Title: The Evidence on Vouchers: How Does the Cleveland Scholarship Program Fit In?
1The Evidence on Vouchers How Does the Cleveland
Scholarship Program Fit In?
- Clive Belfield
- City University of New York
- National Center for the Study of Privatization in
Education, - Teachers College, Columbia University
2New Research on Vouchers
- Economic principles of choice and efficiency
- Evidence from K-12 programs (Milwaukee, Florida,
Washington DC) - Data analyses using exogenous variation and other
school choice reforms - Expansion of voucher programs (e.g., in
pre-school, higher education)
3Evidence on Vouchers
- School choice/demand
- Religious preferences dominate
- Although targeted at low-income families,
educated parents apply for vouchers - Use rate 25 below offer rate
- Use rate lower for minority students
- Racial segregation by choice
- Sub-additive 25 in private school already 5
go to private school without voucher
4Evidence on Vouchers
- School technology/supply
- Various religions participate
- Development of all-voucher schools
- Only religious, for-profit, and adverse selection
schools others charter - Unknown Cost/revenue functions and superior (?)
technologies of private schools - Broader implications
- Home-owners do not like vouchers
- High ability students gain most voucher
cross-subsidy and ability-tracking in public
schools
5Evidence on Vouchers
- Effectiveness
- Small positive impact on achievement
- In Milwaukee, for math (0.1sd) not reading
- In Florida, mainly a function of student
characteristics - In experiments, for African Americans
- Charter school literature and lotteries show
similarly weak results - Other outcomes may be more important
6Key Research Issues
- Voucher status
- User
- Non-user (conditional on offer)
- Rejected applicant
- Public school population
- Biases when comparing across voucher status
- Application eligibility usage school switch
resources - Effects of interest
- Across each status treatment intention to
treat dose-response by race
7Cleveland Scholarship Program
- 5,734 students participate in K-12 grades
- Low-income families given priority
- Voucher worth 2,750-3,000 (lt40 of public
school ave. cost) - Data on 3,000 students from K in 1997
- Evaluation by Indiana University
8Advantages of research on CSP
- The silent program
- Recent data in context of growing school choice
- Religious schools participate
- Students in many different schools
- High proportion of African American students
- Longitudinal data across voucher status
- Relatively ungenerous voucher
9Same findings for CSP
- Religious preferences 40 of 42 chosen schools
- Voucher schools 50 of Catholic school
enrollments - Non-profit voucher schools become charter schools
- Use rate below offer rate non-users 40 of users
- Use rate lower for minority and low-income
students - Very high rates of sub-additivity (applications
from private sector)
10Frequencies for Second Grade
Scholarship user (2nd grade) Rejected applicant Public school comparison group Non-user (incl. user in prior grades)
Terra Nova math scores (1st grade) 518 (32) 522 (39) 519 (38) 506 (38)
Black 56 58 48 75
Free lunch 58 50 49 83
Female 52 49 48 49
Class size (2nd grade) 23.1 23.7 23.7 23.3
Teacher exper. (2nd grade) 13.7 12.9 12.4 12.6
N 624 438 971 326
11Achievement Effect Sizes vs. Public School
Comparison Group 2nd Grade
Reading Math Language
User -0.06 (0.05) -0.09 (0.04) 0.10 (0.05)
Non-user -0.02 (0.09) -0.02 (0.08) 0.16 (0.09)
Rejected 0.08 (0.06) 0.03 (0.05) 0.14 (0.05)
R-squared 0.27 0.46 0.35
N 1733 1786 1736
- Notes OLS. SE in parentheses. Equation
includes subject-specific first grade scores
constant black Hispanic female free lunch
unsubsidized lunch class size and teacher
experience measured in years. significant at
10 significant at 5 significant at 1.
12Achievement Effect Sizes vs. Public School
Comparison Group 4th Grade
Reading Reading Math Language Science
User User 0.04 (0.04) -0.11 (0.04) 0.04 (0.04) 0.03 (0.04)
Non-user Non-user -0.07 0.06) 0.04 (0.06) 0.08 (0.06) 0.09 (0.06)
Rejected Rejected -0.03 0.05) -0.06 (0.05) -0.06 (0.05) 0.05 (0.05)
R-squared R-squared 0.46 0.44 0.41 0.45
N N 2089 2102 2085 2096
- Notes OLS. SE in parentheses. Equation
includes subject-specific first grade scores
constant black Hispanic female free lunch
unsubsidized lunch class size and teacher
experience measured in years. significant at
10 significant at 5 significant at 1.
13Additional Results for Users
- 1. Unadjusted for prior scores
- Math penalty -0.12sd 2nd/4th grade
- Language gain 0.12sd 4th grade
- 2. Winners (users and non-users)
- Math penalty -0.08sd 2nd grade
- Math penalty -0.06sd 4th grade
- Language gain vs public 0.12sd 4th grade
- 3. Treatment group (users vs. rejected)
- Reading penalty -0.14sd 2nd grade
- Math penalty -0.11sd 2nd grade
- Language gain 0.11sd 4th grade
14Additional Results for Users
- 4. Dose-response (users split by years of use)
- Weak evidence
- for 2nd grade, all significant effects are for
3-year users - for 4th grade, significant effects across any
durations - 5. African-American sub-sample
- Same results, except
- Reading penalty -0.14sd 2nd grade
- Math penalty disappears 2nd grade
- Reading gain 0.14sd 4th grade
15Economic Analysis
- Cleveland Program outcomes not much worse than
existing public schools - Voucher value 2,400 x1 for transportation
x2 for assessment x3 for private school
support/exemption from regulations - Sub-additivity 33, i.e. 3 vouchers for every
marginal user - CMSD average cost 8800 in 2001
- (incl. special education incl. less motivated
families) - Even with a low face-value voucher efficiency
could be equal
16Conclusions
- Cleveland Program similar in effect to other
voucher programs - Can be justified as promoting freedom of choice
- State-wide vouchers
- Most of the equity arguments apply only to
small-scale programs - With charter school option, religious schools
must be supply-responsive