https://orcid.org/0000-0001-6490-9764
![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
ABSTRACT
The rate of students switching from public schools to private schools because they received a scholarship and would not have switched without the subsidy (“switcher rate”) is an integral factor for reliably estimating the net fiscal impact of private school choice programs. Switcher rates observed among students in control groups in random assignment studies can provide useful information for inferring switcher rates among program participants when conducting fiscal analyses. This paper identifies 27 estimates of switcher rates from eight lottery-based evaluations of six private school choice programs in the United States that report information about which types of school students enroll in after they apply to a choice program and do not win a lottery. Switcher rates for both full samples and subgroups of students range from 52 percent to 98 percent. Lower bound weighted average and median switcher rates from these studies are 84 percent. Upper bound weighted average and median switcher rates are 90 percent and 89 percent, respectively. Switcher rates are slightly higher for African American students participating in three privately funded school choice programs, where the average switcher rate for these students is 93 percent and the median is 94 percent. Switcher rates observed in these studies are remarkably stable across time and states, and they can provide useful information for inferring switcher rates in fiscal analyses.
Acknowledgments
This manuscript benefited from very helpful feedback from Corey DeAngelis and Benjamin Scafidi. Paul DiPerna provided useful comments and guidance. I’m also grateful for comments and insights from five individuals who anonymously reviewed the manuscript. Any errors are the author’s.
Disclosure Statement
No potential conflict of interest was reported by the author.
Notes
1. Valerie Strauss, “Why it matters who governs America’s public schools,” The Washington Post, November 4, 2018, https://www.washingtonpost.com/education/2018/11/04/why-it-matters-who-governs-americas-public-schools/.
2. For example, two different analyses for a tax-credit scholarship bill in Kentucky, one conducted by the Legislative Research Commission and one by the state’s budget director were presented during testimony for the House Appropriations and Revenue Committee.“Private school tax bill: Here’s what legislators, Bevin are saying,” by Mandy McIaren, Louisville Courier Journal, March 5, 2019, https://www.courier-journal.com/story/news/politics/ky-legislature/2019/03/05/hb-205-kentucky-what-legislators-say-private-school-tax/3064086002/.
3. The term “public school,” as used throughout the present paper, refers to both traditional (district) public schools and public charter schools.
4. The same point can be applied to students who leave school districts for any reason, such as moving into a homeschool environment. The fiscal impact on the state will depend on the state’s cost-sharing scheme. The fiscal impact on public schools will depend on the amount of costs that are variable in the short-run.
5. The rate at which students who lost the lottery go on to enroll in nonpublic school environments is the same as the percentage of students who won the lottery but who didn’t need a voucher to enroll in a private school – they would have been enrolled in a private school or a homeschool environment anyway. This number is important in a fiscal analysis because these students may have previously been in nonpublic school environments and so their costs wouldn’t be part of the public school budget. Their presence in the program constitutes a net cost rather than a net savings to taxpayers because in the absence of the program their parents would have paid private tuition for them.
6. Although average cost is what we usually observe in fiscal impact calculations, ideally analyses would incorporate the marginal cost of educating students in the public school system.
7. Note that the equation can be rewritten as:
The first term on the right hand side represents the estimated savings from students receiving vouchers who would likely enroll in public schools if without financial assistance from the choice program. The second term on the right represents the estimated cost to the state to fund vouchers. Setting NFI = 0 and solving for p yields the break-even switcher rate.
8. Costrell scales the proportion of lotter losers who choose public schools by 1/(proportion of students offered a scholarship who use a voucher). For example, suppose 80 percent of students offered a voucher use it, and suppose that 10 percent of students losing the lottery enroll in a nonpublic school. Then the lower bound estimate for the switcher rate is 87.5 percent (= 1– 0.10/0.80) while 90 percent represents the upper bound estimate. Please see footnote 15 on p. 11 in Costrell (Citation2008) for further details.
9. For example, switcher rates that treat students who are never observed in a public school prior to participating in the program as switchers would be biased downward if they do not accurately reflect students in grades not served by public schools, such as kindergarten. In this case, all first grade voucher students would be counted as non-switchers when it’s likely that at least some would be enrolled in public schools if not for financial assistance from the choice program.Unobserved structural switches could also yield biased estimates, though the direction is unclear. For example, switcher rate estimates would be biased upward if they include eighth grade public school students as switchers when some of those students would enroll in private schools even without the choice program. Estimates would be biased downward if they count eighth grade private school students as non-switchers when some of these students would enroll in public schools (perhaps due to cost considerations).
10. Generally speaking, random assignment offers the best research methods for answering questions about the causal effects of education programs and policies (Cook & Payne, Citation2002). These studies test for mean differences for observed characteristics of the treatment and control groups to evaluate whether random assignment was successful in generating “apples-to-apples” comparison groups.Data were not reported in any studies that would allow comparisons of observed characteristics between switchers and non-switchers within control groups. Studies without random assignment that reported switcher rates also did not report differences between switchers and non-switchers. Studying the determinants of switching would be a fruitful avenue for future research.
11. One factor that can pose challenges to fiscal policies and potentially influence the success of choice programs overall is the degree to which students switching to private schools switch back to public schools. This “back-end sorting” or “churn” may develop for a number of reasons, including differences in school discipline policies and choice program’s non-renewal policies. Switchers who stay in private schools short-term will generate different fiscal effects than switchers who stay long-term. The random assignment studies reviewed in the present paper do not include information about the extent to which “back-end sorting” occurs in private school choice programs. This avenue is fruitful for future research.
12. For each year of the program, the Alabama Department of Revenue reports the total number and percentage “of first time recipients continuously enrolled in a public school for the entire previous year.” Alabama Department of Revenue, Alabama Accountability Act, retrieved 5/22/2019 from: https://revenue.alabama.gov/legal/alabama-accountability-act/.
13. For FY 2015, the average voucher in Louisiana’s School Choice Program for Certain Students with Exceptionalities was $2,213 while the estimated average per-pupil variable cost was $15,012. The average voucher amount for the Racine Parental Choice Program was $7,324, more than three times the average voucher amount for Louisiana’s program. The Racine program, however, was not restricted to only students with special needs, and its estimated per-student variable cost was considerably lower, $8,691. In this way, break-even switcher rates for the two programs differed dramatically.
14. The LSP was not actually eliminated and currently remains in operation.
15. Costrell’s estimate is based on evidence from random assignment studies of the DCOSP and privately funded voucher programs in New York City, Washington, D.C., and Dayton, OH. Page 11 in Costrell (Citation2008) provides a summary of this evidence.
16. The switcher rate was based on information suggesting that financial need represented the full tuition at the private school where 40 percent of tuition grant families enrolled. Thus, the authors argued that 40 percent of tuition grant recipients would attend public school in the absence of the tuition grant. They set on the lower switcher rate of 30 percent to be more cautious in their analysis.
17. Buschman and Sjoquist engaged in an exchange with Benjamin Scafidi, who critiqued their methods. For a summary of this exchange, please see Lueken (Citation2018b).
18. Assumptions about switcher rates were based on data by the U.S. Census Bureau indicating that 5 percent of school-aged children living in households with income below 185 percent of FPL attended private school.
19. SGOs that responded to the survey represented 13 percent of scholarships awarded. Several reasons were suggested for why the switcher rate was likely low. Of scholarships awarded by these SGOs, 36 percent went to former public school students. Most SGOs also indicated they exerted limited effort to attract new reasons in order to keep overhead costs low. Moreover, most SGOS at the time were run by people in their spare time who likely had limited capacity to market themselves to attract new students. The program was in its infancy at the time of the survey. As the program expanded and SGOs matured, more scholarships were awarded to new students, and it may be the case that the share of participants who were switchers increased as well.
20. Here, the term “study” refers to an evaluation by a team of researchers of a private school choice program rather than individual publications. Thus, evaluations of one program by different research teams would be considered two studies (as with the LSP and DCOSP programs).
21. Notably, Rouse (Citation1998) used random assignment methods to study the effects of the Milwaukee Parental Choice Program and noted that 1 percent of students not selected to attend a choice school in the MPCP enrolled in a choice school anyway. Private schools participating in the MPCP (“choice schools”), however, represent a very small portion of all private schools in Milwaukee. If 99 percent of students in the control group not enrolling in choice schools all enrolled in district schools, then the switcher rate would be 99 percent. To the extent, however, that some students enrolled in religious private schools that were not enrolled in the MPCP, the switcher rate would be less than 99 percent. Because of this uncertainty, I do not include this study in the analysis.
22. I report switcher rates from the most recent publication. For example, if switcher rates observed in the second year of a program were reported in a year two report and a final report, and if they differ, then I report estimates from the final report. Any observed differences from different reports of the same program were very small.
23. Among students in the control group, between 11 percent and 14 percent had unknown school enrollments. Removing these students yields estimated switcher rates between 86 percent and 93 percent.
24. Removing the 4 percent of students with unknown school enrollments yields a switcher rate estimate of 95 percent.
25. This estimate represents an upper bound. Information about student crossover among lotter winners was not available to estimate a lower-bound switcher rate for the program.
26. The proportions of students in the control group that attended private school during the first three consecutive years in the New York City and D.C. programs were reported by Howell and Peterson (Citation2002).
27. This estimate is equivalent to about $40,000 in 2019 USD. (Bureau of Labor Statistics, Inflation Calculator, accessed 5/21/2019 at https://data.bls.gov/cgi-bin/cpicalc.pl).