Three Essays on Economics of Public Policy

Abstract

My dissertation focuses on developing and applying program evaluation techniques to better understanding how public policies affect low-income households and social mobility. In general, my research attempts to address three questions regarding the effect of public policies: (1) What is the long-term effect of the policy? (2) Does the policy foster social mobility? (3) Is there an unintended consequence of the policy? In my view, equality of opportunity is one of the pillars of a free society. I favor the idea that poor children have equal opportunities for success. Since children from low-income families grow up in a relatively disadvantaged environment, public policies that redistribute resources to poor children can foster social mobility. However, as well-documented in the literature, redistribution policies lead to a change in incentives. In some cases, these unintended consequences offset the ``benevolence" of the policy. As a labor economist interested in policy analysis, I focus on evaluating a policy from these three perspectives. In the three essays in my dissertation, I answer the policy-relevant questions using different econometric approaches. When an exogenous policy change is available, a simple econometric model with few assumptions can provide credible answers. If we do not have a natural experiment in the context of the question, I model the selection process so that we can still learn from the data. In the first essay, I investigate whether exam preparatory programs in Taiwan are effective. I set up a Bayesian selection model that formalizes the selection process and explicitly takes into account parameter heterogeneity. In the second essay, I study the effect of the expansions of Medicaid in 1990 on childhood obesity. The Omnibus Budget Reconciliation Act 1990 expanded eligibility to children who were born after September 30, 1983 from families below the poverty line. I employ the birth date discontinuity to study the policy effect. In the third essay, I develop a new test based on the empirical distribution functions of the compliers in the Local Average Treatment Effect (LATE) model. This method tests the validity of the LATE model, which is a common empirical strategy when endogeneity is an issue. In my first essay, I estimate the impact of attending exam preparatory programs, in particular “cram schools,” on students’ academic performance. These programs are the product of market system and the Joint Entrance Exam System, which has been in place for decades in Taiwan. I measure the outcome by admission to a public high school and an “elite” high school. Focusing on the problem that students are not randomly assigned to “cram schools,” I approach the issue using propensity score matching and a Bayesian simultaneous-equations model. Using data from a survey of Taiwanese junior high school students in the Taiwan Youth Project, I find evidence that there is an insignificantly negative sorting into exam preparatory programs and attending an exam preparatory program improves a student’s high school placement. Both approaches indicate similar positive treatment effects. The second essay studies the effect of Medicaid expansions on childhood obesity and finds robust evidence of ex-ante moral hazard induced by public insurance. I establish this result by estimating two reduced-form models and a structural model. My reduced-form identification strategy exploits eligibility discontinuity created by the Omnibus Budget Reconciliation Act 1990, which extended Medicaid eligibility to children from families below the federal poverty threshold and born after October 1983. Drawing on the MEPS, I find offering low-income children public insurance leads to an approximately 10-percentage-point increase in the chances of obesity. Combining the MEPS and the SIPP, I am able to investigate the effects of insurance take-up. I estimate a fuzzy regression discontinuity design using Angrist-Krueger two-sample IV estimator (Angrist and Krueger 1992). The results suggest that early insurance take-up induced by the expansions of Medicaid leads to a roughly 5-percentage-point increase in chances of obesity. I also develop and estimate a two-period structural model that quantifies moral hazard, net-wealth effect, and risk preferences. I use the estimates to study the relative importance of income effect and moral hazard in the childhood obesity problem. The estimates of the choice model suggest that net-wealth effect is a statistically significant avenue to the observed policy effect. In the third essay, I develop a method to test the validity of the Local Average Treatment Effect (LATE) model. The LATE model is widely applied to evaluating policies when randomized experiments are impossible. The model relies on two critical assumptions: (1) the existence of a randomly assigned instrument that affects the outcome variable only through the treatment; and (2) the instrument only affects the treatment selection in one direction. The basis for the test is an estimator for the distribution function of the compliers. If the CDFs decrease more than the derived bound, then we reject the assumption of the exclusion restriction. If the CDFs are not completely non-decreasing, then either one of the assumptions fail to hold. To show the applicability, I apply the test to three datasets.