Abstract
We analyse the effect of geographic competition between schools on academic performance in Chile. The analysis controls for prior pupil performance, and a range of school and municipality characteristics. We allow for the endogeneity of voucher school location, using the number of local Catholic churches as an instrument. We find that a larger number of public schools positively affects the quality of education of other schools located in the same area, particularly amongst middle-class families and in middle-ranking schools. However, the number of voucher schools is associated with lower performance in neighbouring schools, which we attribute to pupil sorting.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Public schools were also allowed to top up their public funding but only at the level of secondary education.
2. Although scores on the PISA test have been improving since the early 2000s, Chile still scores the lowest of all 35 OECD countries, with the exception of Mexico and Turkey (OECD, Citation2015).
3. In more recent years, English and physical education have also been added.
4. For this research, private schools are dropped from the analysis, since they were never part of the voucher reform and tuition is fully paid by families, with almost no control from the government.
5. Alternatives to using such an output measure (pupil performance) as an indicator of quality of education provision, are input measures such as class size, expenditures, or measures of teachers’ skills (Hanushek, Citation1986). We prefer to use the output measure as capturing the effects of all inputs, rather than focus on a specific input.
6. The distance was selected using the average distance that pupils travel from their residence to their school presented by Chumancero, Gomez Caorsi, and Paredes (Citation2009).
7. Information on which schools received a SNED premium is available at: http://datos.mineduc.cl/dataviews/235866/VISTA-SNED-2004-2005/ .
8. The proportion of voucher schools within regions with catholic links varies in 2016 between around 30 per cent and 50 per cent, with an average value across regions. Source: Ministry of Education (Citation2017).
11. Any level of improvement has been considered.
12. For example, the Metropolitan Region public transportation service has experienced a thorough modernisation and expansion in its coverage, since the ‘TranSantiago’ plan was first implemented in 2005. However, massive chaos was faced by commuters and the new system was largely rejected by popular opinion.
13. The same is observed when the SNED variables are entered separately. Note that the school level SNED indicator attracts a positive and significant coefficient, while the variable measuring the proportion of schools in the municipality in receipt of SNED has an insignificant effect.
14. See in Appendix D.
15. F(19,148) = 36.21, Prob> F = 0.000.
16. The first stage was estimated manually because the instrumented competition variable was interacted with other variables in the second stage regression (column b). Without using bootstrapping (300 iterations) the standard errors in the second stage would be wrong (Wooldridge, Citation2002).
17. In this way, a school’s characteristics within the same group are similar and different to a school’s characteristics in other groups.