Research and work in progress

with Matthew Jacob

Link to paper

How much local taxes affect the spatial distribution of economic activity within metropolitan areas remains unclear. We study the wage tax, a local tax on labor earnings that applies to residents and to non-residents working within city limits. The wage tax has long been argued to have driven firms and workers to relocate from large cities to the suburbs. Guided by a quantitative spatial model, we test this hypothesis using granular data on residence, workplace, and commuting in the Philadelphia metropolitan area from 1940 to 2018. We argue that the elasticity of commuting flows to the net-of-tax rate can be identified from changes in neighborhood commuting shares at the city boundary with a spatial regression discontinuity design. We validate our design and show that suburban commuting to the city discontinuously fell as the wage tax rose, and discontinuously rose as the wage tax fell. Our preferred estimate of this elasticity is 6.39, which we validate with complementary estimates from a gravity equation and from the structure of the model. In a model counterfactual that eliminates the wage tax, we estimate a stock elasticity of the tax base of 0.86. Finally, we use counterfactuals to show that the wage tax raises substantial revenue despite distorting the location of economic activity and acting as a fiscal externality that reduces other tax revenues.

with Gordon Hanson

Link to paper | Link to slides

During the middle of the 20th century, U.S. manufacturing reorganized itself. A substantial share of industrial jobs were reallocated from the U.S. North to the U.S. South, presaging the exodus of manufacturing from the global North to the global South that would occur several decades later. We study how the rise of the South affected labor market outcomes in the North after 1940, instrumenting for the South's industrial growth using insights from the literature on the region's structural transformation. Northern regions more exposed to the rise of the South saw larger declines in manufacturing employment, small offsetting gains in service employment, and decreases in overall employment rates that persisted out to 1980. These changes were accompanied by reductions in wage and salary income among low-wage workers, especially Black men, which reached their peak impact in 1960 and had mostly recovered by 1980. Impacts on real earnings may have been offset by differential reductions in the cost of housing, which also reached their peak impact in 1960 but remained depressed through 1980. Consistent with the hypothesis of Wilson (1987) on the origins of urban decline in former industrial cities after 1960, adverse changes in Northern labor markets were followed by more intense rioting during the social upheaval of the late 1960s, increased uptake of government transfers to low-income families following the launch of Great Society social programs, and reduced economic mobility of children born to low-income parents in the 1980 birth cohort relative to the 1940 birth cohort.

with Prottoy Aman Akbar and Allison Shertzer

Commuting technologies transformed American urban form in the half-century before 1930. Yet those transformations remain largely unmeasured because data on commuting began only in the 1960 Census. Using Philadelphia’s city directories over 1887-1930, we use large language models to extract name, occupation, industry, employer, home address, and work address for roughly one million directory entries. We then study the response of establishment location and individual commuting to changes in transportation technologies over that period. In preliminary work, we find that the electrification of streetcars 1) centralized service establishments downtown, 2) had no effect on the location of manufacturing establishments, and 3) decentralized residences of workers regardless of industry or occupation. 

with Alfonso Cebreros, Aldo Heffner-Rodríguez, and Daniela Puggioni

Bank of Mexico Working Paper 2020-04

We use estimates for the probability of automation of occupations in Frey and Osborne (2017) together with household survey data on the occupational distribution of employment to provide a risk assessment for the threat that automation may pose to the Mexican labor market. We find that almost two thirds of total employment is at high risk of automation; slightly more than half if we only consider employment in the formal sector. We argue that, while these estimates provide a useful benchmark to start thinking about the impact that automation may have on the labor market, they should be interpreted with care as they are solely based on the technical feasibility to automate and do not reflect the economic incentives, or other factors such as the accumulation of human capital through education, to adopt automation technologies.