Socioeconomic school segregation in a market-oriented educational system. The case of Chile

Abstract

This paper presents an empirical analysis of the socioeconomic status (SES) school segregation in Chile, whose educational system is regarded as an extreme case of a market-oriented education. The study estimated the magnitude and evolution of the SES segregation of schools at both national and local levels, and it studied the relationship between some local educational market dynamics and the observed magnitude of SES school segregation at municipal level. The main findings were: first, the magnitude of the SES segregation of both low-SES and high-SES students in Chile was very high (Duncan Index ranged from 0.50 to 0.60 in 2008); second, during the last decade, SES school segregation tended to slightly increase in Chile, especially in high schools (both public and private schools); third, private schools – including voucher schools – were more segregated than public schools for both low-SES and high-SES students; and finally, some market dynamics operating in the Chilean education (like privatization, school choice, and fee-paying) accounted for a relevant proportion of the observed variation in SES school segregation at municipal level. These findings are analyzed from an educational policy perspective in which the link between SES school segregation and market-oriented mechanisms in education plays a fundamental role.

Keywords:

• equity/social justice
• research

Acknowledgments

Paulina Sepulveda, Claudio Allende, and Amanda Telias worked as research assistants. The research was partially funded by the Ministry of Education of Chile, through FONIDE Fund. Bellei and Valenzuela also thanks the support of CONICYT-PIA project CIE-05. Previous versions of the paper benefitted from comments made by Francisco Gallego, Harald Beyer, Juan Eduardo Garcia Huidobro, and Gregory Elacqua; we also thank four anonymous reviewers for their valuable comments and suggestions.

Notes

1. Note that we are focusing on educational policy variables; additional factors – like cultural aspects linked to family preferences- are not discussed here. Anyway, there is no rigorous evidence about whether or not Chilean parents prefer segregated schools.

2. This changed in 2008, when the Preferential Voucher Law was passed: this differentiated the cost of the public voucher according to the socioeconomic level of students. Our empirical analyses covered until 2008, so they are not affected by this policy change.

3. Shared Financing schools apply a monthly charge per student. Depending on the amount charged, a partial discount of the public subsidy is applied. Even though parental contributions to state-funded schools were approved in 1988, the present framework of the co-payment system was set up in 1993. Although the co-payment system was initially restricted to voucher private schools (both primary and secondary schools), later it was extended to public secondary schools.

4. We also studied the exposure dimension of Chilean schools’ segregation by estimating an Isolation Index (not reported here). The findings were highly consistent with those discussed in this paper. Results are available from the authors.

5. D-Index is calculated by: $\text{D}=\frac{1}{2}{\sum }_{i=1}^I\left|\frac{{\text{EV}}_i}{\text{EVT}}-\frac{{\text{ENV}}_i}{\text{ENVT}}\right|$ where i represents a school within the territory of analysis (country, region or municipality); EV represents the disadvantaged students and ENV the non-disadvantaged ones; while EVT is total disadvantaged students and ENVT is total non-disadvantaged ones.

6. Since D-Index is symmetrical, its value is identical for the group of disadvantaged and non-disadvantaged students, so what is relevant is the definition of the dichotomous separation of the school going population. For a technical discussion about the D-Index see Allen and Vignoles (2007), Hutchens (2004), and Cortese, Falk, and Cohen (1976).

7. For example, the Hutchens index (2004) is sensitive to population size but it is much more difficult to interpret and relate with the available research on this issue, and it tends to display low values when the level of segregation is moderate (Allen and Vignoles 2007).

8. Missing data on father’s education was imputed from the mother’s educational level.

9. SIMCE is the Chilean national testing system, which evaluates all students in 4th, 8th, and 10th grade in alternate years. Since 1998, SIMCE also applies a complementary survey to the families of students being assessed (the response rates of these surveys range from about 80 to 90%).

10. In order to have an estimate of the level of SES school segregation at a national level based on a completely different data source, we calculated the Dissimilarity Index using the information provided by JUNAEB (the public institution that coordinates the food program within the educational system), which is the percentage of students considered disadvantaged at the level of each school. Elacqua (2012) estimated school segregation in Chile by using this dataset. Broadly, our results were consistent with those reported there, but they were highly unstable because the quality of these data is lower. In particular, JUNAEB index is less appropriate for our purposes, mainly because – due to its link with the School Food Program, the proportion of schools that send information varied significantly over time and the value ‘0’ is assigned to schools that do not send information (mainly all non-subsidized private schools and a significant but variable share of voucher private schools).

11. These databases include nearly 780,000 individuals for 2002 and 693,000 for 1992.

12. Also, controlling by size of enrollment and number of schools at municipal level corrects insensibility of Duncan-Index to population size (Cortese, Falk, and Cohen 1976).

13. Only municipalities with at least five urban primary schools were included, since a minimum number is required for families and schools to be able to choose.

14. Since the last Chilean Population Census was conducted in 2002, we decided to use the closer available data on school segregation for conducting regression analyses that controlled for residential segregation. Nevertheless, as a sensitivity analysis, we carried out similar regressions using 2006 data, and confirmed that the conclusions obtained for 1999 were robust.

15. Note that – ceteris paribus – the D-Index tends to be higher as the proportion of students considered in the reference category out of the total population is lower; in our study, the extreme case of this situation are the private fee-paying schools, in which nearly all students are non-disadvantaged. Nevertheless, as we will discuss, the multiple regression analysis showed that the level of school segregation was not a mechanical reflection of the existing proportion of disadvantaged students.

16. Detailed results are available from the authors.

17. Note that – obviously – both measures (level of enrollment and number of schools) are highly correlated.