UC Berkeley

Mexico City's public high schools use a competitive, choice-based assignment system to allocate students to seats. Similar mechanisms are used in many countries and cities throughout the world. This dissertation explores two aspects of Mexico City's schools: the importance of peer effects on school choice behavior and the consequences of being admitted to a "better" school.

Students have incomplete information about the schools in their choice set, which may make choice difficult. Chapter 1 argues that while new information about a school allows students to update their beliefs about student-school match quality, which may make students more or less likely to choose the school, it also acts through channels that strictly increase demand for the school. Two such channels are a reduction in uncertainty facing risk-averse students and a direct effect of information on returns to attending that school. Peer networks, then, influence choice by providing students with information about some schools but not others. The expected effect of peer-provided information on demand for the peer's school is thus positive. This hypothesis is tested using exogenous variation in older peers' school assignment generated by the allocation mechanism. The average effect of a peer signal on the probability of choosing both the peer's school and observably similar schools is positive, consistent with information increasing expected utility on average. An alternative explanation, that students simply want to go to school with their peers, does not explain the empirical findings. The results suggest that incomplete information has a large impact on school choice even in a relatively information-rich environment, and that social networks partially overcome this problem while encouraging selection into schools attended by peers.

Chapter 2, which is joint work with Alain de Janvry and Elisabeth Sadoulet, explores an important and high-profile question in Mexico City: is there an academic benefit to elite high school admission? Winning a seat in an elite high school both promises modest rewards and imposes substantial risks on many students. We find that admission raises end-of-high school test scores by an average of 0.11 standard deviations for the marginal admittee. On the other hand, elite school admission in Mexico City increases the probability of high school dropout by 8.5 percentage points. Students with weaker middle school grades experience a much larger rise in dropout probability as a result of admission, suggesting that the additional dropout risk is a result of increased academic rigor. We introduce a new "penalized imputation" method to show that the effect on exam scores is robust to accounting for differential dropout.

Chapter 3 explores the effect of marginal admission to a school with higher-ability peers on dropout probability and exam scores, extending some of the results of Chapter 2 to the full set of high schools that fill up during the assignment process. The average impact of admission on dropout and exam scores is negligible for students who barely score high enough to be admitted. One possible explanation for this finding is that relative ability matters for academic performance, which is consistent with the empirical finding larger jumps in peer ability due to admission predict greater increases in probability of dropout. Motivated by this empirical fact, a simple model of school choice accounting for incomplete information about one's own ability and the dependence of academic performance on relative ability is presented. The model shows that under these two conditions, the optimal choice strategy is much more complicated than under standard models of school choice.