If workers are uncertain about their labor market productivity, this uncertainty could have crucial implications for occupational sorting and wage dynamics. This paper starts by documenting the key features in the US labor market that can be explained by the presence of uncertainty about workers’ own skills and the learning process that unfolds depending on specific experiences. A structural model is built on these empirical patterns and characterizes that workers in their initial occupations decide each year whether to stay or quit in response to the wage outcomes while accumulating and learning about their skills. The dynamic discrete choice model, extending conventional discrete-time duration analysis, is estimated using the Kalman filter and a conditional choice probability estimator. I find supporting evidence that workers experience significant uncertainty about their skills. The results depict that the difference in the dynamics of wage distribution across occupations can be attributed to workers’ sorting as well as human capital accumulation in different occupations. Counterfactual analysis shows that, in occupations with non-trivial probabilities of mismatch due to the choices based not on the true skills but on the beliefs, an information provision policy has a sorting effect. By reducing the share of the mismatched, the policy influences employment duration in the initial occupations, resulting in higher wage growth and lower wage dispersion.

This paper provides structural estimates of a dynamic discrete choice model of education and labor market decisions and studies heterogeneity in returns to higher education. The model explicitly accounts for different alternatives for higher education and the mismatch between the education level attained and that required for jobs, called overeducation. In the segmented labor markets, overeducation is the option inherent in college education, providing college graduates with access to the non-college sector, but not the other way around. A conditional choice probability (CCP)-based estimation method enables the model to be identified and estimated using recent Korean data, which is a short panel without assumptions about far into the later lifecycle not observed in the data. The estimation result shows that overeducation makes the wages among college graduates more dispersed, while the option secures their employment. A counterfactual policy of more stringent academic-ability-based sorting into vocational and general high schools is found to reduce academic college graduation driven by its consumption value, with those directed to vocational education lowering the overeducation rate and yielding larger total log wage income by their late twenties.

We study educational investment decisions when students are uncertain about their own ability and preferences. We estimate a dynamic discrete choice model in which we allow agents to be uncertain about their own unobserved type. We estimate the degree of uncertainty by exploiting commonly available data on agents' self-reported most likely outcome. We apply this model to the Korean context and find large uncertainty before important high school track and effort decisions are made, particularly among the (overconfident) low-ability students. Providing more information would lead to stronger sorting as it disincentivizes low-ability students to exert effort and choose academically focused tracks, while it incentivizes high-ability students more. We also find effects in the long run, leading them to more different types of college degrees.

This paper investigates spillover effects from the U.S. stock market to the Korean stock market by considering a functional regression model. Instead of using stock market indices, we consider cross-sectional distributions of all stock returns that comprise the KOSPI index and the S&P 500 index. We use daily data from January 2005 to December 2013 and consider three subsamples: pre-crisis, crisis and post-crisis period. We estimate a functional regression model and adopt novel econometric tools recently proposed by Hu et al. (2016) to analyze the relationship between cross-sectional distributions of two stock markets. We use response functions to examine how each moment and tail probability of the Korean stock market distribution react to shocks in the US stock market distribution. We also conduct variance decomposition and provide interpretations on spillover effect from the U.S. stock market to the Korean stock market. The spillover effects to mean, variance and left tail probability were higher in the post-crisis period than in the pre-crisis period and were the highest during the crisis period. On the contrary, the spillover effect to right tail probability was the highest in the post-crisis period.