Working Papers




Internal migration and extended families in China
download the paper here

         In this paper, I study how rural-to-urban migration in China affects households’ inter-generational behavior, and the effects of policies targeting migrant households on their welfare. Internal migration from rural to urban areas can have significant welfare effects on migrants and their extended families. In China, migration is often temporary, and most family members of migrant workers are left behind. In these households, many left-behind grandparents look after the children of migrating parents. However, the behavioral and welfare effects of government policies directed towards rural households with potential migrants remain unknown. 
         Using five Chinese data sets on the migration patterns, education choices, financial transfers, and health of multi-generational families, I first present a rich set of stylized facts about migration and household behavior. The evidence shows that in many rural households, parents migrate to urban areas for work when healthy grandparents are able to provide childcare. When the grandparents are sick, migrating parents return to the rural area to provide elder care and pay for their parents’ healthcare. With the facts as a guide, I develop and estimate a structural model of the behavior of migrants and their families. The model features an informal limited-commitment contract over child care, financial transfers, and elder care. Parents and grandparents play a sequential game by choosing migration status, informal contract status, remittances, children’s education, and grandparents’ healthcare. The estimates suggest that poor households adopt the informal contract so that rural consumption, education, and healthcare are funded by the migrants’ remittance.
         I then use the model to evaluate the effects of a set of hypothetical government policies. An urban education subsidy promotes children’s education, increases the migrants’ consumption in the urban area, and does not affect the grandparents’ welfare. But it does not alleviate the problem of children left behind as the government had hoped. An improved insurance coverage that lowers out-of-pocket healthcare costs would reduce the grandparents’ demand for the informal contract. It would generate a welfare gain to the grandparents, discourage parents’ migration, and increase children’s education. The policy counterfactual outcomes imply that policies intended to improve the welfare of one family member can affect the welfare, consumption behavior and migration decisions of all three generations through intra-household cooperation. The design of these policies should account for intra-household responses.



The Labor Market Returns to Advanced Degrees (with Joseph G. Altonji)
download the paper here

         Graduate study has grown rapidly in the U.S. and in other developed countries, but causal evidence on the value of particular degrees is in short supply. In this paper, we estimate the returns to a broad set of graduate degrees. The analysis employs a large number of NSF surveys that provide rich education histories and multiple observations for some individuals. The data allow us to study the interplay among undergraduate field, graduate education, and occupation choice in determining earnings. To control for heterogeneity in preferences and ability, we use fixed effects for combinations of field specific undergraduate and graduate degrees obtained by the last time we observe an individual. The key assumption is that beliefs about talents, preferences, and labor market returns do not change between when a person chose the observed pre-graduate school job and when she started graduate school. We obtain four main results. First, we find large differences across graduate fields in effects on earnings. Second, the incremental value of a given advanced degree often depends on the undergraduate field. Third, the contribution of occupational upgrading to the earnings gain varies with the BA-advanced degree combination. Finally, multivariate regression methods that do not account for selection bias in who chooses particular undergraduate field and graduate fields are often highly misleading.



Estimating the value of educational quality in China
using Beijing school district assignment policies
download the paper here

         I exploit a change in Beijing’s primary school admissions policy in 2014 to estimate parents’ willingness to pay for primary school quality. The new policy required primary schools to admit students from their assigned neighborhoods without admission exams. I estimate the magnitude of the parents’ willingness to pay for educational quality through the change in relative housing prices. I use three identification strategies to distinguish the value of school quality from the value of other local public goods. First, I employ a difference-in-differences analysis to estimate the changes in prices using all houses in Beijing. Second, I combine the diff-in-diff approach and Black (1999)’s boundary discontinuity design to estimate the value of education quality using the subsample of houses near the school district boundaries. Third, I compare the policy effects on sales prices with effects on rents, exploiting the fact that the location of a rental unit does not determine school assignment. I find that parents’ additional willing to pay for a flat in the top 5% of school districts ranges between $24,452 and $54,186. These revealed valuations for high-quality schools are large, particularly compared to the average monthly income of $1,068 in 2014.


Works in progress




The dynamics of college major, graduate degrees and occupation choices (with Joseph G. Altonji)



         We study the links among choices of college major, graduate field, and occupation. It is a multiple-stage decision process, in which the links between stages need to be studied in depth. First, we are interested in the decision to obtain a graduate degree. An analysis of the determinants of whether and when individuals go to graduate school will help us understand (1) the extent to which the probability of going to graduate school depends on individual characteristics and (2) how much of the decision is related to the nature of the field of study, the occupation choice after college, and prior labor market success.
         Second, we study how the choice of type of graduate education depends on college major and work experience before entering graduate school. Our work in progress shows that both college major and past occupation play roles in the choice of graduate field. After completion of education, occupation choices again depend on all past educational and occupational choices. Third, for those who already hold an advanced degree, we are interested in the role of past experience, including college major, occupation prior to graduate school, and graduate degree and field in shaping the current occupation. With panel data constructed using the restricted use NSF data sets, we are able to look at a longer period of an individual’s education/labor market career in greater detail.

         We will build on the above analysis by specifying and estimating a Markov transition model to describe the path of education and work experience. To estimate the transition matrices of this Markov process, we will use information about individual characteristics, undergraduate education, graduation education, and occupation over time. This model will show us how people who start with, for instance, a college degree in engineering, end up in a variety of occupations and associated distributions of earnings.

         As part of the analysis, we will study the role of preference for working conditions and personal life in the choice of college major, graduate degree and field, and occupation. The surveys ask participants about their reasons for attending graduate school, choosing an advanced degree and choosing an occupation. We will use this information to assess how education and job choices are influenced by preferences and job attributes.

         We will also explore gender difference in the links between undergraduate major, graduate degree and major, and occupations. All of the patterns of the transitions we describe above may differ by gender; summarizing them in detail would complement the study of gender differences in education and in the labor market.

Surname discrimination in Chinese justice system


         How do local surnames influence court rulings? I link the Chinese Court Record Data to data on the geographical distribution and density of Chinese surnames to investigate the partiality of the justice system for people with locally powerful last names. While legal and economics research have shown racial and gender discrimination in the justice system, the impact of social power and other potential factors implied by a surname has not been studied. In China, where 92% of the population is of the Han ethnicity, names reveal little about ethnicity. The fairly homogenous population provides a good source of identification to estimate the legal privileges of influential surnames. For example, in the province where Confucius was born, a person who shares his surname has a large chance of being a member of the prestigious Kong family, who are descendants of the ancient philosopher. I will look at how having such a last name affects likelihood to sue, to be sued, and the outcome of court judgments.