Publications and Active Research Projects
The labor market reflects both the institutional structure of the economy and the decisions of individuals within it. Between globalization, new technologies, and the introduction of new policies, work has changed. My research is focused on worker decisions and the institutional environment that surrounds them.
I am particularly interested in how nonstandard work arrangements interact with the changing nature of work, and policies like the Fair Labor Standards Act, Unemployment insurance, the Earned Income Tax Credit, and the Affordable Care Act.
I use survey and administrative data sources to inform research designs for policy analysis. I am also interested in the use of machine learning methods in social science research as it relates to predictive problems and impact estimates. This includes the linking of large administrative data sources and predictive trends in survey data for unobserved time periods.
The Minimum Wage, Self-Employment, and the Online Gig Economy (Forthcoming at the Journal of Labor Economics)
- This paper estimates the response of work that is not covered by minimum wage laws to minimum wage increases. I find minimum wage increases in the early 2000s resulted in small reductions in engagement in traditional self-employment. Following the development of the online gig economy in the 2010s, a 10% increase in the minimum wage increased the number of nonemployer establishments classified as transportation and warehousing services by approximately 2.7%. The counties most likely to exhibit a positive relationship between the minimum wage and participation in uncovered work are those with low labor market concentration and active Uber marketplaces.
Medicaid Expansion and Tax Evasion Among the Self-Employed (Under Review)
- This paper tests whether the expansion of Medicaid following the Affordable Care Act impacted reported self-employment. Using administrative tax data, I find evidence of a negative effect of Medicaid expansion on nonemployer establishments. I estimate that states which expanded Medicaid see a reduction in the number of nonemployer establishments of 2.17%, and a reduction in total declared receipts of 1.43%. Using data from 2013, before Medicaid expanded, this was equivalent to a $9.6-billion reduction in declared earnings by the self-employed and roughly 300,000 fewer declared nonemployer establishments among states which expanded Medicaid. Using data on Uber, an informational reporting platform, I find evidence that the reduction in declared self-employment is significantly manipulated in reference to the means-tested Medicaid expansion.
- Early studies have established that the expanded Child Tax Credit (CTC), which provides monthly cash payments to most families with children in the United States, has substantially reduced poverty and food hardship since its introduction in July 2021. Some researchers posit, however, that the CTC payments may generate negative employment effects that could offset its potential poverty-reduction effects. Scholars have simulated various employment scenarios using different assumed labor supply elasticities, but no study to date has empirically assessed how the CTC payments to date have affected employment outcomes using real-world data. To evaluate actual employment effects, we follow previously-established methodology used to estimate other actual CTC impacts, applying a series of difference-in-differences analyses using data from the monthly Current Population Survey files from April 2021 through August 2021 and the Census Household Pulse Survey microdata collected from April 14 through September 13, 2021. Across both samples and several model specifications, we find very small, inconsistently signed, and statistically insignificant impacts of the CTC both on employment in the prior week and on active participation in the labor force among adults living in households with children. Further, labor supply responses to the change in CTC do not differ for households previously earning within the phase-in range of the prior CTC, in striking contrast to the predictions of the simulation work. Thus, our analyses of real-world data do not support claims that the CTC has negative employment effects that offset its documented reductions in poverty and hardship.
- Online Appendix
Multiple Job Holding and the Minimum Wage
- This paper tests whether the increase in state minimum wage policies from 2013 to 2016 impacted multiple jobholding and the balance of work hours between jobs. Using the Survey of Income and Program Participation, I find minor evidence of an effect of the minimum wage on the multiple jobholding status. A $1 increase in the minimum wage is shown to increase the probability of transitioning into multiple jobholding over a 12 month period post treatment by 0.6%. I do not find evidence of a significant effect on hours worked or earnings across primary or secondary sources of income. In total, minimum wage policies appear to have little impact on multiple jobholding. These findings highlight that minimum wage policies are not an effective way to reduce the necessity of multiple jobholding among low-wage workers.
Measuring Seasonal Poverty
- Using large representative surveys collected at uniform intervals over time and across Cote d’Ivoire, Ghana, Malawi, and Peru, we demonstrate that poverty rates vary widely in different periods of the year. We use random forests to develop a feasible method of measuring seasonal influences on consumption which does not rely on a priori knowledge about ecologies or the functional form of an equation governing consumption. Our method of seasonality can be used to incorporate seasonality and timing of poverty into poverty measurement to assess welfare. Applications of these measures could be used in targeting applications to differentially target households who are chronically poor vs only expected to be poor in lean seasons or in impact evaluations of programs expected to reduce exposure to seasonality.
Low Wage Worker Identification in Administrative Data: The Case for Machine Learning
- With the expansion in availability of merged administrative data sets, finding ways of leveraging new data sources can prove enlightening. When considering policies and research targeted at low wage workers, one hurdle is the identification of who is low wage. Many data sources include quarterly and annual earnings for workers, but without information on hours worked, it is impossible to distinguish with certainty between workers with low hourly wage and a significant number of hours, and high hourly wage workers with few hours. Using Washington State unemployment insurance data, I test methods of identifying low wage workers in datasets where hours data is missing. I replicate the methods used by Jardim et al. (2018) in their paper “Minimum Wage Increases and Individual Employment Trajectories” using parametric and nonparametric machine learning methods to test the reliability of low wage cohort selection. I find that random forests are a more reliable method of predicting low wage status in administrative data than parametric logit models or industry subsets. This paper establishes two primary results: (1) machine learning methods may help bridge administrative datasets with similar characteristics but different focuses, expanding research opportunities, and (2) the use of industry cohorts as indicative of low wage status for minimum wage research may produce misleading estimates of the aggregate effect of policy changes.