School Choice, Stratification, and Information on School Performance: Lessons from Chile

In the early 1980s, Chile implemented a nationwide school choice system, under which the government finances education via a flat per-student subsidy (or voucher) to the public or private school chosen by a family. At present, about 94 percent of all schools (public, religious, and secular private) are voucher funded. More than half of urban schools are private, and most of these operate as for-profit institutions.¹ Since the early 1990s, Chile has also publicized information on school performance and increased per pupil expenditure substantially.

Despite these and other reforms, Chile has found it challenging to improve students’ learning outcomes.² Hsieh and Urquiola find that the country’s relative performance in international tests did not change much between 1970 and 1999.³ Its performance on the 2000 Programme for International Student Assessment (PISA) is not only much lower than the OECD average but is similar to that of other Latin American countries and low relative to countries with similar income per capita.⁴

Recent discussions and events in Chile reflect disappointment consistent with such findings. For instance, Engel and Navia observe that Chile is doing relatively well in terms of education quantity, but confronts problems with respect to quality.⁵ In 2006, student protests resulted in renewed government commitments to address education quality. Additionally, there was surprising agreement among candidates in the last presidential election that education policy should address high levels of inequality.

Chile’s experience suggests that we have much to learn about the consequences of education reform, especially school choice, and why it has not resulted in learning gains of the magnitude one might have predicted. This paper addresses these issues in two ways. First, we review the previous literature on the impact of choice in Chile, focusing on the effects on average student achievement and stratification. The reform was implemented nationwide, without an experimental design. Naturally, identifying its causal effect is very difficult and the literature has yet to reach a consensus in this area. To shed additional light on this debate, we present new evidence from a regression-discontinuity design that supports previous findings that school choice—at least as it has been implemented until now—has increased stratification while having little effect on average achievement.

Second, we explore some factors that may explain why the evolution of school quality in Chile has been disappointing. We focus on the difficulties researchers have encountered in generating and interpreting information on school performance. Chile has been at the forefront of measuring, analyzing, and disseminating data on student achievement. Our research suggests three cautionary tales regarding such efforts. First, schools’ average test scores are a very good proxy for average student income in Chile, so they are of limited use to either parents or policymakers in terms of identifying especially effective or high value added schools. Second, commonsense approaches to control for parent and student characteristics—such as analyzing school gains between years—yield volatile school rankings. That is, assessments about whether a given school is “good” or “bad” can change substantially between years, misleading parents and policymakers. Third, the presence of test-score volatility complicates the evaluation of education programs that are assigned on the basis of test scores. We show how this volatility in test scores produced misleading estimates of the impact of one of Chile’s signature education programs in the early 1990s.

We conclude by discussing the implications of our findings for current education policy in Chile. While we touch on multiple issues, we highlight three conclusions. First, over the past three decades, Chile’s school system has improved its users’ welfare along many dimensions. At the very least, choice is likely to have raised the welfare of some households by letting them attend schools of their preference. The evolution of aggregate learning outcomes has not been as satisfactory, however. Second, despite concerns regarding the use of test score information for accountability purposes, we argue that the Chilean government should continue to collect and improve the use of student performance data. Third, Chile would be...