Selection, Structural Transformation, and the Cost Disease of Services
Job Market Paper [paper]
Over the last 30 years, labor productivity has risen in manufacturing while stalled in services. The large-scale reallocation of labor from manufacturing into services might explain the different productivity trends when the reallocating workers are less skilled in both sectors. This paper explores the contribution of selection and labor reallocation to the sectoral labor productivity trends in the U.S. Utilizing the panel structure of the CPS, I document that professional services workers reallocating from manufacturing earn more than incumbent professional services workers. The opposite happens to education, health services, and public administration, where workers who move from manufacturing earn less than incumbent workers. These observations from micro data point to a richer pattern of selection that cannot be explained by conventional Roy models. Therefore, I build a quantitative multi-sector Roy model that explicitly allows for new workers to have higher earnings compared to incumbent workers. I calibrate the model to worker-level panel data and show that selection explains 22% of the gap in measured labor productivity growth between manufacturing and aggregate services. All of the effect goes through selection out of manufacturing, since the effect for the two service sectors is much smaller and offsets each other.
The Role of Micro Data in Understanding Structural Transformation (with David Lagakos)
This paper reviews the use of micro-level data for research on structural transformation. We survey the literature on the topics of cross-country productivity gaps, within-country gaps, labor markets, land markets, and infrastructure, and summarize how the use of micro-level data enhances our understanding of structural transformation that is otherwise hard to achieve with aggregate-level data. We suggest several areas that may benefit from more use of micro-level data. Our recommendations on data effort include collecting more panel data over longer years, especially from developing countries, complementing current time use surveys with data from developing countries, and improving the measurement of non-agricultural output. Relatedly, better measurement of physical and business capital is desired. Lastly, we note the rising trend of joining experimental data with structural models, and encourage more studies to takes advantage of exploiting the strength of both approaches.
We refine the classification of search goods and experience goods by the probability of the consumer conducting pre-purchase inspection. If the consumer conducts a lot of search before deciding to purchase, then the product is said to be more of a search good. If the consumer rarely or never searches before purchase, then the product is said to be more of an experience good. In a dynamic model where a monopolist firm sells to a sequence of short-lived consumers, reductions in search costs through a critical value induce consumers to search for information on product quality and encourages the firm to start producing high-quality products with positive probability. Further reductions in search costs decrease the frequency of consumer search and increase the probability of the firm producing high-quality goods. The firm earns more profits if it absorbs all search costs instead of letting the consumers bear all costs.
Work in Progress
Selection, Labor Share, and Income Inequality
While the labor share of income declines in manufacturing and distribution services, it is increasing in a different set of services industries. The heterogeneity of sector-specific skills among workers leads to sector-specific elasticity of labor supply, which contributes to the sectoral trends of labor share through reallocation of workers. At the same time, prices increase most in industries with inelastic labor supply, which leads to more top earners in these industries and worsens income inequality. I show in this project that the selection mechnism contributes to these trends in both the labor share and income inequality.
Workers are more likely to enter their parents' occupations than random. Poor countries are more dynastic than rich countries, and becomes less so as they develop. Elite occupations tend to be more dynastic across countries. This project studies the mechanism of occupation dynasties with a model of selection and explore the implications for economic development.