The effectiveness of private school franchises in Chile's national voucher program

Abstract

There is persistent debate over the role of scale of operations in education. Some argue that school franchises offer educational services more effectively than small independent schools. Skeptics counter that large centralized operations create hard-to-manage bureaucracies and foster diseconomies of scale and that small schools are more effective at promoting higher quality education. We can gain insight into this debate by examining Chile's national voucher program. This paper uses 4th-grade data to compare achievement in private franchises, private independent, and public schools in Chile. Our findings suggest that franchises have a large advantage over independent schools, once student and peer attributes and selectivity are controlled for. We also find that further disaggregating school franchises widens the larger franchise advantage. We conclude that policies oriented to create incentives for private school owners to join or start up a franchise may have the potential for improving educational outcomes.

Keywords:

• school effectiveness
• private school franchises
• Chile's national voucher program

Acknowledgements

For generous feedback, we would like to thank Cristian Aedo, Herald Beyer, Francisco Gallego, Harry Patrinos, Paul Peterson, and Daniel Ortega. This paper also benefited from discussions with John Londregan, Martin Gilens, Henry Levin, Mark Schneider, Jack Buckley, Elif Calki, Andrew Owen, and David Glick. We would like to thank the Chilean Ministry of Education for providing the data. We also thank the funding granted by Iniciativa Cientifica Milenio “Centro de Microdatos” Proyecto P07S-023-F. The contents of the present study are the sole responsibility of the authors.

Notes

 1. Only two significant modifications have been introduced since 1980. The first was in 1994, when the Ministry instituted a shared financing scheme that allowed all private voucher schools – both elementary and secondary – and public secondary schools to charge limited tuition (Montt, Elacqua, González, & Razyinski, 2006). The second was in 2008, when the Chilean legislature enacted the Adjusted Voucher Law (Ley de Subvención Escolar Preferencial or SEP). The SEP law recognizes that it is more costly to educate disadvantaged students by introducing an extra per-pupil subsidy (50% over the base voucher) for students classified as priority in the Ministry of Education's socioeconomic status classification system and for schools with a high concentration of priority students. This system determines whether a student is “priority” based on individual and household surveys collected by the Chilean government. See Elacqua, Mosqueira, and Santos (2009) for details on the decision tree the Ministry of Education uses to classify “priority” students.

 2. Catholic voucher schools are operated by religious orders, parishes, archdioceses, and religious foundations.

 3. Protestant church schools include Methodist, Baptist, Seventh-Day Adventist, Anglican, Lutheran, and Presbyterian churches. There are four private voucher schools of other religious orientations.

 4. Most of the secular non-profit schools are branches of foundations that were created for other specific tasks, such as the Rural Social Development Corporation.

 5. We do not include the private non-voucher schools in this analysis. This set of schools charge high tuition fees, do not receive per-pupil subsidies, and are mainly focused on high income students. In a previous version of this paper, we included private non-voucher schools in our analysis. The results (available upon request) do not change the substantive conclusions reported here.

 6. We include peer-group controls because a body of literature has documented the positive spillover effects of having high-ability peers and the negative effects of being surrounded by disadvantaged students (e.g., Zimmer & Toma, 2000).

 7. It is possible to test whether it is more suitable to estimate a single equation model (which only allows the intercept to vary between different types of schools) or a multiple equations model (where both the intercept and the slope can vary between different types of schools). We estimate a model with a single equation, which includes a set of dummies identifying the type of school, a set of control variables, and a set of interactions between school type dummies and the control variables. In this model, we test the null hypothesis that the coefficients of all interaction terms are jointly equal to 0. This hypothesis can be rejected, which justifies the use of independent equations. These results are available upon request.

 8. In order be able to control for “school choice bias”, information on school selection practices would be required.

 9. For example, according to Elacqua and Fabrega (2007), 62% of parents know the religion of their children's school, 83% know whether or not the schools is public or private, but only 24% know the principal's name. The latter demonstrates the low level of parental knowledge about the administrative features of schools.

10. In a previous iteration of this research, we used the two-stage selection bias procedure developed by Lee (1983) for cases where school choice is among more than two school categories. Here, we use only two categories because it is unlikely that parents can distinguish between an independent private school and a private school that belongs to a franchise. However, our results are substantively similar. These results are available upon request.

11. The results of the selection equation are presented in Appendix 1. This equation is estimated with a Logit model for predicting the probability of attending a private voucher school.

12. Over 80% of primary school students go to school in their home municipality. Thus, the density measure provides a good proxy for local neighborhood schooling options.

13. This assumption was tested. Following McEwan (2001), Neal (1997), and Evans and Schwab (1995), we re-estimated the achievement regressions by including density measurements as an independent variable. The results, which are available upon request, suggest that the exclusion restrictions are reasonable.

14. In addition, the proposed instruments have other potential problems. For example, the instrument is likely correlated with parent preferences for quality schools due to student mobility and thus is correlated with the error term in the student achievement equation, making the school choice estimates biased. Thus, school density may partially reflect unobserved characteristics of families or communities. In other words, school density cannot properly be excluded from the achievement regressions.

15. The authors present a method for decomposing the selection bias. This approach permits an analysis to develop an intuition and makes a judgment about the sources, severity, and direction of sample selection bias.

16. Although the strength of this relation is not well established (see Sirin, 2005; White, 1982).

17. We use private independent schools as the omitted reference category because we are interested in comparing private franchise and independent school outcomes.

18. In a separate analysis not reported here, we excluded the largest private voucher school franchise in Chile that has 147 schools to make sure it was not confounding our findings. The results (available upon request) do not change the substantive conclusions reported here.

19. In addition, our findings are substantively similar when we only consider urban schools in the Metropolitan Region of Santiago. These results are available upon request.

20. An additional topic for future research that is relevant for other countries, which is beyond the scope of this paper, is to compare the effectiveness of smaller and larger public school districts (municipalidades).