"Effects of Cohort Size on College Premium: Evidence from China's Higher Education Expansion" (Job Market Paper) - Download
This paper examines heterogeneous trends of the college earnings premium by age groups from 1995 to 2013 in China. Specifically, based on China Household Income Projects, the college premium for the younger group (age 25-34) stagnated, while the college premium for the older group (age 45-54) increased substantially. Following the model by Card & Lemieux (2001), I attribute the stagnation for the younger group to the fast-growing relative supply of younger college workers due to China's higher education expansion. Holding the age cohort and survey year constant, a one unit increase in log relative size of college workers leads to 10.3 log points decrease in college premium and 17.8 log points decrease in four-year college premium. I further find that the negative cohort size effect is much more substantial among the new entrants (age 25-29) than among the experienced workers (age 30-54). By this pattern, I not only demonstrate that the new labor market entrants are more sensitive to their cohort size but also argue that the confounding ability composition effect should not be a serious issue.
This paper examines heterogeneous trends of the college earnings premium by age groups from 1995 to 2013 in China. Specifically, based on China Household Income Projects, the college premium for the younger group (age 25-34) stagnated, while the college premium for the older group (age 45-54) increased substantially. Following the model by Card & Lemieux (2001), I attribute the stagnation for the younger group to the fast-growing relative supply of younger college workers due to China's higher education expansion. Holding the age cohort and survey year constant, a one unit increase in log relative size of college workers leads to 10.3 log points decrease in college premium and 17.8 log points decrease in four-year college premium. I further find that the negative cohort size effect is much more substantial among the new entrants (age 25-29) than among the experienced workers (age 30-54). By this pattern, I not only demonstrate that the new labor market entrants are more sensitive to their cohort size but also argue that the confounding ability composition effect should not be a serious issue.
"China's Higher Education Expansion and College Premium: IV Estimations" with Christopher Bollinger, R&R China Economic Review Download
China's higher education expanded substantially after 1977 when the National College Entrance Examination (NCEE) was restored. The rapidly changing access to college provides a rare opportunity to estimate a local treatment effect (LATE) of college education on earnings by utilizing the drastic exogenous increase in college admission rate in 1999. We also utilize the yearly admission rate as an instrumental variable and to construct an additional interaction instrumental variable for the endogenous college education. Using China Household Income Project 2013, the two IV estimates of college premium are 75.7 and 57.5 log points respectively . Given that the IV estimates are the weighted average effects for the group of individuals who are affected by the IVs, both estimates provide strong policy implication that China's higher education expansion has substantially improved the earnings for those induced into college by the expansion. Our study contributes to the literature exploiting natural experiment as instrumental variables to identify the earnings effect of education.
China's higher education expanded substantially after 1977 when the National College Entrance Examination (NCEE) was restored. The rapidly changing access to college provides a rare opportunity to estimate a local treatment effect (LATE) of college education on earnings by utilizing the drastic exogenous increase in college admission rate in 1999. We also utilize the yearly admission rate as an instrumental variable and to construct an additional interaction instrumental variable for the endogenous college education. Using China Household Income Project 2013, the two IV estimates of college premium are 75.7 and 57.5 log points respectively . Given that the IV estimates are the weighted average effects for the group of individuals who are affected by the IVs, both estimates provide strong policy implication that China's higher education expansion has substantially improved the earnings for those induced into college by the expansion. Our study contributes to the literature exploiting natural experiment as instrumental variables to identify the earnings effect of education.
“Education-Occupation Matching and the Stagnation of College Premium for China’s Younger Workers”
This paper further explores the channel through which the cohort size affects the college premium, focusing on education-occupation matching. The higher education expansion has led to an increasing number of college educated workers. Workers can be placed into jobs which have typically been held by people with less education. The literature on education-occupation matching calls these OE (Over-education). These over-educated individuals can face a lower rate of return to education. This paper examines the extent to which the observed stagnation of the college premium for younger workers can be explained by the increasing over-education among them. However, the basic OLS estimation might be biased due to endogenous occupational selection. To address this issue, this paper employs the logit-OLS two-stage estimation proposed by Lee (1983).
This paper further explores the channel through which the cohort size affects the college premium, focusing on education-occupation matching. The higher education expansion has led to an increasing number of college educated workers. Workers can be placed into jobs which have typically been held by people with less education. The literature on education-occupation matching calls these OE (Over-education). These over-educated individuals can face a lower rate of return to education. This paper examines the extent to which the observed stagnation of the college premium for younger workers can be explained by the increasing over-education among them. However, the basic OLS estimation might be biased due to endogenous occupational selection. To address this issue, this paper employs the logit-OLS two-stage estimation proposed by Lee (1983).