Assistant Professor, Department of Economics
Sam M. Walton College of Business
University of Arkansas
Fayetteville, AR 72701, USA
Office: Room 419, Walton College of Business (WCOB)
Research Interests: Macroeconomics, Labor Economics, Development
Email: gsun@uark.edu | Curriculum Vitae (PDF)
Pronunciation: “Ge” /gə/ is pronounced like “guh” (soft g, as in “get”)
Abstract: This paper examines how employers’ expectations about women’s future fertility increase the gender wage gap in contract-based labor markets, which are standard in many occupations involving complex, long-horizon tasks. In such environments, salaries are set in advance based on expected match productivity rather than contemporaneous output. If employers expect women’s productivity to decline more than men’s after childbirth, they offer women lower wages today. Exploiting China’s relaxation of the One-Child Policy as a quasi-experiment, I implement a difference-in-differences design and find that women’s wages declined by 15.3% immediately after the reform, despite no short-term increase in actual births. To interpret these findings, I develop a search-and-matching model with on-the-job human capital accumulation, integrated with a household framework in which noncontractible, fertility-driven effort choices are made. Effort links the two components by governing human capital growth and, in turn, long-run labor-market productivity. Estimating the model using Chinese data, I find that gender differences in expected productivity, rooted in the unequal division of household labor, explain nearly the entire pre-reform wage gap and approximately 80% of its post-reform widening. The policy implication is stark: rules intended to protect women’s employment through contractual provisions may not reduce the wage gap. By reinforcing employers’ present-value pricing, they may instead be offset by ex ante wage markdowns applied to all women.
Abstract: Work experience is an important source of human-capital accumulation and economic growth, especially in advanced economies. Yet poorer countries exhibit lower measured returns to experience. Understanding this pattern is important for explaining why some countries remain poor. This paper proposes a new mechanism that contributes to limited wage growth over the life cycle in developing countries: rapid technological progress can accelerate the obsolescence of previously accumulated human capital, lowering the relative earnings of older workers. Using repeated cross-sectional data across countries and decades, we show that periods of faster GDP per capita or TFP growth are associated with lower relative earnings for older workers, especially among more educated workers. We then develop and estimate an overlapping-generations model in which stochastic technological innovations raise productivity and skill prices but also increase human-capital depreciation. The estimates imply sizable technology-driven depreciation: a 1.5 percent increase in the technological frontier raises the depreciation rate by about 4 percentage points in the skilled sector and 2.78 percentage points in the unskilled sector. Counterfactual simulations show that replacing the U.S. technology process with a faster process calibrated from Indian TFP data flattens the experience profile. The results suggest that rapid development can have an overlooked distributional cost: technology-induced skill obsolescence reduces the relative earnings of older workers and increases inequality across cohorts over the life cycle.
Abstract: Sampled network data are common in empirical research because collecting complete network information is costly, but analyses based directly on sampled networks can produce biased estimates. We propose a flexible method for imputing missing links in sampled networks and show that downstream empirical analyses based on the imputed networks yield consistent parameter estimates. Our nonparametric procedure combines a projection onto observed covariates with a local two-way fixed-effects regression. It avoids parametric assumptions, does not rely on low-rank restrictions, and flexibly accommodates both observed covariates and unobserved heterogeneity. We establish entrywise convergence rates for the imputed adjacency matrix and prove the consistency of GMM estimators based on the imputed network. We also derive the convergence rate of the corresponding estimator in a linear-in-means peer-effects model. Simulations demonstrate strong performance in both imputation accuracy and downstream empirical analysis. We illustrate the method using the microfinance network data of Banerjee et al. (2013).
Abstract: A student’s choice of major is consequential for academic progress and subsequent labor-market outcomes. This paper studies an overlooked determinant of major choice: unexpected grade shocks during the early years of college. Using administrative records from Purdue University, we combine course-evaluation data reporting students’ expected letter grades shortly before final exams with their realized grades to construct grade shocks, defined as actual minus expected performance. We find that unexpected negative grade shocks significantly increase the probability that students leave their current major in the following term. The response is stronger among women: a one-standard-deviation negative shock raises the probability that a female student leaves her major by approximately 2.5 percentage points relative to female students who do not experience such a shock. These findings highlight the role of early academic signals in major sorting and reveal substantial gender heterogeneity in responses to adverse performance feedback.
University of Arkansas
University of Notre Dame
Peking University
Last updated: