This dissertation contributes towards our understanding of Labor Economics and Applied Econometrics. It consists of three chapters. The first two chapters shed light on the determinants of female labor supply behavior by connecting theory to household-level data. The third chapter studies the nonlinear generalized method of moments (GMM) in dynamic panels and its application to value-added models of learning.

In Chapter 1, I propose that the rising sex ratio (number of males per female) imbalance has been an important factor in the recent feminization of rural-to-urban migration in China. To establish this connection, I first develop a three-player noncooperative household model in which both the parents and the daughter contribute time or money to improve the well-being of sons. The local sex ratio can affect the players' choices via two channels: either by influencing the preference towards sons, or by imposing negative impact on sons' welfare due to intensified marriage market competition. My model predicts that daughters are more likely to participate in migratory work when the local sex ratio is higher. Drawing on data from Rural-Urban Migration in China Survey, I then test the hypothesis by comparing unmarried rural women with brothers and those without brothers when conditioning on family size. My identification strategy exploits the exogenous variation in the number of brothers a rural woman has that comes from the randomness in parental sibling structure. I show that an increase in the local sex ratio significantly raises the probability of becoming a migrant worker for unmarried rural women who have brothers, while no significant effect is observed among those without brothers. The positive link is stronger for rural women who have a larger number of brothers or whose brothers are relatively younger. I also discover that around 40% of the increase in rural female labor migration rate from 1990 to 2000 could be explained by the changes in the sex ratio. I further find evidence in favor of the marriage market pressure mechanism.

Chapter 2 (joint work with Zhongda Li) examines the intergenerational determinants of women's labor force participation decision. Existing studies have established a positive correlation between a married woman's work behavior and her mother-in-law's. Such linkage is attributable to the profound influence of maternal employment on son's gender role preferences or household productivity. In this chapter we investigate the relative importance of the two potential mechanisms using the Chinese survey data. We show that a substantive part of the intergenerational correlation is left unexplained even if we control for the husband's gender role attitudes. Instead, we find that the husband's household productivity is more crucial in the wife's work decision, suggesting the dominance of the endowment channel over the preference channel.

Chapter 3 develops a novel framework for constructing nonlinear moment conditions in dynamic panel data models. I demonstrate that the nonlinear GMM estimator considerably mitigates the classical weak identification problem arising from two data generating processes: (i) the autoregressive parameter is close to the unit circle; (ii) the ratio of variances of individual heterogeneity and idiosyncratic errors diverges to infinity. I further derive analytical expressions for the bias term of the linear and nonlinear GMM estimators, and show that the use of nonlinear moments results in smaller finite sample bias. In simulation studies, the nonlinear GMM estimator performs well compared to both the difference and system GMM estimators. As an empirical illustration, I estimate the effect of class size reduction and private school attendance on student academic achievement using a value-added model with learning dynamics.

This dissertation consists of two essays on job search and job choice. In Chapter 1, I examine how on-the-job search (OJS) effort of employed workers varies over the business cycle. First, I document new evidence that aggregate OJS effort rises during a recession, as more workers start searching on-the-job and average search intensity increases when unemployment rises. Next, I account for compositional changes in the pool of employed workers and job seekers over the business cycle and find that workers change their search behavior in response to changing economic conditions. In particular, workers are more likely to search due to fear of job loss and search for an additional job when unemployment is higher. In addition, I find that job seekers increase their search intensity when unemployment rises during a recession. In Chapter 2, I estimate the value that workers place on non-wage job characteristics and assess their impact on compensation inequality in Germany. First, I evaluate the incidence of four key job attributes and find large disparities in the prevalence of job attributes by gender, education, and age. Next, I analyze an experiment given as part of a national survey in Germany to estimate the value that workers place on eight non-wage job attributes. In particular, I find that workers are willing to pay 31% of their wage to have a permanent employment contract, 13% of their wage for good promotion opportunities, 10% of their wage for schedule flexibility, and 8% of their wage to avoid overtime work requirements. Finally, I find that accounting for the incidence and valuation of non-wage job characteristics widens compensation inequality in Germany.

The California Policy Lab found that a majority (74%) of people who enrolled to receive homeless services in Los Angeles between 2010 and 2018 had worked in California before enrolling for services. Over one third (37%) worked in the two years prior to receiving homeless services, and about one in five (19%) of individuals were working in the same calendar quarter that they enrolled to receive services. This report provides an in-depth analysis of employment dynamics for people before, during, and after homelessness. The report includes an analysis of quarterly and annual employment rates, earnings, and differences among various groups, with a focus on “recent workers” who had worked within three or four years prior to enrolling for services and who tended to have higher employment rates and earnings than the rest of the sample.

This work has been supported, in part, by the University of California Multicampus Research Programs and Initiatives grant MRP-19-600774.

The first chapter of this dissertation studies the causes of rising sorting between high-skill workers and high-paying firms. Despite accounting for a substantial share of rising wage inequality, little is known about how or why sorting is rising. To understand how, I develop a novel decomposition method to measure the relative importance of different worker flow channels. I find that labor market entry of young workers accounts for about half of the total rise in sorting. To understand why sorting is rising, I use exogenous variation induced by the fall of the Soviet Union to estimate the effect of trade liberalization on rising sorting within German local labor markets. I find that export exposure can account for 14% of the rise in sorting. I then apply the decomposition method to the export-induced changes in employment to confirm an important role for labor market entry in rising sorting.

The second chapter studies the effect of temporary employment shocks on the future earnings of professional golfers. Although a large literature documents the persistent effects of temporary employment shocks on the earnings of wage-and-salary workers, we have little evidence on the effects on self-employed workers. I exploit entry rules of the PGA TOUR to estimate the long-term effects of temporary employment shocks using a regression discontinuity design. Although, I find large earnings differences in first year after an employment shock, these differences quickly dissipate. Furthermore, I find no effects of employment shocks on performance. Golfers have less job stability than typical workers and these differences likely explain why the earnings losses of golfers are less persistent than of wage-and-salary workers.

The third chapter studies the evolution of wages at large firms. Although large firms have paid significantly higher wages for over a century, we document that the large-firm wage premium has declined over the last thirty years. Decomposing pay into worker and firm fixed effects, we show that the decline is due to a reduction in firm effects at large firms, while worker composition has changed little. We also find that the majority of the change occurs within industries.

At the start of the pandemic, Congress temporarily expanded Unemployment Insurance (UI) programs through the federal CARES Act In this report, we mainly focus on Pandemic Emergency Unemployment Compensation (PEUC), which expanded the weeks of UI eligibility This and other emergency provisions were set to expire on September 4, 2021, which is commonly referred to as a “benefts clif ” However, policymakers in some states chose to end these programs before that date This policy report evaluates the impact the expansion and expiration of these programs had on the US labor market using data from the U S Department of Labor, the Current Population Survey, and California’s Employment Development Department Using national data, we fnd the benefts clif caused a dramatic drop in the proportion of unemployed workers that are covered by UI, which was three times as large as a similar clif at the end of the Great Recession To measure coverage, we focus on the UI recipiency rate, which is the share of jobless workers who received UI benefts In comparison, gains to employment were modest at most Our primary measure of employment is the percentage of the adult population that is employed in each month Data on the demographics of extension program benefciaries is only available in California, and it suggests that workers from disadvantaged backgrounds were disproportionately afected by the expirations In California, UI claimants who were relying on the PEUC extension program when it expired were more likely to identify as Black or women Older and less educated workers were also more impacted by these programs expiring As a whole, our results paint a positive picture of the efects of the federal UI expansions on the U S labor market Our fndings of a dramatic decline in UI recipiency without a meaningful rise in employment imply the pandemic UI expansions increased coverage and bolstered incomes of unemployed workers without a substantial efect on employment As in many other studies, these results rely on comparisons of changes in outcomes between states over time The assumptions underlying our research design are discussed in further detail in the report.

This work has been supported, in part, by the University of California Multicampus Research Programs and Initiatives grants MRP-19-600774 and M21PR3278

Every year, over a million Californians receive workforce support and training from state and federally funded programs. In an efort to learn more about the impact of these programs and to improve them, an inter-agency partnership led by the California Workforce Development Board (CWDB) created Cross-System Analytics and Assessment for Learning and Skills Attainment (CAAL-Skills). The CAALSkills partnership facilitates data-sharing across seven California state agencies that deliver thirteen workforce programs. Bringing this data together signifcantly improves the state’s ability to observe who is enrolled in these programs and makes it possible, for the frst time, to measure the impacts these programs have on participants’ employment and earnings.

This policy brief highlights fndings from the frst causal study to estimate the impacts of ten training programs that report to CAAL-Skills, performed by the California Policy Lab at the University of California. The causal impact measures the efect of receiving training on participants’ employment and earnings, relative to what those same workers would have experienced without training. Training participants are compared to matched individuals with similar demographic and earnings histories who did not receive training, and the impact of training is measured as the diference between the trainees’ outcomes and their matched “twins.” This strategy successfully identifes training impacts for most of the training programs. However, there are a few programs — particularly those that serve distinctive populations with specifc employment barriers, such as Vocational Rehabilitation programs — for which the strategy is not successful at identifying valid comparisons. Of the ten programs that provide training, there was evidence on causal impact estimates available for six programs, suggestive evidence on impacts for two programs, and no evidence available for two programs. Full methodological details and fndings, along with improved research options for all programs and the CAAL-Skills partnership are detailed in the report.

This work has been supported, in part, by the University of California Multicampus Research Programs and Initiatives grants MRP-19-600774 and M21PR3278

This study examines how the first two years of the COVID-19 pandemic affected life expectancy in California and the relationship between census tract income and life expectancy relative to prepandemic years. In this retrospective analysis of 1, 988, 606 deaths in California during 2015 to 2021, life expectancy declined from 81.40 years in 2019 to 79.20 years in 2020 and 78.37 years in 2021. Life expectancy differences between the census tracts in the highest and lowest income percentiles increased from 11.52 years in 2019 to 14.67 years in 2020 and 15.51 years in 2021. This ecological study of deaths in the state of California demonstrated that life expectancy declines in 2020 increased in 2021 and that the life expectancy gap by income level increased during the first 2 years of the COVID-19 pandemic relative to the prepandemic period.

This work has been supported, in part, by the University of California Multicampus Research Programs and Initiatives grants MRP-19-600774 and M21PR3278.

Historically, the share of unemployed workers receiving regular UI benefits (recipiency rate) in California has been relatively low (as has also been the case in other states).This Data Point combines administrative data from California’s Employment Development Department (EDD) with monthly Current Population Survey (CPS) data to construct an improved recipiency rate to measure the extent to which unemployed and underemployed Californians are receiving regular UI benefits.

This work has been supported, in part, by the University of California Multicampus Research Programs and Initiatives grants MRP-19-600774 and M21PR3278.