984). A place called school: Prospects for the future. New York: McGraw-Hill. Goss, D. (1994). Principles of human resource manage- ment. New York: Routledge. Guthrie, J. (1979). Organizational scale and school suc- cess. Educational Evaluation and Policy Analysis, 1(1), 17-27. Kenny, L. (1982). Economies of scale in schooling. Eco- nomics of Education Review, 2(1), 1-24. Lee, V. E., & Bryk, A. S. (1988). Curriculum tracking as mediating the social distribution of high school achievement. Sociology of Education, 61(2), 78-94. Lee, V. E., & Bryk, A. S. (1989). A multilevel model of the social distribution of high school achievement. Sociology of Education, 62, 172-192. Lee, V. E., Bryk, A. S., Smith, J. B. (1993). The organi- zation of effective high schools. In L. Darling- Hammond (Ed.), Review of research in education (Vol. 19, pp. 171-267). Washington, DC: American Edu- cational Research Association. Lee, V. E., & Smith, J. B. (1993). Effects of school re- structuring on the achievement and engagement of middle-grade students. Sociology ofEducation, 66(3), 164-187. Lee, V. E., & Smith, J. B. (1995). The effects of high school restructuring and size on gains in achievement and engagement for early secondary school students. Sociology of Education, 68(4), 271-290. Lee, V. E, & Smith, J. B. (1996). Collective responsibil- ity for learning and its effects on gains in achieve- ment for early secondary school students. American Journal of Education, 104(2), 103-147. Lee, V. E., Smith, J. B., & Croninger, R. G. (1997). How high school organization influences the equitable dis- tribution of learning in mathematics and science. So- ciology of Education, 70(2), 129-152. Michelson, S. (1972). Equal school resource allocation. Journal of Human Resources, 7, 283-306. Monk, D. (1987). Secondary school size and curricu- lum comprehensiveness. Economics of Education Review, 6, 137-150. Monk, D., & Haller, E. J. (1993). Predictors of high school academic course offerings: The role of school size. American Educational Research Journal, 30, 3- 21. Morocco, J. C. (1978). The relationship between size of elementary schools and pupils' perceptions of their environment. Education, 98, 451-454. Morrison, C. (1993). A microeconomic approach to the measurement of economic performance: Productive growth, capacity utilization, and related performance indicators. New York: Springer-Verlag. National Association of Secondary School Principals (NASSP). (1996). Breaking ranks: Changing an American institution. Reston, VA: Author, in partner- ship with the Carnegie Foundation for the Advance- ment of Teaching. National Center for Education Statistics (NCES). (1995). National Education Longitudinal Study of 1988: Psy- chometric report for the NELS:88 base year through second follow-up (NCES-95-382). Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement. Oakes, J. (1985). Keeping track: How schools structure inequality. New Haven, CT: Yale University Press. Powell, A. G., Farrar, E., & Cohen, D. K. (1985). The shopping mall high school: Winners and losers in the educational marketplace. Boston: Houghton-Mifflin. Raudenbush, S. W., Eamsukkawat, S., Di-Ibor, I., Kamali, M., & Taoklam, W. (1993). On-the-job improvements in teacher competence: Policy options and their ef- fects on teaching and learning in Thailand. Educa- tional Evaluation and Policy Analysis, 15(3), 279- 297. Rosenthal, R., & Rosnow, R. L. (1984). Essentials of behavioral research: Methods and data analysis. New York: McGraw-Hill. Sizer, T. R. (1984). Horace's compromise: The dilemma of the American high school. New York: Houghton- Mifflin. Weber, M. (1947). Theory of social and economic orga- nization (A. M. Henderson & T. Parsons, Trans.). New York: Macmillan. Willet, J. B. (1994). Change, measure of. In T. Husen & T. N. Postlethwaite (Eds.), The international encyclo- pedia of education (2nd ed.) (Vol. 2, pp. 671-678). London: Pergamon. Authors VALERIE E. LEE is a professor of education at the School of Education, University of Michigan, 610 East University, 4220-A, Ann Arbor, MI 48109 (or e-mail velee@umich.edu). Her specialties are the sociology of education and statistical methodology. JULIA B. SMITH is an assistant professor in the De- partment of Learning and Leadership at Western Michi- gan University, Sangren Hall, WMU, Kalamazoo, MI 49008. She specializes in equity and teacher efficacy. Received May 24, 1996 Revision received May 8, 1997 Accepted May 12, 1997 227 Section 9-23 American Educational Research Association High School Size: Which Works Best and for Whom? Author(s): Valerie E. Lee and Julia B. Smith Source: Educational Evaluation and Policy Analysis, Vol. 19, No. 3 (Autumn, 1997), pp. 205-227 Published by: American Educational Research Association Stable URL: http://www.jstor.org/stable/1164463 Accessed: 14/06/2010 12:35 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aera. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. American Educational Research Association is collaborating with JSTOR to digitize, preserve and extend access to Educational Evaluation and Policy Analysis. http://www.jstor.org Section 9-24 Teachers, Schools, and Academic Achievement Author(s): Steven G. Rivkin, Eric A. Hanushek, John F. Kain Source: Econometrica, Vol. 73, No. 2 (Mar., 2005), pp. 417-458 Published by: The Econometric Society Stable URL: http://www.jstor.org/stable/3598793 Accessed: 14/06/2010 12:40 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=econosoc. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. The Econometric Society is collaborating with JSTOR to digitize, preserve and extend access to Econometrica. http://www.jstor.org Section 9-25 This page is intentionally blank Section 9-26 Econometrica, Vol. 73, No. 2 (March, 2005), 417-458 TEACHERS, SCHOOLS, AND ACADEMIC ACHIEVEMENT BY STEVEN G. RIVKIN, ERIC A. HANUSHEK, AND JOHN E KAIN1 This paper disentangles the impact of schools and teachers in influencing achieve- ment with special attention given to the potential problems of omitted or mismeasured variables and of student and school selection. Unique matched panel data from the UTD Texas Schools Project permit the identification of teacher quality based on stu- dent performance along with the impact of specific, measured components of teachers and schools. Semiparametric lower bound estimates of the variance in teacher qual- ity based entirely on within-school heterogeneity indicate that teachers have power- ful effects on reading and mathematics achievement, though little of the variation in teacher quality is explained by observable characteristics such as education or experi- ence. The results suggest that the effects of a costly ten student reduction in class size are smaller than the benefit of moving one standard deviation up the teacher quality distribution, highlighting the importance of teacher effectiveness in the determination of school quality. KEYWORDS: Student achievement, teacher quality, school selection, class size, teacher experience. 1. INTRODUCTION SINCE THE RELEASE of Equality of Educational Opportunity (the "Coleman Re- port") in 1966, the educational policy debate in the United States and else- where has often been reduced to a series of simplistic arguments and assertions about the role of schools in producing achievement.2 The character of this de- bate has itself been heavily influenced by confusing and conflicting research. While this research has frequently suffered from inadequate data, imprecise formulation of the underlying problems and issues has been as important in obscuring the fundamental policy choices. This paper defines a series of basic issues about the performance of schools that are relevant for current policy de- bates and considers how observed student performance can be used to address 1While John Kain participated fully in this project, he sadly died before its publication. We are grateful to Kraig Singleton, Jaison George, and Dan O'Brien for excellent research assis- tance, and we thank Eric French, Caroline Hoxby, Jessica Wolpaw Reyes, Finis Welch, Geoffrey Woglom, and a co-editor, along with seminar participants at UC Berkeley, UC Davis, UC San Diego, the National Bureau of Economic Research, the Public Policy Research Institute, Stan- ford University, University of Texas, and Texas A&M University for their many helpful comments. The arguments and estimation were considerably strengthened by the comments of anonymous referees. Hanushek and Rivkin thank the Donner Foundation, the Smith Richardson Founda- tion, and the Packard Humanities Institute for funding, and Kain thanks the Smith-Richardson Foundation and the Spencer Foundation. 2The original Coleman Report (Coleman et al. (1966)) was subjected to considerable criticism both for methodology and interpretation; see, for example, Hanushek and Kain (1972). The en- suing controversy led to considerable new research, but this new work has not ended the contro- versy; see Hanushek (1996, 2003) and Greenwald, Hedges, and Laine (1996). Those discussions represent the starting point for this research. 417 Section 9-27 418 S. RIVKIN, E. HANUSHEK, AND J. KAIN each. It then employs a unique panel data set of students in Texas to identify the sources of differences in student achievement and the relevance of a broad class of policies related to school resources. Some very basic questions that have arisen from prior work command a cen- tral position in most policy discussions. First, partly resulting from common misinterpretations of the Coleman Report, do schools "make a difference" or not? While a surprising amount of controversy continues over this issue, it comes down to a simple question of whether or not there are significant and systematic differences between schools and teachers in their abilities to raise achievement. Second, how important are any differences in teacher qual- ity in the determination of student outcomes? Finally, are any quality differ- ences captured by observable characteristics of teachers and schools including class size, teacher education, and teacher experience? If so, how large are the effects? This third issue is in fact the genesis of the first, because the Cole- man Report reported relatively small effects of differences in the measured attributes of schools on student achievement-a finding that has frequently been interpreted as indicating that there are no systematic quality differences among schools. An extraordinarily rich data set providing longitudinal information on indi- vidual achievement of students in the State of Texas permits analyses that yield quite precise answers to each of these questions. The data contain test scores spanning grades 3 through 7 for three cohorts of students in the mid-1990s. The multiple cohorts and grades coming from repeated observations on more than one-half million students in over three thousand schools permit the clear identification and detection of even very small teacher and school effects. A primary objective of the initial empirical analysis is to obtain estimates of differences in teacher contributions to student learning that eliminate the ma- jor sources of possible contamination from student selection or teacher assign- ment practices. Because family choice of neighborhood and school depends on preferences and resources, students are nonrandomly distributed across schools (Tiebout (1956)). Schools also use student characteristics including as- sessments of ability and achievement to place students into specific programs and classes. Such nonrandom selection may easily contaminate estimates of school or teacher effects with the influences of unmeasured individual, family, school, and neighborhood factors. Repeated performance observations for individual students and multiple co- horts provide a means of controlling explicitly for student heterogeneity and the nonrandom matching of students, teachers, and schools through the use of fixed effects models. The models control for fixed student, school-by-grade, and in some cases school-by-year effects and then relate remaining differ- ences in achievement gains between grades and cohorts to differences in school characteristics or teachers. This variation in academic performance cannot be driven by unchanging student attributes such as ability or motivation or by un- changing school characteristics and policies that are either common across all Section 9-28 ACADEMIC ACHIEVEMENT 419 grades at a point in time or unique to specific grades. Moreover, the empirical models also account for potentially important time varying influences not cap- tured by the student or school fixed effects. Therefore we are able to identify the impacts of schools and teachers uncontaminated by the many unobserved family and other influences that have plagued past research. The results reveal large differences among teachers in their impacts on achievement and show that high quality instruction throughout primary school could substantially offset disadvantages associated with low socioeconomic background. These differences among teachers are not, however, readily mea- sured by simple characteristics of the teachers and classrooms. Consistent with prior findings, there is no evidence that a master's degree raises teacher ef- fectiveness. In addition, experience is not significantly related to achievement following the initial years in the profession. These findings explain much of the contradiction between the perceived role of teachers as the key determinant of school quality and the body of research showing that observed teacher char- acteristics including experience and education explain little of the variation in student achievement. Students also appear to benefit from smaller classes, particularly in grades 4 and 5. In comparison to the gains from higher teacher quality, however, the estimates indicate that even a very costly ten student reduction in class size such as that undertaken in some U.S. states produces smaller benefits than a one standard deviation improvement in teacher quality. The next section provides an overview of patterns of achievement gains that suggests the presence of substantial within school variation in teacher quality. Section 3 describes the empirical approach used to generate a lower bound estimate of the within school variation in teacher quality. Section 4 provides a detailed description of the Texas data on students and teachers. Section 5 re- ports estimates of the variance in teacher quality based on the method devel- oped in Section 3, and Section 6 presents an extension of traditional analyses of the effects of measured resources: class size, teacher education, and teacher experience on achievement. The final section considers the policy implications of the findings, particularly the importance of measured resources relative to the overall contribution of teachers. 2. SCHOOLS AND TEACHERS Students and parents refer often to differences in teacher quality and act to ensure placement in classes with specific teachers. Such emphasis on teachers is largely at odds with empirical research into teacher quality. There has been no consensus on the importance of specific teacher factors, leading to the com- mon conclusion that the existing empirical evidence does not find a strong role for teachers in the determination of academic achievement and future acad- emic and labor market success. It may be that parents and students overstate the importance of teachers, but an alternative explanation is that measurable Section 9-29 420 S. RIVKIN, E. HANUSHEK, AND J. KAIN characteristics such as teacher experience, education, and even test scores of teachers explain little of the true variation in quality. To motivate the concentration on teacher quality, we begin with aggregate statistics on the variation in student achievement. Table I displays correla- tions of school average annual mathematics and reading achievement gains in grades 5, 6, and 7 between two cohorts of students for all public elemen- tary schools in Texas.3 The diagonal elements report correlations for the same grade in adjacent years, while the off-diagonal elements report correlations for adjacent grades in the same year. The striking difference in magnitudes of the diagonal and off-diagonal el- ements suggests the existence of substantial within-school heterogeneity in school quality. Remarkably, the correlation of between-cohort average gains in different grades in the same year (the off-diagonal terms) is quite small despite the homogeneity of family backgrounds and peers within most schools and de- spite the common school organization, leadership, and resources for the two cohorts. Indeed for comparisons of 6th and 7th grade reading performance, the correlation is -0.01. In contrast, the correlations of between-cohort aver- age gains in the same grade in adjacent years (the diagonal terms) are much larger. A number of factors may explain this pattern, but perhaps the most ob- vious explanation is that there will be many common teachers for two cohorts when observed in the same grade, while virtually all of the teachers will be dif- ferent when comparing cohort performance across grades at a single point in time. Table II reports the R2 values from a series of achievement gain regressions for reading and mathematics performance run over the sample of schools and grades in which there is only a single teacher per subject. (As we discuss be- TABLE I PEARSON CORRELATION COEFFICIENTS OF SCHOOL AVERAGE TEST SCORE GAINS IN MATHEMATICS AND READING ACROSS GRADES AND YEARS Mathematics Reading Grade of Cohort II Grade of Cohort II Grade of Cohort I 5 6 7 5 6 7 5 0.32** 0.19** 6 0.13** 0.52** 0.13** 0.43** 7 0.05 0.46** -0.01 0.44** Notes: Cohort I attended 4th grade in 1994; Cohort II attended 4th grade in 1995. Thus, for example, Cohort I is attending the 6th grade during the same academic year that Cohort II is attending the 5th grade. All calculations are weighted by the average enrollment of the pairs. *significant at 10% level; **significant at 1% level. 3These data, subsequently used in the detailed empirical analyses, are described in detail in Section 3. All correlations relate just to students in schools that have both of the relevant grades. Section 9-30 ACADEMIC ACHIEVEMENT 421 TABLE II COMPARISON OF THE EXPLANATORY POWER OF TEACHER EXPERIENCE, EDUCATION, AND CLASS SIZE WITH TEACHER FIXED EFFECTS IN EXPLAINING ACHIEVEMENT GAINS Mathematics Reading Included explanatory variables Student covariates yes yes yes yes yes yes yes yes Teacher characteristics no yes no no no yes no no Teacher fixed effects no no yes no no no yes no School fixed effects no no no yes no no no yes R squared 0.0151 0.0182 0.1640 0.0949 0.0085 0.0093 0.0903 0.0507 Observations 89,414 81,897 Notes: Dependent variables are mathematics and reading test score gains; sample includes only grades in a school with a single teacher for that subject. low, these are the only schools in which students can be matched to their ac- tual teachers.) The first column for each subject is based on a specification with only student characteristics and year dummies; the second column adds measured teacher and classroom characteristics (teacher experience, teacher education, and class size); the third column substitutes teacher fixed effects for the observable teacher and classroom characteristics; and the final column employs school rather than teacher fixed effects. The results demonstrate quite clearly that the observable school and teacher characteristics explain little of the between-classroom variation in achievement growth despite the fact that a substantial share of the overall achievement gain variation occurs between teachers. Importantly, even though the sample includes just schools with a sin- gle teacher per grade, the inclusion of school rather than teacher fixed effects reduces the explanatory power by over forty percent, suggesting that much of the variation in teacher quality exists within rather than between schools. Tables I and II are consistent with the existence of substantial variation in teacher quality not explained by observable teacher characteristics. However, other factors could clearly enter into these two simple comparisons, making it necessary to utilize more comprehensive methods to identify the variance of teacher quality and importance of observable factors. For example, a high performing 4th grade teacher could leave less room for subsequent gains; the curriculum could affect specific grade levels in differing ways across school districts; test measurement errors could obscure the relationships; there may be nonrandom sorting across schools; or some schools may have more or less effective leadership. The next section develops a comprehensive model of stu- dent learning that provides the analytical framework for the estimation of the variance of teacher quality. Section 9-31 422 S. RIVKIN, E. HANUSHEK, AND J. KAIN 3. THE IDENTIFICATION OF TEACHER EFFECTS In this section we develop an estimator of the variance of teacher quality that avoids problems of student selection and administrator discretion that po- tentially have biased prior attempts. This estimator is based upon patterns of within-school differences in achievement gains and ignores variations in teacher quality across schools, because such variation cannot readily be disen- tangled from student differences and the contributions of other school factors. This strategy yields a lower bound estimator for the importance of teacher quality that relies upon minimal maintained assumptions about the underly- ing achievement process. Importantly, we do not focus solely on measurable characteristics of teachers or schools as is typically done in this literature but instead rely on student outcomes to assess the magnitude of total teacher ef- fects, regardless of our ability to identify and measure any specific components. This semiparametric approach provides both an estimate of the role of teacher quality in the determination of academic achievement and information on the degree to which specific factors often used in determining compensation and hiring explain differences in teacher effectiveness. 3.1. Basic Model of Student Achievement Academic achievement at any point is a cumulative function of current and prior family, community, and school experiences. A study of the entire process would require complete family, community, and school histories, and such data are rarely if ever available. Indeed, the precise specification of what to mea- sure is poorly understood. In the absence of such information, analyses that study the contemporaneous relationship between the level of achievement and school inputs for a single grade are obviously susceptible to omitted variables biases from a number of sources. An alternative approach focuses on the determinants of the rate of learning over specific time periods. The advantage of the growth formulation is that it eliminates a variety of confounding influences including the prior, and often unobserved, history of parental and school inputs. This formulation, frequently referred to as a value-added model, explicitly controls for variations in initial conditions when looking at how schools influence performance during, say, a given school year. While such a value-added framework by no means elim- inates the potential for specification bias, the inclusion of initial achievement as a means to account for past inputs reduces dramatically the likelihood that omitted historical factors introduce significant bias.4 Equation (1) presents a conventional value-added model that describes the gain in student achievement (AAgs) for individual i in cohort c with teacher j 4One restriction of this formulation is that the parameter estimates capture effects only for the specific period, ignoring any continuing impacts of inputs at an earlier age. See Krueger (1999) for a discussion of this issue. However, without detailed information and knowledge of the full Section 9-32 ACADEMIC ACHIEVEMENT 423 in grade g of school s: (1) A = A -Ac - Xg +jgs T - + ? S -S? fi + Cijg s. This gain, measured as the difference between a student's test scores in grades g and g - 1, depends on family background (X); teacher characteristics (T); school characteristics (S); inherent student abilities (f); and a random er- ror (e). Note that the term "inherent abilities" refers to the set of cognitive skills, motivation, and personality traits that affect the rate of achievement growth but that do not change during the school years being considered.5 Each of the inputs can be thought of as a vector of underlying components. Formulations similar to equation (1) have been estimated in a variety of cir- cumstances in order to identify the causal link between a student outcome such as achievement or years of schooling on the one hand and a school character- istic such as class size on the other (see, e.g., Murnane (1975) or Summers and Wolfe (1977)). Much research has focused on the development of methods to eliminate any remaining biases, and we address this concern as well. However, a potentially much more important issue is the possibility that the measured teacher and school factors do not adequately capture important differences in the quality of education. An alternative approach attempts to circumvent the problem of inade- quate measures of quality through the estimation of classroom fixed effects on achievement gains (see, e.g., Hanushek (1971), Armor et al. (1976), Murnane and Phillips (1981)). These analyses of covariance capture all between- classroom differences in achievement gains controlling for any included regres- sors. The resulting classroom differences in average achievement gain have been interpreted as reflecting teacher quality, since the teacher is the most cumulative achievement production process, it is virtually impossible to isolate any continuing effects of specific school factors. The precise estimation approach found in the literature does vary. At times, initial achieve- ment is added to the right-hand side of a regression equation, possibly with corrections for mea- surement error. At other times, simple differences or growth rates in scores are analyzed. The alternative formulations do place different restrictions on the form of the achievement process. See Hanushek (1979) for a discussion of value-added models. Subsequent analysis, relying on expected expansions of our database, will explore alternative specifications. 5The isolation of inherent student abilities does not rely on any presumption about their source (genetic, environmental, or an interaction of these). Any fixed differences that affect the rate of learning will be incorporated in this term. This formulation goes beyond typical discussions that concentrate just on how fixed ability, family, and motivational terms affect the level of achieve- ment at a point in time. Here we explicitly allow for the possibility that ceteris paribus some children will acquire knowledge at different rates even after allowing for variations in initially observed achievement. Further, these differences do not have to be unidimensional. Section 9-33 424 S. RIVKIN, E. HANUSHEK, AND J. KAIN obvious factor differing across classrooms. However, problems from test mea- surement errors and potential school and classroom selection effects may be even more serious for these types of models than in those that use observable measures, making the interpretation of these as direct estimates of the teacher component problematic.' The central estimation problem results from the processes that match stu- dents with teachers, and schools. Not only do families choose neighborhoods and schools, but principals and other administrators assign students to class- rooms. Because these decision makers utilize information on students, teach- ers and schools, information that is often not available to researchers or measured with error, the estimators are quite susceptible to biases from a num- ber of sources. The following section develops an empirical model designed to avoid these problems and to identify the variations in the quality of instruction. 3.2. An Extended Specification of Education Production Rather than attempting to define each variable in the education process, we begin by thinking in terms of the total systematic effect of students, families, teachers, and schools. In this, we depart from the parametric approach of equa- tion (1) that involved measuring a small set of inputs in their natural units and move to a semiparametric approach with inputs measured in achievement, or output, units. Equation (2) describes a decomposition of education production during grade g into a set of fixed and time varying factors: (2) AAigs = Yi + Oj + + Vgs,. Test score gain in grade g is written as an additive function of student (y), teacher (0), and school (8) fixed effects along with a random error (v) that is a composite of time-varying components. The fixed student component captures the myriad family influences including parental education and permanent in- come that affect the rate of learning; the fixed school factor incorporates the effects of stable school characteristics including resources, peers, curriculum, etc. Finally, the teacher component captures the average quality of teacher j over time. Of course families, schools, and teachers all change from year to year, and such changes receive considerable attention in the analysis below. Equation (2) is not intended to be a comprehensive model of the achieve- ment determination process, and moreover we do not attempt to identify each of the separate components. Rather, it provides a framework for the specific models used to study the effects of teacher quality and school resource differ- ences. We have not, for example, distinguished any role for school districts. 6Hanushek (1992) does provide suggestive evidence that teachers are the primary component by showing that classroom gains for individual teachers tend to be highly correlated across time (for different groups of students). Section 9-34 ACADEMIC ACHIEVEMENT 425 Many school policies-hiring, curriculum, school structure, etc.-emanate from school districts and will produce common elements in the teacher and school effects specified in equation (2). While the study of district effects is clearly important, particularly in a policy context, our focus on within-school achievement differences to avoid the difficulties associated with the endogene- ity of school and district choice precludes identification of separate district ef- fects.7 Moreover, school fixed effects also capture any systematic differences across districts and communities, so there is no econometric reason to specify separate district or community components in this estimation. We do, how- ever, address district related issues as they are relevant to the identification of teacher quality and school resource effects.8 3.3. Estimator of the Variance of Teacher Quality In the semiparametric approach of equation (2), the variance of 0 measures the variation in teacher quality in terms of student achievement gains. One could estimate this variance directly using between-classroom differences in average achievement gains. We do not adopt this approach for a number of reasons, not the least of which is the inability to match students to specific teachers. Yet even if students could be matched with teachers and the analy- sis considered only within-school variation in outcomes, both the intentional placement of students into classrooms on the basis of unobservables and the need to account for the contribution of measurement error to the between- classroom variation would introduce serious impediments to the identification of the variance of teacher quality.9 Consequently, we adopt a very different method that makes use of infor- mation on teacher turnover and grade average achievement gains to generate a lower bound estimate of the within-school variance in teacher quality. This approach avoids the need to identify and to estimate separately the test error 7The role of district environment and policies is a topic that we intend to pursue in the future. That analysis however, requires a different estimation strategy that, importantly, does not permit the precise identification of teacher influences that we pursue here. 8The model also imposes the assumption of additive separability in order to simplify the pre- sentation. We explore the possibility that the magnitudes of school resource effects vary by stu- dent characteristics, allowing for the most commonly cited type of potential complementarity. In addition, we recognize that the matching of students and teachers likely affects the average rate of learning in a school, and the subsequent inclusion of school and school-by-grade fixed effects captures any differences that are maintained across our observation period. 9This discussion can be directly linked to prior estimation of classroom fixed effects, which develop classroom gains after conditioning on measurable characteristics of students or schools. See, for example, Hanushek (1971), Armor et al. (1976), and Murnane and Phillips (1981). In such cases, the interpretation of the individual and school components of equation (3) would re- late directly to dimensions not captured by the included characteristics, and the test measurement errors would remain. Section 9-35 426 S. RIVKIN, E. HANUSHEK, AND J. KAIN variance, and the aggregation to the grade level circumvents any problems re- sulting from classroom assignment.10 The cost of this aggregation is the loss of all within grade variation in teacher quality and the inability to trace out the teacher quality distribution. Equation (3) represents average achievement gain in grade g in school s for cohort c as an additive function of grade average student and teacher fixed effects, a school fixed effect, and the grade average error: (3) aAAC = -- + 0. + 5S + vC With two different cohorts of students (c and c'), we can compare average gains in the same grade: (4) AAg AAC YC C O VC - -C Notice in equation (4) that all fixed school components from equation (3) drop out because they exert the same effect for both cohorts. These eliminated fac- tors include fixed aspects of peers, school administration, technology, and in- frastructure as they affect the growth in achievement, even if they are grade specific. They also include systematic (time invariant) sorting of teachers by school or district that comes from a district's salary or general attractiveness along with its standard teacher assignment practices. The difference in cohort average achievement gains is thus a function of the between-cohort differences in teacher quality (0), in fixed student and family factors (y), and an average error component that includes not only measurement errors but time varying individual, family, and school factors. Though we do report estimates of the variance in teacher quality based on simple between-cohort achievement differences for a single grade, cohort aver- age differences in (y) contaminate estimates of the variance in teacher quality. Consequently, we concentrate on the difference between adjacent cohorts in the pattern of average gains in grades g and g'. In order to control fully for student fixed effects, we limit the sample to students who remain in the same school for grades g - 1 and g: (5) (AACs - AAs ) - (AAC' - AA• ,) =[(0 - oi ) - (o. - Os)] + [( - Vs) -(U. - •,,)]. 1oThis estimator assumes that there are not strong complementarities between specific stu- dents and teachers, that is, that the effects of teachers is linear and separable as in equation (2). Yet as long as schools maintain similar assignment practices from year to year, as discussed below, even such complementarities will not contaminate the estimates. Additionally, changes in assign- ment practices will tend to bias estimates of the variance in teacher quality downward, reinforcing our interpretation of the estimator as a lower bound on teacher quality variance. Section 9-36 ACADEMIC ACHIEVEMENT 427 As equation (5) shows, taking the difference between average gains in grades g and g' eliminates all fixed student and family differences, leaving only cohort- to-cohort differences in the grade average difference in teacher quality and time varying student and school factors (contained in v) as determinants of the difference in the pattern of achievement gains. Squaring both sides of equation (5) gives (6) [(AAgs - AAs) - (AA - AAg, )]2 -2 2 --2 ---2 = O +?c +O , - 2( Occ + Ot -c2, ) +?2[(6Vgs 6Ot- s+s + 6 7 ' - 0)+(1,)] +?e. The squared difference leads to a natural characterization of the observed achievement differences between cohorts as a series of terms that reflect vari- ances and covariances of the separate teacher effects plus a component e that includes all random error and cross product terms between teacher and other grade specific effects. We now impose three assumptions that formally characterize the notion that teachers are drawn from common distributions over the restricted time period of our cohort and grade observations: (i) The variance of grade average teacher quality is the same for all cohorts and grades; (ii) the covariance of grade aver- age teacher quality for adjacent cohorts is the same for all grades; and (iii) the covariance of grade average teacher quality for grades g and g' for adjacent co- horts equals the covariance of grade average teacher quality for grades g and g' for each cohort. For ease of exposition, we also make the simplifying assump- tion that each school has one teacher per grade, but this is relaxed later. Applying these assumptions and taking the expectation of equation (6) yields (7) E[(AA, A Ag - - (AAg - AAg,)]2= 4(ro - oocl) + E(es), where o2 is the variance of teacher quality in school s and o-sC, is the covari- ance of teacher quality across cohorts in a school. The key to the identification of the magnitude of the within-school vari- ance of teacher quality comes from the first element on the right-hand side- the within-school variance of grade average teacher quality minus the within- school covariance of quality across cohorts. Consider first schools in which the two cohorts have the same teacher in each grade (i.e., the proportion of teach- ers who are different equals zero). As long as teachers perform equally well in both years, o2- = --?s', and teacher quality contributes nothing to student performance differences across cohorts. On the other hand, consider schools in which cohorts c and c' have dif- ferent teachers in each grade (the proportion of teachers who are different Section 9-37 428 S. RIVKIN, E. HANUSHEK, AND J. KAIN equals one)." In this case the within-school covariance of teacher quality equals zero. Importantly, this is not to say that schools hire randomly, for as we discuss below there can be little doubt that hiring practices and characteristics related to teacher job preferences differ substantially across schools. Rather, it says that the covariance across teachers in the deviation from the mean teacher quality in a school is zero. Equation (7) provides the basis for estimation of the within-school variance of teacher quality. The left-hand side in most regressions is the squared diver- gence of the grade pattern in gains across cohorts, which we regress on the proportion of teachers who are different. Ignoring the possible confounding influences of other factors and maintaining the assumption that teacher qual- ity remains unchanged in the absence of turnover, the coefficient on this pro- portion divided by four will provide a consistent estimate of the within-school variance in teacher quality.'2 One empirical complication arises because most schools do not have a sin- gle teacher for each grade. Rather the number of teachers varies by school, and consequently the coefficient on the turnover variable would not have a straightforward interpretation. Because the achievement gains and the effects of teachers are averaged across the teachers in a grade, we actually have the variation of the mean in each school, and the relationship of turnover to the within-school variance will depend on the number of teachers. For example, in a sample of schools with three teachers per grade, the coeffi- cient on proportion different would provide an estimate of four times one third (i.e., 4o-2 /3) of the within-school variation in teacher quality. This also means that fifty percent turnover in schools with three teachers per grade would lead to the same expected squared cohort difference in grade aver- age difference in gains as one hundred percent turnover in schools with six teachers per grade. In order to account for such differences in the number of teachers and place all schools on a common metric, the proportion differ- "Note that such differences result from both teacher departures and grade changes. There is an extensive related literature on the determinants of teacher turnover, indicating that salary, working conditions, and alternative wage opportunities do affect the probability of ex- iting a school (cf. Dolton and van der Klaauw (1995, 1999), Murnane and Olsen (1989), Stinebrickner (2002), Hanushek, Kain, and Rivkin (2004b)). None, however, suggests that leavers are systematically more effective teachers than stayers, an issue to which we return below. More- over, our analysis of within-school patterns of student performance implicitly controls for the overall determinants of turnover and focuses solely on the implications of turnover for perfor- mance. Regardless of any differences between leavers and stayers, the within-school covariance of grade average quality equals zero in 100 percent turnover schools as long as any changes in hiring procedures are not systematically related with the quality of leavers. 12Note that we use teacher turnover as a method of identifying the variance in teacher quality. Implicitly, we assume teacher turnover does not directly affect student achievement gains except for the possibility of systematic quality differences by teacher experience. We test this assumption within the general production function estimation (below) and cannot reject it. Section 9-38 ACADEMIC ACHIEVEMENT 429 ent must be divided by the number of teachers per grade, and the coefficient on this variable provides an estimate of the within-school variance in teacher quality. Our empirical strategy focuses on the estimation of a lower bound on the variation of teacher quality, and in that regard a variety of factors that sug- gest downward bias in our turnover estimator are not problematic. First is the almost certain violation of the assumption that the variance and covariance terms are equal in schools without turnover. Even in the absence of teacher turnover, there is almost certainly some difference in teacher quality from year- to-year due to changes in pedagogy, personal problems, learning (particularly for beginning teachers), etc., reducing the expected coefficient on the turnover variable below the true within-school variance. Measurement error in the teacher turnover variable would tend to exacer- bate any such downward bias. The administrative data have missing informa- tion on key variables, and it is not always clear who teaches which subjects. Consequently, there is some error introduced into the calculations of both the percentage of teachers who differ from cohort to cohort and in the number of teachers per grade, and the ratio of the two may thus contain a nontrivial amount of noise. More worrisome for our approach, however, is that there are also two poten- tially important sources of upward bias. First is the standard problem of omit- ted variables. Teacher turnover may be precipitated or accompanied by other changes such as a new principal or superintendent or district induced curricu- lum changes (Ingersoll (2001)). If, for example, administrator turnover also leads to teacher turnover, any direct effects of new administrators on achieve- ment growth would introduce an upward bias if they were not accounted for. In the empirical work below, we take a number of steps to control for potentially confounding time-varying factors including controls for the numbers of princi- pal and superintendent changes over the observation period. We also perform various sensitivity analyses directed at these issues. Second is the possibility that teachers who exit are not drawn randomly from the teacher quality distribution. If attrition and quality are systematically re- lated, the average teacher quality in high turnover years will tend to differ sys- tematically from the average quality of new hires. Consider the possibility that high quality teachers are more likely to exit. In this case, schools that obtain a particularly good draw of teachers in one year will tend to experience both a greater turnover following the year and a larger average difference in achieve- ment gains than would be experienced with random attrition. This situation would lead to an upward bias in our estimator, as would the opposite case where low quality teachers are more likely to exit. Even if attrition and qual- ity are uncorrelated, if teachers in the tails of the distribution are more likely Section 9-39 430 S. RIVKIN, E. HANUSHEK, AND J. KAIN to exit, higher turnover schools will tend to have higher cohort differences in achievement gains, again biasing our estimator upward.13 Appendix A demonstrates that a major departure from random exiting in the form of higher probabilities in either or both tails of the distribution can introduce substantial upward bias. In the absence of student/teacher matches, we have little information on the actual distribution of departures. Moreover, the literature on teacher turnover is not very informative on the quality distri- bution of any school attrition.14 A general presumption, particularly in more policy-related analyses, is that union restrictions, the single salary schedule for teachers, and the lack of performance incentives related to student achieve- ment mute any relationship between teacher quality and attrition, but this is clearly speculative.'" Fortunately, we do have student/teacher matches for a single district, and we use that information to provide empirical evidence on the likely magnitude and direction of any nonrandom turnover induced bias. Finally, this framework relies on just the variation in teacher quality that is found within schools and ignores all variation in teacher quality across schools. If all schools were to hire randomly from a common pool, the between-school variance would equal zero, but this is almost certainly not the case. Rather schools able to offer higher salaries or better working conditions choose among a larger pool of applicants and likely enjoy higher average teacher quality, though the difficulty predicting productivity on the basis of education creden- tials and interviews almost certainly allows for substantial within-school het- erogeneity."6 In the extreme, if schools were perfectly arrayed in their hiring, all variations in quality would be between schools. In any event, the between- school differences would have to be added to the estimates reported below to obtain an estimate of the total variation in the quality of instruction. '3Note that heavy attrition in just one tail also implies drift in the average quality of teachers, which would inappropriately add to our estimate of the within-school variance (and which we explicitly assume is not the case). '4Much of the turnover literature (footnote 11) relates to opportunity costs by specialties (e.g., math and science), but these studies are more relevant for secondary schools and do not directly address issues of quality. Another approach investigates attrition by the teacher's own test score (see Murnane et al. (1991)) and finds some relationship suggesting that higher scoring teachers are more likely to leave, but neither this relationship nor the relationship between teacher test scores and student achievement is very strong. The one direct study relating attrition to classroom performance finds that principal evaluations early in the teaching career are positively correlated with continued teaching. At the same time, while teacher value-added based on student achieve- ment is also positively related to retention of teachers, the estimates are statistically insignificant (Murnane (1984)), perhaps because of the small samples. "5For example, The Teaching Commission (2004, p. 46) notes: "once teachers have passed a probationary period, it is notoriously difficult to dismiss those whose performance is inadequate. In 2002, for instance, only 132 of 78,000 teachers in New York City's massive school system were removed for poor performance." However, no analyses of decisions before tenure or of more informal actions are available. '6Hanushek, Kain, and Rivkin (2004b) find that teachers who switch schools tend to move to schools with higher achieving, higher income, and lower proportion minority student bodies. Section 9-40 ACADEMIC ACHIEVEMENT 431 4. THE TEXAS DATABASE The data used in this paper come from the UTD Texas Schools Project, conceived of and directed by John Kain. Data are compiled for all public school students from administrative records in Texas, allowing us to use the universe of students in the analyses. We use data for three cohorts: 3rd through 7th grade test scores for one cohort (4th graders in 1995) and 4th through 7th grade test scores for the other two (4th graders in 1993 and 1994).17 For each cohort there are more than 200,000 students in over 3,000 public elemen- tary and middle schools. (For details on the database, see Appendix B and Table B1; currently available data along with variable definitions and estima- tion programs are found in Rivkin, Hanushek, and Kain (2005).) In compari- son to studies that use only a small sample of students from each school, these data permit much more precise estimates of school average test scores and test score gains. The administrative data contain a limited number of student and family characteristics including race, ethnicity, gender, and eligibility for a free or re- duced price lunch. Students who switch public schools anywhere within the state of Texas can be followed just as those who remain in the same school or district. Although explicit background measures are relatively limited, the panel feature can be exploited as described previously to account implicitly for time invariant individual and school effects on achievement. Beginning in 1993, the Texas Assessment of Academic Skills (TAAS) was administered each spring to eligible students enrolled in grades 3 through 8.18 These tests are designed to evaluate student mastery of the grade-specific sub- ject matter that is prescribed for students in the state.19 We focus on test re- sults for mathematics and reading, derived from tests of approximately fifty questions. Because the number of questions and average percent correct varies across time and grades, we transform all test results into standardized scores with a mean of zero and variance equal to one, though the empirical findings '7Note that, while we have 3rd grade test information, our analysis begins at 4th grade because of the focus on achievement gains. "SMany special education and Limited English Proficiency (LEP) students are exempted from the tests, as are other students for whom the test would not be educationally appropriate. In each year roughly fifteen percent of students do not take the tests, either because of an exemption or because of repeated absences on testing days. This rate of missing tests appears comparable to those for other high quality testing programs such as the National Assessment of Educational Progress. 19The TAAS tests are generally referred to as criterion referenced tests, because they refer directly to pre-established curriculum or learning standards. The common alternative is norm referenced tests that cover general subject matter appropriate for the subject and grade but that are not as closely linked to the specific state teaching standards. In principle, all students could achieve the maximum score on a criterion referenced test with no variation, while norm refer- enced tests focus on obtaining information about the distribution of different skills across the tested population. In practice, scores on commonly available criterion referenced and norm ref- erenced tests are highly correlated across students. Section 9-41 432 S. RIVKIN, E. HANUSHEK, AND J. KAIN are robust to a number of transformations including the raw percentage cor- rect. The bottom one percent of test scores (all less than or equal to expected scores from random guesses) are trimmed from the sample in order to reduce measurement error. Participants in bilingual or special education programs are also excluded from the samples used in estimating teacher quality and resource effects because of the difficulty in measuring teacher and school characteristics for these students.20 Student data are merged with grade average information on teachers by sub- ject. Because student and teacher data come from different reporting systems that are not directly linked, matching students with their specific teachers is not possible. Teacher personnel data provide information on experience, high- est degree earned, and the class size, subject, grade, and population served for each class taught. This information is used to construct subject and grade average characteristics for teachers in regular classrooms. In the early grades teachers tend to teach all subjects, while in junior high most specialize. We consider those who self identify as general teachers as teachers of both mathe- matics and reading. 5. LOWER BOUND ESTIMATES OF THE VARIANCE OF TEACHER QUALITY The estimation of the within-school variance in teacher quality relies on the notion that teacher turnover increases the variance in student outcomes across grades and cohorts in a school. Although we refine the estimation be- low, the pattern can be seen directly by observing the higher correlations in student achievement across cohorts for schools with lower teacher turnover (fewer than twenty five percent of teachers are different) than schools with high turnover (fewer than twenty five percent of teachers are the same). The corre- lations are 0.40 in math and 0.26 in reading respectively for the low turnover schools and 0.22 in math and 0.14 in reading for the high turnover schools. Of course other factors correlated with teacher turnover could also produce this pattern, and it is necessary to turn to our more structured model in order to identify the importance of teacher quality in the determination of achievement gains. Note that on average roughly one third of teachers are new to a grade and subject in any year. This is roughly double the rate of school leaving, mean- ing that incumbent teachers tend to change grades or subjects every five years or so. 20For an explicit analysis of the achievement of special education students, see Hanushek, Kain, and Rivkin (2002). Kain and O'Brien (1998) provide additional analysis of special edu- cation students along with information on the performance of limited English proficiency (LEP) students. These students are included in the calculations of class sizes for the analysis below when they receive instruction in regular classrooms. Section 9-42 ACADEMIC ACHIEVEMENT 433 5.1. Basic Estimates Table III reports basic estimates from the regression of the squared between- cohort difference in gains on the proportion of teachers who are different and other covariates. The sample includes only students who remain in the same school for two successive grades, either 5th and 6th or 6th and 7th, and only grades that have at least five students with valid test scores and nonmissing data on teacher turnover.21 Just grades 5 and 6 are used for the small number of schools with all three grades.22 The final sample has 3,076 schools in the mathematics specifications and 3,086 in the reading specifications. The three left hand columns in Table III report results from the three spec- ifications for mathematics and reading in order to isolate the sensitivity of the estimates to the different fixed components of achievement growth. The first regresses the squared difference in 5th (or 7th) grade gains between co- horts on 5th (or 7th) grade teacher turnover; the second and third regress the squared difference in the difference of 5th (or 7th) and 6th grade gains be- tween cohorts on the turnover of 5th (or 7th) and 6th grade teachers com- bined. As described previously, using the difference in gains between the two grades controls for both student and school fixed effects in gains. Finally, the third specification adds an additional school fixed effect directly into the re- gression, identifying the variance in teacher quality on the basis of the differ- ence in turnover rates between the first and second cohorts and the second and third cohorts. This last estimation, which captures school specific variations in the grade pattern of performance, directly controls for systematic school and grade specific unobservables that may be correlated with turnover. All three specifications also include a dummy variable identifying the precise cohort comparison, the inverse of enrollment (because the variance of measurement error in student performance is inversely proportional to enrollment), the use of 7th grade information, and the numbers of new principals and superinten- dents. The measures of new school and district leadership capture time varying policy factors that could simultaneously affect teacher turnover and student achievement. The results show that differences in mathematics and reading achievement gains among cohorts are strongly related to teacher turnover. All coefficients 21An additional observation in the reading sample was also excluded, because the grade aver- age gain was more than six standard deviations from the mean (higher than any other school). It turned out to be a single teacher whose students' average gain in the previous year was quite close to the mean and who did not teach in the subsequent year. In addition, the average gain in the subsequent grade was roughly four standard deviations below the mean, far different than the positive gain reported for the prior cohort taught by the same teacher. We believe there is overwhelming evidence of either cheating or miscoding. The exclusion of this observation did not have a large impact on the estimates except in the full fixed effect model. 22The majority of students move from elementary to middle school sometime between grades 5 and 7. Roughly fifteen percent of schools with at least two of the three grades in this range have all three. Section 9-43 434 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE III EFFECT OF TEACHER TURNOVER ON THE DIVERGENCE OF MATHEMATICS AND READING TEST SCORE GAINS BETWEEN COHORTS (STANDARD ERRORS IN PARENTHESES) Individual and Individual and Individual and Individual and No Fixed School Fixed School-by-Grade School Fixed School-by-Grade Effectsa Effectsb Fixed Effectsc Effectsb Fixed Effectsc 1. Mathematics Proportion of teachers 0.080 0.090 0.050 0.080 0.045 who are different/number (0.017) (0.015) (0.021) (0.016) (0.021) of teachers Absolute change in 0.033 0.027 proportion of teachers (0.016) (0.023) with no experience 2. Reading Proportion of teachers 0.067 0.082 0.036 0.078 0.029 who are different/number (0.013) (0.014) (0.018) (0.015) (0.018) of teachers Absolute change in 0.015 0.041 proportion of teachers (0.015) (0.020) with no experience Notes: All equations include the inverse of the number of students, numbers of new principals and superintendents in the school during adjacent years, a grade 7 dummy variable, and a cohort dummy variable. The sample includes all students who remain in the same school for grades 5 and 6 (or 6 and 7). Sample size is 3,076 for the mathematics and 3,086 for the reading specifications. Equations have the same structure for mathematics and for reading. (The analyses of gain patterns between grades 6 and 7 take the same form as those for grades 5 and 6 that are shown.) For P = proportion different math (or reading) teachers/#teachers and adjacent cohorts (c and c'), the specifications take the following forms: (a) (A• - AC)2 - ,'( ', ,+'XA C's C+eC' , / / C,/ (b) r(AC - AC) -_(Ac - Ac1z -4c')2 =pCcI +xX CC1 + ec'Cl 6(b) [( 5) • ]2 and 6,s+X and 6,s 5 and 6,s' (c) [(A - ) - (A )]2 5and6, +s+XX5 and,s +e5 and6,s' where 6s is a fixed effect for school s. are positive and significant at the five percent level, and except for the school- by-grade fixed effect specifications, all t-statistics exceed 4.5 in absolute value. The declines in coefficient magnitudes for the full fixed effect specifications are consistent with measurement error induced attenuation bias, but they may also reflect the presence of omitted variables bias in the other specifications. In order to avoid as much as possible the introduction of any upward biases, we concentrate here on the full fixed effect coefficients of 0.050 and 0.036. These imply lower bound estimates of the within school variance of teacher quality (measured in units of student achievement) equal to 0.0125 (0.050/4) and 0.009 (0.036/4) for mathematics and reading respectively. This means that a one standard deviation increase in average teacher quality for a grade raises average student achievement in the grade by at least 0.11 standard deviations of the total test score distribution in mathematics and 0.095 standard devia- tions in reading. Section 9-44 ACADEMIC ACHIEVEMENT 435 These estimates suggest the existence of substantial within school variation in teacher quality, but they combine average differences across the experience distribution with skill differences not related to experience. As we demonstrate in the direct estimation of educational production functions below, the learn- ing curve appears to be quite steep in the first year or two of teaching before flattening out. Because many of the teachers new to a grade are in their first year, the share of the variance due to differences between beginning and ex- perienced teachers might be quite sizeable. Fortunately, we can identify the effects of beginning teachers by including the absolute change in the share of teachers in their first year as an additional variable.23 The final two columns of Table III present estimates from the two fixed ef- fect specifications that include the absolute change in the share of beginning teachers. These estimates suggest that quality differences between new and ex- perienced teachers account for only ten percent of the teacher quality variance in mathematics and somewhere between five and twenty percent of the vari- ance in reading. The addition of the change in the share of teachers with one year of experience (not shown) has virtually no effect on the estimates. 5.2. Specification Checks The consistency of the estimator relies on the assumption that the turnover variable is unrelated to the error. One important threat to the estimation strategy is the possibility that unobserved changes over time in schools may be correlated with teacher turnover. A comprehensive control for other time varying factors in the schools comes from looking at turnover of teachers not involved in the specific subject. Specifically, by looking at schools that use sep- arate teachers for mathematics and English, we can include English teacher turnover as a control variable in the modeling of math performance and math- ematics teacher turnover in the modeling of reading achievement.24 Table IV reports the results for fixed effect specifications that include turnover in the untested subject. These estimates are generated from the smaller subsample of schools with subject specialists (defined as schools that have no teachers in either of the two sampled grades who teach both math and English), which is roughly thirty percent of the full sample. The results for mathematics remain highly significant though somewhat smaller in the first two specifications and are significant only at the ten percent level in the full fixed effects model, which is not that surprising given the substantial reduction in 23Because we are looking at variance in outcomes across cohorts, any significant change either up or down in the proportion of teachers in their initial year of experience has a similar impact, thus making the absolute value appropriate. 24Because teacher turnover in the untested subject is used to identify any concomitant disrup- tion in the school, the number of teachers in that subject will not directly affect the variance in student performance. Therefore this turnover variable is not divided by the number of teachers in the untested subject. Section 9-45 436 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE IV EFFECT OF TEACHER TURNOVER ON THE DIVERGENCE OF MATHEMATICS AND READING TEST SCORE GAINS BETWEEN COHORTS, CONTROLLING FOR TEACHER TURNOVER IN OTHER SUBJECTSa (STANDARD ERRORS IN PARENTHESES) Individual and School Individual and School-by-Grade Fixed Effectsb Fixed Effectsb 1. Mathematics Proportion different math 0.059 0.058 0.069 0.034 0.034 0.035 teachers/number of teachers (0.015) (0.015) (0.016) (0.021) (0.021) (0.021) Absolute change in proportion -0.029 -0.005 math teachers with no experience (0.013) (0.020) Proportion of same English teachers -0.006 -0.008 0.002 0.002 (0.010) (0.010) (0.014) (0.014) 2. Reading Proportion different English 0.027 0.024 0.010 0.001 -0.001 -0.005 teachers/number of teachers (0.016) (0.016) (0.016) (0.021) (0.021) (0.022) Absolute change in proportion 0.042 0.013 English teachers with no experience (0.015) (0.021) Proportion of same -0.017 -0.016 -0.020 -0.020 mathematics teachers (0.011) (0.011) (0.013) (0.013) aThe sample includes all students who remain in the same school for grades 5 and 6 (or 6 and 7) in schools with no teacher offering both English and math instruction. All equations include the inverse of the number of students, numbers of new principals and superintendents in the school during adjacent years, a grade 7 dummy variable, and a cohort dummy variable. The sample size is 855. bTable III notes describe the estimation specifications. sample size. In contrast, the English teacher turnover coefficients in the read- ing test score regressions become quite small and insignificant in all specifica- tions, raising concern that confounding factors in this estimation method could be driving the results. In this sample, the impact of inexperienced teachers is very imprecisely estimated. Importantly, comparisons across specifications for a common sample reveal that the inclusion of turnover information for the untested subject has virtually no effect on the other turnover estimate in either fixed effect specification. The question remains as to why the estimates in Table IV are uniformly smaller than those reported in Table III. An important difference between the samples for the respective tables is the balance between 5th and 7th grade classrooms. It is almost always the case that junior high schools use subject specific teachers, while elementary schools use a single teacher for most sub- jects. Consequently the vast majority of schools with subject specific teachers include grades 6 and 7, while the majority of all schools in the sample include grades 5 and 6. Systematic differences by grade in the effects of teachers on test scores could therefore account for the observed pattern of results. Section 9-46 ACADEMIC ACHIEVEMENT 437 Table V reports estimates that allow the effect of turnover to vary by grade combination based on the full sample used in Table III. The coefficients sug- gest that the variance in teacher quality declines in mathematics as students progress through school, though the interaction term becomes insignificant in the full fixed effect model. On the other hand, it appears that within school dif- ferences in teacher quality are quite substantial in reading in elementary school but explain little or none of the variation in outcomes in junior high. In both subjects the pattern of estimates in Table V explain the differences between Tables III and IV. Interestingly, this pattern of diminishing effects will repeat itself in the production function estimates below, suggesting either that school TABLE V GRADE DIFFERENCES IN THE EFFECTS OF TEACHER TURNOVER ON THE DIVERGENCE OF MATHEMATICS AND READING TEST SCORE GAINS BETWEEN COHORTS (STANDARD ERRORS IN PARENTHESES)a Individual and Individual and Individual and Individual and School Fixed School-by-Grade School Fixed School-by-Grade Effects Fixed Effects Effects Fixed Effects 1. Mathematics Proportion of teachers 0.113 0.063 0.096 0.036 who are different/number (0.018) (0.026) (0.019) (0.027) of teachers 6th & 7th grades interaction -0.075 -0.036 -0.068 -0.024 with proportion differentb (0.031) (0.044) (0.032) (0.047) Absolute change in 0.026 0.052 proportion of teachers (0.022) (0.032) with no experience 6th & 7th grades 0.018 -0.035 interaction with absolute (0.033) (0.048) changeb 2. Reading Proportion of teachers 0.115 0.066 0.114 0.059 who are different/number (0.017) (0.022) (0.018) (0.023) of teachers 6th & 7th grades interaction -0.092 -0.081 -0.101 -0.083 with proportion differentb (0.028) (0.037) (0.029) (0.038) Absolute change in 0.004 0.048 proportion of teachers (0.020) (0.027) with no experience 6th & 7th grades 0.030 -0.011 interaction with absolute (0.031) (0.039) changeb aTable III notes describe the sample and estimation specifications. bInteraction between an indicator for the grade 6 and 7 observations and specified variable. Section 9-47 438 S. RIVKIN, E. HANUSHEK, AND J. KAIN and teacher quality differences have much smaller effects on achievement in junior high or that the test results do a poor job of capturing differences in school quality in those grades. There remains one other potential source of bias that must be addressed. Although controls for any concomitant changes to teacher turnover address the problem of omitted variables, they do not resolve the potential problem of nonrandom teacher attrition described above. As noted previously, the estima- tion relies upon the assumption that turnover is uncorrelated with quality and is not drawn heavily from either of the tails of the quality distribution. Since our estimator is identified by the assumption of random departures, we cannot readily test this assumption within our model and data. Fortunately, for one large Texas school district we have developed some ad- ditional data that link student test score gains with individual teachers.25 Al- though we cannot account for unobservable selection into classes, sampling error, and the other factors that we explicitly worry about in this paper, we can use these data to compute a within-school measure of quality: average student achievement gains for each teacher minus the average for all teachers in the same school that year. We can then calculate attrition probabilities based on this quality measure and use these probabilities to estimate the impact of any nonrandom attrition on our estimator of the variance of teacher quality. Table VI describes the distribution of teachers placed into twenty quality categories along with the probabilities of exit for each group. We create these categories by dividing the range of teacher average gains relative to the school average into twenty intervals of equal length. (Because of concerns about out- liers, we drop the top and bottom one percent of gains, but the results are invariant to this sampling procedure as we show below.) Within each category we use the mean gain as the index of quality. Since the division into twenty categories is arbitrary, we examine the sensitivity of the results to changes in the number of intervals. With random departures there would be no systematic differences in the probability of exiting. This does not appear to be the case in Table VI, as at- trition clearly declines with quality, probably in part due to the fact that first year teachers have the highest attrition. On the other hand, attrition does not appear to be concentrated in the tails of the distribution, the key element de- scribed in Appendix A. (Note that there are very few teachers in the lowest quality category that is an outlier in the exit rate at 42.9 percent.) We now use the method developed in the simulations in Appendix A to es- timate the bias introduced by deviations from random departure of the type observed in Table VI. Table C1 shows that the nonrandom attrition leads to a very slight increase (less than one percent) in the estimated standard deviation of teacher quality. This result also holds if the number of quality intervals is doubled or tripled or if observations in the tails of the distribution are retained 25These data are described in Hanushek et al. (2005). Section 9-48 ACADEMIC ACHIEVEMENT 439 TABLE VI TEACHER EXIT RATES BY QUALITY OF INSTRUCTION RELATIVE TO OTHERS IN THE SCHOOL FOR TEACHERS IN A LARGE TEXAS DISTRICT Quality Index Frequency (Percent) Exit Rate (Percent) -1.56 0.17 42.9 -1.41 0.20 11.8 -1.27 0.45 23.7 -1.11 0.56 23.4 -0.94 1.17 30.6 -0.79 1.73 26.2 -0.63 2.86 22.2 -0.48 5.08 22.6 -0.32 9.58 21.3 -0.16 15.29 20.6 -0.01 21.35 20.2 0.14 16.65 17.65 0.29 10.58 18.51 0.45 6.51 18.35 0.60 3.55 12.79 0.76 2.07 17.34 0.92 0.96 25.00 1.07 0.62 13.46 1.22 0.43 13.89 1.38 0.19 0.00 Notes: The sample includes all teachers in grades 4-8 in one large Texas district. The mea- sure of quality is the difference between average student gain in mathematics for a teacher and the average gain for all other teachers in the school. These relative gains are divided into twenty equal intervals, and the index for each interval is the interval mean. Frequency is the percentage of all teachers in the city in the category, and exit rate is the percentage of teachers who leave the school at the end of the year. in the sample. Therefore, even if attrition is not random for the sample as a whole, as long as it is not far more concentrated in the tails than is observed for this single large district, it is extremely unlikely that it would introduce much if any upward bias.26 A final robustness check examines only schools with a single teacher per grade. This quite select sample generates large, positive, and statistically sig- nificant estimates in both mathematics and reading for the first two specifica- tions (see Table C2). Not surprisingly given the extremely small sample sizes, the estimates for the full fixed effect specification remain positive but are quite imprecise. 26Note that the estimates of within school variation in quality based on individual teachers are three times as large as our lower bound estimates in Table III. Of course, these estimates do not deal with the selection effects that are the heart of the estimation here. They also include potentially important measurement error. Section 9-49 440 S. RIVKIN, E. HANUSHEK, AND J. KAIN Importantly, the true magnitudes of the variances in mathematics and read- ing teacher quality are likely to be larger than the estimates presented here. First, the identifying assumptions are likely to be violated in ways that bias downward the extent of actual teacher quality differences within schools. Sec- ond, the measures of teacher turnover and number of teachers likely contain some error, and the ratio of the two may in fact have substantial measurement error that would likely attenuate the coefficients. For example, the exclusion of schools with large changes in the number of teachers in a grade from year to year, an indicator of problematic data, tends to increase coefficient magni- tudes and the precision of virtually all estimates. Finally, we focus on just one component of the variance in teacher quality, the within-school variance. All between-school variation in teacher quality is ignored-not because of a be- lief it is small, but rather because it cannot be readily separated from other factors. Thus, there can be little doubt that teacher quality is an important de- terminant of reading and mathematics achievement in elementary school and mathematics achievement in junior high school. 6. EDUCATION PRODUCTION FUNCTION ESTIMATES The frequently employed implicit assumption that schools are homogenous institutions is clearly contradicted by the finding of substantial within-school heterogeneity in teacher quality. These results also contrast sharply with the much smaller estimated differences in teacher and school quality that comes from studies investigating the impacts of specific school or teacher character- istics. Nevertheless, because teacher salaries are closely linked with experience and formal education and because class size reductions have been a widely dis- cussed and often used policy tool, a better understanding of the effects of these specific factors remains important. From a policy viewpoint, a comparison of .the costs and benefits of smaller classes or more educated and experienced teachers with those of improved general teacher quality would be particularly informative. The results from the existing large body of literature on the effects of school resources on a variety of outcomes remain highly variable, in large part, we be- lieve, because of difficulty of controlling for other relevant achievement inputs due to both conceptual and data limitations." The main concern is that ei- ther explicit resource allocation rules-such as the provision of compensatory funds for poor achievers-or simple omitted variables problems could mask 27For summaries of the education production function literature, see Hanushek (1986, 2003), Leva'id and Vignoles (2002), and Woessmann (2004). This work has been quite varied and con- troversial (Burtless (1996)). While concentrated on analyses of test score performance, contin- uing attention has also turned to longer run impacts on labor market outcomes (see, e.g., Card and Krueger (1992), Betts (1995), Heckman, Layne-Farrar, and Todd (1996), Dearden, Ferri, and Meghir (2002), and Dustmann, Rajah, and van Soest (2003)). Section 9-50 ACADEMIC ACHIEVEMENT 441 or distort true causal impacts. A set of more recent studies focuses specifically on identifying factors leading to exogenous variation in class size in order to uncover causal impacts.28 Unfortunately, identification of truly exogenous de- terminants of class size, or resource allocations more generally, is sufficiently rare that other compromises in the data and modeling are frequently required. These jeopardize the ability to obtain consistent estimates of resource effects and may limit the generalizability of any findings. As described in Section 3, our framework eliminates directly the most trou- bling potential endogeneity problems that are the focus of the alternative in- strumental variables approaches. The large samples also permit detection of small effects that may differ by grade or student demographic characteristics, allowing us to distinguish between low power of tests and the true lack of a relationship. 6.1. Empirical Specification of Resource Models Equation (8) describes the value-added empirical model that forms the basis of our examination of school resource effects on achievement. This is a modi- fied version of equation (2) that adds a vector of school resource characteristics (SCH) measured at the grade level and a set of observable, time varying family characteristics (X): (8) AA A =SCHs A + Xig+3+yi+8S ++ Wrgs + Us. composite error The family characteristics include indicator variables for students who switch schools and students who are eligible to receive a free or reduced price lunch. Teacher and school characteristics are computed separately for each grade and subject, and they include the average class size in regular classrooms,29 the pro- portion of teachers with a master's degree, and the proportion of teachers who 28A variety of different approaches have been applied to sort out the causal influence of school resources including instrumental variables approaches relying upon various circumstances of the schooling institutions (e.g., Angrist and Lavy (1999), Feinstein and Symons (1999), Hoxby (2000), Woessmann and West (forthcoming), Dobbelsteen, Levin, and Oosterbeek (2002), Robertson and Symons (2003), and Bonesronning (2004)) and direct consideration of potential pre-treatment selection factors (e.g., Dearden, Ferri, and Meghir (2002)). 29As Boozer and Rouse (1995) and others have pointed out, it is important to separate reg- ular and special education students, because class size and possibly other characteristics differ dramatically by population served and because special education students are much less likely to take tests. If the proportion of students in special education classes or the gap between reg- ular classroom and special education class size differs across schools, estimates of the effect of class size based on the entire school average will be biased. Our measure of class size is the aver- age class size for regular classrooms in specific grades and subjects. Both special purpose classes and student achievement for special education and Limited English Proficiency (LEP) students are eliminated from this estimation. At the same time, special education students in regular class- room instruction are included in the calculation of class size because they will affect the resources Section 9-51 442 S. RIVKIN, E. HANUSHEK, AND J. KAIN fall into four experience categories: zero years, one year, two years, and three or four years (with the omitted category being five years and above).30 The composite error terms should be reinterpreted as the unobserved components of students and schools. Note that we have added two additional error terms: school-by-year fixed effects (8,y) and school-by-grade fixed effects (wgs). These absorb the school fixed effects previously considered. Unlike most educational studies, we concentrate specifically on the actual class sizes reported by regular classroom teachers rather than the more com- mon pupil-teacher ratios for a school. Further, considerable attention was given to the elimination of measurement error in the school variables. We have access to longitudinal information on key data and can therefore adjust reports for inconsistencies that occur over time. Data Appendix B describes in detail the construction of the school characteristics and sample selection criteria. Virtually all prior analyses of school resource effects have estimated specifi- cations similar to equation (8) in either level or growth form, but none has been able to account for all of the fixed components of the composite error term. The elimination of these factors in the estimation of equation (8) addresses virtually all of the concerns typically raised about estimation of educational production functions. For example, arguments about simultaneity arising from compensatory resource allocations based on student performance are directly eliminated, since the level and expected rate of gain of achievement for each student are explicitly dealt with through the investigation of AA and the esti- mation of the individual yi's. The removal of school fixed effects would also control for time invariant school characteristics that might be related to the included teacher and school characteristics. Though the removal of simple school fixed effects (5,) would eliminate the confounding influences of fixed school factors including stable curriculum, neighborhood factors, peer characteristics, school and district leadership, and school organization, changes over time in other school factors may be corre- lated to changes in the included teacher and school characteristics. Consider the possibility that other events in a school-leadership changes, curricular de- velopments, student perceptions and flows, or the like-influence achievement directly and are correlated with changes in school and teacher characteristics. Importantly, the availability of a number of cohorts permits the inclusion of school-by-year fixed effects (S,) rather than simple school fixed effects in some allocated to regular instruction students in those classrooms. Separate analysis of special educa- tion is found in Hanushek, Kain, and Rivkin (2002). 30Including the percentages of teachers with five to nine and twenty or more years of experi- ence as separate categories did not change any of the results, and the hypotheses that teachers with five to nine or twenty or more years of experience had a different impact from those with ten or more years of experience was rarely rejected at any conventional significance level. The class size and teacher education estimates also remained unchanged if average experience was used in place of the experience categories. Section 9-52 ACADEMIC ACHIEVEMENT 443 specifications in order to account for any such systematic year-to-year changes in school factors. Any pattern of events or policies common to the neighbor- hood and school will be eliminated, and the estimates are identified solely by within-school-by-year differences across grades.31 We believe an extremely strong case can be made that the remaining dif- ferences in class size and other teacher characteristics emanate from two un- contaminated sources: random differences between cohorts in the number of students who transfer in or out of the school as students age (i.e., changes in enrollment);32 and school or district induced changes in class size policies that are unlikely to be systematically related to the time varying error components of individual students, controlling for student and school-by-year fixed effects in achievement gains.33 This approach to estimation goes well beyond what has been possible even with the specialized effects of institutional structure that have entered into past instrumental variables estimation. A concern, however, is that the signal to noise ratio falls with the removal of the multiple fixed effects, thus making it difficult to estimate the remaining elements of the specification. We consider this possibility below. 6.2. Impact of Teacher and School Characteristics Table VII reports the full range of estimates obtained from value-added models that progressively contain no fixed effects; student and school fixed ef- fects; student and school-by-year fixed effects; and, finally, student, school-by- year, and school-by-grade fixed effects.34 Based on preliminary findings, class size effects are further allowed to differ by grade. Robust standard errors that account for the correlation of unobservables within a school are reported for all coefficients.35 Table B1 presents descriptive statistics for the school charac- teristics and achievement gain. 31Less substantively, we also allow for changes in the tests over time through inclusion of a fixed effect for year for each subject-grade test (rgy). 32Note that the estimation explicitly controls for the effects of moving on the moving students' achievement growth; see Hanushek, Kain, and Rivkin (2004a). 33The availability of multiple cohorts also permits the inclusion of school-by-grade fixed effects, though at a cost of losing the ability to identify variable effects in the single 4th grade cohort. This may be important if, as suggested to us by Caroline Hoxby, school average achievement and class size change in a systematic way as students progress through school. However, the lack of systematic differences in class size by student demographic composition in any grade suggests that such problems are very minor if they exist at all. In the most complete model, coefficients are identified by school-by-grade-by-year differences in characteristics and achievement gains. 34Related to the work in the prior section, we also included (not shown) the level of teacher turnover in each year but found that it never had a systematic influence on student achievement. Stable differences in teacher turnover for each school are removed with the school fixed effects. 35Robust standard errors in Tables VII-IX are clustered at the school level to correct for gen- eral autocorrelations among the errors across cohorts of students attending the same school; for a discussion of the issue in a related context, see Bertrand, Duflo, and Mullainathan (2004). Section 9-53 444 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE VII EFFECTS OF TEACHER AND SCHOOL CHARACTERISTICS ON 4TH-7TH GRADE GAINS IN MATHEMATICS AND READING TEST SCORES (ROBUST STANDARD ERRORS IN PARENTHESES; n = 1,336,903 FOR MATHEMATICS AND 1,330,791 FOR READING) Student and Student and Student, School-by-Grade No Fixed School Fixed School-by-Year and School-by-Year Effects Effects Fixed Effects Fixed Effects 1. Mathematics Class size 4th grade -0.0049 -0.0106 -0.0107 n.a. (0.0023) (0.0040) (0.0037) 5th grade -0.0043 -0.0085 -0.0081 -0.0055 (0.0010) (0.0017) (0.0024) (0.0018) 6th grade -0.0014 -0.0037 -0.0041 -0.0027 (0.0010) (0.0017) (0.0020) (0.0013) 7th grade 0.0002 0.0025 0.0032 0.0011 (0.0009) (0.0020) (0.0024) (0.0023) Experience Proportion -0.085 -0.103 -0.128 -0.073 0 years (0.012) (0.021) (0.028) (0.023) Proportion -0.043 -0.066 -0.055 -0.002 1 year (0.013) (0.022) (0.028) (0.023) Proportion -0.018 -0.045 -0.055 -.002 2 years (0.013) (0.021) (0.030) (0.022) Proportion -0.012 -0.031 -0.030 -0.017 3-5 years (0.010) (0.018) (0.022) (0.018) Education Proportion with -0.025 -0.018 -0.023 -0.021 graduate degree (0.009) (0.017) (0.021) (0.020) 6.2.1. Class size The results reveal statistically significant effects of class size on both math- ematics and reading achievement gains, but the impact declines markedly as students progress through school and tends to be smaller and less significant in reading than in mathematics. The discussion concentrates on the model that removes school-by-year fixed effects, because 4th grade estimates cannot be produced for models that contain school-by-grade fixed effects with only the single available 4th grade cohort. The estimated effects of class size are quite similar quantitatively and qual- itatively across specifications that include student and either school or school- by-year fixed effects.36 Both the 4th and 5th grade class size coefficients are 36However, the addition of school-by-grade fixed effects substantially reduces the magnitudes and significance levels of estimates in mathematics though not in reading. Nevertheless, class size continues to exert a significant effect on mathematics achievement in grades 5 and 6. It is Section 9-54 ACADEMIC ACHIEVEMENT 445 TABLE VII-CONTINUED Student and Student and Student, School-by-Grade No Fixed School Fixed School-by-Year and School-by-Year Fixed Effects Effects Fixed Effects Effects 2. Reading Class size 4th grade -0.0031 -0.0090 -0.0092 n.a. (0.0017) (0.0031) (0.0029) 5th grade 0.0000 -0.0033 -0.0032 -0.0043 (0.0007) (0.0012) (0.0018) (0.0016) 6th grade 0.0021 0.0000 -0.0003 -0.0021 (0.0009) (0.0013) (0.0019) (0.0013) 7th grade -0.0046 -0.0022 -0.0028 -0.0013 (0.0008) (0.0017) (0.0024) (0.0020) Experience Proportion -0.041 -0.045 -0.064 -0.026 0 years (0.010) (0.019) (0.023) (0.021) Proportion -0.037 -0.042 -0.070 -0.002 1 year (0.010) (0.018) (0.023) (0.020) Proportion -0.004 -0.006 -0.018 0.002 2 years (0.010) (0.019) (0.025) (0.020) Proportion 0.001 0.014 0.002 0.018 3-5 years (0.009) (0.015) (0.020) (0.017) Education Proportion with -0.014 -0.004 0.001 0.010 graduate degree (0.007) (0.014) (0.018) (0.017) Note: All specifications include a full set of grade-by-year dummies and indicators for subsidized lunch eligibility and a change of school prior to or during year. Robust standard errors in Tables VII-IX are clustered at the school level to correct for general autocorrelations among the errors across cohorts of students attending the same school; for a discussion of the issue in a related context, see Bertrand, Duflo, and Mullainathan (2004). highly significant in both subjects, though the magnitude of the 5th grade ef- fect is roughly three-fourths as large as that for 4th grade in mathematics and less than half as large in reading. The 6th grade effects are quite small, and by 7th grade class size appears to have little systematic effect on achievement. We discuss the magnitude of these estimates below. Note that the very large samples permit the precise estimation of quite small effects of less than 0.004 standard deviations. The pattern of estimated class size effects also reveals the importance of controlling for student fixed effects. The inclusion of student fixed effects not possible to know for certain the extent to which change with the addition of school-by-grade fixed effects results from the elimination of further biases as opposed to the exacerbation of any problems with measurement error. Section 9-55 446 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE VIII EFFECTS OF CLASS SIZE ON TEST SCORE GAINS, BY FAMILY INCOME (ROBUST STANDARD ERRORS IN PARENTHESES) Mathematics Reading Disadvantaged Not Disadvantaged Disadvantaged Not Disadvantaged Students Students Students Students Class size 4th grade -0.0118 -0.0103 -0.0111 -0.0087 (0.0038) (0.0037) (0.0030) (0.0029) 5th grade -0.0077 -0.0079 -0.0027 -0.0033 (0.0025) (0.0024) (0.0019) (0.0018) 6th grade -0.0044 -0.0040 -0.0022 -0.0007 (0.0021) (0.0020) (0.0019) (0.0017) 7th grade 0.0036 0.0031 0.0012 -0.0037 (0.0026) (0.0024) (0.0023) (0.0022) Note: Estimates come from a single mathematics regression and a single reading regression. The models include student and school-by-year fixed effects, separate class size, and teacher experience variables for students eligible for a subsidized lunch (disadvantaged) and those not eligible during a given school year, proportion of teachers with a graduate degree, full sets of grade-by-year dummies, and indicators for subsidized lunch eligibility and a change of school prior to or during year. triples the 4th grade coefficient and more than doubles the coefficient for 5th grade.37 An important and often studied question is whether lower income students receive larger benefits from class size reduction. In order to examine this claim we relaxed the restriction that class size effects were the same by income (mea- sured by subsidized lunch eligibility). The results in Table VIII generally do not support the belief that class size effects are substantially larger for disadvan- taged (subsidized lunch eligible) students. Class size effects are roughly 20 per- cent larger for disadvantaged students in 4th grade but actually smaller in 5th grade. Both the grade pattern and the comparable mathematics and reading results are very similar to the results in Table VII. One potential perspective on these estimates comes from Project STAR, the random assignment experiment in class size reduction conducted in Tennessee (Word et al. (1990)).38 While these experimental results are not directly com- parable because they consider just grades K to 3, they indicate that a reduction 37The progressively more stringent estimates found across the columns does introduce some instability in the estimates, particularly in the final column. The smaller though still significant coefficients in the full fixed effects model for mathematics are consistent with the possibility that the school-by-grade and school-by-year fixed effects together aggravate problems associated with measurement error, but the results for reading go in the opposite direction. 38Project STAR randomly assigned a large group of kindergarten students to regular sized classes (22-25 students), regular sized classes with an aide, or small classes (13-17 students). It was designed to follow these students through grade 3, but there were significant attrition problems and subsequent additions of students to the experiment. Achievement tests were given Section 9-56 ACADEMIC ACHIEVEMENT 447 of eight students per class yields kindergarten achievement gains in math and reading of 0.17 standard deviations, which is roughly 60 percent larger than our 4th grade result for mathematics and reading. However, the deeper inconsis- tency that cannot be resolved here is that the experimental results indicate that virtually all of the achievement gain in STAR is associated with the first year in a small class-generally kindergarten or 1st grade-and not subsequent small class treatments (Krueger (1999)), while we find that smaller classes still have an effect in 4th and 5th grade. The STAR experiment also reveals very large variation in student perfor- mance across individual classrooms. Specifically, all randomization occurred within each experimental school, and students in the large classes outper- formed schoolmates in smaller classes in almost half of the schools (Hanushek (1999b)). This experimental finding is consistent with the conclusions here that differences in teacher quality within schools are quite large. The school-by-year fixed effect estimates in column 3 of Table VII provide the basis for a simple comparison of policy alternatives. While it is difficult to estimate the cost of improving teacher quality, our lower bound estimates of the variation in quality found just within schools indicate that one stan- dard deviation in quality is worth at least 0.11 standard deviations higher an- nual growth in mathematics achievement and 0.095 standard deviations higher annual growth in reading in elementary school. This magnitude of change is equivalent to a class size reduction of approximately ten students in 4th grade and thirteen or more students in 5th grade, and an implausibly large number in 6th grade. In 7th grade there appears to be no significant benefit from smaller classes in mathematics, while in reading neither class size nor teacher quality appears to exert a substantial effect on achievement. Note that these compar- isons assume both no accompanying changes in teacher quality and linearity in class size effects, the latter of which appears reasonable based on semi- parametric estimates for class sizes between 10 and 35 students (results not reported). 6.2.2. Teacher characteristics The results for teacher experience generally support the notion that begin- ning teachers and to a lesser extent second and third year teachers in mathe- matics perform significantly worse than more experienced teachers. There may be some additional gains to experience in the subsequent year or two, but the estimated benefits are small and not statistically significant in both mathemat- ics and reading in any of the fixed effect specifications. Similar to the case for class size, the results in the full fixed effect model in column 4 are much weaker at the end of each grade, and a comparison showed that students in small classes outperformed those in regular classes in their first experimental year (K or 1) but that no additional gains were made. See Hanushek (1999b) and Krueger (1999). Section 9-57 448 S. RIVKIN, E. HANUSHEK, AND J. KAIN than in the other fixed effects models, consistent with the view that multiple fixed effects can exacerbate problems with measurement error. The addition of school-by-grade fixed effects reduces the magnitude of all coefficients, and only the estimated effect of proportion of new teachers on math achievement gain is significant. Importantly, the teacher experience effect conceptually combines two very distinct phenomena. First, new teachers may need to go through an adjustment period where they learn the craft of teaching along with adjusting to the other aspects of an initial job. Second, a number of the early teachers discover that they are not well matched for teaching and subsequently leave the profession within the first few years. Between entry and the end of two years, 18 percent of teachers will leave the Texas public schools, and another 6 percent will switch districts (Hanushek, Kain, and Rivkin (2004b)). The estimated parameters in Table VII combine the effects of on-the-job learning and of selective exit and mobility. Table IX presents the basic estimates of first year teaching on achievement (with individual and school fixed effects) for samples that exclude those who immediately leave teaching or switch schools. The close similarity of the esti- mates across the samples compared to those in Table VII for both mathematics and reading indicates that on-the-job learning is the dominant element of the experience effect. Importantly, these results also suggest that the average qual- ity of those who quit teaching after one year is similar to the average quality of those who remain, providing additional support for the validity of the estimates of the variance in teacher quality. TABLE IX EFFECTS OF PROPORTION OF TEACHERS WITH ZERO YEARS OF EXPERIENCE ON MATHEMATICS AND READING TEST SCORE GAINS, BY NEW TEACHER TRANSITIONS (ROBUST STANDARD ERRORS IN PARENTHESES) Excluding Teachers Outcome Who Exit Teaching or Excluding Teachers Measure Switch Schools Who Exit Teaching All Teachers 1. Mathematics Proportion of teachers with 0 -0.105 -0.114 -0.103 years experience (0.030) (0.028) (0.021) Observations [1,185,329] [1,210,155] [1,336,903] 2. Reading Proportion of teachers with 0 -0.040 -0.040 -0.045 years experience (0.024) (0.023) (0.019) Observations [1,181,611] [1,206,139] [1,330,791] Note: Estimates come from a model that includes student and school fixed effects. Specifications also include the percentage of teachers with a graduate degree, full sets of class size variables, and grade-by-year dummies and indicators for subsidized lunch eligibility and a change of school prior to or during year. Section 9-58 ACADEMIC ACHIEVEMENT 449 Finally, consistent with previous work, there is little or no evidence that a master's degree raises the quality of teaching. All estimates are small (or neg- ative) and statistically insignificant. 7. CONCLUSIONS Prior investigations of school and teacher effects have raised as many ques- tions as they have answered, in large part because of the difficulties introduced by the endogeneity of school and classroom selection and in part because of the failure of observable teacher characteristics to explain much of the varia- tion in student performance. The models and data used in this paper permit us to draw a number of sharp conclusions about public elementary education and to provide clear answers for the questions raised in the Introduction. (i) Teachers and therefore schools matter importantly for student achieve- ment. The issue of whether or not there is significant variation in school quality has lingered, quite inappropriately, since the original Coleman Report. This analysis identifies large differences in the quality of instruction in a way that rules out the possibility that the observed differences are driven by family fac- tors. The Coleman Report also popularized the issue of whether family influences are "more important" than school influences. This is not the relevant question for policy, which should focus on whether the benefits produced by any in- tervention justify the costs. Though our analysis does not consider the costs of raising teacher quality, the estimated variation in the quality of instruc- tion clearly reveals an important role for schools and teachers in promoting economic and social equality. Even if none of the between-school variation in achievement is attributed to schools or teachers, it is clear that school policy can be an important tool for raising the achievement of low income students and that a succession of good teachers could, by our estimates, go a long way toward closing existing achievement gaps across income groups. At the very least, more must be known about the feasible means of providing such consis- tently high quality teachers. (ii) Achievement gains are systematically related to observable teacher and school characteristics, but the effects are generally small and concentrated among younger students. This analysis used a fixed effects approach to identify the causal relationship between achievement and key school resources. Four major conclusions emerge from this work. * Similar to most past research, we find absolutely no evidence that having a master's degree improves teacher skills. * There appear to be important gains in teaching quality in the first year of experience and smaller gains over the next few career years. However, there is little evidence that improvements continue after the first three years. * Class size appears to have modest but statistically significant effects on math- ematics and reading achievement growth that decline as students progress through school. Section 9-59 450 S. RIVKIN, E. HANUSHEK, AND J. KAIN * Any differences in school resource effects by family income are small. Partially consistent with recent experimental and statistical efforts to identify class size effects, we find that lowering class size has a positive effect on math- ematics and reading achievement, though the magnitude of the effect is small, particularly following 5th grade. The costs of class size reduction have not been well estimated, but they are likely to exceed the proportional increase in the number of teachers needed to staff the smaller classes. First, class size reduc- tion almost certainly leads to more support expenditure, increased building re- quirements, and the like. Second, and more directly relevant to this discussion, it is highly unlikely that the supply of teacher quality is perfectly elastic, so that expansion of the teacher work force, at least in the short run, is likely to lead either to increased salary demands or a reduction in teacher quality. More- over, the potential tradeoff between teacher quality and class size is probably most acute in difficult to staff schools serving largely disadvantaged student populations (Hanushek (1999a), Jepsen and Rivkin (2002)). (iii) The disjuncture between estimates of the variation of teacher quality and the explanatory power of measured teacher characteristics creates a clear dilemma for policy makers. Though it is tempting to tighten standards for teachers in an effort to raise quality, the results in this paper and elsewhere raise serious doubts that more restrictive certification standards, education levels, etc. will succeed in raising the quality of instruction. Rather the substantial differences in quality among those with similar observable backgrounds highlight the im- portance of effective hiring, firing, mentoring, and promotion practices. Re- search shows that principals can, when asked, separate teachers on the basis of quality (Murnane (1975), Armor et al. (1976)), but the substantial varia- tion documented in this paper strongly suggests that personnel practices in the Texas public schools are very imperfect. One dimension of policy does, nonetheless, deserve special attention. Eco- nomically disadvantaged students systematically achieve less than more advan- taged students, on average falling some 0.6 standard deviations behind.39 While we find little reason to believe that school resources have a larger impact on disadvantaged students, we do know that low income and minority students face higher teacher turnover and tend to be taught more frequently by be- ginning teachers (Hanushek, Kain, and Rivkin (2004b)). Because beginning teachers, regardless of their ultimate abilities, tend to perform more poorly, policies should be developed to both keep more senior teachers in the class- rooms of disadvantaged students and to mitigate the impact of inexperience. These may include improved mentoring of new teachers and policies designed specifically to cut down teacher turnover. Of course, it goes without saying that 39The measure of family income is eligibility for a free or reduced price school lunch. This measure, while quite commonly used because of its availability in administrative records, is an imprecise categorization of economic circumstances. Section 9-60 ACADEMIC ACHIEVEMENT 451 effective policies will pay particular attention to the substantial variation in teacher quality. The desirability of specific policy changes remains quite speculative be- cause of the limited experience with alternative organizational forms, incen- tives, and accountability policies. A very appealing though untested approach to raising teacher quality would move the focus away from the state legisla- tures and schools of education and toward principals and other administrators (Hanushek and Rivkin (2004)). In the presence of incentives such as expanded choice, school report cards, or other types of accountability systems, admin- istrators would likely alter their behavior and personnel policies in ways that benefit students. In particular, there would likely be much more focus on stu- dent outcomes of interest. Not only would improved personnel policies likely raise the performance level of existing teachers, there is strong reason to be- lieve that a closer link between rewards and performance would improve the stock of teachers. Of course inappropriate incentives likely lead to adverse out- comes, and it is imperative that schools learn from their mistakes and evolve toward more effective systems of school governance. Dept. of Economics, Amherst College, Amherst, MA 01002, U.S.A.; sgrivkin @amherst.edu, Hoover Institution, Stanford University, Stanford, CA 94305, U.S.A.; hanushek@stanford.edu; http://www.hanushek.net, and University of Texas at Dallas (deceased). Manuscript received July, 2002; final revision received October, 2004. APPENDIX A: THE EFFECT OF NONRANDOM TEACHER ATTRITION ON TURNOVER-BASED ESTIMATOR OF TEACHER QUALITY VARIATION The estimator of teacher quality derived from equation (7) assumes that the error term (e) is uncorrelated with teacher turnover. If, however, there is sys- tematic teacher attrition that varies by quality, the estimator may no longer be a lower bound but may in fact overestimate the variance in quality. This specif- ically would be the case if attrition is concentrated in the tails of the quality distribution. It is most natural to think of this as a problem of sample selection where teachers who depart have a different distribution in terms of quality than those who remain. Thus, schools with turnover would tend to have a different quality distribution for teachers. The nature of the problem with selective attrition using our estimator is eas- iest to see in the simpler comparison of the squared difference in grade g gains for successive cohorts, although it would easily generalize to the full estimator. The subtraction of 5th grade average gain from 6th grade average gain for a cohort removes any student and school fixed effects (including overall hiring practices) but does not address problems related to nonrandom teacher depar- tures. Section 9-61 452 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE Al UNDERLYING DISTRIBUTION OF TEACHER QUALITY FOR NEW HIRES Relative Teacher Quality (q) Frequency: f(q) -1 0.25 0 0.50 1 0.25 The potential impact of selective attrition is directly seen from a simple sim- ulation using a trinomial quality distribution. Table Al describes a distribution of new hires that has a variance of quality equal to 0.5. With this distribution of new hires, it is possible to simulate the estimator of school quality both with random departures and with systematic departures that differ across the distri- bution. First, consider the turnover-based estimator of the variance in teacher qual- ity when there are random departures. Table A2 begins with the distribution of teacher quality in Table Al and then assumes that teachers leave randomly (and are replaced by a random selection of teachers according to the distrib- ution in Table Al). Consequently there are nine possible transitions, three for each of the period 0 quality categories. In this simple one grade example, the expected period 0/period 1 difference in quality is two times the variance in teacher quality (instead of four times the variance as derived in the full estimator that considers deviations across grades and cohorts). Table A2 shows that the estimator yields the true variation in quality when there is random hiring and departures. Consider, however, the identical estimator with strongly nonrandom depar- tures characterized by probabilities of departure of 0.5, 0.0, and 0.5 for the TABLE A2 TRANSITION MATRIX AND VARIANCE ESTIMATE WITH RANDOM ATTRITION Relative Teacher Relative Teacher Transition Frequency: Squared Quality Quality (qo) Period 1 Quality (ql) Period 2 f(q1, qo) Difference (ql - q0)2 -1 0.0625 0 -1 0 0.125 1 +1 0.0625 4 -1 0.125 1 0 0 0.250 0 +1 0.125 1 -1 0.0625 4 +1 0 0.125 1 +1 0.0625 0 Notes: Weighted sum of squared differences = 1.0; estimated variance = 1/2 squared differences = 0.5. Section 9-62 ACADEMIC ACHIEVEMENT 453 TABLE A3 TRANSITION MATRIX AND VARIANCE ESTIMATE WITH NONRANDOM ATTRITION CONCENTRATED IN THE TAILS OF THE QUALITY DISTRIBUTION Relative Teacher Relative Teacher Transition Frequency: Squared Quality Quality (qo) Period 1 Quality (ql) Period 2 f(q1, qo) Difference (ql - q0)2 -1 0.125 0 -1 0 0.250 1 +1 0.125 4 -1 0.0 1 0 0 0.0 0 +1 0.0 1 -1 0.125 4 +1 0 0.250 1 +1 0.125 0 Notes: Weighted sum of squared differences = 1.5; estimated variance = 1/2 squared differences = 0.75. three quality groups in Table Al. Table A3 describes the transition probabili- ties, sum of squared quality differences, and the simulated variance estimates. If departures were as concentrated in the tails of the distribution as they are in this example, our method would overstate the variance in teacher quality by 50 percent: 0.75 instead of 0.5. Note that this upward bias would also arise if all departures were concentrated in only one of the tails of the distribution. In general, if attrition is weighted toward the tails of the quality distribution the turnover-based estimator will tend to overestimate the variance of quality, and the opposite will hold if attrition is concentrated in the center of the quality distribution. APPENDIX B: TEXAS SCHOOL DATA The data that are used in this paper come from the data development activ- ity of the UTD Texas Schools Project of the University of Texas at Dallas; see Kain (2001). Working with the Texas Education Agency (TEA), this project has combined a number of different data sources to compile an extensive data set on schools, teachers, and students. Demographic information on students and teachers is taken from the PEIMS (Public Education Information Man- agement System), which is TEA's statewide educational data base. Test score results and a limited amount of student demographic information are stored in a separate data base maintained by TEA and must be merged with the student data on the basis of unique student IDs. Data are compiled for all public school students in Texas, allowing us to use the universe of students in the analyses. In this paper all of the information on students comes from the test score data base, and we combine student information from the Texas Assessment of Aca- demic Skills (TAAS) data base with teacher and school information contained in the PEIMs data base for three student cohorts: 3rd through 7th grade test Section 9-63 454 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE B1 VARIABLE MEANS AND STANDARD DEVIATIONS Teacher Characteristics Math Test Reading Test Class % with % 0 Years % 1 Year Score Gain Score Gain Size Graduate Degree Experience Experience Observations 4th grade -0.01 -0.02 19.5 23.7 6.1 5.9 143,314 (0.70) (0.73) (2.3) (24.3) (12.4) (12.5) 5th grade 0.01 0.01 22.6 25.1 5.9 6.0 438,561 (0.64) (0.68) (3.6) (26.2) (13.7) (13.6) 6th grade 0.02 0.02 22.1 24.5 7.4 6.9 455,438 (0.61) (0.68) (3.9) (27.4) (16.6) (15.7) 7th grade -0.02 -0.01 21.5 22.0 9.2 8.9 299,590 (0.55) (0.66) (4.2) (26.7) (18.4) (18.0) scores for one cohort (4th graders in 1995) and 4th through 7th grade test scores for the other two (4th graders in 1993 and 1994).40 Beginning in 1993, the Texas Assessment of Academic Skills (TAAS) was ad- ministered each spring to eligible students enrolled in grades 3 through 8. We focus on test results for mathematics and reading. The bottom one percent of test scores are trimmed from the sample in order to reduce measurement error. Participants in bilingual or special education programs are also excluded from the sample, because of the difficulty in measuring school and teacher charac- teristics for students who split time between regular classrooms and special programs. Student data are merged with information on teachers using unique school identifiers. The personnel data provide information on all Texas public school teachers for each year. Experience and highest degree earned are reported, as are the class size, subject, grade, and population served for each class taught. Although the currently available data do not permit linking individual students with specific teachers, the available information is used to construct subject and grade average characteristics for teachers in regular classrooms. In an effort to reduce problems associated with measurement error, a num- ber of observations are excluded from the data set. The following paragraphs describe in detail the construction of the variables and the sample selection procedures. Measurement error in the teacher characteristics is an important issue. In many cases reported teacher experience in one year does not correspond with reported teacher experience for other years. If the experience sequence is valid except for one or two years that do not follow from the others, we correct ex- 40Note that, while we have 3rd grade test information, our analysis begins at 4th grade because of the focus on achievement gains. Section 9-64 ACADEMIC ACHIEVEMENT 455 perience for those years. If experience data are inconsistent for all the years, if there are two consistent patterns, or if correction would impute negative years of experience, no corrections are made. In any case, no teachers are excluded from the final sample on the basis of inconsistent experience data, though the results are not sensitive to their inclusion, possibly because we used discreet experience categories. The case of average class size is somewhat more complicated. Teachers were asked to report the average class size for each class they taught that was of a different size. Unfortunately, many teachers appear to have reported the total number of students taught per day. This becomes particularly problematic for schools that move from general to subject specific teachers. Consider a school with two 4th grade classes of twenty students in which the two teachers each teach all subjects. If the school switches to math and reading specialists for 5th grade and each teaches one subject for each class, they will report class sizes of forty if they report total number of students served. It will appear that class sizes doubled as students aged, when in fact they remain the same. In order to reduce problems introduced by measurement, all reported class sizes that fall below ten or above twenty five in 4th grade (thirty five in higher grades) are set to missing prior to the computation of school averages for each grade. By statute, 4th grade classes are not supposed to exceed twenty two stu- dents, though some schools receive waivers to provide slightly larger classes. It is our understanding that very few elementary schools in Texas have actual class sizes in later grades that exceed thirty five students during this period. Es- timates of class size effects increased in magnitude following these exclusions, suggesting that class size was measured with error for these schools. Access to the administrative data on student performance is currently re- stricted by U.S. federal law. Further information on data access along with the specific variable definitions, data construction, and data that may currently be released are found in Rivkin, Hanushek, and Kain (2005). APPENDIX C: ALTERNATIVE TEACHER QUALITY ESTIMATES TABLE C1 TEACHER QUALITY STANDARD DEVIATION ESTIMATES CALCULATED FROM SQUARED DIFFERENCE IN QUALITY FOR PERIODS 0 AND 1, BASED ON OBSERVED DISTRIBUTIONS OF TEACHER QUALITY AND DEPARTURE RATES Number of Teacher a Assuming Random a Assuming Empirical Distribution Quality Intervals Departures of Departures 20 (Table VI) 0.395 0.399 40 0.397 0.401 60 0.397 0.402 30 with tails 0.422 0.427 Section 9-65 456 S. RIVKIN, E. HANUSHEK, AND J. KAIN TABLE C2 EFFECT OF TEACHER TURNOVER ON THE DIVERGENCE OF GAINS IN MATHEMATICS AND READING TEST SCORES BETWEEN COHORTS FOR SCHOOLS WITH ONE TEACHER PER GRADE (STANDARD ERRORS IN PARENTHESES) No Fixed Individual and Individual and Effects School Fixed Effects School-by-Grade Fixed Effects 1. Mathematics Proportion different 0.124 0.117 0.042 math teachers/number (0.039) (0.039) (0.047) of teachers 2. Reading Proportion different 0.181 0.180 0.061 English teachers/number (0.037) (0.049) (0.042) of teachers Notes: All equations include the inverse of the number of students, numbers of new principals and superintendents in the school during adjacent years, and a cohort dummy variable. Sample size is 294 for the mathematics and 300 for the reading specifications. Table III notes describe the estimation specifications. REFERENCES ANGRIST, J. D., AND V. LAVY (1999): "Using Maimondides' Rule to Estimate the Effect of Class Size on Scholastic Achievement," Quarterly Journal of Economics, 114, 533-575. ARMOR, D. J., P. CONRY-OSEGUERA, M. COX, N. KING, L. MCDONNELL, A. PASCAL, E. PAULY, AND G. ZELLMAN (1976): Analysis of the School Preferred Reading Program in Selected Los Angeles Minority Schools. Santa Monica, CA: Rand Corp. BERTRAND, M., E. DUFLO, AND S. MULLAINATHAN (2004): "How Much Should We Trust Differences-in-Differences Estimates?" Quarterly Journal of Economics, 114, 249-275. BETTS, J. R. (1995): "Does School Quality Matter? Evidence from the National Longitudinal Survey of Youth," Review of Economics and Statistics, 77, 231-247. BONESRONNING, H. (2004): "The Determinants of Parental Effort in Education Production: Do Parents Respond to Changes in Class Size?" Economics of Education Review, 23, 1-9. BOOZER, M. A., AND C. ROUSE (1995): "Intraschool Variation in Class Size: Patterns and Impli- cations," Working Paper 5144, National Bureau of Economic Research, Cambridge, MA. BURTLESS, G. (ED.) (1996): Does Money Matter? The Effect of School Resources on Student Achievement and Adult Success. Washington, DC: Brookings. CARD, D., AND A. B. KRUEGER (1992): "Does School Quality Matter? Returns to Education and the Characteristics of Public Schools in the United States," Journal of Political Economy, 100, 1-40. COLEMAN, J. S., E. Q. CAMPBELL, C. J. HOBSON, J. MCPARTLAND, A. M. MOOD, E D. WEINFELD, AND R. L. YORK (1966): Equality of Educational Opportunity. Washington, DC: U.S. Government Printing Office. DEARDEN, L., J. FERRI, AND C. MEGHIR (2002): "The Effect of School Quality on Educational Attainment and Wages," Review of Economics and Statistics, 84, 1-20. DOBBELSTEEN, S., J. LEVIN, AND H. OOSTERBEEK (2002): "The Causal Effect of Class Size on Scholastic Achievement: Distinguishing the Pure Class Size Effect from the Effect of Changes in Class Composition," Oxford Bulletin of Economics and Statistics, 64, 17-38. DOLTON, P. J., AND W. VAN DER KLAAUW (1995): "Leaving Teaching in the UK: A Duration Analysis," The Economic Journal, 105, 431-444. Section 9-66 ACADEMIC ACHIEVEMENT 457 (1999): "The Turnover of Teachers: A Competing Risks Explanation," Review of Eco- nomics and Statistics, 81, 543-552. DUSTMANN, C., N. RAJAH, AND A. VAN SOEST (2003): "Class Size, Education, and Wages," Eco- nomic Journal, 113, F99-120. FEINSTEIN, L., AND J. SYMONS (1999): 'Attainment in Secondary Schools," Oxford Economic Papers, 52, 300-321. GREENWALD, R., L. V. HEDGES, AND R. D. LAINE (1996): "The Effect of School Resources on Student Achievement," Review of Educational Research, 66, 361-396. HANUSHEK, E. A. (1971): "Teacher Characteristics and Gains in Student Achievement: Estima- tion Using Micro Data," American Economic Review, 60, 280-288. (1979): "Conceptual and Empirical Issues in the Estimation of Educational Production Functions," Journal of Human Resources, 14, 351-388. - (1986): "The Economics of Schooling: Production and Efficiency in Public Schools," Journal of Economic Literature, 24, 1141-1177. (1992): "The Trade-Off Between Child Quantity and Quality," Journal of Political Econ- omy, 100, 84-117. (1996): '"A More Complete Picture of School Resource Policies," Review of Educational Research, 66, 397-409. (1999a): "The Evidence on Class Size," in Earning and Learning: How Schools Matter, ed. by S. E. Mayer and P E. Peterson. Washington, DC: Brookings Institution, 131-168. (1999b): "Some Findings from an Independent Investigation of the Tennessee STAR Experiment and from Other Investigations of Class Size Effects," Educational Evaluation and Policy Analysis, 21, 143-163. (2003): "The Failure of Input-Based Schooling Policies," Economic Journal, 113, F64-F98. HANUSHEK, E. A., AND J. E KAIN (1972): "On the Value of 'Equality of Educational Opportunity' as a Guide to Public Policy," in On Equality of Educational Opportunity, ed. by E Mosteller and D. P. Moynihan. New York: Random House, 116-145. HANUSHEK, E. A., J. E KAIN, D. M. O'BRIEN, AND S. G. RIVKIN (2005): "The Market for Teacher Quality," Working Paper, National Bureau of Economic Research, Cambridge, MA. HANUSHEK, E. A., J. E KAIN, AND S. G. RIVKIN (2002): "Inferring Program Effects for Special- ized Populations: Does Special Education Raise Achievement for Students with Disabilities?" Review of Economics and Statistics, 84, 584-599. (2004a): "Disruption versus Tiebout Improvement: The Costs and Benefits of Switching Schools," Journal of Public Economics, 88/9-10, 1721-1746. (2004b): "Why Public Schools Lose Teachers," Journal of Human Resources, 39, 326-354. HANUSHEK, E. A., AND S. G. RIVKIN (2004): "How to Improve the Supply of High Quality Teach- ers," in Brookings Papers on Education Policy 2004, ed. by D. Ravitch. Washington, DC: Brook- ings Institution Press, 7-25. HECKMAN, J. J., A. LAYNE-FARRAR, AND P. TODD (1996): "Human Capital Pricing Equations with an Application to Estimating the Effect of Schooling Quality on Earnings," Review of Economics and Statistics, 78, 562-610. HoxBY, C. M. (2000): "The Effects of Class Size on Student Achievement: New Evidence from Population Variation," Quarterly Journal of Economics, 115, 1239-1285. INGERSOLL, R. M. (2001): "Teacher Turnover and Teacher Shortages: An Organizational Analy- sis," American Educational Research Journal, 38, 499-534. JEPSEN, C., AND S. G. RIVKIN (2002): "What Is the Trade-Off Between Smaller Classes and Teacher Quality?" National Bureau of Economic Research, Cambridge, MA. KAIN, J. E (2001): "The UTD Texas Schools Microdata Panel (TSMP): Its History, Use and Ways to Improve State Collection of Public School Data," Paper Prepared for The Secre- tary's Forum on Research and Value-Added Assessment Data, U.S. Department of Education; http://utdallas. edu/research/tsp/index. htm. Section 9-67 458 S. RIVKIN, E. HANUSHEK, AND J. KAIN KAIN, J. E, AND D. M. O'BRIEN (1998): 'A Longitudinal Assessment of Reading Achieve- ment: Evidence from the Harvard/UTD Texas Schools Project," University of Texas at Dallas, Dallas, TX. KRUEGER, A. B. (1999): "Experimental Estimates of Education Production Functions," Quarterly Journal of Economics, 114, 497-532. LEVAdti(, R., AND A. VIGNOLES (2002): "Researching the Links Between School Resources and Student Outcomes in the UK: A Review of Issues and Evidence," Education Economics, 10, 313-331. MURNANE, R. J. (1975): Impact of School Resources on the Learning of Inner City Children. Cam- bridge, MA: Ballinger. - (1984): "Selection and Survival in the Teacher Labor Market," Review of Economics and Statistics, 66, 513-518. MURNANE, R. J., AND R. OLSEN (1989): "The Effects of Salaries and Opportunity Costs on Length of Stay in Teaching: Evidence from Michigan," Review of Economics and Statistics, 71, 347-352. MURNANE, R. J., AND B. PHILLIPS (1981): "What Do Effective Teachers of Inner-City Children Have in Common?" Social Science Research, 10, 83-100. MURNANE, R. J., J. D. SINGER, J. B. WILLETr, J. J. KEMPLE, AND R. J. OLSEN (1991): Who Will Teach? Policies That Matter. Cambridge, MA: Harvard University Press. RIVKIN, S. G., E. A. HANUSHEK, AND J. E KAIN (2005): "Variable Definitions, Data, and Pro- grams for 'Teachers, Students, and Academic Achievement'," Econometrica Supplementary Ma- terial, 73, 2, www.econometricsociety.org/ecta/supmat/4139data.pdf ROBERTSON, D., AND J. SYMONS (2003): "Do Peer Groups Matter? Peer Group versus Schooling Effects on Academic Attainment," Economica, 70, 31-53. STINEBRICKNER, T. R. (2002): '"An Analysis of Occupational Change and Departure from the Labor Force," Journal of Human Resources, 37, 192-216. SUMMERS, A. A., AND B. L. WOLFE (1977): "Do Schools Make a Difference?" American Eco- nomic Review, 67, 639-652. THE TEACHING COMMISSION (2004): Teaching at Risk: A Call to Action. New York, NY: The Teaching Commission. TIEBOUT, C. M. (1956): 'A Pure Theory of Local Expenditures," Journal of Political Economy, 64, 416-424. WOESSMANN, L. (2004): "Educational Production in Europe," Paper Presented at 40th Meeting of Economic Policy in Amsterdam Ifo Institute for Economic Research at the University of Munich. WOESSMANN, L., AND M. R. WEST (forthcoming): "Class-Size Effects in School Systems Around the World: Evidence from Between-Grade Variation in TIMSS," European Economic Review, forthcoming. WORD, E., J. JOHNSTON, H. P. BAIN, B. D. FULTON, J. B. ZAHARIES, M. N. LINTZ, C. M. ACHILLES, J. FOLGER, AND C. BREDA (1990): Student/Teacher Achievement Ratio (STAR), Tennessee's K-3 Class Size Study: Final Summary Report, 1985-1990. Nashville, TN: Tennessee State Department of Education. Section 9-68 THEEFFECTSOFCLASSSIZEONSTUDENT ACHIEVEMENT:NEWEVIDENCEFROM POPULATIONVARIATION* CAROLINE M.HOXBY Iidentifytheeffectsofclasssizeonstudentachievementusinglongitudinal variationinthepopulationassociatedwitheachgradein649elementaryschools.I usevariationinclasssizedrivenbyidiosyncraticvariationinthepopulation.Ialso usediscretejumpsinclasssizethatoccurwhenasmallchangeinenrollment triggersamaximumorminimumclasssizerule.Theestimatesindicatethatclass sizedoesnothaveastatisticallysignicanteffectonstudentachievement.Irule outevenmodesteffects(2to4percentofastandarddeviationinscoresfora10 percentreductioninclasssize). I.INTRODUCTION Classsizereductionisprobablythemostpopularandmost fundedschoolimprovementpolicyintheUnitedStates.In1996 theCalifornialegislaturededicatedonebilliondollarsperyearto classsizereduction.The1999federalbudgetcontained12billion dollars(oversevenyears)forthesamepurpose.Classsize reductionsareenactedoftenbecausetheyarepopularwithnearly everyconstituencyinterestedinschools.Parentslikesmaller classesbecausetheirpersonalexperiencesuggeststhatthey themselvesgivemoretoeachchildwhentheyhavefewerchildren tohandle.Evenifparentsinaschooldisagreebitterlyabout educationalmethods,theycanagreethatclasssizereductionis good:smallerclassesgiveteacherstheopportunitytopractice moreof each parent’sfavorededucationalmethod.Teachers, teachers’unions,andadministratorslikesmallerclassesforthe samereasonsparentsdo,buttheymayalsolikesmallerclasses forreasonsthatspringfromself-interest.Teachersmaylike *TheauthorgratefullyacknowledgesthesupportofNationalScienceFoundationgrantSBR9511343.ShegratefullyacknowledgestheConnecticutDepartmentofEducation,whosestaffwereextremelyefficientandknowledgeable.NeithertheNationalScienceFoundationnortheConnecticutDepartmentof Educationisresponsibleforanystatementsorerrorsinthepaper.Theauthorgratefullyacknowledgeshelpfulideasfromtheeditors,theanonymousreferees,HenryFarber,RichardFreeman,EdwardGlaeser,DanielHamermesh,EricHanushek,LawrenceKatz,KevinLang,RobertLalonde,RichardMurnane, StevenRivkin,DouglasStaiger,andseminarparticipantsattheNationalBureauofEconomicResearch,theUniversityofRochester,theUniversityofToronto,theUniversityofWisconsin,andthemeetingoftheAssociationofPublicPolicyandManagement.ThisresearchwasablyassistedbyBridgetTerryandIlyana Kuziemko. r 2000bythePresidentandFellowsofHarvardCollegeandtheMassachusettsInstituteof Technology. TheQuarterlyJournalofEconomics,November2000 1239 Section 9-69 smallerclassesbecausetheyreducetheeffortthattheymust expendinordertodeliverinstruction.Teachers’unionsmaylike classsizereductionsbecausetheyincreasethedemandfor teachers.Administratorsmaylikeclasssizereductionsbecause theyincreasethesizeoftheirdomain.Asaresultofthepolicy’s popularity,thetwentiethcenturyhasbeenaperiodofcontinuous declineinclasssize,tothepointwhereAmericanelementary schoolshad,onaverage,18.6studentsperteacherinthe1997– 1998schoolyear.1 Nevertheless,therearebotheconomicandempiricalprob- lemswithclasssizereductionpolicies.Ontheeconomicfront, classsizeisaprimaryexampleoftheeducationproduction functionfallacy.Itisconventionaltoestimatetherelationship betweeneducationalinputs(likeclasssize)andoutputs(achieve- ment)andtocalltherelationshipan‘‘educationproduction function.’’Thisnomenclaturesuggeststhatinputstranslatesys- temicallyintoachievement,astheydointheproductionfunctions ofprot-maximizingrms.Theanalogyisafalseone,however, becauserms’productionfunctionsarenotjustaresultoftheir abilitytoturninputsintooutputs.Arm’sproductionfunctionis theresultofmaximizinganobjective(prots),givenaproduction possibilitiesset.Itisnotobviousthatschoolshavestringent achievementmaximizationobjectivesimposedonthem.Asde- scribedabove,classsizereductionscanfulllavarietyofobjec- tives,notallofwhicharerelatedtoachievement.Thus,while classsizereductionalwaysaffords opportunities forincreased investmentineachchild’slearning,itisnotobviousthatevery schooltakesupsuchopportunities.Theactualeffectofreducing classsizewilldependontheincentivesaschoolfaces.Putanother way,ifapolicy-makerwantstopredicttheeffectthataproposed classsizereductionwouldhave,sheshouldrelyonevidencefrom schoolsthatfaceincentivesthataresimilartotheincentivesthat schoolswouldfaceundertheproposedpolicy. Ontheempiricalfront,classsizeisdifficulttostudy.2 The vastmajorityofvariationinclasssizeistheresultof choices made byparents,schoolingproviders,orcourtsandlegislatures.Thus, 1.SeeNationalCenterforEducationStatistics[1999].Therearedifferencesbetweenthestudent-teacherratioandclasssize,butthedifferencesarelessofaconcernforelementaryschoolsthanforsecondaryschools.Inanycase,the differencesarenotrelevanttotheempiricalworkinthispaper,becauseIuseclasssizeasreportedbyschools.2.SurveysoftheevidenceonclasssizeincludeHanushek[1996,1986],CardandKrueger[1996],andBetts[1995]. QUARTERLYJOURNALOFECONOMICS1240 Section 9-70 mostoftheobservedvariationinclasssizeiscorrelatedwithother determinantsofstudentachievementandislikelytoproduce biasedresults.Thismayappeartobeanobviouspoint,butthough researchersoftenclaimthatthevariationtheyuseisnotendoge- noustostudentachievement,theyrarelygoontoexplainwhere thevariation does comefrom.Theprocessesbywhichschool inputsaredeterminedshouldmakeusdoubtthatvariationin schoolinputsisexogenousunlessthereissomeexplicitreason whyweshouldthinkitis. Thiscriticismdoesnotapplytoexplicitexperimentsthat randomlyassignsomestudentstosmallclassesandotherstu- dentstolargeones.ProjectStarisanexperimentofthistype,and evidencebasedonithasmanifestadvantages.3 Theseadvantages, however,areoffsetbyafewdisadvantages.Explicitexperiments arerare(temptinginterpreterstoextrapolatetheresultsunduly), manyexperimentstakeplaceindevelopingcountries(sothatthe rangeofinputsisnotrelevantfortheUnitedStates),and—most importantly—theactorsinanexperimentareawareofit.For instance,theschoolsinaclasssizeexperimentmayrealizethatif theexperimentfailstoshowthatthepolicyiseffective,thepolicy willneverbebroadlyenacted.Insuchcasestheschoolshave incentivesthatthefullyenactedpolicywouldnotgive.Thatis,the experimentalterstheincentiveconditions,sothattheproduction functionbeingestimatedisnottheproductionfunctionthatwould beineffectifthepolicywerefullyenacted.Inaddition,some individualstemporarilyincreasetheirproductivitywhentheyare beingevaluated.Thisphenomenon,knownasthe‘‘Hawthorne effect,’’canmakepoliciesappeartohaveproductivityeffectsthat theywouldnothaveiffullyenacted.Finally,individualssome- timestrytoundotherandomnessoftheexperiment.Forinstance, someadministratorsmaytrytollthesmallclasseswithchildren whoaremostinneedofindividualattention(generatingresults thatarebiasedagainstndingthatclasssizereductionworks). Otheradministratorsmayassigntheirbestteacherstothesmall classesormonitorthesmallclassesmore(generatingresultsthat arebiasedtowardndingthatclasssizereductionworks). InthisstudyIattempttoaddresstheempiricalandeconomic problemswithtwoidenticationstrategies,bothofwhichuse variationinclasssizethatcomesfrompopulationvariation.The 3.ProjectStarwasanexplicitexperimentinclasssizereductioninTennesseeelementaryschools.SeeKrueger[1999]foradescriptionoftheProjectStarresults. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1241 Section 9-71 rststrategyusesnaturalrandomnessinthepopulation,andits logicisstraightforward.Consideraschoolattendanceareathat hasapopulationthatisinsteadystate.Thereisstillnatural randomnessinthetimingofbirthssuchthattheentering kindergartencohortvariessomewhatinsize.Thisvariationisnot fullysmoothedbecausethereisdiscretenessinschoolentryrules (forinstance,childrenbornbetweenJanuary1andDecember31 inyear t mustenrollinrstgradeinyear t 1 6)andbecausethe numberofclassroomsineachschoolisaninteger.Ifonethinksof aschoolwithoneclassroompergrade,thennaturalrandomness inthepopulationtranslatesdirectlyintodifferencesinclasssize betweencohorts.Forinstance,supposethataschoolattendance areahasanunusuallysmallnumberofve-yearoldswith birthdaysinNovemberandDecember1985buthasthe‘‘decit’’ madeupbyanunusuallylargenumberofve-yearoldswith birthdaysinJanuaryandFebruary1986.Thesesmalltiming differenceswouldtypicallymakeforanunusuallysmallkindergar- tencohort(say,15students)inthe1990–1991schoolyearandan unusuallylargekindergartencohort(say,25students)inthe 1991–1992schoolyear.Therstcohortmightpersistentlyexperi- encesmallclassesingradeskindergartenthrough6,whilethe subsequentcohortmightpersistentlyexperiencelargeclasses. Essentially,thetwocohortsarerandomlyassigneddifferentclass sizes. Iimplementtherstidenticationstrategybyisolatingthe randomcomponentofpopulationvariationusinglongpanelsof dataonenrollmentandkindergartencohortsinConnecticut schooldistricts.Thelongpanelsallowmetoeliminatenearlyall smoothchangesinpopulation.Iuseresidualsthatremainafter ttingaquarticfunctionoftime separately foreachgradeineach school. InthesecondstrategyIusethefactthatclasssizejumps abruptlywhenaclasshastobeaddedtoorsubtractedfroma gradebecauseenrollmenthastriggeredamaximumorminimum classsizerule.Returningtothepreviousexample,supposethat the1992–1993kindergartencohortwere26studentsandthe district’smaximumkindergartenclasssizewere25.Thenthere wouldbetwokindergartenclassesof13studentseachin1992– 1993.Althoughthedifferenceincohortsizebetween1991–1992 and1992–1993wouldbeonlyonestudent,thedifferenceinclass sizewouldbetwelvestudents.Thelogicoftheidentication strategyisthatthereisadiscontinuousrelationshipbetween QUARTERLYJOURNALOFECONOMICS1242 Section 9-72 classsizeandenrollmentatcertainknownlevelsofenrollment whilethereisasmoothrelationshipbetweenachievementandthe determinantsofenrollment.Iusethepanelsofdatatoobserve smallchangesinenrollmentassociatedwithchangesinthe numberofclassesineachgradeineachschool.Iuseinformation oneachdistrict’sclasssizerulestodeterminewhetherchangein thenumberofclasseswaspurelytheresultofthesmallchangein enrollmenttriggeringarule.Iimplementthesecondidentica- tionstrategybycomparingtheclasssizeandachievementof adjacentcohortswhoimmediatelyprecedeandsucceedeachsuch event. Thetwoidenticationstrategiesareindependentofone anotherandprovideacheckononeanother’sresults.Iprovidea numberofotherspecicationtestsaswell. Oneniceconsequenceofusingpopulationvariationisthat therangeofclasssizeforwhichIobtainestimatesistherange thatisrelevantforpolicy.AnotherniceconsequenceisthatI observeschoolsfunctioningundertheincentiveconditionsthat theynormallyexperience.Theonedisadvantageofnaturalpopu- lationvariationisthatateachermayadjustherteachingmethods moreovertwoorthreeyearsofsmallclasssizethanshedoesover oneyearofsmallclasssize(evenifsheperiodicallyexperiences smallclasses).InProjectStar,mostoftheeffectofsmallclasssize occurredafteroneyear,withoutteachers’beingtrainedtoalter theirteachingmethods.Thesefactssuggestthatteacherscan adjustquicklyandwithoutspecialtraining,iftheyhavean incentivetodoso.Inshort,thesefactssuggestthatthetransitori- nessofsmallclasssizeduetopopulationvariationshouldnotbea problem,butIdiscusstheissuecarefullyininterpretingmy results.4 II.SOURCESOF VARIATIONIN SCHOOL INPUTSANDTHE POTENTIALFOR BIAS Parents’choosingschoolsbychoosingtheirresidencesis probablythesinglelargestsourceofvariationinschoolinputs. Between-districtvariationinschoolinputsgeneratedbyparents’ choicesislikelytogenerateupwardbiasedestimatesofthe 4.Populationvariationmakesteachersexperiencesmallandlargeclassesrepeatedly,butnotpredictably.Notethatclasssizeis not transitoryfromthepointofviewofacohortofstudents.Acohortinaschooltendstoexperienceeithersmallorlargeclassesconsistently. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1243 Section 9-73 efficacyofinputs.Thesamemaybesaidforsystematicvariation withinadistrictovertime.Forinstance,classsizereductionswill appeartobemoreefficaciousthantheyreallyareifparentswho contributemoretotheirchildren’slearningalsochooseschool districtsthatoffersmallerclasssizes.Whenwemakesimple comparisonsofschoolsincross-sectiondataortime-seriesdata, thereislikelytobebiasinfavorofclasssizereductions. Ifweidentifyparentswhohavesimilarattributes,thereis amplebutsomewhatdifferentpotentialforbias.Parentschoose schoolinputsendogenously,basedontheirchild’sabilityandprior achievementinschool.Theseendogenouschoicesmaybecompen- satory(greaterinputsforchildrenwhoexhibitpoorachievement), reinforcing(greaterinputsforgiftedchildren),orboth.Thus, whenwecomparestudentswithsimilarfamilies(using,say, cross-sectiondatawithextensivecontrolsforfamilybackground), thesignofthebiasisambiguous. Similarly,wecannotpredictthesignofthebiasgeneratedby thechoicesofschoolingproviders,suchasadministratorsand teachers.Ifprovidersattendmoretothedemandsofparentswho contributemoretotheirchildren’slearning,inputsandparental contributionswillbepositivelycorrelated,generatingupward biasedestimatesoftheefficacyofinputs.Ontheotherhand,if providersattendmoretochildrenwithlearningproblems,esti- mateswillbebiaseddownward. Thenalplayerswhodetermineschoolinputsarestateand federaljudgesandlegislators,whomandateandfundincreased schoolinputsforcertainstudents.Policy-makerspursueboth compensatoryandreinforcingpolicies,buttheytendtodevotethe majorityoftheresourcesattheirdisposaltocompensatory policies.5 Thenegativebiasresultingfromtheuseofcompensa- torypolicies,however,isoftenoffsetbypositiveomittedvariables biascausedbypolicy-makers’simultaneouspursuitofcomplemen- tarypolicies.Forexample,policiesthatdecreasedracialdiscrimi- nationinschoolinputswereimplementedsimultaneouslywith policiesthatdecreasedracialdiscriminationinemployment.Both typesofpoliciescouldleadminoritystudentstohavehigher achievement. 5.SeeSalmonetal.[1995]forevidenceontheprevalenceofcompensatory policiesinstateschoolnance.Morethan80percentoffederalmoneyforelementaryandsecondaryeducationisdevotedtocompensatorypolicies:TitleI,bilingualeducation,specialeducation,andthefreeandreduced-pricelunchprogram. QUARTERLYJOURNALOFECONOMICS1244 Section 9-74 Inshort,itisnotsurprisingthatempiricalresultsdiffer(that is,sufferfromdifferentbiases)dependingonthesourceof variationinschoolinputsthattheyuse. Thereisadifferencebetweenvariationthatisnotobviously biasedandvariationthathasanexplicitreasontoberandom.The systematiclinksbetweenschoolinputsandotherdeterminantsof studentoutcomesmaybe obscure withoutthevariationininputs being exogenous.Explicitlyarticulatingasourceofexogenous variationispreferabletosimplyeliminatingallapparentsources ofbias.ThisiswhatIattempttodointhispaper. III.EMPIRICAL STRATEGY Considertheachievementofstudentsingrade i ofschool j in district k incohort t.Itisdeterminedbyclasssizeaswellas unobservedattributeslikestudentabilityandparentalcontribu- tionstolearning.Ageneral‘‘educationproductionfunction’’that subsumesmostcommonspecicationsis (1)Aijkt 5b 1 log(Cijkt )1 It b 2 1 Ij b 3 1 Xijkt b 4 1e ijkt , where Aijkt isachievement;log(Cijkt )isthenaturallogofclasssize; It isavectorofcohortindicatorvariables;Ij isavectorofschool indicatorvariables;Xijkt isavectorofobservedstudent,parent, andcommunitycharacteristics;and e ijkt isallotherdeterminants ofachievement,includingtheunobservedattributesofthestu- dents,parents,andcommunity.Fewstudiesactuallyincludeallof thetermsinequation(1),butmoststudiesincludesomesubsetof them. Ifthemeasureofachievementisatestscore,itisoften dividedbythestandarddeviationofstudents’scoresonthetest. Thiscommonpractice(whichIfollowinthispaper)facilitates understandingofunfamiliartestscoresandallowscomparisons tobemadeacrossstudiesthatusedifferenttests.Itisnow commontousethenaturallogofclasssizetotakeaccountofthe factthataone-studentreductionisproportionatelylargerfroma baseof17students,say,thanfromabaseof35students.The vectorofcohortindicatorvariablesisincludedtoallowfortests thatchangeslightlyfromyeartoyearandtoallowforteachers whoadjusttheirteachingtoatest’scontent.Forinstance,ifatest wereslightlyeasierforallfourthgradersin1996thanitwasfor allfourthgradersin1995,theslighteasingwouldbepickedupby thecohortindicators.Thevectorofschoolindicatorvariablesis THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1245 Section 9-75 includedtocontrolforanyattributesofschools’populationsthat areconstantacrosstime—especiallyrelativelyunchangingat- tributesofthecommunityinwhichtheschoolislocated.The vector X typicallyincludesvariablesthatdescribetheracial compositionandfreeluncheligibilityofstudents.Italsoincludes variablesthatdescribetheracialcomposition,educationalattain- ment,andincomeoflocalhouseholds(althoughschoolxed effects,ifincluded,absorbanysuchvariablesthatareconstant acrosscohortsofstudents). 1.TheFirstIdenticationMethod Bydenition,thelogofaverageclasssizeisequaltolog(E )2 log(n),where E isregularenrollment,and n isthenumberof classes.Fornow,letusfocusonenrollmentandsupposethatthe numberofclassesisxedforagivengradeinagivenschool.I returntothechangesinthenumberofclassesbelow. Enrollmentisafunctionofstudent,parent,andcommunity characteristics(observedandunobserved).Inaddition,thereis randomvariationinthepopulationofchildrenwhoareintheage rangeappropriateforagivengradeinagivenyear.Thatis,actual enrollmenthasadeterministiccomponent,Eï (X,e ),whichiswhat enrollmentwouldbeifthetimingandnumberofbirthswerea deterministicfunctionofthepopulation’sobservedandunob- servedcharacteristics;andithasarandomcomponent u,whichis thevariationinenrollmentthatresultsfromthefactthatbiology causesrandomvariationinthetimingandnumberofbirths.One expectsthat u affects E proportionally,soonecanwrite6 (2)Eijkt 5 Eïijkt (Xijkt ,e ijkt )·uijkt , orlog(Eijkt )5 log(Eïijkt (Xijkt ,e ijkt ))1 log(uijkt ). Log(u)isnotcorrelatedwith X and e ,whicharedeterminants ofachievement,butlog(u)isadeterminantoflog(E ),soa consistentestimateoflog(u)isagoodinstrumentforclasssize. Thatis,ifanestimateoflog(u)isconsistent,thenitfulllsthe twobasicinstrumentalvariablesconditions:itiscorrelatedwith log(E )anduncorrelatedwith e .Iattempttogetaconsistent estimateoflog(u)usingthefactthatthedeterministicpartof enrollmentchangesmuchmoresmoothlythanactualenrollment 6.Itisnaturaltosupposethattheshare,nottheabsolutenumber,of‘‘deviantly’’timedbirthsisconstantacrosspopulationsofdifferentsizes. QUARTERLYJOURNALOFECONOMICS1246 Section 9-76 doesforaparticulargradeinaparticularschoolinaparticular year.Consideraschoolattendanceareathathas,forvarious reasons,apositivetrendinthenumberofhouseholdswith school-agedchildren.(Thetrendcouldbenonlinear.)Thedetermin- isticpartofenrollmentineachgradewouldbearelativelysmooth functionofthetrend.Butactualenrollmentineachgradewould deviatefromthisrelativelysmoothfunctionandmightnoteven beamonotonictransformationofthetrendinthenumberof householdswithschool-agedchildren.Recalltheexampleofthe schoolattendanceareathathadanunusuallysmallnumberof childrenwithbirthdaysinNovemberandDecember1985andan unusuallylargenumberwithbirthdaysinJanuaryandFebruary 1986.Thesetimingdifferenceswouldtypicallygenerateasmall kindergartencohortin1990–1991andalargeonein1991–1992, andthelatercohortcouldexperiencelargerclasssizes evenif the schoolattendanceareahadanegativetrendinthenumberof householdswithschool-agedchildren.Moreover,ifonewereto de-trendenrollment,onewouldndthatthe1990–1991cohort hadanegativeresidualandthe1991–1992cohorthadapositive residual. Anylog(Eï )thatchangessmoothlyovertimecanbeapproxi- matedbyagrade-school-specicinterceptandagrade-school- specicpolynomialintime.Thatis,wecanwrite (3)log(Eïijkt )5a 0ijk 1a 1ijk t 1a 2ijk t 2 1a 3ijk t 3 1a 4ijk t 4 1 ···, orlog(Eijkt ) 5a 0ijk 1a 1ijk t 1a 2ijk t 2 1a 3ijk t 3 1a 4ijk t 4 1 ···1 log(uijkt ). Iestimatesuchanequationseparatelyfor each gradein each school.Itypicallyhave24yearsofenrollmentdataforeach regression.Ishowresultsthatuseuptoaquarticintimebecause quarticsappeartocaptureallofthesmoothvariationovertimein enrollmentwithinagradewithinadistrict.7 Theestimated residualshouldbeaconsistentestimateoflog(u),whichisthe instrumentweneedforclasssize. Inshort,therstidenticationstrategyhasthreeintuitive steps:one,obtainestimatesoftherandompartofenrollment variation;two,usetherandomvariationinenrollmenttoidentify randomvariationinclasssize;three,seehowachievementis 7.Infact,theestimatedresidualshardlychangeinthemovefromacubictoaquartic,quintic,orsixth-orderpolynomial. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1247 Section 9-77 affectedbyrandomvariationinclasssize.Formally,therst identicationstrategyrequiresthefollowingprocedure.First, estimateequation(3)separatelyforeachgradeineachschooland obtaintheestimatedresiduals.Stacktheestimatedresidualsto getavectoroftheestimatedresidualsforeachschool:log(uˆijkt ). Second,estimatethefollowingrst-stageequationforeachgrade: (4)log(Cijkt )5d 1 log(uˆijkt )1 It d 2 1 Ij d 3 1 Xijkt d 4 1n ijkt , andobtainpredictedlog(Cijkt ).Third,estimateequation(1)by TwoStageLeastSquares(2SLS),usingpredictedlog(Cijkt ). Calculatecorrectstandarderrorsforthe2SLSprocedure.8 Notice thattheprocedureuses within-school comparisonsofenrollment, classsize,andachievement.Aschoolxedeffectistakenoutof enrollmenttoformlog(uˆijkt );aschoolxedeffectisestimatedin therst-stageequation;andaschoolxedeffectisestimatedin thesecond-stageequation.Onecohortinaschoolisbeing comparedwithothersinthesameschool,wherethedifference betweenthecohortsisthatoneislargerthantheothersdueto (whatappeartobe)purelyrandomcircumstances. Themethodjustdescribedexploitsthefactthataggregate characteristicsthataffectachievement,X and e ,changemuch morecontinuouslythanenrollmentinaspecicgrade-school-time does.Yet,becauseparentscanresponddirectlytotheclasssize theyobservetheirchildexperiencing,themethodleavesopena smallrouteforbias.Consideraparentwhoobservesthathis child’sclassisunusuallylarge.Evenifthecauseofthelargeclass israndompopulationvariation,theparentmightdecidetohave hischildtransferredtoanotherschoolinthesamedistrict,might movetoanotherdistrict,mightsendhischildtoaprivateschool, ormightattempttohavehischildheldbackagradeoradvanceda grade.Suchreactions,althoughprobablyrare,wouldhavethe potentialtomake X and e endogenousto u.Aparentwhowould reactthiswaywouldhavetobeunusuallyconcernedabout education,abletopayforamove,abletopayforprivateschooling, orabletoconvinceschooladministratorstoallowatransfer.One expectsthatsuchaparentwould,inanycase,makeanunusually largecontributiontohischild’seducation,sothattheendogeneity underconsiderationwouldprobablymakeusoverestimatethe efficacyofclasssizereductions.Inotherwords,classesthatwere 8.Thatis,calculatethestandarderrorsusingtheactualdataonclasssize,notthepredicteddata. QUARTERLYJOURNALOFECONOMICS1248 Section 9-78 randomlylargewouldendupwithadisproportionatelysmall shareofeducation-concernedparents.Fortunately,onecando betterthanspeculateaboutthesizeandsignofthisbias:asimple modicationoftheestimationmethodeliminatestheproblem. Ratherthancarryouttheinstrumentalvariablesprocedure attheschoollevel,onecanaggregateequations(1)and(3)tothe districtlevelandcarryouttheprocedureatthehigherlevelof aggregation.Atthedistrictlevel,transfersamongschoolswithin thedistrictwillcancelout,soresidualsfromthedistrict-level versionofequation(3)— (5)log(Eikt )5a ˜0ik 1a ˜1ik t 1a ˜2ik t 2 1a ˜3ik t 3 1a ˜4ik t 4 1 ···1 log(uikt ) —giveusacrediblyconsistentestimatorforlog(u)thathasno potentialtobecorrelatedwith X or e throughparents’reactingto largeclasssizebytransferringachildtoanotherschoolwithinthe district.Eikt isenrollmentingrade i indistrict k forcohort t, summedoveralloftheschoolsinthedistrict.Carryingoutthe procedureatthedistrictleveleliminatesbiascausedbytransfers; italsohasadvantagesbecausemoreyearsofachievementdata areavailableatthedistrictlevel.Ontheotherhand,carryingout theprocedureatthedistrictlevelreducestheexplanatorypower oftheprocedure.Inparticular,theexplanatorypowercontributed bylargeschooldistrictsisreducedbecauserandompopulation variationaveragesouttoagreatextentwithineachcohortovera largedistrict.(Elementaryschoolsinlargedistricts,however,are smallenoughthatlargedistrictsdocontributesignicantlyin school-levelestimation.) Thedistrict-levelproceduredoesnotentirelyeliminatethe potentialforbiascausedbyparents’reactingtoclasssize.Parents couldstillshifttheirchildtoaprivateschool,havetheirchildheld backoradvancedagrade,ormoveoutofthedistrictoncethey observedthattheirchild’sclasswasgoingtobeunusuallylarge. Fortunately,onecanremovethispotentialforbiasbyusingdata onthenumberofchildrenineachdistrictwhowereageveatthe schoolentrydate.Inotherwords,onecanobservethepotential kindergartencohortatthedistrictlevel(‘‘K5’’)anduseitasthe sourceofrandomvariationinclasssize.Onesimplyestimatesa versionofequation(3)withthepotentialkindergartencohortas THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1249 Section 9-79 thedependentvariable: (6)log(K5ikt ) 5a ï 0ik 1a ï 1ik t 1a ï 2ik t 2 1a ï 3ik t 3 1a ï 4 ik t 4 1 ...1 log(uikt ). Equation(6)givesusacrediblyconsistentestimatorforlog(u) thathasnopotentialtobecorrelatedwith X or e throughparents reactingtoidiosyncraticallylargeclasssizebymovingtoanother district,sendingachildtoprivateschool,orshiftingachildtoa differentgrade.Inadditiontothedisadvantagesdiscussedabove fordistrict-levelestimation,thedisadvantageofusingkindergar- tencohortresidualsisthattheywillbestrongerinstrumentsfor classsizeinearlyelementarygradesthaninlaterelementary gradesbecauseexogenousstudentmobilityweakensthecorrela- tionbetweenkindergartencohortsizeandlatergrades’cohort sizes. Thusfar,Ihavenotdiscussedchangesinthenumberof classes n inagradeinaschool.Mysecondidenticationmethod exploitsthesechanges,buttheyaresimplyanuisanceformyrst identicationmethod.Thecostsandbenetsofaddinganother classdependnotonlyonhowmuchlocalparentscareabout schoolingbutalsoonactualenrollmentinanygivenyear(evenif theriseorshortfallinactualenrollmentcomesfromrandom variation).Thus,ifonecarriesouttheprocedurefortherst identicationmethodandignoreschangesinthenumberof classes,themonotonicityconditionforinstrumentalvariableswill occasionallybeviolated:anincreaseinenrollmentwill reduce classsizeifittriggersanincreaseinthenumberofclasses.9 There isasimplewaytoadjusttherstidenticationmethodsothatthe monotonicityconditionisneverviolated.Theproceduredescribed aboveisvalidsolongasthevariationinenrollmentdoesnot triggerachangeinthenumberofclasses.Therefore,Iuse variationinenrollmentthatisnotjustwithin-schoolbutiswithin anexpectednumberofclasses.Inotherwords,insteadofhaving schoolindicatorvariablesintherst-andsecond-stageequations, thereisanindicatorvariableforeachcombinationofaschooland expectednumberofclasses.Thatis,thereisavectorofindicator variablesforcombinationslikethefollowing:theschoolis j and itssecondgradeisexpectedtohavetwoclasses,theschoolis j and itssecondgradeisexpectedtohavethreeclasses,andsoon.The 9.SeeAngrist,Imbens,andRubin[1996]foradiscussionofthemonotonicityconditionforinstrumentalvariables. QUARTERLYJOURNALOFECONOMICS1250 Section 9-80 logicisstraightforward.Ifenrollmentinaschool’ssecondgradeis randomlyhigherthisyearthanitwaslastyearbutissuchthat therearetwosecondgradeclassesinbothyears,thenthe regressioncomparesthedifferenceinachievementbetweenthe twoyearswiththedifferenceinclasssize.Ifenrollmentina school’ssecondgradeisrandomlyhigherthisyearandittriggers amaximumclasssizerulesothatthisyearthereare three second gradeclasses,thentheregressiondoesnotcomparethetwoyears. Noticethatthe expected numberofclassesiswhatmatters.Iuse districts’maximumandminimumclasssizerulestodetermine whenanenrollmentchangewouldbeexpectedtotriggerachange inthenumberofclasses,sinceitisatthesetimesthatthe monotonicityconditionwouldbeviolated.(Ifaschoolchangesthe numberofclassesforreasonsunrelatedtoenrollmentbutrelated to,say,changesinparents’preferences,themonotonicitycondi- tionisnotviolated.)Calculationoftheexpectednumberofclasses isdiscussedinthenextsubsection. Summingup,therstidenticationstrategyproceedsas follows.First,estimateequation(3)separatelyforeachgradein eachschool,andobtaintheestimatedresiduals,log(uˆijkt ).Second, estimatethefollowingrst-stageequation,inwhichthereisa xedeffectforeachschool-expectednumberofclassescombina- tion: (7)log(Cijkt )5d 1 log(uˆijkt )1 Itd 2 1 Ij,njd 3 1 Xijkt d 4 1n ijkt . Ij,nj isvectorofindicatorvariablesforcombinationsofschoolsand expectednumberofclasses.Third,estimatethefollowingachieve- mentequation,inwhichthereisaxedeffectforeachschool- expectednumberofclassescombination: (8)Aijkt 5b 1 log(Cijkt )1 It b 2 1 Ij,njb 3 1 Xijkt b 4 1e ijkt . Calculatecorrectstandarderrorsforthe2SLSprocedure.Repeat theprocedurewithdistrict-levelenrollmentandwithdistrict- levelkindergartencohorts. 2.TheSecondIdenticationMethod Thesecondidenticationmethoddoesnottreatchangesin thenumberofclassesasanuisance;itexploitsthem.Itmakesuse ofthefactthatchangesinthenumberofclassesinagradecan produceabruptchangesinclasssize.Thesimplestwaytouse thesediscontinuitiesisthecross-sectionmethodofexploiting maximumclasssizethresholds.AngristandLavy[1999]illus- THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1251 Section 9-81 tratethismethodusingIsraelischools.(Israelischoolshavea maximumclasssizeof40;mostAmericandistrictshavemaxi- mumclasssizesintherangeof20to30students.)Forinstance,if aschoolhasamaximumclasssizethresholdof25,itputs studentsintooneclassuntilenrollmentis25,putsstudentsinto twoclassesuntilenrollmentis50,andsoon.Itsrulecanbe writtenas (9)Cijkt 5 Eijkt int[(Eijkt 2 1)/C max]1 1 , where C max is25andint(z)isthegreatestintegerlessthanor equalto z.Classsizevariesabruptlyandpredictablywhen enrollmentisatamultipleof25.Thesediscontinuitiesprovide identicationbecausethedifferenceintheunderlyingpopulation thatproducesenrollmentof25versus26isverysmall(andshould haveacorrespondinglysmalleffectonachievement),butthe differenceinclasssizeforenrollmentof25versus26islarge(and shouldhaveasignicanteffectonachievementifreductionsin classsizeareefficacious).Thus,thechangeinthepredictedclass sizebetweenenrollmentof25andenrollmentof26basedsolelyon therulegivenbyequation(9)isagoodinstrumentfortheactual differenceinclasssizesbetweenschoolswithenrollmentof25and 26.Thesameistruefor50and51,75and76,andsoon. Therearethreeessentialthingstounderstandaboutthis methodofidentication.First,theidenticationisindependentof theidenticationthatcomesfromusinglog(u)asaninstrument forclasssize,sothetwomethodscanbeusedaschecksonone another. Second,betweenthediscontinuities,predictedclasssize varieswithactualenrollment,whichis,ofcourse,afunctionof X and e .Therefore,predictedclasssizeis not avalidinstrument except whentheruletriggersachangeinthenumberofclasses. Putanotherway,theestimateswillbeconsistentonlyifidentica- tionrelies solely onthediscontinuitiesinequation(9).All variationinpredictedclasssizethatisnotgeneratedbya rule-triggeredchangeinthenumberofclassesissuspectand mustbediscardedifbiasistobeeliminated.Incross-sectiondata onedoesnotobserveactualchangesinthenumberofclasses,so theonlynonsuspectvariationisthevariationatmultiplesof maximumclasssize—thedifferenceinachievementforenroll- mentof25versus26,forenrollmentof50versus51,etcetera.In crosssectiondataothervariationinenrollmentissuspectbecause QUARTERLYJOURNALOFECONOMICS1252 Section 9-82 itislikelytobebetween-districtvariationthatreectsdifferences intheunderlyingpopulations(X and e )andcouldevenbe endogenoustorealizationsofclasssize.Someschools routinely havelargerclasssizesthanothersbecauseofthewaytherules function,andparentscanendogenouslychooseschoolstaking realizedclasssizeintoaccount.Discardingallsuspectobserva- tions,however,placesgreatdemandsoncross-sectiondata,since theresultswilldependontherebeingsufficientoccurrencesof enrollmentatmultiplesofmaximumclasssize.AngristandLavy [1999],forinstance,areabletodoonlysomeofthedesirable discardingbecausetheircross-sectiondatacontaintoofewoccur- rencesofenrollmentintherightranges.Below,Ipresentcross- sectionresultsthatdemonstratewhathappensasonediscards moreandmoreofthesuspectobservations.Sincemydataare actuallypaneldata,Iamabletoemployawithin-districtmethod (describedbelow)thatismorepowerfulandlesssubjecttobias thanthecross-sectionmethod. Third,identicationarisesonlywhentherulebinds,soifone usesarulethatbindsonlyinsomeschools,onelearnsaboutthe effectsofclasssizeonlyforthoseschools.Forinstance,inAngrist andLavy’s[1999]data,themaximumclasssizeruledoesnotbind indistrictsthatservewell-offhouseholds.Itisusefultoestimate theeffectofclasssizeonlyforless-well-offstudents,butonemust becarefultointerprettheresultsappropriately.Ifbetter-off districtsactuallyhavemaximumclasssizerulesoftheirownthat theyfollow,thenusingastatewiderulethatdoesnotbind everywhereisthrowingawayusefulvariation.Sincethereis typicallynotmuchusefulvariationanywayfordiscontinuity- basedidenticationstrategies,itisimportanttouseallthat exists. Giventheseissuesaboutidenticationbasedondiscontinui- ties,Iusechangesinthenumberofclassesthataregeneratedby small within-school changesinenrollmentthattriggera district’s maximumorminimumclasssizerule.Thismethodismore accurateandlesspronetobiasthanthecross-sectionmethod becauseonecanfollowenrollmentinagradeinaschoolovertime andactuallyseeeveryoccasiononwhichtherulesaretriggered bysmallchangesinenrollment.Themethodalsoproducesmore preciseestimatesbecauseitcomparesadjacentcohortswithina school,whoarelikelytobesimilar except fortheirdifferentclass sizeexperiences.Finally,thismethodhastheadvantagethatit usesvariationfromallsortsofdistricts.Districtschooserules THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1253 Section 9-83 thatarerelevantforthem,andaslongastherulesarestable,they generateusefuldiscontinuitiesinclasssize.Thewithin-school regressiondiscontinuitymethodrequiresdistrict-by-districtinfor- mationonclasssizerules,whichisoneroustocollect.Iobtained informationoneachdistrict’srulesbysurveyingsuperintendents (seebelow).Notethat,aslongaseachdistrict’sruleisstableover theperiodinquestion,therulescouldbeendogenoustothe underlyingcharacteristicsofthedistrictsandthesecondidenti- cationmethodwouldstillproduceconsistentresults.Thisis becausethesecondidenticationmethodrelieson within-district variationinclasssize.10 Thesecondidenticationmethod(‘‘within-schoolregression discontinuity’’)hasaverysimpleprocedure.First,Iidentifyallof theeventsinwhichaschoolincreasedordecreasedthenumberof classesinoneofitsgrades.Second,withinthisgroupIidentifyall theeventsinwhichthechangeinthenumberofclasseswas predictable,givenjustthechangeinenrollmentandthedistrict’s maximumandminimumclasssizerules.Ikeepthissubset(which is,inpractice,78percentofalleventswherethenumberofclasses changes).Thatis,theexpectednumberofclassesisgivenby (10)E (nijkt )5 nijk,t2 1 1 Iijktadd 2 Iijktsubtract, where Iijktadd 5 1if Eijkt nijk,t2 1 .C max;0otherwise, and Iijktsubtract 5 1if Eijkt nijk,t2 1 ,C min;0otherwise. Ikeepthesubsetofeventswherethenumberofclassesactually increasedand Iijktadd 5 1andwherethenumberofclassesactually 10.Thereisacaveattothisstatement.Sincedistrictscansetlowerorhighermaximumclasssizes,districtswillgenerateusefulclasssizevariationinslightlydifferentranges.Forinstance,onedistrict’susefulvariationinclasssizemaytend tobeintherangefrom16studentsto25students,whileanother’smaytendtobeinrangefrom18to27students.Ifoneweretondthatareductioninclasssizewas,say,moreefficaciouswhenitoccurredabovesomeclasssize,thenonewould beunsurewhetherthegreaterefficacywasduetodecreasingreturnstoreductionsinclasssizeorgreaterefficacyinthesortofschoolsthattypicallychoosehighermaximumclasssize.Onecouldthentrytosortouttheexplanationsbyexaminingthecharacteristicsofdistrictswithlowerandhighermaximumclasssizes.This problemdoesnotarise,inpractice,inthisstudy. QUARTERLYJOURNALOFECONOMICS1254 Section 9-84 decreasedand Iijkt subtract 5 1.Third,withinthesubsetIkeepthe eventsinwhichthechangeinenrollmentthattriggeredthe changeinclasssizewassmallerthan20percent.Forinstance,if enrollmentrosefrom40to48,andittriggeredachangeinthe numberofclasses,Ikeeptheevent.However,ifenrollmentrose from40to54,Idiscardtheevent.Thereasonisthatregression discontinuitymethodsdependona modest changeinacontinuous variable,likeenrollment,triggeringa large changeinadiscrete variable,likethenumberofclasses.Ifsomechangeinthe underlyingcircumstancesofaschoolweretomakebothenroll- mentandthenumberofclassesjumpbyalargeamount,theevent wouldbeinappropriateforregressiondiscontinuitymethods.In practice,Ikeep94percentofthesubsetatthisstage. Havingidentiedasetofeventswherethenumberofclasses changesbecauseamodestchangeinenrollmenttriggersamaxi- mumorminimumclasssizerule,Iestimatearst-differenced versionoftheachievementequation— (11)Aijkt 2 Aijk,t2 1 5b 1 log(Cijkt )2 log(Cijk,t2 1) 1 Itb 2 1 Xijkt 2 Xijk,t2 1 b 4 1 e ijkt 2e ijk,t2 1 —usingjustthecohortsimmediatelybeforeandaftereachevent. Intuitively,ifschool j ’sthirdgradeenrollmentismodestlyhigher thisyearthanitwaslastyear,andtheenrollmentincrease triggersamaximumclassrulesothatthenumberofclassesrises andclasssizefalls,thenIcomparetheachievementofthisyear’s thirdgradecohortwiththatoflastyear’sthirdgradecohort.11 11.Onemayworryabout‘‘nonevents’’—occasionsonwhichasmallchangeinenrollmentshouldhavetriggeredachangeinthenumberofclassesbutdidnot.Itturnsoutthatonly9percentofwould-betriggereventsareactuallynotassociated withachangeinthenumberofclasses.Moreover,discussionswithsuperinten-dentssuggestthatalertparentstendtomakesurethatmaximumclasssizerulesareenforcedbuttrytopreventtheenforcementofminimumclasssizerules.Thus,ignoringnoneventsmayproduceasmallbiasinfavorofclasssizebeingefficacious (smallerclasssizesareassociatedwithmorealertparents).Suchbiasisnotaconcern,giventheresults.Onecancarryoutadistrict-levelversionoftheprocedureforthesecondidenticationmethod.Thedistrict-levelprocedureeliminatesthepossibilitythat theresultsareduetoparents’responding(tolargeclasssizes)bytransferringtheirchildrentootherschoolswithintheschooldistrict.Relativetotheschool-levelprocedure,thedistrict-levelprocedurehasallthesameadvantagesand disadvantagesasithasintherstidenticationmethod.Ratherthancarryoutthedistrict-levelprocedureforthesecondidenticationmethod,Isimplyexam-inedeacheventtodeterminewhethertherewereoffsettingenrollmentchangesinotherschoolsinthesamedistrict.Ididnotndanysuchoffsettingchanges. Event-by-eventexaminationispossiblebecausethenumberofeventsislimited. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1255 Section 9-85 3.ANoteonSingle-YearversusMultiple-YearEffectsofa ChangeinClassSize Sofar,Ihavewrittenalloftheequationsasthoughachange inclasssizethisyeargeneratesachangeinachievementbythe endoftheyear.Thisisbecauserecentempiricalresultssuggest thatsuchsingle-yeareffectsaretheimportanteffects(seeKrue- ger[1999]andAngristandLavy[1999]).Itmaybe,however,that suchspecicationsarenotafairtestofclasssizebecausea studentneedstobeinsmallerclassesforafewgradesbeforethere isanyeffect.Intheempiricalworkthatfollows,Idoprovide resultsforclasssizeinthemostrecentyear,butImainlyfocuson specicationsthatusetheaverageclasssizethatacohorthas experiencedupuntilthetimeittakesthetest. Ifocusonthespecicationsthatuseaverageclasssize becausetheyfavorndingthatclasssizeisefficacious.Thereason isthatacohortusuallyexperiencessmallorlargeclassesconsis- tently,12 sothatalmostnoneofthedifferencebetweentheaverage classsizeexperiencedbyonecohortandthenextwithinaschoolis likelytobecausedbymeasurementerror.Ifthereweremeasure- menterror,itwouldwashoutoncetheclasssizeexperiencedbya cohortwasaveragedoverafewgrades.Itisimportantto rememberthattheidenticationstrategiesrelyon cohort-to- cohort differencesinclasssizeandthateachcohortexperiences relativelyunchangingclasssize. IV.DATA Thetwoidenticationmethodscreateanumberofdata requirements.First,becausetheintegernatureofteachersand classroomsisusefulformakingnaturalpopulationvariation translateintovariationinclasssizeandcomposition,dataonthe elementarygradesisneeded.Elementaryclassesarelessdivis- iblethansecondaryschoolclassesbecausethestandardmethodof elementaryschoolinstructionisoneteacherspendingthemajor- ityofeachschool-daywitharegulargroupofstudentsinone classroom.13 Also,classsizeiswell-denedinelementaryschools butpoorlydenedinmiddleandhighschools,wherestudents mayexperiencedifferentclasssizesindifferentsubjects.The 12.Thispointisdemonstratedbelow.13.Aclassisthegroupofstudentswhospendthemajorityoftheschooldaywithoneteacher.Themeasureofclasssizeexcludespull-outinstructionbyspecialeducationteachersoraides. QUARTERLYJOURNALOFECONOMICS1256 Section 9-86 resultingemphasisonelementaryclasssizetstheempiricaland pedagogicaldebates,whichhavefocusedonclasssizeinearly grades.Anotherreasonforfocusingonelementaryclasssizeis thatelementaryschoolsarenotlarge.Inverylargeschools, naturalpopulationvariationaveragesouttoagreatextentwithin eachcohort.14 Finally,sinceschoolcohortsaredenedbybirth date,oneneedsdataonpopulation-by-ageattheschoolentry cutoffdate(whichisDecember31inConnecticut).15 Connecticutschooldataareparticularlyappropriateforthe empiricalstrategy.Thestatehas649elementaryschoolsthatbelong to146elementarydistricts.16 Overall,25percentoftheschoolshave typicalcohortsizessmallerthan46students;50percenthavesmaller than63students;and75percenthavesmallerthan92students.17 DistrictsareessentiallytownsinConnecticutand,formanyyears,the townscollectedannualEnumerationsofChildren(population-by-age dataasofJanuary1forallschool-agedchildren).Inthelastfewyears, similardatahavebeencompiledbyClaritasIncorporated.18 The EnumerationofChildrenandClaritasarethesourceofthepotential kindergartencohortdata.Finally,everyyearsince1986,Connecticut hasadministeredstatewidetestsinthefourth,sixth,andeighth grades.19 From1986to1991,testdataareavailablebydistrict.From 1992onward,testdataareavailablebyschool(aswellasbydistrict).I usesixyearsofschool-leveltestdata(from1992–1993to1997–1998) 14.Forthedistrict-levelversionsoftheprocedures(whichareessentiallyspecicationtests),itisusefultohavesomedistrictsthataresmall(thatcontainonlyonetothreeelementaryschools). 15.InConnecticutachildisordinarilyenrolledinkindergartenifhewillbevebyDecember31oftheschoolyear.16.Elementaryschoolsareschoolsthatcontainsomecombinationofgrades1to6.MostelementaryschoolsinConnecticutcontaingrades1through6,butsome districtshaveseparateschoolsforthelowerelementarygradesandupperelementarygrades.SeenotestoTableI.17.Amongschoolsthatareindistrictswithmedianhouseholdincomebelowthetwenty-fthpercentileforConnecticut,thedistributionofcohortsizesisas follows:43 5 25thpercentile,57 5 50thpercentile,78 5 75thpercentile.Amongschoolsthatareindistrictswithmedianhouseholdincomeabovetheseventy-fthpercentileforConnecticut,thedistributionofcohortsizesisasfollows:45 5 25thpercentile,66 5 50thpercentile,89 5 75thpercentile.Amongschoolsthatareless than5percentAfrican-American,thedistributionofcohortsizesisasfollows:46 525thpercentile,63 5 50thpercentile,93 5 75thpercentile.Amongschoolsthataremorethan10percentAfrican-American,thedistributionofcohortsizesisasfollows:44 5 25thpercentile,61 5 50thpercentile,79 5 75thpercentile.Thelast groupofschoolsisalmostexclusivelyurban.18.Somedistrictscombinetwosmalltowns.Insuchcases,thetowns’population-by-agestatisticsareaggregated.19.Between1979and1985,Connecticutadministeredstatewideachieve- menttestsintheninthgrade.Ninthgradescoresarenotidealforexaminingtheeffectsofelementaryclasssizeandcomposition,butpreviousversionsofthispapercontainresultsbasedontheearliertests.Theseresultsareavailablefromtheauthor. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1257 Section 9-87 andtwelveyearsofdistrict-leveldata(from1986–1987to1997–1998). Imainlyshowresultsforthefourthandsixthgradetestssincethey arecloselylinkedtoelementaryclasssize,butsimilareighthgrade resultsareshowninHoxby[1998]andareavailablefromtheauthor. Inmostyears,classsizeisreportedbymultiplesources,andcross- checksofthosesourcessuggestthatitisaccurate.20 Averageclasssize inConnecticutisabout21students,anditsstandarddeviationis about5.5students,butclasssizerangeswidely.Therstpercentileis 8students,andtheninety-ninthpercentileis34students.While Connecticutisnotuniqueinhavingappropriatedata,fewotherstates havesimilarlypropitiousconditionsandlongpanelsoftherelevant data. TableIshowsthestructureoftheConnecticutdatabycohort. Eachcohortisdescribedbyitslikelygraduatingclass—forinstance, oneexpectsthatchildrenwhoentersixthgradeinthefallof1991will beintheJune1998graduatingclass.Enrollment,classsize,andsome oftheachievementdataareavailablebyschool,bygrade,andby cohort.Thekindergartenpopulationdataandsomeoftheachieve- mentdataareavailablebydistrict,bygrade,andbycohort.Ihave24 yearsofenrollmentdata,soIestimatetheenrollmentresidualsbased onall24yearsofdata.Thelargenumberofyearsallowsmetoget morepreciseestimatesoftheresiduals. Thetestsareadministeredatthebeginningofeachschool year(September).Thus,thefourthgradetestsmaybeaffectedby classsizesintherstthroughthirdgrades,buttheyareunlikely tobeaffectedbyfourthgradeclasssize.Similarly,classsizesin rstthroughfthgradesarerelevantforthesixthgradetests. Eachequationhas,asexplanatoryvariables,theclasssizesthat couldhaveaffectedthedependentvariable.However,notethat 20.Onemightworryabouterrorinthemeasureofclasssize,especially becausemeasurementerrorcanbeexacerbatedbyrst-differencing.Ihaveveried,however,thatthereislittlemeasurementerrorinclasssizebyexaminingmultiple,independentreportsonclasssizes.SeethenotestotheAppendix.Moreimportantly,acohortusuallyexperiencessmallorlargeclasssizeforseveralyears running,soalmostnoneofthedifferencebetweentheaverageclasssizeexperi-encedbyonecohortandthenext(withinaschool)islikelytobecausedbymeasurementerror.Ifthereweremeasurementerror,itwouldaverageoutoverafewgradesforacohort.Finally,averageclasssizeisinstrumentedbypredicted averageclasssize,andthisshouldremedymeasurementerrorbias.Measurementerrorinthedependentvariablewouldshowupinthestandarderrors,whichareverysmall.ThedependentvariablesaremeasuredwitherrorbecausethetestsareadministeredinSeptember,butfortunately,thereislow studentturnover.Thisisbecausefamiliestimetheirmovestocoincidewithschoolchangeovers.Thus,adistrictwithmoderateturnoverhaslowwithin-schoolturnover.InConnecticutin1997–1998,themeanelementaryschoolhad93percentofitsstudentsreturn. QUARTERLYJOURNALOFECONOMICS1258 Section 9-88 mostcohortsexperiencesimilarclasssizesintherstthrough sixthgrades.Unusuallylargecohortstendconsistentlytoexperi- encelargeclasssizes,andunusuallysmallcohortstendconsis- tentlytoexperiencesmallclasssizes.21 21.Statisticalevidenceforthelaststatementcanbeobtainedbyexaminingthecorrelationbetween,say,acohort’srstgradeenrollmentresidualanditsfth TABLEI STRUCTUREOFTHE DATA SET Grad- uating class Kinder. cohort District-gradeleveldataSchool-gradeleveldata Enroll- ment Class size Testsingrade Enroll- ment Class size Testsingrade 4689 468 1983 333 33 1984 333 33 1985 333 33 1986 333 33 1987 333 33 1988 333 33 1989 333 3 1990 333 3 1991 333 33 1992 333 33 1993 333333 1994 333333 1995 3333333 1996 3333333 1997 3333333 3 1998 3333333 3 1999 333333333 2000 333333333 2001 3333333333 2002 3333333333 2003 3333 3333 2004 3333 3333 2005 333 333 2006 333 333 Acohort’s‘‘graduatingclass’’isthecalendaryearinwhichitwouldbeexpectedtoobtainitshighschool diplomaifitsmembersgraduatedontime.Forinstance,ifastudentobtainshishighschooldiplomainJune 1998,thenhisgraduatingclassis1998.Theschool-levelpanelisslightlyunbalancedbecausegradesare occasionallymovedbetweenschoolswithinadistrict.Thereare3504school-levelobservationsofrstand secondgrades(6yearstimesapproximately584schools);3464school-levelobservationsofthirdgrades(6 yearstimesapproximately577schools);3404school-levelofobservationsoffourthgrades(6yearstimes approximately567schools);3071school-levelobservationsoffthgrades(6yearstimesapproximately511 schools);and1150school-levelobservationsofsixthgrades(6yearstimesapproximately192schools). Connecticuthas146elementarydistricts,andthereare1752observationsindistrict-levelregressions(12 yearstimes146districts). THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1259 Section 9-89 EveryschooldistrictinConnecticutwassurveyedaboutits maximumandminimumclasssizerules,teachers’aides,and mixed-gradeclasses.Acopyofthesurveyisavailablefromthe author.Responsesweregatheredbymail,e-mail,telephone,and fax,andtheresearchersspoketomultiplepeopleinmostdistricts, althoughthemostcommonrespondent(byfar)wasthedistrict superintendentandthesecondmostcommonwasarepresenta- tiveoftheschoolboard.Thekeyfeaturesoftheresponsesareas follows.Informationwasobtainedfromeverydistrict,andsuper- intendentswerequeriedaboutrulesintheirdistrictoverthepast decade.Bothmaximumandminimumclasssizerulesvaried amongthedistricts,butthemodalmaximumclasssizewas25 andthemodalminimumclasssizewas15.Onlyvedistricts reportedachangeintheirrules,andthechangeswerevery modest(frommaximumclasssizeof27to25,forinstance).22 The lackofchangeswasexplainedbyanumberofsuperintendents, whoreportedthattheirdistricts’ruleshadbeensetduringthe early1980swhenstudentpopulationsinConnecticutwereat theirnadir.Duringsubsequentyears,whenstudentpopulations begantogrowrapidly,mostdistrictsfoundthat maintaining their ruleswassufficientlychallenging.Aboutone-thirdofdistricts claimedthattheydidnothaveaminimumclasssizerulebecause sucharulewouldneverbind.Empirically,itturnedouttobetrue thatdistrictsthatclaimedthattheydidnotneedtherulewere districtsthathadsteadyincreasesintheirschool-agedpopulation fortheentireperiod.Insuchdistricts,minimumclasssizerules neverneedtobeused.23 Districts’answerstothequestionsaboutteachers’aidesand mixed-gradeclasseswererelativelyuniform.Althoughteachers’ aidesandmixed-gradeclassesaresometimesusedforpedagogical purposes(aidesareusedespeciallyforspecialeducation),theyare rarelyusedasamethodofmanagingtoo-largeclasses.24 gradeenrollmentresidual.(Anypairofgradesissuitable.)Theenrollmentresidualsarecomputedusinggrade-school-specicregressionsbasedonequation (3)withaquarticintime.Theenrollmentresidualsforanypairofgradesarecomputedindependently.Nevertheless,thereisacorrelationofabout0.85betweenpairsofresidualsforacohort. 22.Idonotusethesechangesinrules.Forthevedistrictsinquestion,Ieffectivelydivideeachdistrictintotwo:a‘‘beforechange’’districtandan‘‘afterchange’’district.23.Twodistrictsstatedthattheydidnothaveanymaximumclassrules becausetheywouldneverneedtobeapplied.Thesestatementswereconrmedbytheempiricalevidence:thetwodistrictsinquestionhavesmallcohortsizesforeachgrade(almostalwaysunder20).24.Mostdistrictswerevehementlyopposedtotheuseofaidesormixed-grade classesasaregularremedyfortoo-largeortoo-smallclasses. QUARTERLYJOURNALOFECONOMICS1260 Section 9-90 TherawscoresforConnecticut’stestsarenotintuitive,soI formdependentvariablesfortheregressionsbydividingeachtest scorebythestandarddeviationofschools’scoresonthattestin Connecticut.25 Forthepurposeofinterpretation,itisconvenient thatastandarddeviationoneachtestcorrespondsroughlytothe state’sideaofamasterylevel.Forinstance,onthemathtestthe differencebetweenbeing‘‘atthestate’sgoal’’and‘‘slightlybelow thestate’sgoal’’islittlemorethanonestandarddeviation. Similarly,thedifferencebetweenbeing‘‘slightlybelowthestate’s goal’’and‘‘belowthestate’sgoal’’isaboutonestandarddeviation, andthedifferencebetweenbeing‘‘belowthestate’sgoal’’and‘‘well belowthestate’sgoal’’isaboutonestandarddeviation. Allthedatausedarepubliclyavailableandwereobtained fromtheConnecticutDepartmentofEducationoritspublica- tions.TheAppendixshowsunweighteddescriptivestatisticsof thedataset,whereanobservationisaschool. V.SOME ILLUSTRATIVE GRAPHS Graphsforindividualschoolscanprovideintuitionaboutthe empiricalstrategyandtheresults.Iconsiderthreeschoolsin Connecticut,chosenfortheirillustrativevalueratherthantheir representativeness.SchoolAhasoneclassroompergrade;school Bhaseitheroneortwoclassroomspergrade,dependingon enrollment;andschoolChaseithertwoorthreeclassroomsper grade,dependingonenrollment.26 EachofFiguresIthroughIII showsaschool’senrollmentandclasssizeinthefthgrade,by cohort.Iselectedthefthgradebecausestudentsaretestedatthe beginningofthesixthgradeyear,butitwouldnothavemattered muchifIhadselectedanothergrade.FigureIshowsthat,in schoolA,enrollmentandclasssizewereidenticallyequalfor 25.Thestandarddeviationsinschools’scoresthatIusecomefromtechnicalreportswrittenbythetestmakers[Harcourt-BraceEducationalMeasurement]anddistributedbythestate’sBoardofEducation.Thestandarddeviationin students’scoresonatestareabout25percentgreaterthanthestandarddeviationsinschools’scores.IfIweretousethestandarddeviationinstudents’scores,theestimateswouldappeartobeevenmoreprecise.Foreachtest,IusedthemedianstandarddeviationamongtheyearsforwhichIhavetestdata.The standarddeviationonatestdoesnotdiffermuchfromyeartoyear,however,andtheresultsarenotsensitivetodividingeachtestscorebythestandarddeviationforitsyear.Thetechnicalreportsareagoodguidetothescoringofthetests,whichwaschangedonce(‘‘rstgeneration’’versus‘‘secondgeneration’’intheterminology oftests).IntheregressionsIdonotusedataacrossyearsinwhichthescoringonatestchanged,andtheyeareffectsintheregressionspickupidiosyncraticchangesinthetestfromyeartoyear.26.ThenamesofschoolsA,B,andCareavailablefromtheauthor. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1261 Section 9-91 FIGURE I EnrollmentandClassSizeinSchoolA FIGURE II EnrollmentandClassSizeinSchoolB QUARTERLYJOURNALOFECONOMICS1262 Section 9-92 everycohort,andclasssizevariedbetween10and23.FigureII showsthat,inschoolB,enrollmentandclasssizewereidentically equalfortherstvecohorts,uptothegraduatingclassof2001. Fortheserstvecohorts,classsizevariedbetween16and29 students.Thegraduatingclassof2002hadenrollmentof30 students,however,andschoolBisinadistrictwithamaximum classsizeis29.Therefore,schoolBaddedasecondfthgrade classroomforthegraduatingclassof2002.Thereafter,even thoughenrollmentfellbackbelow30students,schoolBmain- tainedtwofthgradeclassroomsbecauseenrollmentneverfellso farthatthedistrict’sminimumclasssizerulewastriggered.The graduatingclassesof2002–2005experiencedclasssizesranging fromthirteentofteen.FigureIIIshowsschoolC,whichbegan withtwofthgradeclassroomsandclasssizeof24.However, therewere56studentsinthegraduatingclassof1998,andschool Cisadistrictwithmaximumclasssizeof25.Therefore,schoolC addedafthgradeclassroomforthatcohortandkeptthethird classroomuntilthegraduatingclassof2003,whichhadenroll- mentofonly40students.TheminimumclasssizeinschoolC’s districtisfourteen,sotherulewastriggered,andschoolCwent backtohavingonlytwofthgradeclassrooms.Thenextcohort, however,had59students,andthethirdfthgradeclassroomwas FIGURE III EnrollmentandClassSizeinSchoolC THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1263 Section 9-93 reinstated.Overall,classsizevariedfrom14to24studentsin schoolC. Allthreeguresillustratethevariationthatisusefulforthe rstidenticationmethod,whichusesthevariationinenrollment thatdoesnotappeartobepartofatrendandthatdoesnottrigger achangeinthenumberofclassrooms.SchoolAisanice,simple examplebecause,althoughitappearstohaveanupwardtrendin enrollment,itisobviousthatmuchoftheyear-to-yearvariationin enrollmentisnotsystematic.FiguresIIandIIIillustratethe variationthatisusefulforthesecondidenticationmethod, whichusesthechangesinclasssizethatoccurwhenwithin- schoolvariationinenrollmenttriggersamaximumorminimum classsizerule.Allthreeguresshowthatclasssizevariesovera rangethatcoversthepolicyrangeveryfully.Justinthesethree schools,classsizevariesfrom10to29. FiguresIVthroughVIsuperimposeeachschool’saverage sixthgradereadingscoresonitsfthgradeclasssize.Ifreducing classsizeimprovedreadingscores,thenwewouldexpecttosee thetwolinesgenerallymoveinoppositedirections,likemirror imagesofoneanother.But,itisdifficulttodiscernanypattern FIGURE IV ClassSizeandReadingScoresinSchoolA QUARTERLYJOURNALOFECONOMICS1264 Section 9-94 linkingreadingscoresandclasssize.Thesamecanbesaidfor mathscoresandwritingscores,whicharenotshownhere. Lookingatthesethreeschools,however,ishardlyasystematic wayofdeterminingwhetherthereisasignicantrelationship betweenachievementandclasssize.Thereisaneedforregression analysis. VI.RESULTS InthissectionIexaminetheeffectsofclasssizeonachieve- ment.Beforeshowingtheresultsforthetwoidentication methodsdescribedabove,Ishowresultsforafewmethodsthat arecommonlyuseddespitehavingtheidenticationproblems describedinSectionII.Theseresultsgiveoneasenseofwhatthe datawouldshowifoneweretoapplytypicalmethodsnaively. 1.ResultsfromCommonlyUsedMethodsofIdentication EachcellinTableIIshowstheestimatedcoefficientonclass sizefromaseparateregression.Thecolumnsdenethespecica- tionoftheregression,andtherowsshowresultsfordifferent FIGURE V ClassSizeandReadingScoresinSchoolB THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1265 Section 9-95 dependentvariables.Forinstance,thenumberintheupper-left- handcellistheeffectoflogaverageclasssizeingrades1through 3onfourthgrademathscoresusingaspecicationthatpools observationsacrossschoolsandcohorts(withcohortxedeffects). Thisnaivespecicationislikelytoproduceestimatesthatare biasedbycorrelationbetweenclasssizeandunobservedparent andcommunityattributes.Parentswithunobservedgoodcharac- teristicsarelikelytochooseschoolswithsmallclasssizesand communitieswithunobservedgoodqualities.Infact,theesti- matesintherstsixrowsofcolumnIareallnegativeandhighly statisticallysignicant.(Idiscussthebottomthreerowsbelow.)If oneweretogivetheestimatescredence,onewouldsaythatthe coefficientintherstcellindicatesthata10percentreductionin classsizeingrades1through3improvesfourthgrademathscores by0.1468(about15percent)ofastandarddeviation.Other coefficientsincolumnIaresimilar:a10percentreductioninclass sizeingrades1through5appearstoimprovesixthgrademath scoresbyabout13percentofastandarddeviation. IncolumnII,Iaugmenttheequationbyaddingdistrict-level demographicvariablesfromthe1990census:medianhousehold FIGURE VI ClassSizeandReadingScoresinSchoolC QUARTERLYJOURNALOFECONOMICS1266 Section 9-96 income,thepercentageofthepopulationinpoverty,thepercent- ageofadultswhoarehighschoolandcollegegraduates,andthe percentagesofthepopulationwhoareAfrican-Americanand Hispanic.(Thesevariablesareobservedatthedistrictlevelonly indecennialcensusyears.)Thesecontrolsforobservedparentand communitycharacteristicsgreatlyattenuatetheestimatedeffect ofclasssizeontestscores,buttheestimatesarestillallnegative insign,andtwoofthesixestimatesintherstsixrowsare statisticallysignicantatthe10percentlevel.Infact,the TABLEII NAIVE ESTIMATESOFTHE EFFECTSOF CLASS SIZEON STUDENT TEST SCORES Eachcellcontainstheestimatefromaseparateregression(anditsstandard errorinparentheses). Dependent variable Independent variable I Cohort xed effects II Cohortxed effects& demographic controls fourthgrademathscorelogavgclasssizethrough grade3 2 1.4675 2 0.1028 (0.2067)(0.0994) fourthgradereading score logavgclasssizethrough grade3 2 1.1532 2 0.1338 (0.1450)(0.0752) fourthgradewritingscorelogavgclasssizethrough grade3 2 0.5872 2 0.0301 (0.0919)(0.0578) sixthgrademathscorelogavgclasssizethrough grade5 2 1.3141 2 0.1364 (0.2788)(0.1209) sixthgradereadingscorelogavgclasssizethrough grade5 2 1.4043 2 0.1821 (0.2771)(0.1162) sixthgradewritingscorelogavgclasssizethrough grade5 2 0.5571 2 0.0497 (0.1409)(0.0907) sixth—fourthgrademath score avgclasssizeinfourth andfthgrds 0.1081 2 0.1335 (0.0829)(0.0722) sixth—fourthgrade readingscore avgclasssizeinfourth andfthgrds 2 0.2645 2 0.1572 (0.0581)(0.0498) sixth—fourthgrade writingscore avgclasssizeinfourth andfthgrds 2 0.1980 2 0.2950 (0.0848)(0.0968) Sourceisauthor’scalculationsbasedonConnecticutdataset.TheregressionsareOLS,areweightedby numberofstudentsoverwhomthedependentvariableisaveraged,andincludeaxedeffectforeachcohort. Standarderrorsareinparenthesesandadjustedforthegroupednatureofthedata(multipleobservationson eachschool).Thenumberofobservationsintheregressionsforgrades1through6is,respectively,3504,3504, 3464,3404,3071,1150(seethenotestoTableI).Thedependentvariablesareformedbydividingtheaverage testscorebythestandarddeviationofConnecticutstudents’scoresonthattest.Thus,thecoefficientsshow howtestscores,measuredinstandarddeviations,changewiththelogofclasssize.Thedemographiccontrols incolumnIIaremedianhouseholdincome,percentageofthepopulationinpoverty,percentageofadultswho arehighschoolgraduates,percentageofadultswhoarecollegegraduates,percentageofthepopulationwho areAfrican-American,andpercentageofthepopulationwhoareHispanic. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1267 Section 9-97 estimatesshownincolumnIIaresimilartomanyoftheestimates thathavegeneratedempiricalcontroversy.Theyareofmixedor marginalstatisticalsignicance,andtheeffectsare small.The equationscontrolforsomeobserveddemographics,butitisnot clearthattheremainingvariationinclasssizecomesfrom exogenoussources.Atleastsomeoftheremainingvariationis likelytobeduetounobserveddemographicsthatarecorrelated withclasssizeinmuchthesamewayastheobserveddemograph- icsarecorrelatedwithclasssize:parentswithdemographicsthat arebenecialforachievementchoosedistrictswithsmallerclass sizes,producingresultsbiasedtowardndingthatclasssize reductionsareefficacious.Ontheotherhand,otherdemographic controlsbeingequal,theschoolswithlowerclasssizemaybe thosethatarereceivingcompensatoryfundstoreduceclasssize because theirstudentshaveunusuallylowachievement.This wouldproduceresultsbiasedagainstndingthatclasssize reductionsareefficacious. ThebottomthreerowsofTableIIshowwhatisusuallycalled avalue-addedspecication.Thedifferencebetweenacohort’s sixthgradeandfourthgradetestscoresisregressedonthelogof theaverageclasssizesthattheyexperiencebetweenthetwotests: fourthandfthgrade.Suchspecicationsareoftenthoughtto controlforalltheeffectsoffamilybackgroundandneighborhood, throughtheearliertestscore.Itisfarfromobvious,however,that suchclaimsarevalid.Unobservedbackgroundmayaffectthe growth ofastudent’sachievement;unobservedbackgroundneed notbefullyincorporatedbythe level ofastudent’spriorachieve- ment.Inotherwords,parentswhoprovidealotoflearning resourcesathomearelikelytohelptheirchildrenlearnmorein everygrade,foreverybundleofresourcesthatthechildgetsat school.Thesameparentsarelikelytoputtheirchildreninschools withsmallclasssize.27 Inshort,value-addedestimatesmaybe biasedeithernegativelyorpositively,butitislikelythatthe preponderanceofthebiasfavorsclasssizeappearingtobe efficacious.Infact,veofthesixestimatesinthebottomthree rowsofTableIIarenegativeandstatisticallysignicantly differentfromzeroatthe5percentlevel.Ifoneweretogivethe 27.TheproblemwouldnotbealleviatedifIregressedthechangeintestscoresonthechangeintheclasssize(averageclasssizeinfourthandfthgrades minusaverageclasssizeinrstthroughthirdgrades).Mostofthechangesinclasssizethatacohortexperienceswouldnotberandom.Itwouldbetheresultofreactionstothecohort’sownachievementortheresultofsystematicchangesintheschool’senvironment. QUARTERLYJOURNALOFECONOMICS1268 Section 9-98 estimatescredence,onewouldsaythata10percentreductionin classsizeingrades4and5makesreadingscoresrise(betweenthe fourthandsixthgrades)byabout16percentofastandard deviation,controllingforobservabledemographics. Thefundamentalproblemwithallofthespecicationsin TableIIisthattheyeliminateonesourceofsuspectvariation,only tohavemoreobscuresourcesbecomedominant.Whenconsider- ingapolicyvariablelikeclasssize,wherethevastmajorityofthe variationcomesfromsuspectsources,itismoreeffectivetostart withsourcesofvariationthatareknowntobeexogenousand workfromthere.Thisisthelogicbehindthetwoidentication methodsadvancedinthispaper. 2.ResultsfromtheFirstIdenticationMethod Therstidenticationmethodattemptstouserandom variationintheschool-agedpopulation,andthestrategyis implementedbyinstrumentingforclasssizewithenrollment residualsorkindergartencohortresiduals.TableIIIshowscoeffi- cientestimatesfromtherst-stageequation(equation(7)).Each cellrepresentsaseparateregression,andeachcontainstheeffect oflog(u)onthelogofclasssize.Eachcolumnheadingdescribes thespecication,andeachrowlabeldescribesthegradelevelfor whichclasssizeisbeingestimated.Forinstance,theupper-left- handcellcontainstheestimatedcoefficientonlog(u)froma regressionthatisbasedonschool-levelobservationsofrstgrade classsizeandschool-levelobservationsofrstgradeenrollment residuals,wheretheenrollmentresidualsarecalculatedusinga versionofequation(3)thathasaninterceptanda linear time trend.Theestimatedcoefficientsintherstcolumnrangefrom 0.8566to0.9773.Theysuggestthatarandom10percentincrease inenrollmentraisesclasssizebybetween8.6and9.7percent—in otherwords,alittlelessthan1-for-1.Theprobablereasonwhythe coefficientsarenotevencloserto1isthattheenrollment residualsaremeasuredwitherror(theyarejustestimates,after all).Thereis,thus,alittleattenuationbias.ColumnIIcontains theestimatedcoefficientsonlog(u)fromregressionsthatare basedonschool-levelobservationsofclasssizeandenrollment residuals,wheretheenrollmentresidualsarecalculatedusinga versionofequation(3)thathasaninterceptanda quartic time trend.TheestimatesshownincolumnIIareverysimilartothose shownincolumnI:theyrangefrom0.8546to0.9799.The estimatesincolumnIII,whichisestimatedatthedistrictlevel, THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1269 Section 9-99 TABLEIII COEFFICIENTSFROM FIRST-STAGE REGRESSIONSFOR IDENTIFICATION METHOD 1 Eachcellcontainstheestimatefromaseparateregression(anditsstandard errorinparentheses). Dependent variable I II IV V Explanatoryvariable Residuallog enrollment (school-level, linear timetrend removed) Residuallog enrollment (school-level, quartic timetrend removed) Residuallog enrollment (district-level, quartic timetrend removed) Residuallog kindergarten cohort (district-level, quartic timetrend removed) logrstgradeclass size 0.85660.85460.76200.6186 (0.0110)(0.0210)(0.0349)(0.0452) logsecondgradeclass size 0.72940.72750.66790.6416 (0.0129)(0.0183)(0.0346)(0.0423) logthirdgradeclass size 0.89370.87170.78540.4557 (0.0105)(0.0164)(0.0360)(0.0459) logfourthgradeclass size 0.94190.89200.78870.3786 (0.0098)(0.0309)(0.0365)(0.0480) logfthgradeclass size 0.90390.86780.70270.3953 (0.0128)(0.0197)(0.0345)(0.0434) logsixthgradeclass size 0.97730.86690.83560.3099 (0.0111)(0.0290)(0.0499)(0.0623) xedeffectsforeach ‘‘school·expected numberofclassesin thegrade’’combina- tion yes yes xedeffectsforeach ‘‘district·expected numberofclassesin thegrade’’combina- tion yes yes Sourceisauthor’scalculationsusingConnecticutdataset.IdenticationMethod1attemptstouse randomvariation,overtime,inthepopulationofstudentswhobelongtoagradeinaschoolasaninstrument forclasssizeinthatgradeinthatschool.Eachrst-stageregressionhas,asitsdependentvariable,thelogof classsizeinagrade.Eachregressionhas,asitskeyexplanatoryvariable,anestimateofthepartofthegrade’s populationthatisduetorandomvariation.Forinstance,eachexplanatoryvariableincolumnIistheresidual fromaregressionofenrollmentinagradeinaschoolonaconstantandalineartimetrend.Theresidualscome from separate regressionsforeachgradeineachschool.Eachrst-stageregressioncontainsaxedeffectfor eachschool-expectednumberofclassescombination.Thesexedeffectsensurethatthemonotonicity conditionforinstrumentalvariablesisfullled.SeeSectionIIIforfurtherexplanation.Intheschoollevel regressions,thenumberofobservationsis3404infourthgraderegressions,and1150insixthgrade regressions.Inthedistrictlevelregressionsthereare1752observations(seethenotestoTableI).Ifthe independentvariableisclasssizeinthemostrecentgrade,insteadofaverageclasssizeingradesthatprecede thetest,thentheresultsforthespecicationincolumnIIare 2 0.1304(0.0980),2 0.1204(0.0747),0.1550 (0.0901),0.0304(0.1167),2 0.0330(0.1084),and0.0925(0.1537). QUARTERLYJOURNALOFECONOMICS1270 Section 9-100 arealsosimilar:theyrangefrom0.7027to0.8356.Thereismore attenuationbiasinthedistrict-levelregressionsbecausethe district-levelenrollmentresidualsarealessprecisemeasureof therandomuctuationsinenrollmentexperiencedbyanygiven schoolinthedistrict.Finally,columnIVofTableIIshowsresults basedondistrict-levelkindergartencohortresiduals.Asone expects,thecoefficientsaresomewhatlowerthanthoseofcolumn IIIbecausesomechildreninthekindergartencohortgotoprivate school.Moreover,a10percentincreaseinkindergartencohort sizeproducesthebiggestincreaseinrstgradeclasssize,a smallerincreaseinsecondgradeclasssize,andsoondownto sixthgradeclasssize.Oneexpectsthisbecausemobilityintoand outofthedistrictmakeskindergarten-cohortsizemoreimportant fortheearlierelementarygrades. Overall,therst-stageregressionssuggestthatenrollment residualsarestronginstrumentsforclasssize:the t-statisticsin thersttwocolumnsareallgreaterthan40.District-leveland kindergarten-cohortresidualsarelessstrongasinstruments,but stillstrongenough:the t-statisticsaregenerallymuchgreater than10.Also,thecoefficientsaccordwithexpectations,which suggeststhattheresidualsarebeingestimatedinareasonable fashion.Residualsfromschool-specicinterceptsand quartic timetrendsaremypreferredsetofresiduals.Movingfroma quarticpolynomialtoahigh-orderpolynomialsaddsanegligible amountofexplanatorypowerandproducesresidualsthatarenot discerniblydifferent. Iusepredictedclasssizebasedontheequationsshownin TableIIItoformindependentvariablesforthesecond-stage equations.Forinstance,toformapredictionofthelogaverage classsizethatacohortinaschoolexperiencesingrades1through 3,Icomputetheaverageofthatcohort’spredictedclasssizein grades1,2,and3,andItakethelogoftheresult.Icomputethe logoftheaveragepredictedclasssize,nottheaverageofthelog predictedclasssizes.Itakeaccountof‘‘feederschools’’—for instance,studentfromtworst-to-fourthgradeschoolsmay attendthesamefth-sixthgradeschool. TableIVcontainsthemainclasssizeresultsfortherst identicationmethod.Eachcellcontainsanestimatefroma separateregression.ColumnIusesrst-stageregressionsin columnIofTableIII,columnIIusesrst-stageregressionsin columnIIofTableIII,andsoon. Beforeconsideringtheestimatedcoefficients,notethatthe THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1271 Section 9-101 standarderrorsaresmall.Intheschool-levelregressions(col- umnsIandII)thestandarderrorsaresosmallthatifa10percent reductioninclasssizeweretochangetestscoresbyjust2to4 percentofastandarddeviation,thechangewouldbestatistically signicantatthe5percentlevel.Inthedistrict-levelregressions (columnsIIIandIV)thestandarderrorsareslightlyhigher,butif a10percentreductioninclasssizeweretochangetestscoresby just3to4percentofastandarddeviation,thechangewouldbe TABLEIV BASIC RESULTSFROM IDENTIFICATION METHOD 1:2SLSESTIMATESOFTHE EFFECTS OF CLASS SIZEON STUDENT TEST SCORES Eachcellcontainstheestimatefromaseparateregression(withitsstandard errorinparentheses). Dependent variable Independent variable isthe prediction of: IIIIIIIV Classsizeispredictedusingrst stageregressionsfrom Column Iof TableIII Column IIof TableIII Column IIIof TableIII Column IVof TableIII fourthgrademath score logavgclasssize throughgrade3 0.0664 2 0.08450.12450.2203 (0.1069)(0.1227)(0.2100)(0.1537) fourthgrade readingscore logavgclasssize throughgrade3 2 0.0736 2 0.1027 2 0.15130.1260 (0.0759)(0.0870)(0.1643)(0.1084) fourthgrade writingscore logavgclasssize throughgrade3 0.13640.1871 2 0.01980.0332 (0.1085)(0.1214)(0.1472)(0.1061) sixthgrademath score logavgclasssize throughgrade5 0.04960.0394 2 0.05220.2059 (0.1367)(0.1578)(0.1346)(0.1714) sixthgradereading score logavgclasssize throughgrade5 2 0.01740.1288 2 0.01520.0843 (0.1247)(0.1462)(0.1063)(0.1410) sixthgradewriting score logavgclasssize throughgrade5 0.06750.0494 2 0.1384 2 0.1003 (0.1769)(0.2077)(0.1166)(0.1562) xedeffectsforeach‘‘school·expected numberofclassesinthegrade’’group yesyes xedeffectsforeach‘‘district·expected numberofclassesinthegrade’’group yesyes xedeffectsforeachcohort yesyesyesyes Standarderrorsarecorrectfor2SLS.Theregressionsareweightedbythenumberofstudentsoverwhom thedependentvariableisaveraged.Intheschoollevelregressionsthenumberofobservationsis3404in fourthgraderegressions,and1150insixthgraderegressions.Inthedistrictlevelregressionsthereare1752 observations(seethenotestoTableI).Predictedclasssizeiscomputedusingtherst-stageregressionsshown inTableIII.Thedependentvariablesareformedbydividingtheaveragetestscorebythestandarddeviation ofstudents’scoresonthattestinConnecticut.Thus,thecoefficientsshowhowtestscores,measuredin standarddeviations,changewiththelogofclasssize. QUARTERLYJOURNALOFECONOMICS1272 Section 9-102 statisticallysignicantatthe5percentlevel.Inotherwords,if reducingclasssizeby10percentmadestudentsmovejust2to4 percentclosertomasteringthestate’snextlevelofprociency,the improvementwouldbestatisticallysignicant.Therandomvaria- tioninclasssizehasconsiderablepowertoidentifyachievement gains. Despitethispropitioussituation,theestimatesincolumnsI throughIVdonotshowthatsmallerclasssizesproduceachieve- mentgains.Theestimatesaremixedinsign,andnoneis statisticallysignicantatthe5percentlevel.Onewouldnotwish forsmallerstandarderrorsbecauseasmanyresultswiththe ‘‘wrong’’aswiththe‘‘right’’signwouldbecomestatistically signicant.ThesimplestinterpretationofTableIVisstraightfor- ward:giventhestandarderrors,theeffectofreducingclasssizeis ratherpreciselyestimatedtobeclosetozero.Becauseallofthe estimatesareclosetozero,thefourspecicationsdonotseemvery different,butinfactweshouldrememberthattheyusedifferent enrollmentresidualsastheestimatesaslog(u).Inparticular,the columnIIIestimatesdonotallowparents’movingstudentswithin thedistricttoproducebias,andthecolumnIVestimatesdonot allowparents’movingstudentstootherdistrictsortoprivate schoolstoproducebias. GiventhatTableIVpresents‘‘well-estimatedzeros,’’oneis naturallydrawntoestimateavarietyofalternativespecications toseeifandwhenclasssizematters.Icanshowonlyafractionof thespecicationsIestimated.ThenotestoTableIVpresent resultsforclasssizeinthemostrecentgrade.InHoxby[1998]I explorenumerousotherspecicationsanddemonstratethat resultsmuchlikethoseinTableIVareobtainediftheindepen- dentvariableisclasssizeingrade1,classsizeingrades1and2, classsizewithaspline(withabreakatclasssizeof23),an indicatorforstudents’everhavingexperiencedclasssizebelow 15,oranindicatorforstudents’everhavingexperiencedclasssize above30.InTableV,Ishowtwoalternativespecicationsthatare especiallylikelytobeinteresting,giventheempiricalliterature onclasssize.Krueger[1999],Hanushek,Kain,andRivkin[1998], andFerguson[1998]arguethatthereductionsinclasssizeare moreefficaciousinschoolsthatservestudentswhoarelow- incomeorminorities.African-Americansarethemostimportant minoritygroupinConnecticut,soIexamineresultsthatdifferby whethertheschool’sstudentscomefromlow,medium,orhigh THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1273 Section 9-103 TA B L E V ADD I T I O N A L RES U L T S F R O M IDE N T I F I C A T I O N MET H O D 1: 2 S L S E ST I M A T E S USI N G SPL I N E SPE C I F I C A T I O N S Ea c h c e l l c o n t a i n s t h e e s t i m a t e f r o m a s e p a r a t e r e g r e s s i o n ( w i t h i t s s t a n d a r d e r r o r i n p a r e n t h e s e s ) . De p e n d e n t v a r i a b l e 2 In d e p e n d e n t v a r i a b l e Sp l i n e : e a c h r o w i s a r e g r e s s i o n S p l i n e : e a c h r o w i s a r e g r e s s i o n Lo w i n c M e d i n c H i g h i n c H i g h % b l k M e d % b l k L o w % b l k fo u r t h g r a d e m a t h s c o r e l o g a v g c l a s s s i z e t h r o u g h g r a d e 3 2 0. 0 4 5 5 2 0. 0 8 7 7 2 0. 1 4 9 0 2 0. 1 8 3 7 2 0. 1 2 5 0 0 . 1 6 1 8 (0 . 1 4 2 5 ) ( 0 . 1 2 9 9 ) ( 0 . 1 4 0 9 ) ( 0 . 1 4 5 5 ) ( 0 . 1 4 4 5 ) ( 0 . 2 8 6 3 ) fo u r t h g r a d e r e a d i n g s c o r e l o g a v g c l a s s s i z e t h r o u g h g r a d e 3 2 0. 1 2 1 8 2 0. 1 6 3 8 2 0. 2 0 5 2 * 2 0. 1 5 0 0 2 0. 0 8 5 5 2 0. 3 5 5 7 (0 . 0 9 9 7 ) ( 0 . 1 0 0 9 ) ( 0 . 0 9 8 6 ) ( 0 . 1 0 1 6 ) ( 0 . 1 0 0 9 ) ( 0 . 2 0 0 0 ) fo u r t h g r a d e w r i t i n g s c o r e l o g a v g c l a s s s i z e t h r o u g h g r a d e 3 0 . 2 2 1 3 0 . 2 0 0 2 0 . 1 3 2 3 0 . 2 2 5 3 0 . 2 0 0 2 0 . 0 8 9 9 (0 . 1 3 8 6 ) ( 0 . 1 2 7 7 ) ( 0 . 1 4 5 1 ) ( 0 . 1 4 2 3 ) ( 0 . 1 4 1 9 ) ( 0 . 2 9 3 1 ) si x t h g r a d e m a t h s c o r e l o g a v g c l a s s s i z e t h r o u g h g r a d e 5 0 . 3 1 2 1 2 0. 0 0 7 2 2 0. 0 2 0 3 0 . 0 9 4 3 0 . 1 4 7 8 0 . 1 5 0 1 (0 . 2 6 2 7 ) ( 0 . 2 6 3 5 ) ( 0 . 4 8 4 7 ) ( 0 . 3 0 3 5 ) ( 0 . 2 5 2 3 ) ( 0 . 3 9 4 5 ) si x t h g r a d e r e a d i n g s c o r e l o g a v g c l a s s s i z e t h r o u g h g r a d e 5 0 . 2 8 1 1 0 . 0 8 4 0 0 . 0 0 2 8 0 . 5 3 7 8 0 . 1 4 9 3 0 . 2 6 1 9 (0 . 2 4 1 6 ) ( 0 . 2 4 3 7 ) ( 0 . 4 4 8 7 ) ( 0 . 3 6 5 1 ) ( 0 . 2 3 2 4 ) ( 0 . 2 8 1 0 ) si x t h g r a d e w r i t i n g s c o r e l o g a v g c l a s s s i z e t h r o u g h g r a d e 5 0 . 3 0 7 4 0 . 0 4 8 2 2 1. 0 8 4 7 0 . 4 6 8 0 2 0. 2 8 3 5 0 . 0 1 6 0 (0 . 3 3 2 0 ) ( 0 . 3 3 5 0 ) ( 0 . 6 1 6 6 ) ( 0 . 3 8 6 5 ) ( 0 . 3 1 9 7 ) ( 0 . 5 0 2 2 ) Se e n o t e s t o T a b l e I V e x c e p t f o r t h e f o l l o w i n g d e t a i l s s p e c i c t o T a b l e V . T h e r e g r e s s i o n s i n T a b l e V a r e t h e s a m e a s t h e r e g r e s s i o n r e p o r t e d i n c o l u m n I I o f T a b l e I V , e x c e p t t h a t t h e y ha v e s p l i n e s c r e a t e d b y i n t e r a c t i n g t h e c l a s s s i z e v a r i a b l e w i t h i n d i c a t o r v a r i a b l e s f o r a c h a r a c t e r i s t i c o f t h e s c h o o l d i s t r i c t . T h e i n d i c a t o r f o r a l o w i n c o m e d i s t r i c t i s e q u a l t o 1 i f t h e di s t r i c t h a s m e d i a n h o u s e h o l d i n c o m e l e s s t h a n o r e q u a l t o t h e t w e n t y - f t h p e r c e n t i l e o f m e d i a n h o u s e h o l d i n c o m e i n C o n n e c t i c u t d i s t r i c t s ; 0 o t h e r w i s e . T h e i n d i c a t o r f o r a m e d i u m in co m e d i s t r i c t i s e q u a l t o 1 i f t h e d i s t r i c t h a s m e d i a n h o u s e h o l d i n c o m e g r e a t e r t h a n t h e t w e n t y - f t h p e r c e n t i l e a n d l e s s t h a n s e v e n t y - f t h p e r c e n t i l e ; 0 o t h e r w i s e . T h e i n d i c a t o r f o r a hi g h i n c o m e d i s t r i c t i s e q u a l t o 1 i f t h e d i s t r i c t h a s m e d i a n h o u s e h o l d i n c o m e g r e a t e r t h a n o r e q u a l t o t h e s e v e n t y - f t h p e r c e n t i l e ; 0 o t h e r w i s e . T h e i n d i c a t o r s f o r l o w , m e d i u m , a n d h i g h pe r c e n t A f r i c a n - A m e r i c a n a r e c o n s t r u c t e d s i m i l a r l y a r o u n d t h e t w e n t y - f t h a n d s e v e n t y - f t h p e r c e n t i l e s o f t h e p e r c e n t a g e o f t h e p o p u l a t i o n t h a t i s A f r i c a n - A m e r i c a n i n C o n n e c t i c u t di st r i c t s . QUARTERLYJOURNALOFECONOMICS1274 Section 9-104 incomefamiliesandbywhethertheschool’sshareofstudentswho areAfrican-Americanishigh,medium,orlow.28 Specically,inTableV,Iestimatethespecicationfrom columnIIofTableIV,exceptthatIallowclasssizetohave differentcoefficientsforschoolsthattintodifferentgroups.I rstdivideschoolsintogroupswherethe‘‘lowincome’’groupis districtswithpercapitalincomeatorbelowthetwenty-fth percentileofpercapitaincomeinConnecticut(15,454dollarsin 1990),the‘‘mediumincome’’grouphaspercapitaincomeabove thetwenty-fthpercentileandbelowtheseventy-fthpercentile (23,075dollarsin1990),andthe‘‘highincome’’grouphasper capitaincomeatorabovetheseventy-fthpercentile.These divisionsproducetheestimatesincolumnsIthroughIII.Withone exception,noneofthecoefficientsisstatisticallysignicantly differentfromzero,althoughthestandarderrorsarestillsmall enoughtogenerallyidentifyimprovementsinachievementas smallas3to6percentofastandarddeviationfora10percent reductioninclasssize.Innocaseisthepointestimateforhigh incomeschoolsstatisticallysignicantlydifferentfromthepoint estimateforlowincomeschools.Also,thediscerniblepatternof thepointestimatesdoesnotsuggestthatclasssizereductionsare moreefficaciousinschoolsthatservelowincomestudents(if anything,thepatternsuggeststheopposite).Theonlystatisti- callysignicantestimatesuggeststhatclasssizereductions improvefourthgradereadingscoresinschoolsthatservestu- dentsfromhighincomebackgrounds.Perhapsteacherswhowork insuchschoolsaremorelikelytomakegooduseofclasssize reductions,orhighincomeparentsaremorelikelytoensurethat their‘‘slowreader’’getsindividualattentionwhenclasssizeis small. Inextdivideschoolsintogroupswherethe‘‘highpercentage African-American’’groupcontainsdistrictswithpercentAfrican- Americanatorabovetheseventy-fthpercentileofdistricts’ percentblackinConnecticut(17percentin1990),the‘‘medium’’ groupcontainsdistrictswithpercentAfrican-Americanbelowthe seventy-fthpercentileandabovethetwenty-fthpercentile(1 percentin1990),andthe‘‘low’’groupcontainsdistrictswith percentAfrican-Americanatorbelowthetwenty-fthpercentile. InConnecticut,African-Americanhouseholdsareconcentratedin 28.In1990,11percentofConnecticut’sschool-agedpopulationwasAfrican-American,slightlylessthan9percentwasHispanic,andslightlylessthan2percentwasAsian. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1275 Section 9-105 urbandistricts,sothe‘‘high’’groupmayalsobethoughtofasthe urbangroup.ThesedivisionsproducetheestimatesincolumnsIV throughVI.Noneofthecoefficientsisstatisticallysignicantly differentfromzero,althoughthestandarderrorsarestillsmall enoughtogenerallyidentifyimprovementsinachievementas smallas3to6percentofastandarddeviationfora10percent reductioninclasssize.Innocaseistheestimateforhighpercent African-Americanschoolsstatisticallysignicantlydifferentfrom theestimateforlowAfrican-Americanschools,andthereisno discerniblepatterninthepointestimates. Insummary,theestimatesinTablesIVandVsuggestthat classsizereductionsarenotefficaciousforimprovingstudent achievement.Theestimatesdonotconrmthehypothesisthat classsizereductionsaremoreefficaciousindistrictsthatcontain lowincomeorAfrican-Americanstudents. 3.ResultsfromtheCross-SectionRegression DiscontinuityMethod Nowconsiderchangesinclasssizethatoccurwhenaschool changesthenumberofclassesinagrade.InTableVI,Itreatthe dataasthoughtheywerecross-sectiondata,estimatethepre- dictedclasssizefunctionforeachschoolbasedonitsdistrict’s maximumclasssizeandequation(9),andusethelogofpredicted classsizeasaninstrumentforlogclasssize.(Thismethoddoes notlenditselftoexaminingtheeffectsofclasssizeinmultiple grades,soIuseclasssizeinthegradeimmediatelypriortothe test.)Recallthatthecross-sectionapproachislikelytoproduce unbiasedresultsonlywhenthesampleisnarrowedtothe observationsjustoneithersideofamaximumclasssizethresh- old.IncolumnI,Iusetheentiretyofthepredictedclasssize functionandexpecttoproducebiasedresults,sincemostofthe functionreectspermanentcharacteristicsoftheschool.In columnII,Iuseonlytheobservationsthatarewithinfour studentsofadiscontinuity,sotheresultsshouldbelessbiased.In columnIII,Iuseonlytheobservationsthatare at adiscontinuity, andIexpecttheresultstobeunbiased. Considerrstthenumberofobservationsineachregression, shownatthebottomofTableVI.Asonenarrowsinonthe discontinuities,thenumberofobservationsinthefourthgrade regressionsfallsfrom1953incolumnIto76incolumnIII.The numberofobservationsinthesixthgraderegressionsfallsfrom 1011incolumnIto37incolumnIII.Asthenumberofobserva- QUARTERLYJOURNALOFECONOMICS1276 Section 9-106 tionsfalls,thestandarderrorsrise.Thestandarderrorsin columnIaresuchthata10percentreductioninclasssizewould havetoproduceanimprovementof6to16percentofastandard deviationfortheimprovementtobestatisticallysignicant.The standarderrorsincolumnIIIaresuchthata10percentreduction inclasssizewouldgenerallyhavetoproduceanimprovementof 30to50percentofastandarddeviationfortheimprovementtobe statisticallysignicant.Thefallingnumberofobservationsand therisingstandarderrorsdemonstratetheextraordinaryde- mandsthatthecross-sectionmethodputsondatawhenitis appliedsoastoensureunbiasedresults. TheregressionsincolumnIsuggestthatreductionsinclass TABLEVI IVESTIMATESOFTHE EFFECTOF CLASS SIZE,GENERATEDBY CROSS-SECTION REGRESSION DISCONTINUITY Eachcellcontainstheestimatefromaseparateregression. Dependentvariable I II III Thepredictedclasssizefunctionisused: Inits entirety Within4students ofadiscontinuity Solelyatthe discontinuities fourthgrademathscore 2 0.0503 2 0.0972 2 0.0506 (0.0229)(0.0593)(0.1060) fourthgradereadingscore 2 0.0423 2 0.0856 2 0.0746 (0.0166)(0.0454)(0.0821) fourthgradewritingscore 2 0.0137 2 0.0082 2 0.0211 (0.0130)(0.0321)(0.0372) sixthgrademathscore 2 0.0922 2 0.0992 0.0674 (0.0496)(0.0872)(0.1220) sixthgradereadingscore 2 0.1042 2 0.1496 0.0250 (0.0511)(0.0992)(0.0796) sixthgradewritingscore 2 0.0301 2 0.0181 0.0159 (0.0241)(0.0401)(0.0498) numberofobservationsin fourthgraderegressions 1953 703 76 numberofobservationsinsixth graderegressions 1011 374 36 Thesourceisauthor’scalculationsbasedontheConnecticutdataset.Theregressionsareweightedbythe typicalnumberofobservationsoverwhichthedependentvariableisaveraged.Thedependentvariablesare formedbydividingtheaveragetestscorebytheoverallstandarddeviationofscoresonthattestin Connecticut.Theindependentvariableisclasssizeinmostrecentgrade,instrumentedbypredictedclasssize. Theequationcontainsaxedeffectforeachcohort.Thecross-sectionmethodtreatstheConnecticutdataas thoughtheywerecross-sectiondataandactualchangesinthenumberofclasseswerenotobserved.The predictedclasssizefunctionuseseachdistrict’smaximumclasssizeandtheformulagivenbyequation(9).See textforfurtherexplanation. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1277 Section 9-107 sizeimproveachievementsignicantly.Fouroutofthesixcoeffi- cientsarestatisticallysignicantatthe5percentlevel,andallsix coefficientshavethe‘‘right’’sign.Ifweweretointerpretthese resultsnaively,wewouldconcludethata10percentreductionin thirdgradeclasssizeraisesfourthgrademathscoresbyabout12 percentofastandarddeviation.Aswenarrowinonthedisconti- nuities,however,suchresultsdisappear.IncolumnIII,where onlythediscontinuitiesareused,noneoftheresultsiscloseto beingstatisticallysignicant,andfouroutofthesixestimates havethe‘‘wrong’’sign.Therefore,thestatisticallysignicant resultsincolumnIaregenerated not bythediscontinuitiesinthe predictedclasssizefunction,butbythesuspectpartsofthe function. 4.ResultsfromtheSecondIdenticationMethod(the Within-SchoolRegressionDiscontinuityMethod) TableVIshowsresultsfromthesecondidenticationmethod: thewithin-schoolregressiondiscontinuitymethod.Forthismethod Ifocusoneventswherethenumberofclasseschangedbecausea modestchangeinenrollment(smallerthan20percent)triggered amaximumorminimumclasssizerule.Iestimateequation(11), arst-differencedversionoftheachievementequation,usingjust thecohortsimmediatelybeforeandaftereachevent.Thismethod isquitepowerfuldespitethefactthatitreliespurelyondiscontinu- ouschangesinclasssizedrivenbychangesinthenumberof classes.Itspowerderivesfromthefactthatitcomparesadjacent cohortsinthesameschool,whohavelittlereasontobedifferent apartfromtheirdifferentclasssizeexperiences.Infact,the secondidenticationmethodproducesstandarderrorssosmall thatifa10percentreductioninclasssizeweretochangetest scoresbyjust2to4percentofastandarddeviation,thechange wouldbestatisticallysignicantatthe5percentlevel. ColumnIofTableVIIincludesalltheeventsinwhichthe numberofclasseschanged(andaffectedclasssize)inthegrade beforethetest.ColumnIIincludesonlytheeventsinwhichthe numberofclasseschangedinthesameway(andaffectedclass size)inthethreegradesimmediatelybeforethetest.Inother words,columnIIusesthefactthatacohortthatwasbigenoughto haveathirdgradeclassaddedwhenitenteredthirdgradeoften hadaclassaddedinsecondgradeandrstgradeaswell.Despite thesmallstandarderrors,noneoftheestimatesinTableVIIis statisticallysignicantlydifferentfromzeroatthe5percentlevel. QUARTERLYJOURNALOFECONOMICS1278 Section 9-108 TABLEVII ESTIMATESOFTHE EFFECTOF CLASS SIZE,IDENTIFICATION METHOD 2 (WITHIN-SCHOOL REGRESSION DISCONTINUITY ) Eachcellcontainstheestimatefromaseparateregression(withitsstandard errorinparentheses). Dependentvariable I Independent variable ischangein classsize(due totheaddition orsubtractionof aclass)inthe gradepreviousto thetest,forthe2 adjacentcohorts II Independent variable ischangein classsize(due totheaddition orsubtractionof classes)inthe3 gradespreviousto thetest,forthe2 adjacentcohorts Changeinfourthgrademathscore betweentwoadjacentcohortsinthe sameschool 0.0844 2 0.0714 (0.1001)(0.1605) Changeinfourthgradereadingscore betweentwoadjacentcohortsinthe sameschool 0.0468 2 0.0540 (0.0636)(0.1396) Changeinfourthgradewritingscore betweentwoadjacentcohortsinthe sameschool 0.1731 0.1602 (0.0976)(0.1568) Changeinsixthgrademathscore betweentwoadjacentcohortsinthe sameschool 0.0126 2 0.0207 (0.0969)(0.1588) Changeinsixthgradereadingscore betweentwoadjacentcohortsinthe sameschool 2 0.0468 0.0238 (0.0828)(0.1520) Changeinsixthgradewritingscore betweentwoadjacentcohortsinthe sameschool 0.1585 0.1543 (0.1300)(0.1901) numberofobservationsinfourthgrade regressions 147 117 numberofobservationsinsixthgrade regressions 108 86 Thesourceisauthor’scalculationsbasedontheConnecticutdataset.Theregressionsareweightedbythe typicalnumberofobservationsoverwhichthedependentvariableisaveraged.Thedependentvariablesare formedbydividingtheaveragetestscorebytheoverallstandarddeviationofscoresonthattestin Connecticut.Thewithin-districtmethodexploitsthefactthattheConnecticutdataarepaneldataandactual changesinthenumberofclasseswithinagradewithinaschoolareobserved.Theequationisestimatedin rst-differences:thechangeinscoresbetweenback-to-backcohortsisregressedonthechangeinclasssize,if thatchangeinclasssizeistheresultofasmallchangeinenrollmenthavingtriggeredamaximumor minimumclasssizerule. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1279 Section 9-109 Onewouldnotwishforsmallerstandarderrorsbecausemore resultswiththe‘‘wrong’’signwouldbecomestatisticallysigni- cantthanwouldresultswiththe‘‘right’’sign.Thebestinterpreta- tionofTableVIIisthattheestimatedeffectsofclasssize reductionsareratherpreciselyestimatedzeros. Obviously,the20percentcutofffora‘‘small’’changein enrollmentisarbitrary.Ihaveexperimentedwithcutoffsbetween 35and15percent,andtheresultsaresimilar.29 VII.INTERPRETATION Estimatesbasedonbothidenticationmethodsindicatethat classsizereductionshavelittleornoeffectonachievement.The estimatesaresufficientlyprecisethatimprovementsthatare educationallysignicantwouldbeidentiable.Thetwoidentica- tionmethodsareindependentandthusprovidechecksonone another.Theresultsarealsorobusttospecicationchanges,some ofwhichareshownaboveandsomeofwhichareshowninHoxby [1998]. Theestimatesarebasedonvariationinclasssizethatoccurs mainlyintherangeof10to30studentsperclass.Thisisthe relevantrangeforAmericanpolicy,butitwouldbeamistaketo extrapolatetheseresultstoschoolsinwhichclasssizeistypically higherthan30.Sincemostschoolsindevelopingcountrieshave classsizeshigherthan30,theresultsinTablesIVthroughVII neitherconrmnorcontradictmostdevelopingcountrystudies.It wouldalsobeamistaketoextrapolatetheseresultstoclasssizes oflessthan10.Suchtinyclassesaretooexpensiveformost Americandistrictstoconsiderbecausethecostofaone-student reductionincreasesasclasssizegetssmaller(costisroughly linearinthe percentage reductioninclasssize).Ave-student reductionfromabaseof40raisescostsbyonly14.3percent;buta ve-studentreductionfromabaseof15raisescostsby50percent. Krueger[1999]providesevidencethat,inProjectStar,a10 percentreductioninclasssizeforoneyearimprovesscoresby about10percentofastandarddeviation,a10percentreductionin classsizeforthreeimprovesscoresbyabout13percentofa standarddeviation(comparethiswiththefourthgraderesultsin thispaper),anda10percentreductioninclasssizeforveyears improvesscoresbyabout17.5percentofstandarddeviation 29.Theseresultsareavailablefromtheauthor. QUARTERLYJOURNALOFECONOMICS1280 Section 9-110 (comparethiswiththesixthgraderesultsinthispaper).30 Anyof theseimprovementswouldbehighlystatisticallysignicantif theyappearedinthispaper,giventhispaper’sstandarderrors. Howmightoneexplainthecontrastinresultsinthenatural experimentandanexplicitpolicyexperiment?Inboththenatural experimentandpolicyexperiment,teachershadmore opportunity toimproveachievementwithsmallerclasses.Inneitherexperi- mentdidteachersreceivespecialtrainingtotakeadvantageof thesmallerclasssizes.Thedifferenceintheresultsmaybecaused bythefactthatthenaturalexperimentvariedclasssizebutdid notvaryincentives,whilethepolicyexperimentvariedclasssize andcontainedimplicitincentivesforteachersandadministrators tomakegooduseofsmallerclasssizes(becausefullenactmentof thepolicydependedonasuccessfulevaluation).Ifthisisthe correctinterpretationofthedifferenceintheresults,thenthe implicationisthatclasssizereductionpoliciesshouldcontain built-inevaluationandincentives. SinceConnecticutschoolstaffwereunawareofthenatural experiment,theycouldnothavereactedtotheevaluation. Explicitpolicyexperimentsmayworkdifferentlybecauseof Hawthorneeffectsorotherreactivebehavioronthepartof participants. Onemightattributesomeofthedifferenceinresultstothe necessarilytransitorynatureofpopulationvariation(fromthe teachers’,notstudents’,pointofview).Thatis,teachersexperi- encesmallclasssizesrepeatedly,butnoteveryyear.Teachersdo notreceivetrainingtotakeadvantageofsmallerclasssizesina systematicway—inotherwords,theymaynotvarytheirprimary classroomstylemuchwhentheyhavetheopportunitiespresented byasmallerclass.Thisinterpretationwouldsuggestthatreduc- tionsinclasssizeshouldbecombinedwithinstructionfor teachersthathelpsthemmodifytheirteachingtechniques.This cannot,however,betheentireexplanation.Evenifshedoesnot lecturedifferentlytoasmallerclass,ateachercandevotemore efforttoeachstudentduringeveryteachingactivitythathasan individualelement.Manyoftheseactivitiesarepartofateacher’s basicrepertoire:answeringquestions,correctingassignments, dealingwithdisciplinaryproblems,tutoringastudentwhois aheadorbehindtheclass,talkingtoparents,andsoon.Also,the 30.Theaverage‘‘small’’classinProjectStarwas30percentsmallerthantheaverage‘‘regular’’class.Students’scoresincreasedbyabout30percentofastandarddeviationonmathandreadingtestsafteroneyear. THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1281 Section 9-111 ProjectStarresultswereachievedafteronlyoneyearofsmaller classsize,andtheteachersinvolveddidnotreceiveinstruction aboutchangingtheirprimaryteachingtechniques. VIII.CONCLUSIONS InthisstudyIusenaturalvariationintheschool-aged populationtoidentifytheeffectsofclasssizeonstudentachieve- ment.Thisapproachhasthreebenets.First,thevariationin classsizethatIstudyiscrediblyexogenous.Itisnotvariation generatedbyparents’choices—choicesthatareaffectedbypar- ents’incomesandparents’assessmentsoftheattentiontheir childrenneed.Second,theactorsinthenaturalexperimentI examinewerenotawareofbeingevaluatedormindfulofrewards beingcontingentupontheoutcome.Realpoliciesthatreduceclass size,suchasthe1996Californiainitiativeandthe1999federal initiative,rarelyincludeevaluationorrepercussions(suchasthe fundsbeingtakenawayifthepolicyhasnoeffect).Itisimportant thatresearchmimictheincentivesthatexistunderrealpolicies. Third,naturalpopulationvariationgeneratesuctuationsin classsizethatareintherangerelevanttocurrentpolicy. Thisstudydemonstrateshowpopulationvariationcanbe usedtoconsistentlyestimatetheeffectofclasssizeonstudent achievement.Ioutlinetwoindependentmethodsforusingpopula- tionvariation.Therstmethodisbasedonisolatingthecredibly randomcomponentofthenaturalvariationinpopulationfora gradeinaschool.Randomvariationinthepopulationgenerates exogenousvariationinclasssize.Thesecondmethodisbasedon exploitingthediscontinuouschangesinclasssizethatoccurwhen asmallchangeinenrollmenttriggersamaximumorminimum classsizeruleandtherebychangesthenumberofclassesina gradeinaschool.Bothmethodsproduceresultsthatareappropri- ateforconsideringclasssizechangesintherangeof10to30 students. Usingbothmethods,Indthatreductionsinclasssizehave noeffectonstudentachievement.Theestimatesaresufficiently precisethat,ifa10percentreductioninclasssizeimproved achievementbyjust2to4percentofastandarddeviation,Iwould havefoundstatisticallysignicanteffectsinmath,reading,and writing.Indnoevidencethatclasssizereductionsaremore efficaciousinschoolsthatcontainhighconcentrationsoflow incomestudentsorAfrican-Americanstudents. QUARTERLYJOURNALOFECONOMICS1282 Section 9-112 Theresultsjustdescribedarefarlesslikelytosufferfrom omittedvariablesbiasandendogeneitybiasthanaretypical estimatesthatdependonvariationinclasssizethatis(directlyor indirectly)generatedbyparents’decisions,teachers’decisions, administrators’decisions,orpolicy-makers’decisions.Idemon- stratethatmethodsthatrelyonsuspectvariationdisplaythe expectedpatternsofbias. ThemethodsIemployhavetheadvantagethatparticipants arenotawareofbeingevaluated.Inthisway,theexperiments mimicactualclasssizereductionpolicies,whichrarelyinclude evaluationsorincentivesforschoolstomakegooduseofthe opportunitiesprovidedbysmallerclasssizes.Ifonewereconsis- tentlytondthatpolicyexperimentsthatreducedclasssize and containedincentivesproducedgreaterimprovementsintest scoresthannaturalexperimentsthatjustreducedclasssize,one mightconcludethattheincentiveenvironmentisimportant.That is,policiesthatjustprovidemoreresourcesmaybesignicantly lessefficaciousthanpoliciesthatlinkresourcestoperformance. APPENDIX TABLE MeanStd.dev.1st %ile99th %ile enrollmentingrade1 68.88433.11215180 enrollmentingrade2 63.34029.52315161 enrollmentingrade3 62.01328.46115158 enrollmentingrade4 62.00631.35614177 enrollmentingrade5 63.57735.04912200 enrollmentingrade6 82.38567.85710340 classsizeingrade1 21.4145.539832 classsizeingrade2 21.1525.248832 classsizeingrade3 22.2995.4861233 classsizeingrade4 22.6545.7951134 classsizeingrade5 21.8866.2771332 classsizeingrade6 23.5166.4201033 residuallogenrollmentingrade1*0.0000.103 2 0.7071.586 residuallogenrollmentingrade2*0.0000.106 2 0.6501.813 residuallogenrollmentingrade3*0.0000.105 2 0.6960.947 residuallogenrollmentingrade4*0.0000.107 2 1.1471.343 residuallogenrollmentingrade5*0.0000.106 2 0.7270.519 residuallogenrollmentingrade6*0.0000.096 2 0.6730.495 residuallogkindergartencohort*0.0000.080 2 2.2910.627 minimumclasssizegrade1 15.0581.498620 minimumclasssizegrade2 15.0901.542620 minimumclasssizegrade3 15.0991.553620 minimumclasssizegrade4 15.1651.713622 THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1283 Section 9-113 HARVARD UNIVERSITYAND NATIONAL BUREAUOF ECONOMIC RESEARCH REFERENCES Angrist,Joshua,andVictorLavy,‘‘UsingMaimonides’RuletoEstimatetheEffectofClassSizeonScholasticAchievement,’’QuarterlyJournalofEconomics, CXIV(1999),533–575.Angrist,Joshua,GuidoImbens,andDonaldRubin,‘‘IdenticationofCausalEffectsUsingInstrumentalVariables,’’JournaloftheAmericanStatisticalAssociation,XCI(1996),444–455. Betts,Julian,‘‘IsThereaLinkbetweenSchoolInputsandEarnings?FreshScrutinyofanOldLiterature,’’inGaryBurtless,ed.,DoesMoneyMatter?TheLinkbetweenSchools,StudentAchievement,andAdultSuccess (Washington,DC:TheBrookingsInstitution,1995). APPENDIX TABLE (CONTINUED) MeanStd.dev.1st %ile99th %ile minimumclasssizegrade5 15.1741.734622 minimumclasssizegrade6 15.1741.734622 maximumclasssizegrade1 23.3382.7121530 maximumclasssizegrade2 23.7312.8791532 maximumclasssizegrade3 24.6562.4121832 maximumclasssizegrade4 25.3452.3561932 maximumclasssizegrade5 25.5382.3491932 maximumclasssizegrade6 25.6552.2701932 medianhouseholdincome 43,93013,46822,140104,483 percentofthepopulationwhoareAfri- can-American 7.36810.723041.184 percentofthepopulationwhoareHis- panic 5.6407.686031.037 percentofadultswhoareatleasthigh schoolgraduates 83.5796.98562.56695.304 percentofadultswhoareatleast 4-yearcollegegraduates 28.06611.8747.83660.768 percentofpopulationwhoareurban74.56035.6300100 Enrollmentineachgradeistakenfromtheseriesvariouslytitled StrategicSchoolProles,Townand SchoolDistrictProles,and ConditionofPublicElementaryandSecondaryEducationinConnecticut:Town andSchoolDistrictProles.Classsizefortheschoolyears1991–1992through1997–1998istakenfromthe sameseries.ForpreviousyearsitistakenfromunpublisheddatamadeavailablebytheStateofConnecticut DepartmentofEducation.Inmanycasesclasssizehasbeencheckedagainstindividualdistricts’annual reports.Classsizehasalsobeencheckedagainsttheseries ElementaryClasses,bySize,inConnecticutPublic Schools.Thesameseries(StrategicSchoolProles andsoon)containsthedemographicvariableslistedabove (medianhouseholdincomethroughthepercentofthepopulationwhoareurban).Theprimarysourceforthe demographicvariables,however,isthe SchoolDistrictDataBook,whichisaschooldistrictlevelsummaryof the1990UnitedStatesCensusofPopulationandHousing.Thesizeofeachkindergartencohortistakenfrom theseriestitled EnumerationofChildren andfromasimilarseriescompiledbyClaritas,Incorporated.For schoollevelvariables(enrollment,classsize),thereare3504observationsinrstgrade,3504observationsin secondgrade,3464observationsinthirdgrade,3404observationsinfourthgrade,3071observationsinfth grade,and1150observationsinsixthgrade.Fordistrictlevelvariables(maximumandminimumclasssize rules,demographics),thereare1752observations. QUARTERLYJOURNALOFECONOMICS1284 Section 9-114 Card,David,andAlanKrueger,‘‘SchoolResourcesandStudentOutcomes:AnOverviewoftheLiteratureandNewEvidencefromNorthandSouthCarolina,’’JournalofEconomicPerspectives,X(1996),31–40. Claritas,Incorporated,ClaritasUpdate,ElectronicdataonpopulationbyagefortownsinConnecticut(Arlington,VA:Claritas,Incorporated).ConnecticutPublicExpenditureCouncil.ElementaryClasses,bySize,inConnecti-cutPublicSchools (Hartford,CT:ConnecticutPublicExpenditureCouncil, 1975through1988).Ferguson,Ronald,‘‘CanSchoolsNarrowtheBlack-WhiteTestScoreGap?’’inChristopherJencksandMeredithPhillips,eds.TheBlack-WhiteTestScore Gap,(Washington,DC:BrookingsInstitutionPress,1998).Hanushek,Eric,‘‘TheEconomicsofSchooling:ProductionandEfficiencyinPublicSchools,’’JournalofEconomicLiterature,XXIV(1986),1141–1177.,‘‘MeasuringInvestmentinEducation,’’JournalofEconomicPerspectives,X (1996),9–30.Hanushek,Eric,JohnKain,andStevenRivkin,‘‘Teachers,Schools,andAcademicAchievement,’’NationalBureauofEconomicResearchWorkingPaperNo.6691,1998. Harcourt-BraceEducationalMeasurement,ConnecticutMasteryTestsInterpreta-tiveGuide (ConnecticutStateBoardofEducation:variousyears1986to1998).Hoxby,Caroline,‘‘TheEffectofClassSizeandCompositiononStudentAchieve-ment:NewEvidencefromNaturalPopulationVariation,’’NationalBureauof EconomicResearchWorkingPaperNo.6869,1998.Krueger,Alan,‘‘ExperimentalEstimatesofEducationProductionFunctions,’’QuarterlyJournalofEconomics,CXIV(1999),497–532.NationalCenterforEducationStatistics,DigestofEducationStatistics,1998 (Washington,DC:GovernmentPrintingOffice,1999).Salmon,Richard,ChristinaDawson,StephenLawton,andThomasJohns,PublicSchoolFinanceProgramsoftheUnitedStatesandCanada,1993–94edition(Denver,CO:AmericanEducationFinanceAssociation,1995). StateofConnecticutBoardofEducation,ConditionofPublicElementaryandSecondaryEducationinConnecticut:TownandSchoolDistrictProles(Hartford,CT:StateofConnecticutBoardofEducation,1977–1978through 1981–1982).StateofConnecticutStateDepartmentofEducation,EnumerationofChildren(Hartford,CT:ConnecticutStatePrintingOffice,1975through1983).StateofConnecticutBoardofEducation,StrategicSchoolProles (Hartford,CT: StateofConnecticutBoardofEducation,1991–1992through1998–1999).StateofConnecticutBoardofEducation,TownandSchoolDistrictProles(Hartford,CT:StateofConnecticutBoardofEducation,1982–1983through1990–1991). StateofConnecticutDepartmentofEducation,Unpublishedseriesonvariables(enrollment,classsize)thatappearinrecent(1993–1994through1998–1999)editionsof StrategicSchoolProles (Hartford,CT:StateofConnecticutDepartmentofEducation,1995). UnitedStatesDepartmentofEducation,NationalCenterforEducationStatistics,SchoolDistrictDataBook:1990CensusSchoolDistrictSpecialTabulation,Computerle(Washington,DC:NationalCenterforEducationStatistics,1994). THEEFFECTSOFCLASSSIZEONACHIEVEMENT 1285 Section 9-115 This page is intentionally blank Section 9-116 EducationalResearchforPolicyandPractice 2:71–86,2003. ©2003 KluwerAcademicPublishers.PrintedintheNetherlands. ClassSizeandTeacherQuality JenniferBuckingham TheCentreforIndependentStudies,POBox92,StLeonardsNSW1590,Australia Abstract The‘VinsonReport’onPublicEducationinNSWhasbecomereceivedwisdom.Thereport’s recommendationonclasssizeshasattractedmoreattentionthananyother.Thisisunfortunate becauseitisonthisissuethattheReportisweakest.Athoroughappraisaloftheresearch onclasssizesrevealsthatmanystudieshavemethodologicalproblemsthatmaketheir applicationinarealworldcontextdoubtful;manystudieshaveintroducedotherreforms suchascurriculumchangesatthesametimeasclasssizereduction,makingtheirindividual effectsimpossibletodetermine;thelargemajorityofstudieshavefoundnosignificanteffects ofclasssizeonstudentachievement,whiletheremainderhaveshownsmallbenefits,usually onlywhenclasseshavelessthan20students;classsizehaslesseffectwhenteachersare competent;andthesinglemostimportantinfluenceonstudentachievementisteacherquality. Researchshowsunequivocallythatitisfarmorevaluable,bothineducationalandfiscal terms,tohavegoodteachersthanlotsofteachers.Itmustbeensuredthatthecurrentand incomingteachingforceisthebestitcanbe,beforeseekingtoexpandit. KeyWords:childdevelopment,classsize,publiceducation,studentachievement,teacher quality,teachingmethods Introduction In2000,theNewSouthWales(NSW)TeachersFederationinitiatedandfundedan ‘IndependentInquiryIntoPublicEducationinNSW’,chairedbyProfessorTony Vinson.ThisyeartheInquirycommitteepublisheditsreport(hereafterreferredto asthe‘VinsonReport’). Thefindingsoftheinquiryandtherecommendationsmadeinthethreevolumes ofitsfinalreportreceivedagreatdealofmediaandpoliticalattention,andrightly so.Thereportscontainawealthofinformationintheformofinsightsfromstu- dents,teachersandparents,aswellaspreviouslyunpublisheddatafromtheNSW DepartmentofEducation. Thereare,however,twocentralproblems.First,thecommitteeseemstohave madelittleefforttoseekoutandprovideinformationbeyondthesubmissions received,andonlythemostrudimentaryofliteraturereviewsandinternational comparisonsareoffered.Attemptingtocoveralltheresearchonschoolingwould havemadethereportunwieldyandtime-consuming,butthereareimportantrea- Section 9-117 72 JENNIFERBUCKINGHAM sonstobethorough.Issuessuchasclasssize,whereexpertopinionisfarfrom unanimous,requiredetailedanalysisataprimarysourcelevel.Further,themajority ofsubmissionswerefromteachers,whoarearguably(ifunderstandably)biased towardsmallerclasses. Second,theconclusionsdrawnonthebasisoftheinformationpresentedare debatable,andconnectionsbetweenthevarioustroublesinschoolsareoftennot made.Anyonefamiliarwitheducationalresearchandawareofthechallengesthat classroomteachersfaceonadailybasisknowsthatthedifficultiesassociatedwith largeclassesarerelatedtodisciplineproblemsandthewiderangeofabilitiesin eachclassroom.Similarly,whatmattersinaclassroommorethananythingelse, includingthenumberofstudents,isgoodteaching.TheVinsonreportdoesnot maketheseimportantpointsexplicit. Athoroughreviewoftheresearchonclasssizeandstudentachievementshows thatmuchofitisflawedinwaysthatmakeitunreasonabletoexpectthesame resultsinareal-worldsituation.Manystudieshaveintroducedotherreformsatthe sametimeasclasssizereduction,makingtheeffectofclasssizealoneimpossi- bletodetermine.Inmostcasesthepersonsparticipatingintheexperimentwere motivatedtoproducepositiveresults.Onlyasmallminorityofstudiesfoundany positiveeffectofsmallerclassesonstudentachievement,usuallyinclassesofless than20,andfewoftheseeffectswerelarge. Thefindingsonclasssizesuggestthatthereislittleifanyreasontobelievethat reducingclassesfrom25to20,asrecommendedbytheVinsonReport,willhavean effectlargeenoughtowarrantthecost.Researchtellsusthateffectiveteachingis muchmoreimportantthanthenumberofchildrenintheclassroom.Itis,therefore, muchwisertoinvestinthequalityofteachers,ratherthanquantity. GiventheauthoritytheVinsonReportanditsprincipalauthorhavebeenaf- forded,andthelikelihoodthatthereportwillbereferredtoregularlyinthefuture, itisnecessarytopointoutitsflawsandputreservationswithitsfindingsonthe publicrecord. ClassSizeandAchievement Intheareaofschoolreform,classsizereductionseemstoholdalltheaces.Itis popularwithacademics,teachers,studentsandparentsalike.Itseemsintuitivethat tohavefewerchildreninaclassisbetter. Researchappearstoconfirmthis.Severallargescalestudiesandmanysmaller onesfindarelationshipbetweenlearningandclasssize.Butacloserexamination revealscrucialmethodologicalproblemsandgeneralisationsthatmakethefindings farlessthandefinitive,evenmeaningless. Reviewersofthisresearch,whopresentitasevidencefortheimportanceand efficacyofclasssizereduction,ofteneitherignoretheseproblemsoracknowledge theminpassing. Section 9-118 CLASSSIZEANDTEACHERQUALITY 73 TheVinsonReporthadthescopeandexpertisetocovertheissueofclasssize thoroughly,butitrelatesthefindingsofvariousstudies,oftenfromsecondary sources,withouttheimportantcaveats.Thesecaveatsaresuchthatmuchofthere- searchisinapplicableinothercontexts.Thatis,thesameresultscannotbeexpected underdifferentcircumstances. Thereportdismissestheseproblems.Itconcludesthattheevidenceofare- lationshipbetweensmallerclasssizeandbetterlearningoutcomesisstrongand thattheeffectislargeenoughtowarranttheexpenseofsuchreformsinNSW stateschools.ItrecommendsreducingclasssizesinNSWstateschoolsfroma maximumof25toamaximumof20inKindergartenthroughtoYearTwo(Vinson Report1:85). Althoughclasssizereductionisoneofthemostexpensivereformsproposed bytheVinsonReport,onlysevenpagesaredevotedtojustifyingit.Themeritsof smallerclassesareconsideredself-evidentandinarguable,yetthereport’sliterature reviewisincompleteandinsufficienttoconfidentlydrawtheseconclusions. Hundredsofstudiescanbecitedontherelationshipbetweenclasssizeand studentachievement.Ehrenberg,Brewer,Gamoran&Willms(2001a)claimthat: Mosthavefoundsomeevidencethatsmallerclassesbenefitstudents,particularlyintheearlygrades, andespeciallykidsatriskofbeingunderachievers.Unfortunately,mostofthesestudieswerepoorly designed.Teacherandstudentassignmentswererarelysufficientlyrandom;anumberofstudieswere simplytoobriefortoosmall,andtoofewhadindependentevaluation.(p.78) Otherresearchers,suchasHanushek(1998),gofurther,arguingthatmostof thesestudiesarenotonlyflawedbutalsofailtoproduceconvincingevidence thatclasssizehasanysignificanteffectonstudentachievement.Hanushekisnot withouthiscriticsandsomeoftheirpointsofcontentionwithhisresearchare worthconsidering. Hanushek versus Krueger EconomistsEricHanushekofStanfordUniversityandAlanKruegerofPrinceton Universityhaveuseddifferentmethodstoconductmeta-analysesofstudiespro- vidingestimatesofclasssizeeffectsupto1994.Thedebatethathastakenplacein recentyearsbetweenthesetwoeconomistsisveryimportant. Hanushekiswellknownforhisresearchdemonstratingthatthereisnodirect relationshipbetweenfinancialresourcesandschoolperformance.Heclaimsthat onlyasmallminorityofstudiesshowasignificantpositiveeffectofsmallerclasses onstudentachievement. KruegerisbestknownforhisworkonProjectSTAR.Oneofthelargestand mostinfluentialstudiesofclasssizereduction,itsresultsarefrequentlycitedas proofofthebenefitsofsmallerclasses. Inameta-analysisof59studiesyielding277estimatesoftheeffectofclasssize onstudentachievement,Hanushek(1997)foundthat14.8%oftheseestimateswere positiveandsignificant.Thatis,studentsinsmallerclassesshowedsignificantly Section 9-119 74 JENNIFERBUCKINGHAM higherachievementthantheircounterpartsinlargerclasses.Theremainingesti- mateswereeitherinsignificant(nodifferenceinachievement–71.9%)ornegative andsignificant(smallerclasseshadlowerachievement–13.4%). Krueger(2002)arguesthatHanushek’smethodofselectingstudies,extracting andcountingtheestimatesisirrationalandhasproducedabiasedresult.Krueger’s maincriticismsare: •ThestudiesfromwhichHanushekdrewthemostestimatesarethosewhich producedinsignificantornegativeresults. •Whenaninsignificantorunexpectedresultisfoundbyresearchers,itreduces theirchanceofpublicationsotheyoftenlookfordisaggregatedeffects,separating thesampleintosmallersub-samples. •Thishastwoconsequences.First,anover-representationofinsignificantand negativeestimates.Second,theseestimatesarelesspowerfulbecausethesample sizeissmaller. •Itis,therefore,erroneoustocounteachoftheeffectestimatesfrommultiple- estimatestudiesandgivethemequalweightaseffectestimatesfromsingle-estimate studies. Kruegerproposesthreealternativemethodsofanalysis: (a)Estimatesshouldbegivenweightsproportionaltothenumberofestimates yieldedinthestudy.Forexample,asingle-estimatestudyshouldbecountedas one,butanestimatefromastudyyieldingfourestimatesshouldbecountedasone quarter. (b)Sincesomestudiesarebetterdesignedthanothers,theseshouldbegivenmore weightintheanalysis.Hissuggestedmethodiscitationfrequency;thatis,stud- ieswhicharereferredtomoreofteninacademicliteraturewouldbegivenmore weight. (c)Becausethesmallersub-samplesinmultiple-estimatestudiesreducetheirsta- tisticalpower,regressionanalysisshouldbeusedtoestimatewhattheeffectesti- matewouldbeifthestudyhadyieldedoneestimateonly. Onlythefirstoftheseisconvincing.IfKruegeriscorrectthatmultipleestimates fromonesamplearebiasedtowardsinsignificanceandthattheseresultshavea greatermarginoferror,theyprobablyshouldhavelessweightinameta-analysis andthereforelessinfluenceontheresults. Proposedmethods(b)and(c)areproblematic.Regardingthesecond,citation frequencyisnotaprovenindicatorofquality.Itmayjustaseasilybebiased towardstudieswithonetypeofresultortheother.Asforthethird,thefurther astatisticalanalysismovesfromtheoriginaldata,themoreroomforerrorandthe lessmeaningfultheresults. HanushekcountersKrueger’scriticismswell. •Hearguesthatmultiple-estimatestudiesprovidemoreinformationthanasingle estimateandshouldnotbeweightedlessinananalysis. •HerespondstoKrueger’sclaimofover-representationofinsignificantresults frommulti-estimatestudiesbyrestatingKrueger’sownargumentthatinsignificant Section 9-120 CLASSSIZEANDTEACHERQUALITY 75 Table1 Krueger’s(2002)Re-analysisofHanushek’s(1997)Meta-analysis. Result Hanushek:Krueger(1):Krueger(2):Krueger(3): EstimatesEstimatesEstimatesEstimates weightedweightedbyweightedbyderivedfrom equallyinverseofcitationregression numberoffrequencyanalysesof estimatesoriginal instudyestimates Positive&14.8%25.5%30.6%33.5% Significant Insignificant71.9%61.2%62.3%58.4% Negative&13.4%10.3%7.1%8.0% Significant Source:Krueger(2002,p.11.) resultsarelesslikelytobepublished,implyingthatthereisabiastowardpositive significantresultsintheliterature. •Hedismissestheaccuracyofderivingsingleestimatesfrommultipleestimates onthebasisthatdifferentsub-samplesofstudents(forexample,disadvantaged students)willyielddifferentresults.Thisimportantinformationislostwithaggre- gation. WhetheroneispersuadedmorebythecasepresentedbyHanushekorby Krueger,thestrongestevidenceisinthestatisticsproducedbytheirvariousmeth- odsofanalysis. Table1showsthatevenwhenestimatesareweightedandmanipulatedsoas toavoidperceivedbiastowardstudiesshowingnoeffectofclasssize–arguably creatingbiasintheoppositedirection–thestatisticsdonotshowthe‘systematic evidenceofarelationshipbetweenclasssizeandachievement’claimedbyKrueger (2002,p.31). IfweacceptKrueger’sfirstandleastcontroversialproposal–thatmultipleesti- matesfromasinglestudyshouldnotcarryasmuchweightasasingleestimate (whichisdebatableevenso)–onlyoneinfourstudiesfoundthatstudentsin smallerclasseshadachievementratessignificantlyhigherthanstudentsinlarger classes. Section 9-121 76 JENNIFERBUCKINGHAM OtherEvidence Theaboveconclusionisconsistentwiththefindingsofotherliteraturereviews. TheVinsonreportdescribestwonationaldataanalysesandfourliteraturereviews asfollows. Nationaldataanalyses: •Wenglinsky(1997):InYears4and8,‘lowerstudent/teacherratioswereposi- tivelyrelatedtohighermathematicsachievement’(seeVinsonReport1,p.83).In- consistently,thereportdoesnotdismissthisfindingduetotheuseofstudent/teacher ratioinsteadofclasssize,butdoessowithregardtotheworkofEricHanushek. •ReesandJohnson(2000):“...noevidencethatsmallerclassesaloneledto greaterstudentachievement”(asabove). Literaturereviews: •Glass&Smith(1979):‘...themajorbenefitsofreducingclasssizeoccurred wherethenumberofstudentswaslessthan20’(asabove). •Robinson&Wittebols(1986):‘positiveeffectswerelesslikelyifteachersdidnot changetheirmethodsandproceduresinthesmallerclasses’(SeeVinsonReport 1,p.84). •Slavin(1990):Foundthatclassesoflessthan20hada‘smallpositiveeffecton studentsthatdidnotpersistaftertheywereremovedfromthesmallerclass’(as above). •Hanushek(1998):‘Theevidenceaboutimprovementsinstudentachievementthat canbeattributedtosmallerclassesturnsouttobemeagreandunconvincing’ (Hanushek1998,citedinVinsonReport1,p.84). Oftheabovesixstudies,threeconcludethatthereisnolastingbenefittostu- dentsofreducingclasssize,twoconcludethatclassesmusthavelessthan20 studentstomakeadifferenceandonefoundthattheeffectofclasssizewas mediatedbyteachingstyle. Aswellasthesereviews,theVinsonReportdetailsthefindingsofthreemajor studiestheydescribeas‘trialprogrammesandlargefieldexperiments’–Project STAR,SAGEand‘Prime-Time’.Eachoftheseispresentedasproofpositivethat smallerclassesarebeneficialtostudents.Below,theVinsonReport’scomments willbesummarised,followedbyamoreaccuraterepresentationofthestudies’ findings. •ProjectSTAR(StudentTeacherAchievementRatio)inTennessee: AccordingtotheVinsonReport: Thisisthe‘mostscientificallyrigorous’and‘best-designedfieldexperimentever’ (VinsonReport1,p.82).Thefindingsreportedarethatthepositiveeffectsofsmall classes(13-17students)inK-3onachievementlevelsarecumulative(thelonger thetimespentinasmallclass,thelargertheeffect)andpersistent(theeffectlasts intolatergradeswhenstudentsreturntoregularsizeclasses).Italsoreportsthat gainsweregreaterfordisadvantagedstudents. Section 9-122 CLASSSIZEANDTEACHERQUALITY 77 TheVinsonReportacknowledgesthatthenon-randomself-selectionofschools intotheprojectmaybeaproblem,becausesuchschoolsmighthaveagreater interestandenthusiasmforsuchreforms,perhapsinflatingtheresults. MissingfromtheVinsonReport: ThesourceoftheVinsonReport’sinformationonProjectSTARisnotclear,but veryrecentanalysesoftheProjectSTARdatabyitsprincipalresearchersisless straightforward.Ina2001article,JeremyFinnandcolleaguesreportedthatthe gainsmadebysmallclassstudentsontheirregularclasspeersdeclinedwhen theyreturnedtoregularclasses,andthatsignificantenduringeffectsofclasssize occurredonlyforstudentswhohadbeeninasmallclassforthreeorfouryears. Therewasonlyweakandmixedevidenceofalargereffectforminorities(Finn, Gerber,Achilles,Boyd-Zaharias,2001). AnotherstudyfromprincipalresearchersonProjectSTARfoundthatclass- roompracticesdifferedbetweenthesmallclassesthatachievedthelargestand smallestgains(Boyd-Zaharias&Pate-Bain,2000).Thatis,smallclassbenefits weremediatedbythequalityandmethodofteaching. Althoughitmakesanodtoit,theVinsonReportdoesnotexplainthefullrami- ficationsofthefactthatProjectSTARsuffersfromthemethodologicalproblemof the‘HawthorneEffect’.Thisiswheretheparticipantsinanexperimentareaware oftheirroleandthepotentialconsequences.Hoxby(2000)explainsthatthiscauses threeproblems:First,incentiveconditionsarealtered,sothatresultsproduced underexperimentalconditionsmaynotnecessarilybetheresultsinreality.Sec- ond,somepeopletemporarilyincreasetheirproductivitywhilebeingevaluated, especiallyiftheyhaveaninterestintheexperimentsucceeding.Third,people sometimesundotherandomnessoftheexperimentduetoexternalpressures,for instancebyplacingcertainchildreninsmallclassesduetodemandsfromparents. ThemethodologicalproblemsofProjectSTARcannotbedismissedas‘crit- icisms’.TheycreateseriousdoubtoverwhethertheresultsachievedbyProject STARwouldbereplicatedunderdifferentconditions. Evenifthesedoubtscouldbesetaside,thefindingsareinconsistentwiththe recommendationsmadebytheVinsonInquiry.SmallclassesinProjectSTARare 13-17students.BarbaraNyeofTennesseeStateUniversity,whohasstudiedthe resultsindetail,hasbeenquotedassayingthat‘thepublicshouldn’tnecessarily expectthesameresultsfromclassesofaround20asthoseof15.It’stakenalong timetogetthatmessageacross’(Jacobson,2001).Itseemsthemessagestillhasa waytogo. NotonlydoestheVinsonReportrecommendthatclasssizesbereducedtoa numberthathasnotbeenshowntohaveanyeffect,italsorecommendsdoingthis inclassesfromKindergartenthroughtoYear2atthesametime.InProjectSTAR, enduringresultswereonlyfoundforstudentswhohadbeeninasmallclassfor threeorfouryears.Thissuggeststhatitwouldnotbeeffectivetoreduceclasses inallyearlevelsatonce,buttostaggerclasssizereduction,beginningwitha kindergartencohort. Section 9-123 78 JENNIFERBUCKINGHAM GiventhattheProjectSTARfindingscouldbeviewedasirrelevant,itmayseem futiletopointouthowtheyshouldcorrectlybeinterpreted.Butthefactthatthey havebeenreportedinaccuratelyandwithoutsufficientthoughttotheirimplications indicatestheincautiousapproachtakentothisresearchintheVinsonReport. •TheSAGE(StudentAchievementGuaranteeinEducation)inWisconsin: AccordingtotheVinsonReport: UndertheSAGEprogramme,K-3classeswerereducedtoanaverageof15in schoolswhereatleast50%ofstudentswerelivingbelowthepovertyline.Find- ingscitedarethosefroma1999studyshowingthat‘Year1studentsintheSAGE programachievedbettertestresultsthanstudentsincomparisonschoolsinlan- guage,artsandmaths.Resultsfromgradestwoandthreegenerallyfollowthe samepattern’.(VinsonReport1,p.83). MissingfromtheVinsonReport: MorerecentevidencepublishedbyMolnar,Smith,Zahorik,Halbach,Ehrle,Hoff- manandBeverleyCross(2001)confirmsthatstudentsinSAGEschoolsperformed significantlybetterthanstudentsincomparisonschoolsonavarietyofmeasures. Mostimportantly,however,thiscannotbeattributedtoreductionsinclasssize. SchoolsinvolvedintheSAGEprogrammeimplementedavarietyofreformsatthe sametime: 1.class-sizereduction 2.alongerschooldayandincreasedcollaborationwithcommunityorganisations 3.amorerigorousacademiccurriculum 4.staffdevelopmentandaccountabilitymechanisms. Inaddition,thesameteamofresearchersdiscoveredimportantdifferencesin teachingstylesbetweenSAGEandcomparisonschools.InstructioninSAGE schoolswaspredominantlyteacher-centredasopposedtostudent-centred.(Molnar etal.,2001).DifferenceswerealsoidentifiedbetweenclassroomswithinSAGE schools.Teachersinhigherachievingclassroomsshowedapreferenceforstruc- tured,goal-oriented,explicitinstructionandclassroomswithestablishedroutines wherelearningproceedssequentiallyandataquickpace.Teachersinlowerachiev- ingclassestendedtobelievethattheprimaryadvantagesofreduced-sizeclasses aretheopportunitiestodevelopcriticalthinking,topermitstudentstochoosetheir activitiesandtohavemoreactivitiesandproblem-solvinglessons.Theyalsohada morepermissivemanagementstyleandamorerandomlessonstructure. So,asinProjectSTAR,theaptitudeoftheclassroomteacheristhekey,notthe numberofchildren. •Prime-TimeprojectinIndiana: AccordingtotheVinsonReport: Theinitialresultsofatwoyearstudyin24schoolswhereclasseswerereduced toanaverageof18were‘sopromising’(VinsonReport1,p.83)thatK-3class sizeswerereducedinallIndianastateschools.Oneanalysisapparentlyfound ‘substantiallylargergainsinreadingandmathsachievementforstudentsinsmall classes’(McGivern,GilmanandTillitski(1989)citedinVinsonReport1,p.83). Section 9-124 CLASSSIZEANDTEACHERQUALITY 79 TheVinsonReportgivesamoreaccuraterepresentationofthevalueofthis studythanitdoesofSTARandSAGE.Itnotesthatthestudywasnotrandom,that otherchangesinschoolpolicyoccurredatthesametimeandraisesthepossibility thatteachersweremotivatedtoensurethatsmallclassesworked. MissingfromtheVinsonReport: Theextensionofclasssizereductionfromtheoriginal24schoolstoallschools occurredafteronlyoneyear.Evenreviewerswhofavourclasssizereductionhave admitteditwastherefore‘notpossibletocompareresultsforsmallclasseswith acomparablegroupoflargerclasses’(Biddle&Berliner,2002,p.6).Theresults citedintheVinsonReportwereactuallyfromastudyofdatacollectedbefore projectPrimeTimewasinitiated. Severalotherlargescalestudieshavebeenconducted,theresultsofwhichare notpresentedintheVinsonReport.Theyaresummarisedbrieflybelow. •CaliforniaClassSizeReductionInitiative: InspiredbyProjectSTAR,K-3classsizesinallCalifornianschoolswerereduced fromamaximumof33(average29)toamaximumof20.Tomeetthisrequirement, schoolswereforcedtohireunderqualifiedteachers. TheClass-SizeReduction(CSR)ResearchConsortiumconcludedonthebasis offouryearsofdataanalysisthat‘nostrongrelationshipcanbeinferredbetween achievementandCSR’(Stecher&Bohrnstedt,2002,p.2).Jepsen&Rivkin(2002) foundthatthelargenumberofextrateachersdemandedbyCSRledtoadeterio- rationinteacherqualitywhichinsomecasesfullyoffsetanybenefitsofsmaller classes. •Hoxby’s(2000)PopulationVariationStudyinConnecticut: Inthisstudy,Hoxby(2000)lookedattherelationshipbetweenachievementand changesinclasssizeduetonaturalvariationinagecohortsinthepopulation. Thisobservationalapproachavoidspossibleexperimentalmanipulationeffects. Sheusestwodifferentmethodstocomparetheclasssizeandachievementof adjacentcohorts,takingintoaccountenrolmentdataandmaximumclasssizereg- ulations. Neithermethodshowsthatsmallerclassesproducedachievementgains.Even giventheprecisionofthedataanalysis,whichallowedtinyimprovementstobesig- nificantatthe5%level(theimprovementsfoundinProjectSTARwouldhavebeen significantiffoundinthisstudy),theeffectofreducingclasssizewasestimated tobeclosetozero.Further,theresultsdonotsuggestthatclasssizereductionsare moreeffectiveinschoolsthatservelow-incomeorAfricanAmericanstudents(in fact,theonlysignificantresultwasanimprovementinfourthgradereadingscores ofhigh-incomestudents). •ChristchurchHealthandDevelopmentSurvey: AlongitudinalstudyconductedinNewZealand,althoughnotdesignedtostudy classsizeeffects,hasyieldedinformationthatcanbeusedasanobservational study. Section 9-125 80 JENNIFERBUCKINGHAM BoozerandMaloney(2001)firstcomparedtheresultsofchildrenpermanently insmall(19),medium(29.9)andlarge(33.8)classesbetweentheagesof8to13 years.Onlyasmallnumberofstudentswerepermanentlyinclassesofthesesizes overtheageperiod,andtheresultswereinsignificant.Theythencomparedstudents whose average classsizeoverthisageperiodwassmall(21.2),medium(29.7)or large(33.2).Theyfoundsignificanteffectsonlyforchildrenin persistently smaller averageclassesbetweentheagesof8and13,onbothchildhoodtestscoreim- provementsaswellasonearlyadultoutcomessuchascompletededucationand unemployment. •UKNationalChildDevelopmentStudy Inanotherobservationalstudyofexistingdatafromthe1960s,Iacovou(2001) controlledforschooltype/sizeandstreamingtoaccountforthepossibility(and someevidence)thatlessablechildrenaremorelikelytobeallocatedtoasmaller class–whichwouldmakethedifferenceinachievementindifferentsizeclasses internallycreated. Iacovoulookedataverageclasssizeatage7(excludingstudentsinclassesof lessthan20andmorethan45)andfoundthatclasssizewasrelatedtostudent attainmentinreadingbutnotmaths.Asmallereffectpersistedtoage11onlyfor girlsandforchildrenfromlargefamilies.Therewasnoevidenceofgreaterbenefit todisadvantagedstudents. •ThirdInternationalMathsandScienceSurvey(TIMSS) Classsizeeffectsfor18countrieswereestimatedusingmathsandscienceperfor- manceinTIMSSandaverageclasssizedata.WoessmannandWest(2002)found thatclasssizeeffectsvariedgreatlybetweencountries,withlargeeffectsinonly twocountries:GreeceandIceland. Whentheycomparedthesecountrieswiththosewherenoclasssizeeffectwas found,severalthingswereapparent.First,countrieswithlargeclasssizeeffects performedbelowaverageinternationally,whereasthosewithsmallornoclasssize effectsperformedaboveaverageinternationally.Also,countrieswithlargeclass sizeeffectshadlesseducated,lowerpaidteacherscomparedtocountrieswithsmall ornoclasssizeeffects. Fromthistheydrewseveralconclusions.First,classsizeeffectscannotbe imputedfromonecountrytoanotherbecauseschoolsystemsvarysignificantly. Second,classsizeismoreimportantwhenteachersarelesseffective.Investment infewer,morehighlyeducatedandbetterpaidteachersseemstoresultinhigher studentachievement. AustralianResearch Australianresearchonclasssizesisscarce.AstudybyBourke(1986)inMel- bourneinthe1980sfoundthatsmallerclasseswererelatedtohigherachievement inmaths,butKeeves(1995)hasnotedthatanalysisoftheseresultsattheclass levelrevealedthatclasssizewasalsorelatedtostudentability(sorting)andthat Section 9-126 CLASSSIZEANDTEACHERQUALITY 81 controllingforthischangedtherelationshipbetweenclasssizeandachievement. Keevesconcludesthat‘thereislittleclearevidencetosupportthecostlyreductions inclasssize’(Keeves,1995,p.148). ResearchconductedinBrisbanebyJackCambellisoftencitedinsupportof smallerclasses.PublishedinamagazineoftheQueenslandTeachersUnionin 1991,thisstudyisdifficulttoobtain.Secondarysourcesdescribeitasfindingthat reducingclassesfrom35to26studentsincreasedthe‘timeontask’by22days perschoolyear(AustralianEducationUnion,1995).Whetherthisstudycontrolled forthesortingfactorwhichcausedproblemsinBourke’sstudy,andhowincreased timeontaskmighthavetranslatedintoincreasedstudentachievementisnotknown. TheanalysisofTIMSSresultsdescribedabovedidnotleadtoanymeaningful findingsforAustralia.TheresearchersfoundthataverageAustralianclasssizes inmathsandsciencewerenotgoodproxiesforactualclasssizes,sodifferences instudentachievementbetweenclassesofdifferentsizescouldnotbeconfidently attributedtothesizeoftheclass. Implications TheVinsonReportestimatesthatthereductionofclasssizestoamaximumof20in YearsK-2wouldcost$47milliondollarsperannumindisadvantagedschoolsand $225millionperannuminallschools.Thisisthemostexpensiverecommendation made,theallschoolsfigureof$225millionexceedingthetotalcostofallother recommendationsby40%. Eventhisfigureunderestimatesthecostofclasssizereductionasitaccounts onlyforextrastaffingcosts.Eachadditionalteachernecessitatesanadditional classroom,mustbeeducatedandtrained,willneedextraclassroomresourcesand requireon-goingprofessionaldevelopment.Thecostofmoreclassroomshasbeen conservativelyestimatedbytheNSWOppositiontobeintheorderof$140million initially(LiberalPartyofNewSouthWales,notdated). Notonlyisthecostlarge,butthefindingsofthestudiesdescribedaboveare mixedandweakatbestontheissueofclasssize.Onlyonethingcomesthrough loudandclearfromtheresearch:whatgoesonintheclassroomismoreimportant thanhowmanychildrenareinvolved.Thisisnottosaythatclassroomactivity isunaffectedbythenumberofchildren,butthatprovenandappropriateteaching methodsareparamount. WhatthendoestheVinsonReportmakeofthis?Itrecommendsthatlarge scaleclasssizereductiontakesplaceinstateschools,bringingclasssizesinK- 2toamaximumof20.Thereportsaysthatithasbeenguided‘notonlybythe consistencyofthefindings,butalsothequalityoftheresearchyieldingparticular results’(VinsonReport1,p.84). MuchoftheVinsonReport’sinformationonclasssizeresearchcomesfroma shortliteraturereviewbyBiddleandBerliner(2002),includingtheirconclusions, whicharereproducedverbatim.YetBiddleandBerlinerseemjustasconfusedas Section 9-127 82 JENNIFERBUCKINGHAM theauthorsoftheVinsonReport,claimingthat‘Althoughtheresultsofindividual studiesarealwaysquestionable,ahostofdifferentstudies...suggestanumber ofgeneralconclusions’(p.14),namelythatclasssizereductionisbeneficialfor studentsintheshortandlongterminacademicachievementandotheroutcomes. Inotherwords,theseauthorsseemtobesayingthatalargenumberofpoorly designedstudieswithmediocreresultscanbeamassedintostrongevidenceofa significanteffect. EvenlessconvincingistheVinsonReport’sattempttojustifytheirrecom- mendationinthefaceoftheevidencetheyhavepresentedtothecontrary.They arguethatpolicymakersshouldnot‘awaitanunlikelytotalconsensus...butto basepolicyonthebestavailableinformation,afterconsideringthestrengthsand limitationsoftheresearch’(VinsonReport1,p.81).Completeagreementfrom researchersmaybetoomuchtoask,butiftheReport’sauthorsfollowtheirown adviceandseriouslyconsidertheevidencepresented,notwithstandingtheevidence theyneglected,theywouldhavetoconcludethatthebestavailableinformation isthatreducingclasssizebytheamounttheyrecommendwouldnotjustifythe expense. TheoriesandFallaciesofClassSizeReduction Thereareseveraltheoriesastowhysmallerclassesshouldbebeneficial: 1.Increasedindividualattentionandinstruction; 2.Greaterscopeforinnovationandstudent-centredteaching; 3.Increasedteachermorale; 4.Fewerdisruptions. Theideathatateachercandevotemoretimetoeachstudentinasmallerclass, therebyincreasingtheamountstudentslearn,isthemostintuitivelyappealingof allthesetheories.Yetsimplecalculationsshowthisappealtobemisplaced. Inasixhourschoolday,approximatelyfivehoursisspentintheclassroom. Ifhalfthistimeisspentdirectlyaddressingtheclass,andtheotherhalfonindi- vidualattention,eachchildwouldhypotheticallyreceivesixminutesofindividual instructioninaclassof25and7.5minutesofindividualinstructioninaclassof 20.Thatis,anextra$1150perstudentperannum(VinsonReport,p.86)buysan extra1.5minutesperdayofteacher’stime.Iftwo-thirdsofclassroomtimeisspent onindividualattention,studentsgettwominutesmoreinaclassof20than25. Thesecalculationsmaybesimplistic,butindicatetheinsubstantialchangein individualattentionthata20%reductioninclasssizebrings,atconsiderablecost. AnothercountertotheindividualinstructiontheorycomesfromProjectSTAR. Someoftheregularsizeclasseswereassignedateacher’saide.Eventhoughchil- drenintheseclassespresumablyhadtwiceasmuchindividualattention,therewas nodifferenceinachievementlevelsbetweenregularsizeclasseswithandwithout teacher’saides. Section 9-128 CLASSSIZEANDTEACHERQUALITY 83 Thesecondtheory–thatsmallclassesprovidethepotentialformoreeffective teachingstrategies–suggeststhatclasssizemaybeconducivetogreaterstudent achievementbutdoesnotguaranteeit.Italsosuggeststhatsmallclassesalone donotproducegainsinlearning;thattheirbenefitsaremediatedbyteacherqual- ity.Researchdiscussedearlierdemonstratesthattherewerenotabledifferencesin teachingandclassroommanagementstylesbetweenhighandlowachievingsmall classes. Teachersrarelychangetheirteachingandclassroommanagementstyles.Even ProjectSTARdatashowsthis,withfewteachersmodifyingtheirclassroomprac- ticesindifferentsizeclassesafterattendingaprofessionaldevelopmentprogram (Ehrenbergetal.,2001b).Ifthisisthecase,thenreducingclasssizewillhave littleornoeffectwithoutensuringthatteachersadoptinstructionandmanagement practicesproventobeeffectiveinsmallclasses.Thissubstantialinvestmentin professionaldevelopmentonceagainaddstothecostofclasssizereduction,and wouldmorethanlikelybeequallyeffectivewithoutchangingclasssizes. Thelasttwotheoriesofsmallclassbenefitsarerelatedandarethemostconvinc- ing.Smallclassesareoverwhelminglypopularwithclassroomteachersanditisnot difficulttounderstandwhy.Schoolsarebeingforcedtocopewith,andattemptto educate,anincreasingnumberofstudentswhoareuninterestedandbadlybehaved. InsomeareasofSydney,schoolshavedifficultyattractingandretainingteachers primarilyforthisreasonandteachersinallareasarefindingtheirjobsmoreand moredifficultandstressful. Fewerstudentslikethisinaclasswouldmaketeachingmucheasier.Reducing classsizesmightbejustifiableifitcanbeshownthattheincreasedcostofreducing classsizeisoffsetbythedecreasedcostofteacherattrition,stressandsickleave. Itmustbeensured,however,thatanewdemandforteachersdoesnotresultin thesamesituationasinCalifornia,wheretheleastqualifiedandleastexperienced teachersweredisproportionatelyemployedinthemostdisadvantagedschools.The mostsimpleandeffectivewaytoavoidthisistoofferfinancialincentivesforteach- ersindifficult-to-staffschools,whichmeansdepartingfromrigidwagestructures basedonyearsofservice. Teacherquality Commonsensesaysthatitisbettertohaveagreatteacherinfrontofalargeclass thanamediocreteacherinfrontofasmallone. WritingintheBulletinoftheUSNationalAssociationofSecondarySchool Principals,KaplanandOwings(2002)statethat‘Researchaffirmsthatteaching qualityisthesinglemostimportantfactorinfluencingstudentachievement’,and citeawidevarietyofsupportingstudies.AccordingtoRonaldFerguson,aHarvard Universityeconomist,researchshowsthatteacherquality,notclasssize,isthemost importantfactorineducation(Matthews&Strauss,1997).Australianresearchhas Section 9-129 84 JENNIFERBUCKINGHAM alsoshownthatthelargestdifferencesinachievementbetweenstudentsisthat betweenstudentsindifferentclasses(Rowe,2002). The‘RamsayReport’ontheReviewofTeacherEducationinNSW(Ram- say,2000),providesplentyofevidencetosupporttheprimacyofteacherquality, demonstratingtheimpactofteachersonstudentachievementandthebenefitsfrom investinginteachereducation. Althoughmuchhasbeensaidabouttheimportanceofteacherquality,what makesagoodteacherisyettobeadequatelydefined.Weknowthatsometeach- ersbringabouthigherlevelsofachievementfromtheirstudentsthanothers,but consensusonhowisstillelusive. Acertainproportionofgoodteachingcomprisestemperament,charisma,en- thusiasmandotherqualitiesthatcannotbemeasuredortaught.However,several criteriacanbeidentified: 1.masteryofsubjectmatterandcurriculumcontent; 2.awarenessoftheindividualabilitiesandcapabilitiesofstudents; 3.classroommanagementskills; 4.useofteachingstrategiesthatareproveneffective; 5.goodverbalcommunicationskills. Eachofthesecapacitiesisnecessarybutinsufficientonitsown.Strongcontent knowledgeiscrucialbutnotenough–teachingalsorequiresasetofprofessional skillsseparatefrombutrelatedtothesubjectbeingtaught(Darling-Hammond, 2000;Haycock,1998;Goldsmith,2002).Theseskillsaresupposedtobegained fromteachereducationcourses. Whatconstituteseffectivepedagogyisbeyondthescopeofthispaper,but thereseemstobeagreementthatteachereducationinAustralianuniversitiesis inadequateinimpartingbothpedagologicalandbehaviourmanagementskillsto teachers.Thereistoomuchemphasisonthetheoreticaloverthepractical–too muchBloomandnotenoughclassroom.Newteachershaveusuallyspentonly afewweeksinteachingpracticum,andsupportforthemintheextremelydiffi- cultfirstyearinaschoolispatentlyinadequate(Ramsay,2000;VinsonReport3, Chapter11). Anotherproblemisthelackofongoingprofessionaldevelopmentforclassroom teachers.TheNSWDepartmentofEducationundervaluestheneedforteachers tobeawareofnewdevelopmentsinbothcurriculumandpedagogy,andteachers havetoofewincentivestoseekoutprofessionaldevelopmentopportunitiesfor themselves. Improvingthequalityandeffectivenessoftheteachingforceasabodywillnot beachievablethroughbetterpre-serviceandin-servicetrainingalone.Someteach- erswillbeunaffectedbyanyamountofprofessionaldevelopment.Improvingthe teacherforceinvolvesbothenhancingtheskillsofwillingteachersandremoving incompetentandunwillingteachers. Thisisbestachievedbyallowingschoolstohireandfire.Thecentralized staffingofpublicschoolsinNSWisoneoftheirgreatestimpedimentstosuccess. Section 9-130 CLASSSIZEANDTEACHERQUALITY 85 Giventhatteachersarethemostimportantinfluenceoneducationalachievement, theinabilityofpublicschools,whetherthroughprincipalsorschoolboards,to ‘choosetheirteam’,putsthematgreatdisadvantage. Conclusions Whenitcomestoteachers,qualityisfarmoreimportantthanquantity.Therecom- mendationsonclasssizereductionserveonlytoweakenthecaseformoreurgent andsupportableinterventions,suchasimprovedteachereducationandprofessional development. GiventhatgoodAustraliandataonclasssizeeffectsisnon-existent,andthat researchfromothercountriesisinconclusiveonwhetherthereareevenmarginal benefitsfromclasssizereduction,itisprudentthatgovernmentsseekmoreevi- dencebeforeembarkingonwhatwilleventuallybeamultibilliondollarspending spree. Ultimately,however,decisionsaboutclasssizearebestlefttoschools.Given theopportunitytousetheirfundingallocationastheyseefit,someschoolsmight decidetohaveslightlylargerclasseswithbetterqualifiedteachers,orinvestina ‘floating’teachertrainedinspecialneedseducation.Mandatorymaximumclass sizessetatanarbitraryfigureareyetanotherunnecessaryrestrictiononschools’ abilitytousetheirresourcesinwaysthatbestmeettheneedsoftheirstudents. References AustralianEducationUnion(1995).Classsizesdomatter.FactSheetNo.1 [On-line].Available: http://www.aeufederal.org.au/Publications/FactSheet1ClassSize.pdf Biddle,B.J.,&Berliner,D.C.(2002).Whatresearchsaysaboutsmallclassesandtheireffects. PolicyPerspectives [On-line].Available:http://www.WestEd.org/online_pubs/small_classes.pdf Boozer,M.A.,&Maloney,T.(2001).TheEffectsofClassSizeontheLong-RunGrowthinReading AbilitiesandEarlyAdultOutcomesintheChristchurchHealthandDevelopmentStudy,Working Paper01/14.Wellington:NewZealandTreasury. Bourke,S.F.(1986).Howsmallisbetter:Somerelationshipsbetweenclass-size,teachingpractices, andstudentachievement.AmericanEducationalResearchJournal,23,558–571. Boyd-Zaharias,J.,&Pate-Bain,H.(2000).EarlyandnewfindingsfromTennessee’sProjectSTAR’. InM.C.Wang&J.D.Finn(Eds.),HowSmallClassesHelpTeachersDoTheirBest (pp.65–97). Philadelphia:TempleUniversityCenterforResearchinHumanDevelopmentandEducationand theU.S.DepartmentofEducation. Darling-Hammond,L.(2000).Teacherqualityandstudentachievement:Areviewof statepolicyevidence.EducationPolicyAnalysisArchives [On-line],8(1).Available: http://epaa.asu.edu/epaa/v8n1 Ehrenberg,R.G.,Brewer,D.J.,Gamoran,A.,&Willms,J.D.(2001a).Doesclasssizematter? ScientificAmerican,285,78–85. Ehrenberg,R.G.,Brewer,D.J.,Gamoran,A.,&Willms,J.D.(2001b),Classsizeandstudent achievement,PsychologicalScienceinthePublicInterest,2,1–30. Section 9-131 86 JENNIFERBUCKINGHAM Finn,J.,Gerber,S.B.,Achilles,C.M.,&Boyd-Zaharias,J.(2001).Theenduringeffectsofsmall classes,TeachersCollegeRecord,103,145–183. Goldsmith,S.S.(2002).Thepedagogyofthesubjectandprofessionaldevelopment.In AConsumer’s GuidetoTeacherQuality:OpportunityandChallengeintheNoChildLeftBehindActof2001. WashingtonD.C.:NationalCouncilonTeacherQuality. Hanushek,E.A.(1997).Assessingtheeffectsofschoolresourcesonstudentperformance:An update.EducationalEvaluationandPolicyAnalysis,19,141–64. Hanushek,E.A.(1998).TheEvidenceonClassSize,OccasionalPaper98-1.Rochester,NY:W. AllenWallisInstituteofPoliticalEconomy,UniversityofRochester. Haycock,K.(1998).Goodteachingmatters.ThinkingK-16,3(2). Hoxby,C.M.(2000).Theeffectsofclasssizeonstudentachievement:Newevidencefrom populationvariation.TheQuarterlyJournalofEconomics,115,1239–1285. Iacovou,M.(2001).ClassSizeintheEarlyYears:IsSmallerReallyBetter?WorkingPaper 2001-10 [On-line].InstituteforSocialandEconomic Research,EssexUniversity.Available: http://www.iser.essex.ac.uk/pubs/workpaps/pdf/2001-10.pdf Jacobson,L.(2001).Research:Sizingupsmallclasses.EducationWeek [On-line],February28, 2001.Available:http://www.edweek.org. Jepsen,C.,&Rivken,S.(2002)Whatisthetradeoffbetweensmallerclassesandteacherquality? WorkingPaper9205.Cambridge,MA:NationalBureauofEconomicResearch. Kaplan,L.S.,&Owings,W.A.(2002).Thepoliticsofteacherquality:Implicationsforprincipals. NationalAssociationofSecondarySchoolPrincipalsBulletin,86(633),22-41 Keeves,J.P.(1995).ThecontributionofIEAresearchtoAustralianeducation.InW.Bos&R. H.Lehmann(Eds.),ReflectionsonEducationalAchievement:PapersinHonourofT.Neville Postlethwaite.NewYork:Waxmann. Krueger,A.B.(2002).Understandingthemagnitudeandeffectofclasssizeonstudentachievement. InL.Mishel&R.Rothstein(Eds.),TheClassSizeDebate.Washington,D.C.:EconomicPolicy Institute. LiberalPartyofNewSouthWales.(notdated).GettingTheBestStart:LoweringK-2ClassSizes. [On-line].Available:www.barryofarrell.com/misc/Getting_the_Best_Start_policy.pdf. Mathews,J.andStrauss,V.(1997).Shouldclassesbesmaller?Asenrolmentrises,issuedivides educators,WashingtonPost,Monday,December15,1997. Molnar,A.,Smith,P.,Zahorik,J.,Halbach,A.,Ehrle,K.,HoffmanL.M.,&Cross,B.(2001).2000– 2001EvaluationResultsoftheStudentAchievementGuaranteeinEducation(SAGE)Program. CentreforEducationResearch,AnalysisandEvaluation,UniversityofWisconsin-Milwaukee. Ramsay,G.(2000).QualityMatters:RevitalisingTeaching:CriticalTimes,CriticalChoices’,Report oftheReviewofTeacherEducation.Sydney:NSWDepartmentofEducation&Training. Rowe,K.(2001).TheImportanceofTeacherQuality.IssueAnalysis22.Sydney:TheCentrefor IndependentStudies. Stecher,B.M.,&Bohrnstedt,G.W.(Eds)(2002).ClassSizeReductioninCalifornia:Summary ofFindingsfrom1999–2000and2000–0.CSRResearchConsortium,CaliforniaDepartmentof Education. Woessmann,L.,&West,M.R.(2002).ClassSizeEffectsinSchoolSystemsAroundthe World:EvidencefromBetween-GradeVariationinTIMSS,ResearchPaperPEPG/02-02 [On-line],ProgramonEducationalPolicyandGovernance,HarvardUniversity.Available: http://www.ksg.harvard.edu/pepg/pdf/PEPG02-02.pdf) Section 9-132 Class Size Priorities for Changing NCLB: A federal class size reduction program is an NEA priority in rewriting NCLB. Didn't we have a class size reduction program a few years ago? Yes. The class size reduction program, which provided $4.1 billion used to hire some 37,000 teachers to reduce class size, was eliminated under NCLB. NEA supports restoring the class size reduction program. NEA's goal is to win funds specifically for class size reduction. Does research support what teachers know- that class size has a direct impact on student achievement? Yes. The research shows that learning increases as class size is reduced, especially in the early grades. NEA considers 15 students to be the optimum class size, especially in kindergarten (K) and first grade. Researchers have documented benefits from class sizes of 15-18 students in K and of fewer than 20 students in grades 1-3. Studies show that students in smaller classes continue to reap academic benefits through middle and high school, especially minority and low- income students. Does NEA support smaller classes in the upper grades as well as the primary grades? Yes. Even in the upper grades, teachers can be more successful in increasing student learning when they can provide more individualized attention. Closing the achievement gaps requires opportunities to work with students who need greater assistance. Does NEA have a specific class size reduction target? NEA recommends an optimum class size of 15 students in regular programs, especially in the early grades, and a proportionately lower number in programs for students with exceptional needs, including children with disabilities and English language learners. What about space to accommodate smaller classes? NEA has taken space needs into account and supports a combination of federal programs-both grants and tax subsidies to states and school districts -- for school modernization to accommodate smaller classes, as well as allowing for two teachers in one classroom to reduce the student-teacher ratio. Section 9-133 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality A P l a n t o G r o w O n 20 0 4 - 2 0 1 0 Th e Ex c e l l e n c e Factor Th e Ex c e l l e n c e Factor Se c t i o n 10 - 1 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Co n t r a s t e d Th e Na t i o n a l Ed u c a t i o n Ag e n c y ra n k s Te x a s 32 n d Th e Na t i o n a l Ed u c a t i o n Ag e n c y ra n k s Te x a s 32 n d am o n g th e 50 st a t e s in cu r r e n t ex p e n d i t u r e s pe r pu p i l in 20 0 2 ‐03 . Av e r a g e sp e n d i n g pe r pu p i l th t i $7 8 2 9 $6 7 7 95 t ac r o s s th e na ti on wa s $7 ,82 9 —$6 7 7 or 9.5 pe r c e n t hi g h e r th a n in Te x a s . Th e hi g h e s t sp e n d i n g st a t e , Co n n e c t i c u t , spen t $1 1 , 3 7 8 per pupil ; th e lo w e s t p p pp sp e n d i n g st a t e , No r t h Da k o t a , sp e n t on l y $4 , 7 7 3 pe r pu p i l . C f Pb l i Pl i Pi i i ‐‐ Cen t e r for Publ i c Poli cy Prior ities 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 2 ht t p : / / w w w . c p p p . o r g / p r o d u c t s / p o l i c y a na l y s i s / b r f - f a s t - f a c t s - p u b - e d . h t m l Se c t i o n 10 - 2 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Ch a l l e n g e s Ch a l l e n ges fo r Ch a r t e r Sc h o o l s in Te x a s g On l y re c e i v e av e r a g e of 68 % fu n d i n g in co m p a r i s o n to other pu b l i c sc h o o l s Do no t re c e i v e de b t po r t i o n o f pr o p e r t y ta x e s Do no t re c e i v e de b t po r t i o n of pr o p e r t y ta x e s Ma y no t is s u e de b t fo r fa c i l i t i e s Ma y no t ch a r g e a tu i t i o n Mu s t ad m i t ba s e d on lo t t e r y ou t s i d e Pr i m a r y Bo u n d a r y Mu s t ad h e r e to TE A gu i d e l i n e s an d re p o r t i n g re q u i r e m e n t s 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 3 Se c t i o n 10 - 3 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Wh e r e we st a n d to d a y Sc h o o l pr o p e r t y ta x e s as s e s s e d ag a i n s t We s t l a k e Re s i d e n t s an d Bu s i n e s s e s in 20 0 4 7% 46 % $3 , 7 8 3 , 8 9 7 7% $5 4 2 , 6 6 9 47 %% $3 , 9 4 6 , 4 4 0 Ke l l e r IS D Ca r r o l IS D No r t h w e s t IS D 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 4 Se c t i o n 10 - 4 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r We s t l a k e Ac a d e m y re c e i v e s an am o u n t e q u i v a l e n t Th e Ex c e l l e n c e Fa c t o r Wh e r e we st a n d to d a y We s t l a k e Ac a d e m y re c e i v e s an am o u n t eq u i v a l e n t to 16 % of We s t l a k e ta x e s pa i d to ar e a sc h o o l Di s t r i c t s $8 2 7 3 0 0 6 wi l l be pa i d by W e s t l a k e re s i d e n t s an d bu s i n e s s e s to $8 ,27 3 ,00 6 wi l l be pa i d by We s t l a k e re s i d e n t s an d bu s i n e s s e s to ar e a sc h o o l di s t r i c t s in fi s c a l ye a r of 20 0 4 ‐20 0 5 . $1 , 3 2 2 , 7 4 0 es t i m a t e d to be re c e i v e d by We s t l a k e Ac a d e m y from the st a t e of Te x a s in fi s c a l ye a r 20 0 4 ‐20 0 5 . 16 % $1 , 3 2 2 , 7 4 0 84 % 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 5 $8 , 2 7 3 , 0 0 6 Se c t i o n 10 - 5 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Wh e r e we st a n d to d a y Th e WA wi l l re c e i v e $4 , 9 5 4 pe r st u d e n t fr o m th e state in 20 0 4 ‐20 0 5 to op e r a t e th e sc h o o l , an d $9 4 0 pe r st u d e n t in fe d e r a l st a r t u p fu n d s wh i c h en d th i s ye a r . WA to t a l = $5 8 9 4 pe r st u d e n t WA to t a l = $5 ,89 4 pe r st u d e n t . Na t i o n a l av e r a g e sp e n t pe r st u d e n t is $7 , 8 6 0 . To w n pa y s al l de b t se r v i c e on fa c i l i t y ($ 1 4 mi l l i o n / y r ) To w n pa y s al l de b t se r v i c e on fa c i l i t y ($ 1 .4 mi l l i o n / y r ) . Th i s wa s a re s u l t of a tw o ye a r ef f o r t to ac h i e v e th e 1st mu n i c i p a l i t y granted Ch a r t e r in Te x a s hi s t o r y in 20 0 1 . Bo a r d di r e c t i e is to ao i d op e r a t i o n a l su p p l e m e n t to Bo a r d di r e c t i ve is to avoi d op e r a t i o n a l su p p l e m e n t to sc h o o l op e r a t i o n s wh i c h wo u l d re s u l t in pr o p e r t y tax. 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 6 Se c t i o n 10 - 6 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Co m p a r e d St a t e / F e d e r a l Ai d to Ar e a Sc h o o l Di s t r i c t s $7 2 7 7 St a t e / F e d e r a l Ai d to Ar e a Sc h o o l Di s t r i c t s (o n a pe r st u d e n t ba s i s ) $6 , 4 8 6 $6 , 6 0 5 $5 , 8 2 2 $6 , 5 5 7 $7 ,27 7 $6 , 0 1 7 $4,954 $6 , 0 0 0 $8 , 0 0 0 $2 , 0 0 0 $4 , 0 0 0 $0 G C I S D A r g y l e K e l l e r C a r r o l l N o r t h w e s t F o r t Wo r t h W e s t l a k e 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 7 NI f e d s t a r t u p f u n d s *T E A Bu d g e t Re p o r t s 20 0 3 ‐20 0 4 Se c t i o n 10 - 7 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Ho w di d we su r v i v e th e fi r s t ye a r ? /h l ff dl Th e Ex c e l l e n c e Fa c t o r Co m p a r e d To w n /Sc hoo l eff ic i e n c y mo del Pu r c h a s i n g Pe r s o n n e l Ac c o u n t i n gg In f o r m a t i o n Te c h n o l o g y Ma i n t e n a n c e Ad m i n i s t r a t i o n Re c r e a t i o n Se c u r i t y Fe d e r a l Fu n d i n g $3 7 5 , 0 0 0 in st a r t u p fu n d i n g wh i c h be g a n in FY 02 ‐03 an d en d s in 04‐05 No MY P (M i d d l e Ye a r s Pr o g r a m m i n g ) co s t s No MY P (M i d d l e Ye a r s Pr o g r a m m i n g ) co s t s Te a c h e r s pa i d be l o w co m p a r a b l e po s i t i o n s in ot h e r di s t r i c t s (8%‐ 14 % ) 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 8 Se c t i o n 10 - 8 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Qu a n t i f i e d We s t l a k e Ac a d e m y We s t l a k e Ac a d e m y Bu d g e t e d Ex p e n d i t u r e s vs . St a t e Fu n d i n g $3 ,40 0 ,00 0 TotalState $2 , 4 1 4 , 9 1 7 $2 , 5 8 3 , 3 5 7 $2 , 8 1 2 , 6 8 3 $2 , 4 0 0 , 0 0 0 $2 , 9 0 0 , 0 0 0 ,, Total State Funding Budgeted Expenditures $2 , 1 1 8 , 4 9 4 $1 , 9 6 1 , 2 8 9 $1 , 8 0 4 , 0 8 3 $1 , 6 4 6 , 8 7 8 $ $2 , 0 0 9 , 1 2 1 $2 , 1 8 3 , 1 1 9 $1 , 8 2 1 , 2 3 9 $1 , 2 9 4 , 3 1 7 $1 4 0 0 0 0 0 $1 , 9 0 0 , 0 0 0 $2 , 4 0 0 , 0 0 0 $1, 4 8 9 , 6 7 2 $1 , 3 2 2 , 7 4 0 $9 3 7 , 6 7 7 $9 0 0 , 0 0 0 $1 ,40 0 ,00 0 20 0 3 ‐ʹ 04 2 0 0 4 ‐ʹ 05 2 0 0 5 ‐ʹ 06 2 0 0 6 ‐ʹ 07 2 0 0 7 ‐ʹ 08 2 0 0 8 ‐ʹ 09 2 0 0 9 ‐ʹ 10 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 9 Se c t i o n 10 - 9 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Qu a n t i f i e d An n u a l Sh o r t f a l l Be t w e e n St a t e Fu n d i n g an d Ex p e n s e s per St u d e n t $8 0 0 0 $694,189 $6 2 2 , 0 6 8 $6 1 0 , 8 3 4 $5 3 6 , 2 4 1 $5 1 9 , 4 4 9 $4 9 8 , 4 9 9 $4 , 0 0 0 $6 , 0 0 0 $8 ,00 0 $1,626 $1 , 8 6 7 $1 , 5 7 5 $1 , 6 8 3 $1 , 6 2 0 $1 , 7 3 7 $0 $2 , 0 0 0 $4 , 0 0 0 20 0 4 ‐ʹ 05 2 0 0 5 ‐ʹ 06 2 0 0 6 ‐ʹ 07 2 0 0 7 ‐ʹ 08 2 0 0 8 ‐ʹ 09 2 0 0 9 ‐ʹ10 St a t e F u n d i n g p e r S t u d e n t 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 10 Ex p e n s e p e r S t u d e n t Sh o r t f a l l p e r S t u d e n t $ $ $ = T o t a l D i f f e r e n t i a l p e r y e a r Se c t i o n 10 - 1 0 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Il l u s t r a t e d In t e r n a t i o n a l Ba c c a l a u r e a t e Cu r r i c u l u m Th r e e IB te a c h e r s fr o m ab r o a d (f o u r in c l u d i n g Mr s Br i z u e l a ) Th r e e IB te a c h e r s fr o m ab r o a d (f o u r in c l u d i n g Mr s . Br i z u e l a ) Al l cl a s s r o o m te a c h e r s IB tr a i n e d Fi r s t IB Co n s u l t a t i o n to be c o m e IB Ce r t i f i e d in ‘0 4 No de b t To w n of We s t l a k e fu r n i s h e s fa c i l i t y wi t h no pr o p e r t y ta x – eq u a t e s to a co n t r i b u t i o n of ov e r $5 , 2 0 0 pe r st u d e n t pe r ye a r by th e To w n of We s t l a k e . Ma n d a t o r y St r i n g s an d Mu s i c Pr o g r a m Ea c h ch i l d 1st th r o u gh 4th pla ys vi o l i n , vi o l a , ce l l o or ba s s g py Fo r e i g n la n g u a g e re q u i r e d ki n d e r g a r t e n & up Te a c h th e cu r r i c u l u m No “t e a c h i n g th e te s t ” N il d f hl ii i New ex t r a c u r r icu lar pr o g r a m s an d afte r sc hoo l ac t ivities Ma t h Cl u b , Fe n c i n g , Ch e s s , De b a t e , Pa i n t i n g , Po t t e r y , Ph o t o g r a p h y Co m p e t i t i v e sp o r t s of f e r i n g Vo l l e y b a l l , Cr o s s ‐Co u n t r y , Ba s k e t b a l l , So f t b a l l , G o l f , Te n n i s , So c c e r 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 11 Vo l l e y b a l l , Cr o s s Co u n t r y , Ba s k e t b a l l , So f t b a l l , Go l f , Te n n i s , So c c e r Se c t i o n 10 - 1 1 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Il l u s t r a t e d St u d e n t Te a c h e r Ra t i o d fl St u den t to fac u lty ra t i o = 9.3:1 26 7 st u d e n t s ÷2 4 cl a s s ro o m te a c h e r s = 11:1 WA av e r a g e cl a s s si z e = 16 WA cl a s s r o o m ma x i m u m ra t i o = 19:1 *O p t i m u m cl a s s si z e = 15:1 Av g . pu b l i c sc h o o l K‐2 cl a s s si z e = 22 Av g . pu b l i c sc h o o l 3‐7= 24 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 12 *N E A r e c o m m e n d e d o p t i m u m T e a c h e r S t u d e n t R a t i o Se c t i o n 10 - 1 2 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Il l u s t r a t e d Yo u r ki d s Ar e th e y ha p p y ? Ar e th e y su c c e e d i n g? y g Do yo u se e th e i r th i r s t fo r kn o w l e d g e gr o w i n g ? Ar e th e y pr o u d of th e i r sc h o o l ? Do th e y lo v e th e i r te a c h e r s ? Wh e n th e y gr a d u a t e wi l l th e y wa n t to gi v e ba c k to the sc h o o l ? 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 13 Se c t i o n 10 - 1 3 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Il l u s t r a t e d We s t l a k e Ac a d e m y ‐ pa t t e r n e d af t e r th e be s t pr i v a t e sc h o o l s sc h o o l s A ph i l o s o p h y of qu a l i t y wi t h o u t co m p r o m i s e A vi s i o n of ex c e l l e n c e Th h l ki d l h d dd Tea c her s that lov e ki d s, lov e to te a c h an d ar e em p o w e r e d to do so Em p h a s i s on ac a d e m i c ac h i e v e m e n t Ad h e r e n c e to a co d e of di s c i p l i n e , re s p e c t an d ho n o r Ou t s t a n d i n g cu r r i c u l u m Re c r u i t m e n t an d re t e n t i o n of ou t s t a n d i n g st a f f Sm a l l cl a s s e s wi t h lo w st u d e n t te a c h e r ra t i o Hi g h ac h i e v e m e n t sc o r e s Qu a l i t y fa c i l i t i e s Co m m i t t e d an d in v o l v e d co m m u n i t y an d pa r e n t s 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 14 Se c t i o n 10 - 1 4 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Il l u s t r a t e d Pr i v a t e Sc h o o l T u i t i o n Co m p a r i s o n Pr i v a t e Sc h o o l T u i t i o n Co m p a r i s o n (A v e r a g e pe r st u d e n t a n n u a l t u i t i o n ) $1 6 , 6 6 3 $1 6 , 3 5 7 $1 0 , 9 4 3 $1 6 , 0 0 0 $2 0 , 0 0 0 $8 , 2 0 5 $5 , 9 5 6 $6 , 9 8 3 $4 , 8 5 0 $8 , 0 0 0 $1 2 , 0 0 0 $‐ $ ‐ $4 , 0 0 0 S t . M a r k ' s H o c k a d a y L i b e r t y Ch r i s t i a n F o r t W o r t h Ch r i s t i a n Fa i t h Ch r i s t i a n Ci s t e r c i a n H a r v e s t Ch r i s t i a n Ac a d e m y Westlake Academy 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 15 Ac a d e m y Se c t i o n 10 - 1 5 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Il l u s t r a t e d $1 5 5 0 0 $16,400 $2 0 , 0 0 0 IB O Sc h o o l Tu i t i o n Co m p a r i s o n $1 1 , 2 1 0 $1 2 , 8 0 0 $1 4 , 0 0 0 $1 5 ,50 0 $, $1 2 , 0 0 0 $1 6 , 0 0 0 $7 , 4 5 0 $4 , 0 0 0 $4 0 0 0 $8 , 0 0 0 $0 $4 ,00 0 Th e Aw t y In t e r n a t ʹl S c h o o l Ch a r l o t t e C o u n t r y Da y Sc h o o l At l a n t a In t e r n a t ʹ l Sc h o o l At l a n t a , C a r l i s l e Sc h o o l Ax t o n , VA Ne w a r k Ac a d e m y Li v i n g s t o n , NJ Un i t e d Na t i o n s In t e r n a t ʹ l Sc h o o l Wa s h i n g t o n In t e r n a t ʹl School 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 16 H o u s t o n , TX C h a r l o t t e , NC G A N e w Yo r k , NY W a s h i n g t o n , D.C. Se c t i o n 10 - 1 6 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Ac t i o n “I f yo u ha v e bu i l t ca s t l e s in th e ai r , yo u r wo r k ne e d no t be lo s t ; th a t is wh e r e th e y sh o u l d be ne e d no t be lo s t ; th a t is wh e r e th e y sh o u l d be . No w pu t th e fo u n d a t i o n s un d e r th e m . ” ‐‐ He n r y Da v i d Th o r e a u 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 17 Se c t i o n 10 - 1 7 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Ac t i o n Th e Bl a c k s m i t h Ap p r e n t i c e Pr o g r a m Go a l : Co l l e c t do n a t i o n s to t a l i n g $1 , 8 0 0 pe r st u d e n t to su p p o r t an n u a l op e r a t i o n s of th e W e s t l a k e Ac a d e m y su p p o r t an n u a l op e r a t i o n s of th e We s t l a k e Ac a d e m y 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 18 Se c t i o n 10 - 1 8 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Ac t i o n Bl a c k s m i t h “A p p r e n t i c e ” Pr o g r a m Ea c h fa m i l y ad o p t s “a p p r e n t i c e ” st u d e n t s Co n t r i b u t e an an n u a l do n a t i o n an n u a l l y o n b e h a l f o f each Co n t r i b u t e an an n u a l do n a t i o n an n u a l l y on be h a l f of each “a p p r e n t i c e ” ad o p t e d Ad o p t “a p p r e n t i c e s ” eq u a l to th e nu m b e r of ch i l d r e n you have in th hl th e sc hoo l 10 0 % pa r t i c i p a t i o n me a n s ev e r y “a p p r e n t i c e ” is ad o p t e d Ev e r y fa m i l y is ur ged to par t i c i pat e at so m e le v e l y y g pp An y do n a t i o n am o u n t qu a l i f i e s fo r an ad o p t i o n of an “apprentice” Co r p o r a t e ma t c h i n g op p o r t u n i t i e s ar e we l c o m e ! 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 19 Se c t i o n 10 - 1 9 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Fi n a n c e Pl a n 20 0 5 Pr o j e c t e d Br e a k d o w n of We s t l a k e Ac a d e m y 20 0 5 Pr o j e c t e d Br e a k d o w n of We s t l a k e Ac a d e m y Fu n d i n g To w n of St a t e Fu n d i n g , 43 % To w n of We s t l a k e / D e b t Co n t r i b u t i o n s , 33 % 43 % Fe d e r a l Pa r e n t a l Co n t r i b u t i o n s , 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 20 Fu n d i n g / O t h e r Re v e n u e s , 8% , 16 % Se c t i o n 10 - 2 0 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Wh e r e we st a n d to d a y Av e r a g e An n u a l Fu n d i n g pe r S t u d e n t p e r Year Av e r a g e An n u a l Fu n d i n g pe r St u d e n t pe r Year (a f t e r a su c c e s s f u l Bl a c k s m i t h Ap p r e n t i c e Ca m p a i g n ) $1 1 , 4 0 0 $1 2 , 0 0 0 $7 , 8 6 0 $6 , 7 5 4 $7 , 0 8 8 $ 6 ,00 0 $8 , 0 0 0 $1 0 , 0 0 0 Pa r e n t a l Co n t r i b u t i o n s of $ 1 , 8 0 0 $2 , 0 0 0 $4 , 0 0 0 $, Pr i o r t o t a l ($ 4 , 9 5 4 ) $ ‐ We s t l a k e Ac a d e m y * * T e x a s Av e r a g e * N a t l Av g * C o n n e c t i c u t Average* 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 21 *N E A Re p o r t 20 0 2 ‐20 0 3 ** 2 0 0 3 ‐20 0 4 Se c t i o n 10 - 2 1 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Exc ell enc e Fact or e Ee e e ao To p Un i v e r s i t i e s Va r i a b l e s me a s u r e d to as s e s s th e to p un i v e r s i t i e s in the US A fo r 20 0 4 a c c o r d i n g to US N e w s c o m US A fo r 20 0 4 ac c o r d i n g to US N e w s .co m Pe e r As s e s s m e n t Fr e s h m a n Re t e n t i o n Ra t e Gd i R Gra dua t ion Rat e Fa c u l t y Re s o u r c e s % of cl a s s e s un d e r 20 pe o p l e St u d e n t Fa c u l t y ra t i o SA T / A C T Fr e s h m e n in to p 10 % of cl a s s Fi n a n c i a l Re s o u r c e s Ra n k Al u m n i Gi v i n g Ra n k Al u m n i Gi v i n g Ra t e 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 22 Se c t i o n 10 - 2 2 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Wh o is at th e to p an d wh y Ha r v a r d an d Pr i n c e t o n (t i e d fo r 1st ) To t a l Al u m n i Gi v i n g Ra n k i n g Pr i n c e t o n #1 (< 7, 0 0 0 st u d e n t s ) N D (1 1 3 0 0 d) Not r e Dam e #2 (1 1 ,30 0 st u den t s ) Ha r v a r d #3 (1 9 , 6 0 0 st u d e n t s ) Av e r a g e Al u m n i Gi v i n g Ra t e Av e r a g e Al u m n i Gi v i n g Ra t e Pr i n c e t o n #1 No t r e Da m e #2 Ha r v a r d #3 no t e : Pr i n c e t o n ’ s st u d e n t to fa c u l t y ra t i o is 5. 6 : 1 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 23 Se c t i o n 10 - 2 3 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r So l u t i o n In or d e r to ke e p th e sc h o o l fu n c t i o n i n g at it s cu r r e n t le v e l of service: Tw o pr o g r a m s of gi v i n g to th e We s t l a k e Ac a d e m y Fo u n d a t i o n : 1. Bl a c k s m i t h Ap p r e n t i c e pr o g r a m (v o l u n t a r y op e r a t i o n s support fr o m par e n t s )p) $ 1, 8 0 0 pe r st u d e n t in 20 0 4 ‐20 0 5 An n u a l co m m i t m e n t es t i m a t e d $1 , 8 0 0 / y r pe r st u d e n t th r o u g h 2010 2. Sp e c i a l ca m p a i g n s an d fu n d ra i s e r s (T h e We s t l a k e Ac a d e m y d) Fo u n dat i o n ) Ga l l e r y Da y s (N o v e m b e r 11 & 13 , 20 0 4 ) Fo u n d e r s Dr i v e ‐ Sp r i n g 05 Co r por a t e ma t c h i n g ca m pai gn p g pg Gr a n t s NO T E : 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 24 Al l do n a t i o n s wi l l be ma d e to th e We s t l a k e Ac a d e m y Fo u n d a t i o n an d ar e Tax Deductible. Se c t i o n 10 - 2 4 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Ou r Co m m i t m e n t St u d e n t s an d pa r e n t s co n t i n u e to ad d ne w gr a d e s an d ex p a n d th e IB wi t h a d d e d ex t r a cu r r i c u l a r co n t i n u e to ad d ne w gr a d e s an d ex p a n d th e IB wi t h ad d e d ex t r a cu r r i c u l a r ac t i v i t i e s Te a c h e r s hi r e th e be s t an d su p p o r t th e m Ad m i n i s t r a t o r s ma i n t a i n lo w ov e r h e a d ex t e n d He a d of Sc h o o l co n t r a c t Co m m u n i t y op e r a t e th e sc h o o l wi t h o u t a lo c a l pr o p e r t y ta x Vo l u n t e e r s co n t i n u e to id e n t i f y me a n i n g f u l pa r t n e r s h i p s th a t be n e f i t th e sc h o o l Fa c i l i t i e s To w n wi l l ad d fa c i l i t i e s as fu n d s be c o m e av a i l a b l e th r o u g h ne w re v e n u e sources 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 25 Se c t i o n 10 - 2 5 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality Th e Ex c e l l e n c e Fa c t o r Th e Ex c e l l e n c e Fa c t o r Co n c l u s i o n Th e We s t l a k e Ac a d e m y wi l l be th e mo s t sought af t e r pu b l i c sc h o o l in th e st a t e of Te x a s be c a u s e of yo u . Th a n k yo u fo r yo u r pi o n e e r sp i r i t , yo u r pa t i e n c e , yo u r ge n e r o u s in ‐ki n d do n a t i o n s , your vo l u n t e e r ti m e , yo u r pr a y e r s , an d yo u r un e n d i n g su p p o r t of th e We s t l a k e Ac a d e m y an d th e su p p o r t of th e We s t l a k e Ac a d e m y an d th e co m m u n i t y of We s t l a k e ! ! 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 26 Se c t i o n 10 - 2 6 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality The Ex c e l l e n c e Fac t o r Th e Ex c e l l e n c e Fa c t o r Re s u l t s To t a l Pl e d g e d $ 25 7 , 1 8 7 To t a l Re c e i v e d to ‐da t e $ 22 8 , 9 2 6 * 10 0 % of st u d e n t s sp o n s o r e d 14 8 ou t of 17 5 fa m i l i e s par t i c i pat e d ‐ 85 % pp 55 o u t of 55 to w n / a c a d e m y em p l o y e e s pa r t i c i p a t e d ‐ 10 0 % * St i l l re c e i v i n g m o n t h l y p a y m e n t s as we l l as di r e c t de p o s i t s and * St i l l re c e i v i n g mo n t h l y pa y m e n t s as we l l as di r e c t de p o s i t s and mo n t h l y pa y r o l l de d u c t i o n s . 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 27 Se c t i o n 10 - 2 7 We s t l a k e Ac a d e m y a co n v e r g e n c e of quality We s t l a k e Ac a d e m y a co n v e r g e n c e of quality The Ex c e l l e n c e Fac t o r Th e Ex c e l l e n c e Fa c t o r Re s u l t s Co r p o r a t e Ma t c h i n g Pl e d g e s AX A Fo u n d a t i o n Br i n k s In c o r p o r a t e d De l Mo n t e Fo o d s Fi d e l i t y Fo u n d a t i o n Fr e d d i e Ma c Fo u n d a t i o n GA P Fo u n d a t i o n Mo t o r o l a Fo u n d a t i o n Sp r i n t Fo u n d a t i o n UP S Fo u n d a t i o n Ve r i z o n Fo u n d a t i o n We l l s Fa r go Fo u n d a t i o n 12 / 1 0 / 2 0 0 8 ©T o w n o f W e s t l a k e , T e x a s 28 g Se c t i o n 10 - 2 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Block scheduling in large, urban high schools: Effects on academic achieveme... Lois-Lynn Stoyko Deuel The High School Journal; Oct/Nov 1999; 83, 1; Research Library pg. 14 Section 11-1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Section 11-12 WWWeeessstttlllaaakkkeee AAAcccaaadddeeemmmyyy BBBoooaaarrrddd ooofff TTTrrruuusssttteeeeeesss PPPaaarrreeennnttt SSSuuurrrvvveeeyyy FFFiiinnnaaalll RRReeepppooorrrttt SSuubbmmiitttteedd TToo:: WWWeeessstttlllaaakkkeee AAAcccaaadddeeemmmyyy BBBoooaaarrrddd ooofff TTTrrruuusssttteeeeeesss ETC Institute Project Manager: Chris Tatham 725 West Frontier Circle Phone: 913-829-1215 Olathe, Kansas Fax: 913-829-1591 66061 E-mail: ctatham@etcinstitute.com EEETTTCCC IIInnnssstttiiitttuuuttteee June 2009 ...helping organizations make better decisions since 1982 ETC Institute (2009)ETC Institute (2009) Section 12-1 Executive Summary i EXE C U T I V E SUM M A R Y 2009 Westlake Academy Board of Trustees Parent Survey Executive Summary Report Overview and Methodology During May and June of 2009, ETC Institute administered a survey of parents of children who attended Westlake Academy. The purpose of the survey was to gather input from parents to improve the overall quality of education and programs provided by the Academy. The four-page survey was administered by mail and phone to a random sample of 170 parents. The results for the random sample of 170 parents have a 95% level of confidence with a precision of at least +/- 5.0%. This summary report contains: ¾ a summary of the methodology for administering the survey and major findings ¾ charts showing the overall results for most questions on the survey ¾ importance-satisfaction analysis ¾ tabular data that show the results for each question on the survey ¾ a copy of the survey instrument. The major findings of the survey are provided on the following pages. Major Findings • Parents Were Generally Satisfied with the Overall Quality of Education Provided by Westlake Academy. Eighty-two percent (82%) of the parents surveyed were “very satisfied” or “satisfied” with the quality of education they received at Westlake Academy; 8% were “neutral,” 9% were “very dissatisfied” or “dissatisfied” and 1% did not have an opinion. Section 12-2 Executive Summary ii EXE C U T I V E SUM M A R Y • Satisfaction With Westlake Academy Services and Programs. The Westlake Academy services and programs that residents were most satisfied with, based upon a combination of “very satisfied” and “satisfied” responses were: o The IB Curriculum (85%) o Maintenance of the Campus (85%) o Suitability of the campus facilities for learning (85%) o Opportunities for parental involvement (81%) o Administration (76%) o Academic progress of your child (76%) • Westlake Academy Services and Programs Parents Felt Were Most Important. The Academy services and programs that parents felt were most important were: (1) teachers/faculty, (2) the academic progress of children, (3) the IB curriculum and (4) the college preparation process. • Student Safety at Westlake Academy. Ninety-two percent (92%) of parents felt their child is physically safe at school, 6% did not and 3% did not know. Eighty-one percent (81%) of parents felt their child is emotionally safe at school, 16% did not and 3% did not know. • IB Curriculum. Of the 11 items assessed on the survey, parents rated the IB curriculum as the number one reason they originally decided to enroll their child in Westlake Academy. When parents were asked about their understanding of the IB curriculum, 80% of parents felt they had an adequate understanding of the curriculum, 18% did not and 2% did not know. • Special Education Services. Of the parents who had children who used special education services, sixty-four percent (64%) of parents were satisfied with the modifications and services provided by Westlake Academy and 36% were not. • Communication. Some of the major findings from the survey related to communication are listed below: o More than half (61%) of the parents surveyed felt direct e-mail communication was the best way for Westlake Academy to keep them informed. In addition, forty-two percent (42%) of parents indicated they would like to receive e-mails from Westlake Academy to direct them to the WA website. o Eighty-two percent (82%) of the parents surveyed felt a WA newsletter would be valuable. When asked how they would like to receive the newsletter, 71% would prefer to receive it by e-mail, 26% would like to receive it by postal mail and 3% did not provide a response. Section 12-3 Executive Summary iii EXE C U T I V E SUM M A R Y • Planning for the Future. Parents were asked to indicate how important they felt it was for Westlake Academy to expand and/or implement various services and programs assuming the resources are available. The items that parents felt were most important, based upon a sum of “extremely important,” “very important” and “important” responses, were: expand IB options at the Diploma level (94%), expand the art program (87%) and expand foreign language offerings/opportunities (86%). • Blacksmith Apprentice Program. Ninety percent (90%) of the parents surveyed indicated they participated in the Blacksmith Apprentice Program annually, 9% did not and 1% did not provide a response. When asked if they would be willing to increase contributions to the program to fund expanded levels of services or new programs, more than half (61%) of parents indicated they would be willing to increase their contributions, 32% were not willing and 7% did not provide a response. Section 12-4 Section 1: Charts and Graphs Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-5 49% 41% 41% 42% 30% 33% 26% 23% 29% 21% 17% 26% 20% 23% 25% 26% 23% 17% 18% 23% 16% 12% 12% 36% 44% 44% 39% 46% 43% 46% 49% 43% 47% 50% 41% 46% 41% 38% 36% 37% 40% 32% 26% 30% 32% 32% 11% 11% 10% 12% 13% 14% 22% 18% 23% 15% 21% 15% 17% 25% 25% 22% 24% 20% 29% 26% 34% 31% 22% 4% 4% 5% 7% 11% 11% 7% 11% 6% 17% 12% 18% 17% 12% 12% 17% 16% 23% 21% 26% 20% 25% 35% IB curriculum Maintenance of campus Suitability of facilities for learning Opportunities for parental involvement Administration Academic progress of your child Effectiveness of the House of Commons Teachers/Faculty Effectiveness of the WA Foundation Quality of communication from WA Traffic Methods of communication from WA School website Effectiveness of the WAAC The four House system for students Opportunities for parental input Communications regarding issues/problems Grading system Trustees College preparation process Extracurricular sports programs Other extracurricular programs School lunch program 0%20%40%60%80%100% Very Satisfied (5)Satisfied (4)Neutral (3)Dissatisfied (1/2) Q1. Overall Satisfaction With Westlake Academy Services and Programs by percentage of parents who rated the item as a 1 to 5 on a 5-point scale (excluding don't knows and not applicable) Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) 2% 1% 67% 55% 45% 33% 19% 12% 10% 9% 7% 7% 4% 4% 4% 2% 2% 2% 1% 1% Teachers/Faculty Academic progress of your child IB curriculum College preparation process Administration Communications regarding issues/problems Extracurricular sports programs Grading system Quality of communication from WA Other extracurricular programs Suitability of facilities for learning Opportunities for parental involvement Methods of communication from WA Opportunities for parental input Trustees School lunch program Effectiveness of the WA Foundation Maintenance of campus The four House system for students Traffic 0%20%40%60%80% 1st Choice2nd Choice3rd Choice by percentage of parents who selected the item as one of their top three choices Q2. Westlake Academy Services and Programs That Parents Felt Were Most Important Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-6 Q3. Do you feel your child is emotionally safe at school? by percentage of parents Yes 81% No 16% Don't know 3% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q4. Do you feel your child is physically safe at school? by percentage of parents Yes 92% No 6% Don't know 2% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-7 Q5. Overall Satisfaction With the Quality of Education Provided by Westlake Academy by percentage of parents Very satisfied 38% Satisfied 44% Neutral 8% Dissatisfied 8% Very dissatisfied 1% Don't know 1% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q6. Do you believe you have an adequate understanding of the IB curriculum? by percentage of parents Yes 80% No 18% Don't know 2% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-8 Q7. Do you have a child with a learning disability that utilizes special education services? by percentage of parents Q7a. If yes, are you satisfied with the modifications and services provided by Westlake Academy? Yes 13% No 86% Not provided 1% Yes 64% No 36% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q8. Would you find a WA newsletter to be of value? by percentage of parents Q8a. If YES, how you prefer to receive the newsletter? Yes 82% No 17% Not Sure 1% E-mail 71% Postal Mail 26% Not provided 3% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-9 Q9. Which of the following are the best ways for Westlake Academy to keep you informed? by percentage of parents (multiple responses allowed) 61% 42% 15% Direct e-mail communication E-mails that direct you to the WA website Hard copies of items posted on WA website 0%20%40%60%80% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) 82% 33% 21% 11% 9% 9% 9% 5% 3% 2% 2% 2% The IB curriculum Quality of teachers/faculty Reputation of the school Location of the school Quality of the facilities Foreign language offering Small Classroom Size/Environment The Academy's status as a "charter" school Cost Extracurricular sports programs Strings program Other extracurricular programs 0%20%40%60%80%100% by percentage of parents who selected the item as one of their top two choices Q10. Which of the following were most important in your decision to originally enroll your children) at Westlake Academy Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-10 Q11. If your child was not enrolled at Westlake Academy, where would your child most likely be attending school? by percentage of parents A public school 65% A private school 26% Home school 5% Not Sure 4% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q12. Importance of Westlake Academy Expanding Various Academy Programs Assuming the Resources Are Available by percentage of parents who rated the item as a 5, 4 or 3 on a 5-point scale where 5 means "Extremely Important" and 1 means "Not Important" (excluding don't knows and not applicable) 94% 87% 86% 76% 74% 74% 62% 49% 39% Expand IB options at the Diploma level Expand the Arts program Expand foreign language offerings/opportunities Offer online access to student grades Expand the new media educational opportunities Implement a Spanish immersion program Improve the website Continue the Strings program Expand residential trip offerings 0%20%40%60%80%100% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-11 Q13. The Blacksmith Apprentice Program is critical to the financial health of the Academy. Do you annually participate in the program? by percentage of parents Yes 90% No 9% Not provided 1% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q14. Would you be willing to increase your contributions to the Blacksmith Apprentice Program to fund an expanded level of services or new programs? by percentage of parents Yes 61% No 32% Not provided 7% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-12 Q18. How many years have you had at least one child attending Westlake Academy by percentage of parents 1 year 22% 2 years 13% 3 years 8% 4 years 12% 5 years 13%6 years 31% Not provided 1% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q19. In which grades do you currently have children enrolled at Westlake Academy? by percentage of parents K-2 19% Grades 3-4 16% Grades 5-6 16% Grades 7-8 19% Grades 9-10 21% Grades 11-12 9% Not provided 1% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-13 Q20. Do you have additional school age children enrolled in other schools? Q20a. If YES, what grades? by percentage of parents Yes 27% No 72% Not provided 1% K-6th 51% 7th-10th 27% 11th-12th 20% Not provided 2% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Q21. How many adults in your home are employed full time outside of the home? by percentage of parents None 2% One 55% Two 42% Not provided 1% Source: ETC Institute (June 2009 - Westlake Academy Board of Trustees Parent Survey) Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-14 Section 2: Importance-Satisfaction Analysis Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-15 Importance-Satisfaction Analysis Westlake Academy Board of Trustees Parent Survey Overview Today, public officials have limited resources to access activities that are of the most benefit to their customers. Two of the most important criteria for decision making are (1) to target resources toward services of the highest importance to customers; and (2) to target resources toward those services where customers are the least satisfied. The Importance-Satisfaction (I-S) rating is a unique tool that allows public officials to better understand both of these highly important decision making criteria for each of the services they are providing. The Importance-Satisfaction rating is based on the concept that organizations will maximize overall customer satisfaction by emphasizing improvements in those service categories where the level of satisfaction is relatively low and the perceived importance of the service is relatively high. Methodology The rating is calculated by summing the percentage of responses for items selected as the first, second, and third most important Academy services and programs. This sum is then multiplied by 1 minus the percentage of parents that indicated they were positively satisfied with the Academy's performance in the related area (the sum of the ratings of 4 and 5 on a 5-point scale excluding “don't know” responses). “Don't know” responses are excluded from the calculation to ensure that the satisfaction ratings among service categories are comparable. [I-S=Importance x (1-Satisfaction)]. Example of the Calculation. Parents were asked to identify the Academy services and programs they thought were most important. Thirty-three percent (33%) of parents ranked the college preparation process as the most important Academy service. With regard to satisfaction, the college preparation process was ranked twentieth overall with 49% rating the college preparation process as a “4” or a “5” on a 5-point scale, excluding “don't know” responses. The I-S rating for college preparation process was calculated by multiplying the sum of the most important percentages by 1 minus the sum of the satisfaction percentages. In this example, 33% was multiplied by 51% (1-0.49). This calculation yielded an I-S rating of 0.1683, which was ranked second out of the twenty three services and programs accessed on the survey. Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-16 The maximum rating is 1.00 and would be achieved when 100% of the parents selected the service or program as one of the three most important areas and 0% indicate that they are positively satisfied with the delivery of the service. The lowest rating is 0.00 and could be achieved under either one of the following two situations: • if 100% of the parents were positively satisfied with the delivery of the service • if none (0%) of the parents selected the service as one of the three most important areas. Interpreting the Ratings Ratings that are greater than or equal to 0.20 identify areas that should receive significantly more emphasis. Ratings from .10 to .20 identify service areas that should receive increased emphasis. Ratings less than .10 should continue to receive the current level of emphasis. • Definitely Increase Emphasis (IS>=0.20) • Increase Current Emphasis (0.10<=IS<0.20) • Maintain Current Emphasis (IS<0.10) The results for the Westlake Academy Survey are provided on the following page. Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-17 Importance-Satisfaction Rating Westlake Academy Board of Trustees Parent Survey Satisfaction With Westlake Academy Programs and Services Category of Service Most Important % Most Important Rank Satisfaction % Satisfaction Rank Importance- Satisfaction Rating I-S Rating Rank High Priority (IS .10-.20) Teachers/Faculty 67%172%80.18761 College preparation process 33%449%200.16832 Academic progress of your child 55%276%60.13203 Medium Priority (IS <.10) IB curriculum 45%385%10.06754 Extracurricular sports programs 10%746%210.05405 Communications regarding issues/problems 12%660%170.04806 Administration 19%576%50.04567 Other extracurricular programs 7%1044%220.03928 Grading system 9%857%180.03879 Quality of communication from WA 7%968%100.022410 Methods of communication from WA 4%1367%120.013211 School lunch program 2%1644%230.011212 Trustees 2%1550%190.010013 Opportunities for parental involvement 4%1281%40.007614 Opportunities for parental input 2%1462%160.007615 Suitability of facilities for learning 4%1185%30.006016 Effectiveness of the WA Foundation 2%1772%90.005617 The four House system for students 1%1963%150.003718 Traffic 1%2067%110.003319 Maintenance of campus 1%1885%20.001520 School website 0%2166%130.000021 Effectiveness of the House of Commons 0%2272%70.000022 Effectiveness of the WAAC 0%2364%140.000023 Note: The I-S Rating is calculated by multiplying the "Most Important" % by (1-'Satisfaction' %) Most Important %: The "Most Important" percentage represents the sum of the first, second, and third most important responses for each item. Respondents were asked to identify the items they thought should receive the most emphasis over the next two years. Satisfaction %:The "Satisfaction" percentage represents the sum of the ratings "4" and "5" excluding 'don't knows.' Respondents ranked their level of satisfaction with the each of the items on a scale of 1 to 5 with "5" being very satisfied and "1" being very dissatisfied. © 2009 DirectionFinder by ETC Institute Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-18 Importance-Satisfaction Matrix Analysis. The Importance-Satisfaction rating is based on the concept that public agencies will maximize overall customer satisfaction by emphasizing improvements in those areas where the level of satisfaction is relatively low and the perceived importance of the service is relatively high. ETC Institute developed an Importance-Satisfaction Matrix to display the perceived importance of the programs and services that were assessed on the survey against the perceived quality of service delivery. The two axes on the matrix represent Satisfaction (vertical) and relative Importance (horizontal). The I-S (Importance-Satisfaction) matrix should be interpreted as follows. • Continued Emphasis (above average importance and above average satisfaction). This area shows where Westlake Academy is meeting customer expectations. Items in this area have a significant impact on the customer’s overall level of satisfaction. Westlake Academy should maintain (or slightly increase) emphasis on items in this area. • Exceeding Expectations (below average importance and above average satisfaction). This area shows where Westlake Academy is performing significantly better than customers expect the Academy to perform. Items in this area do not significantly affect the overall level of satisfaction that parents have with Westlake Academy services. The Academy should maintain (or slightly decrease) emphasis on items in this area. • Opportunities for Improvement (above average importance and below average satisfaction). This area shows where Westlake Academy is not performing as well as parents expect the Academy to perform. This area has a significant impact on customer satisfaction, and Westlake Academy should DEFINITELY increase emphasis on items in this area. • Less Important (below average importance and below average satisfaction). This area shows where Westlake Academy is not performing well relative to the Academy’s performance in other areas; however, this area is generally considered to be less important to parents. This area does not significantly affect overall satisfaction with Westlake Academy services because the items are less important to parents. The Academy should maintain current levels of emphasis on items in this area. A matrix chart showing the results for Westlake Academy is provided on the following page. Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-19 Satisfaction Rating , , , , , , , , , , , , ,, , , , ,,,,,, mean satisfaction Op p o r t u n i t i e s f o r I m p r o v e m e n t We s t l a k e A c a d e m y B o a r d o f T r u s t e e s P a r e n t S u r v e y Im p o r t a n c e - S a t i s f a c t i o n A s s e s s m e n t M a t r i x -S a t i s f a c t i o n W i t h W e s t l a k e A c a d e m y P r o g r a m s a n d S e r v i c e s - (p o i n t s o n t h e g r a p h s h o w d e v i a t i o n s f r o m t h e m e a n i m p o r t a n c e a n d s a t i s f a c t i o n r a t i n g s g i v e n b y r e s p o n d e n t s t o t h e s u r v e y ) me a n i m p o r t a n c e Im p o r t a n c e R a t i n g Lo w e r I m p o r t a n c e Higher Importance lo w e r i m p o r t a n c e / h i g h e r S a t i s f a c t i o n hi g h e r i m p o r t a n c e / higher Satisfaction lo w e r i m p o r t a n c e / l o w e r S a t i s f a c t i o n hi g h e r i m p o r t a n c e / l o w e r S a t i s f a c t i o n Ex c e e d e d E x p e c t a t i o n s Le s s I m p o r t a n t Co n t i n u e d E m p h a s i s So u r c e : E T C I n s t i t u t e ( 2 0 0 9 ) Su i t a b i l i t y o f f a c i l i t i e s Ex t r a c u r r i c u l a r sp o r t s p r o g r a m s Ot h e r e x t r a c u r r i c u l a r p r o g r a m s Ac a d e m i c p r o g r e s s of y o u r c h i l d Co l l e g e p r e p a r a t i o n p r o c e s s Gr a d i n g s y s t e m Te a c h e r s / F a c u l t y Sc h o o l l u n c h p r o g r a m Co m m u n i c a t i o n s r e g a r d i n g i s s u e s / p r o b l e m s Ad m i n i s t r a t i o n Ef f e c t i v e n e s s o f t h e W A A C Fo u r H o u s e s y s t e m Op p o r t u n i t i e s f o r p a r e n t a l i n p u t Tr u s t e e s IB C u r r i c u l u m Op p o r t u n i t i e s f o r p a r e n t a l i n v o l v e m e n t Tr a f f i c Sc h o o l w e b s i t e Ma i n t e n a n c e o f c a m p u s WA F o u n d a t i o n Ho u s e o f C o m m o n s Qu a l i t y o f c o m m u n i c a t i o n Me t h o d s o f c o m m u n i c a t i o n We s t l a k e A c a d e m y B o a r d o f T r u s t e e s P a r e n t S u r v e y : F i n a l R e p o r t ET C I n s t i t u t e ( 2 0 0 9 ) Se c t i o n 12 - 2 0 Section 4: Survey Instrument Westlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-21 Westlake Academy Board of Trustees Parent Survey 1. Satisfaction with Westlake Academy. Using a scale of 1 to 5 where 5 means “very satisfied” and 1 means “very dissatisfied,” please rate your satisfaction with the following items at Westlake Academy: How Satisfied are you with the: Very Satisfied Satisfied Neutral Dissatisfied Very Dissatisfied Don’t Know N/A A. Teachers/faculty 5 4 3 2 1 9 0 B. Administration 5 4 3 2 1 9 0 C.Trustees 5 4 3 2 1 9 0 D.Maintenance of the campus 5 4 3 2 1 9 0 E. Suitability of the campus facilities for learning 5 4 3 2 1 9 0 F. IB (International Baccalaureate) curriculum 5 4 3 2 1 9 0 G.School lunch program 5 4 3 2 1 9 0 H.Extracurricular sports programs 5 4 3 2 1 9 0 I. Other extracurricular programs 5 4 3 2 1 9 0 J. School web site 5 4 3 2 1 9 0 K. Academic progress of your child 5 4 3 2 1 9 0 L. Communications regarding issues/problems related to your child 5 4 3 2 1 9 0 M.Opportunities for parental involvement 5 4 3 2 1 9 0 N.Opportunities for parental input 5 4 3 2 1 9 0 O.Quality of communication from WA 5 4 3 2 1 9 0 P. Methods of communication from WA 5 4 3 2 1 9 0 Q.The four House system for students (Thoreau, Wheatley, Whitman, Keller) 5 4 3 2 1 9 0 R.Effectiveness of the House of Commons 5 4 3 2 1 9 0 S. Effectiveness of the WA Foundation 5 4 3 2 1 9 0 T. Effectiveness of the WAAC 5 4 3 2 1 9 0 U.Traffic (carpools, entry/exit, drop off) 5 4 3 2 1 9 0 V. Grading system 5 4 3 2 1 9 0 W College preparation process 5 4 3 2 1 9 0 2. Which THREE of the items listed above are most important to you? [Write in the letters below using the letters from the list in Question 1 above]. 1st:____ 2nd:____ 3rd:____ 3. Do you feel your child is emotionally safe at school? ___(1) Yes ___(2) No 4. Do you feel your child is physically safe at school? ___(1) Yes ___(2) No ETC Institute (2009)Section 12-22 5. Overall, how satisfied are you with the quality of education provided by Westlake Academy? ___(5) Very satisfied ___(4) Satisfied ___(3) Neutral ___(2) Dissatisfied ___(1) Very Dissatisfied ___(9) Don’t know 6. Do you believe you have an adequate understanding of the IB curriculum? ___(1) Yes ___(2) No 7. Do you have a child with a learning disability that utilizes special education services? ___(1) Yes ___(2) No 7a. If yes, are you satisfied with the modifications/services provided by Westlake Academy? ___(1) Yes ___(2) No 7b. If “No”, why not? 8. Would you find a WA newsletter to be of value? ___(1) Yes ___(2) No 8a. If yes, would you prefer to receive the newsletter via ____ (1) email ____(2) postal mail 9. Which of the following are the best ways for Westlake Academy to keep you informed? ___(1) E-mails that direct you to the WA website ___(2) Hard copies of items posted on the WA website delivered in the HOC folder or similar System ___(3) Direct email communication 10. Which TWO of the following were most important in your decision to originally enroll your child(ren) at Westlake Academy? ___(1) Quality of teachers/faculty ___(2) Quality of the facilities ___(3) The IB (International Baccalaureate) curriculum ___(4) Extracurricular sports programs ___(5) Other extracurricular programs ___(6) Reputation of the school ___(7) Location of the school ___(8) Cost ___(9) The Academy’s status as a “charter” school __(10) Strings program __(11) Foreign language offering ___(0) Other: ___________________________ 11. If your child was not enrolled at Westlake Academy, where would your child most likely be attending school? ___(1) A public school in the community where you live ___(2) A private school ___(3) Home school ETC Institute (2009)Section 12-23 12. Planning for the Future. Westlake Academy’s operating budget for its academic programs is funded by State public education funds and private donations. The Academy’s private donations are raised through the Westlake Academy Foundation’s Blacksmith Apprentice Program. State funding is expected to remain at current levels. Thus, expanding the Academy’s programs would require additional private contributions. With funding limitations understood, we would still like to gauge your interest in the following programs and/or expansion opportunities. Using a scale of 1 to 5 where 5 means “extremely important” and 1 means “not important at all,” please rate how important it is for Westlake Academy to implement the following, assuming resources are available in the future: How important is it for Westlake Academy to: Extremely Important Very Important Important Somewhat Important Not Important Don’t Know N/A A. Expand IB options at the Diploma level 5 4 3 2 1 9 0 B.Expand the Arts program (music, drama) 5 4 3 2 1 9 0 C.Continue the Strings program 5 4 3 2 1 9 0 D.Expand residential (overnight) trip offerings 5 4 3 2 1 9 0 E. Expand the new media educational opportunities (i.e., video production, on-line journalism) 5 4 3 2 1 9 0 F. Improve the web site 5 4 3 2 1 9 0 G.Expand foreign language offerings and opportunities 5 4 3 2 1 9 0 H.Implement a Spanish immersion program 5 4 3 2 1 9 0 I. Offer online access to student grades 5 4 3 2 1 9 0 13. The Blacksmith Apprentice Program is critical to the financial health of the Academy. Do you annually participate in the program? ___(1) Yes ___(2) No 14. Would you be willing to increase your contributions to the Blacksmith Apprentice Program to fund an expanded level of services or new programs (e.g. examples of the services/programs can be found in Question 12)? ___(1) Yes ___(2) No 15. What do you like BEST about Westlake Academy? 16. What ONE thing would you like to improve most at Westlake Academy? estlake Academy Board of Trustees Parent Survey: Final Report ETC Institute (2009)Section 12-24 DEMOGRAPHICS 17. If you live in the Town of Westlake, how important was Westlake Academy in your decision to move to Westlake? ___(4) Very important ___(3) Somewhat Important ___(2) Not sure ___(1) Not important 18. How many years have you had at least one child attending Westlake Academy? __________ years 19. In which grades do you currently have children enrolled at Westlake Academy? ___(1) K-2 ___(2) grades 3-4 ___(3) grades 5-6 ___(4) grades 7-8 ___(5) grades 9-10 ___(6) grades 11-12 20. Do you have additional school age children enrolled in other schools? ___(1) Yes ___(2) No If so, what grade(s): ____K - 6th ____7th - 10th ____11th – 12th 21. How many adults in your home are employed full time outside the home? ___(0) None ___(1) One ___(2) Two 21a. How can we better accommodate the needs of a working parent or working parents? OPTIONAL: If you have any other comments please write them in the space provided below. THE WESTLAKE ACADEMY BOARD OF TRUSTEES THANKS YOU FOR COMPLETING THIS SURVEY. Please Return Your Completed Survey in the Enclosed Postage Paid Envelope Addressed to: ETC Institute, 725 W. Frontier Circle, Olathe, KS 66061 ETC Institute (2009)Section 12-25