Return to search

Three Essays on Human Capital

Human capital investment is of prime interest for many countries at varying stages of development. Knowing both the determinants and the impact of schooling is central for well- designed policy. This dissertation addresses both respects by examining the determinants of secondary school enrollment in Indonesia, and the impact of higher education in South Korea.
In Chapter 1, I begin from the observation that many countries spend substantial resources inducing individuals to attend school. Despite this, high dropout rates are common, particularly when students transition between education levels. To explain this pattern, previous research has focused on supply side factors, such as decreased number of school slots or longer commute times. In contrast, this paper explores a demand side reason for high dropout rates between schooling levels: a nonlinear increase in wage returns from completing the final grade of an education level - a sheepskin effect. I investigate whether schooling decisions in Indonesia are consistent with perceived sheepskin effects. Using four types of income shocks that range from idiosyncratic to systemic (unemployment, crop loss, drought, and financial crises), I test if negative shocks affect enrollment differentially across different grade levels. As in the previous literature, negative shocks reduce children's enrollment probabilities on average. However, consistent with perceived sheepskin effects, this impact is strongly mitigated for students who enter the final grades of junior or senior high school. Moreover, even poor households exhibit this behavior indicating that even the poor are able to continue investments in education when they perceive returns to be sufficiently high.
The remainder of the dissertation begins from the observation that in low income countries, most gains in education attainment have come from expansions at the primary or secondary level. In contrast, middle and higher income countries have seen rapid increases in higher education enrollments. The pace of growth varies considerably, with historically low attainment countries such as South Korea, Belgium and France experienced more than a 40% point increase in the percentage of population with some tertiary education. Despite the salience of these trends, there is limited credible empirical evidence on their impact due to the difficulty in finding a credible exogenous variation.
To address this question, chapters 2 and 3 utilize an unusual policy change in South Korea; the 1980 education reform, which mandated an increase in the freshman enrollment quota by 30 percent nationwide.
Chapter 2 (joint work with Wooram Park) estimates the impact of higher education on labor market outcomes and saving behavior of the household. We use the discrete change in the opportunity to obtain higher education across adjacent cohorts to implement a regression discontinuity design. We find that college education has a substantial positive effect on labor income, employment probability as well as on household savings. We also find that college education reduces the probability of job loss during the 1997 Asian Financial Crisis.
Chapter 3 (joint work with Jisun Baek and Wooram Park) estimates the causal effect of higher education on health related outcomes. Also using a regression discontinuity design, we confirm that the cohorts that are more likely to be affected by the policy have a higher fraction of individuals with college education. However, we do not find evidence of positive health returns to higher education. In particular, we find that the cohorts with higher proportion of college graduates are not less likely to experience disease or report poor health status. Moreover, we find that higher education has limited effects on health behaviors such as smoking and drinking.

Identiferoai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/D8NP22MG
Date January 2014
CreatorsSon, Hye Lim
Source SetsColumbia University
LanguageEnglish
Detected LanguageEnglish
TypeTheses

Page generated in 0.0024 seconds