Working papers
"Distributional Effects of Ability Learning and Career Choice"
Abstract: This paper examines a new channel between wealth and earnings inequality, namely, the opportunity for workers to experiment and learn about their own capacities at work. Selecting into occupations that appropriately match their skills can lead to higher earnings levels for workers. Yet risk averse workers might be reluctant to experiment and to discover their highest earnings potential by moving between jobs because of the downside risk involved. In this paper, I provide a model and empirical evidence of how workers learn about their skills by changing jobs that require different combinations of tasks. Using data from the NLSY combined with data from the Dictionary of Occupational Titles, I structurally estimate a dynamic model that accounts for the career and savings choices of workers who update their beliefs about their own abilities from their productivity realizations in jobs requiring different tasks. The results show that occupational choices reflect constant relative risk aversion implying that lower wealth leads to less experimentation. Using this model, I find that a recently proposed Baby Bonds policy will likely have a large, long-term effect in reducing income inequality over the life-cycle.
Abstract: This paper examines a new channel between wealth and earnings inequality, namely, the opportunity for workers to experiment and learn about their own capacities at work. Selecting into occupations that appropriately match their skills can lead to higher earnings levels for workers. Yet risk averse workers might be reluctant to experiment and to discover their highest earnings potential by moving between jobs because of the downside risk involved. In this paper, I provide a model and empirical evidence of how workers learn about their skills by changing jobs that require different combinations of tasks. Using data from the NLSY combined with data from the Dictionary of Occupational Titles, I structurally estimate a dynamic model that accounts for the career and savings choices of workers who update their beliefs about their own abilities from their productivity realizations in jobs requiring different tasks. The results show that occupational choices reflect constant relative risk aversion implying that lower wealth leads to less experimentation. Using this model, I find that a recently proposed Baby Bonds policy will likely have a large, long-term effect in reducing income inequality over the life-cycle.
"Do Greater Unemployment Benefits Lead to Better Matches? Evidence from Emergency Unemployment Compensation Programs"
Abstract: This paper explores a new aspect of post-unemployment search outcomes, namely occupational choice. Workers' career choice and potential improvement of job match quality upon occupational change is largely studied in the context of on-the-job-searches rather than unemployment. Despite its potential benefits, changing occupations may involve various risks. In particular, this paper documents that those who switch occupations stay unemployed on average 6 more weeks compared to those who stay in the same occupation. In this paper, I examine whether Unemployment Insurance benefits provide additional security in the face of such risks by leveraging cross-state variation in unemployment benefits as part of the Extended Unemployment Compensation program during the early 1990s. Using Matching methods and classic Difference-Differences approach, I find that additional UI benefits as well as household wealth levels and poor match quality with pre-unemployment jobs are deciding factors for workers to experiment and switch job tasks upon re-employment.
Abstract: This paper explores a new aspect of post-unemployment search outcomes, namely occupational choice. Workers' career choice and potential improvement of job match quality upon occupational change is largely studied in the context of on-the-job-searches rather than unemployment. Despite its potential benefits, changing occupations may involve various risks. In particular, this paper documents that those who switch occupations stay unemployed on average 6 more weeks compared to those who stay in the same occupation. In this paper, I examine whether Unemployment Insurance benefits provide additional security in the face of such risks by leveraging cross-state variation in unemployment benefits as part of the Extended Unemployment Compensation program during the early 1990s. Using Matching methods and classic Difference-Differences approach, I find that additional UI benefits as well as household wealth levels and poor match quality with pre-unemployment jobs are deciding factors for workers to experiment and switch job tasks upon re-employment.
"Employer Learning and Multi-Dimensional Ability" with Kyungmin Kang
Abstract: The literature on employer learning has mostly been concerned with testing whether employers learn about true productivity of workers in various education groups, assuming that the rate of employer learning is independent of the type of job task. In this paper, thus, we ask whether the role of employer learning varies by worker task type. We build a model in which workers and firms learn about workers' multi-dimensional skills from productivity signals, where signal accuracy depends on a job's task intensity. To test the model implication that the flow of information is larger at the jobs with intensive tasks, we focus on two task measures, abstract and social tasks, constructed using data from the Occupational Information Network (O*NET). To address the endogeneity associated with job mobility, we take an instrumental variable approach, using the Markov property of Bayesian learning, where the previous periods' occupation and occupational aspiration are used as instruments. We find that employer learning depends on task intensity, especially for cognitive skills. Moreover, the degree of task-based employer learning varies across educational groups. In particular, our analyses show that cognitive tasks play a key role in employer learning for college graduates, while social tasks are more important for high school graduates.
Abstract: The literature on employer learning has mostly been concerned with testing whether employers learn about true productivity of workers in various education groups, assuming that the rate of employer learning is independent of the type of job task. In this paper, thus, we ask whether the role of employer learning varies by worker task type. We build a model in which workers and firms learn about workers' multi-dimensional skills from productivity signals, where signal accuracy depends on a job's task intensity. To test the model implication that the flow of information is larger at the jobs with intensive tasks, we focus on two task measures, abstract and social tasks, constructed using data from the Occupational Information Network (O*NET). To address the endogeneity associated with job mobility, we take an instrumental variable approach, using the Markov property of Bayesian learning, where the previous periods' occupation and occupational aspiration are used as instruments. We find that employer learning depends on task intensity, especially for cognitive skills. Moreover, the degree of task-based employer learning varies across educational groups. In particular, our analyses show that cognitive tasks play a key role in employer learning for college graduates, while social tasks are more important for high school graduates.
Work in progress
"Signaling vs. Learning and Human Capital Accumulation"
Abstract: One important feature of most signaling theory models is asymmetric information, whereby workers know their skill levels but employers do not. In this paper I relax this assumption that workers have perfect information about their ability and replace it with the assumption that workers need to learn about their skills by making human capital investment decisions whose payoffs reveal information about their ability. I construct a learning model of human capital accumulation and analyze the equilibria in such a learning game, focusing on whether it yields different implications than the classic signaling model. I analyze the human capital accumulation decision, the size of the college premium, and student loan policies.
Abstract: One important feature of most signaling theory models is asymmetric information, whereby workers know their skill levels but employers do not. In this paper I relax this assumption that workers have perfect information about their ability and replace it with the assumption that workers need to learn about their skills by making human capital investment decisions whose payoffs reveal information about their ability. I construct a learning model of human capital accumulation and analyze the equilibria in such a learning game, focusing on whether it yields different implications than the classic signaling model. I analyze the human capital accumulation decision, the size of the college premium, and student loan policies.
