Working Papers

There are two major disability insurance programs in the United States -- the Social Security Disability Insurance (SSDI) and the Supplemental Security Income (SSI) program. Beneficiaries of these programs receive both cash and health insurance benefits via Medicare and Medicaid, respectively. This paper focuses on the relative importance of the two types of benefits -- cash and health insurance -- on participation in the two disability insurance programs and participation in the labor market. Most prior work has focused on the role of cash benefits in labor supply and participation, but health insurance benefits could be equally or more important to the disabled. I estimate a multinomial logit choice model of labor supply and program application decision using data from the Health and Retirement Study coupled with administrative F831 records which provide data on applications to the two programs. I find that individuals value the cash and the health insurance benefits differently for SSI and SSDI; further there is heterogeneity across education and gender. While the SSI benefits predict participation decisions for high school dropouts who are drawn to the program primarily by the cash benefits, the SSDI benefits predict participation decisions for high school graduates for whom the health insurance benefits drive participation. The magnitude of these results vary by gender. The relative disincentive effects of the two types of benefits on labor market participation is heterogeneous akin to their relative effects on program participation. Based on these findings, I suggest policies that reduce caseloads and costs while increasing work incentives, without depriving the needy of assistance.

  • Using Subjective Beliefs Data to Characterize Heterogeneity: A Machine Learning Approach (with Daniel Barth, Nicholas W. Papageorge, and Kevin Thom) (draft in preparation)

This paper applies machine learning (ML) methods to data on macroeconomic expectations collected in the Health and Retirement Study (HRS) to better understand heterogeneity in household beliefs. ML techniques allow us to study the patterns of belief formation in the data without taking a priori positions on them. We identify five belief clusters which classify individuals based on their systematic formation of subjective beliefs. One of the clusters, which we refer to as the ``objective cluster", reports subjective beliefs that are consistent with objective probabilities of macroeconomic events. Individuals in the objective cluster have higher levels of education, genetic endowments that are linked to education attainment, higher wealth, and are less risk averse. Other clusters can be broadly categorized as consisting of either those who randomly report a probability from the available distribution or those who consistently report a certain belief probability, irrespective of the question and the year in which the question is asked. Finally, we describe how our results can be used to better inform macroeconomic models of household behavior.

Work in Progress

  • Cohort Trends in Women's Employment: The Role of Family Structure (with Robert Moffitt)

We explore new analysis of the reasons for the turnaround in women's employment in the late 1990s and early 2000s. Two features distinguish it from prior work. First, the analysis focuses on differences in the turnaround for women by marital status and the presence of children. A new finding is that the turnaround was most dramatic for unmarried women without children. The paper explores possible reasons for the concentration of effect in this group. The second feature is that we extend Goldin and Mitchell's (2017) analysis of cohort trends in life cycle patterns of employment to cohort trends in life cycle pattern of marriage and childbearing. The analysis yields new insights into which cohorts experienced the most decline in employment at the time of the turnaround, at what marital-status-childbearing states and how employment trends are connected to family structure trends.

  • Understanding Dynamics in Subjective Mortality Beliefs

In this paper, I aim to better understand the dynamics in subjective mortality beliefs. The HRS collects information on subjective mortality beliefs in each wave of the survey. Using the Jensen-Shannon divergence, I calculate the distance between the prior and posterior beliefs about own mortality between every two waves. This distance could reflect private information unobservable to the econometrician, systematic pattern of belief formation, or some combination of the two. The welfare impact of policies of forced savings plans, for example, depends on an accurate interpretation of subjective beliefs. Responses to beliefs questions about macroeconomics events are arguably free from private information. Belief clusters identified in my previous work (Barth et al, 2020) therefore provide independent information about the pattern of belief formation, and are used to inform how to interpret the dynamics in subjective expectations about mortality, including their relationship to individual characteristics.