Research Papers


House Price Contagion and U.S. City Migration Networks (Job Market Paper)

Latest Draft: Download here.

Abstract: Why do national trends in house prices spread more to some cities than to others?  This paper proposes an explanation of house price contagion based on migration spillovers between U.S. cities: Increases in house prices as a result of local economic shocks and housing supply constraints drive out-migration to other cities. These migration flows are more likely to affect cities with stronger pre-existing migration links to the origin cities, and increase house prices at these destinations. I use the network structure of inter-city migration to develop an instrument for identifying causal spillover effects between cities: An increase in other cities' house prices by 10% in the long run causes a 6.3% house price move in a city exposed to the shock through migration links. I show that migration spillovers from the effect of interest rate declines on house prices in other cities can explain 32% of the cross-sectional variation in house price growth during the run-up  to the housing boom of the 2000s. To quantify the effect of changes in migration costs and housing supply constraints on these house price spillovers, I develop and estimate a dynamic spatial equilibrium model that incorporates forward-looking migration choices. After estimating this model with U.S. data, I show that lower migration costs substantially reduce the dispersion in house price growth: without worker mobility, the spread in house price growth across cities in response to wage shocks would be 65-70% larger. Moreover, declines in worker mobility increase the impact of housing policy on the distribution of house price growth across cities.

Employer Concentration and Outside Options (w. Anna Stansbury & Bledi Taska)

[Working paper PDF]  [Occupational Transition Data]

Abstract: This paper studies the effect of employer concentration on wages in the United States. We make two primary new contributions. First, we develop an instrument for employer concentration, based on differential local exposure to national firm-level trends. We use the instrument to estimate the effect of plausibly exogenous variation in employer concentration on wages across the large majority of U.S. occupations and metropolitan areas. Second, we adopt a flexible “probabilistic” approach to labor market definition, identifying relevant job options outside a worker’s own occupation using new occupational mobility data constructed from 16 million resumes, and estimate the effect of these outside-occupation options on wages. We find that moving from the median to the 95th percentile of employer concentration reduces wages by 3%. But we also reveal substantial heterogeneity: the effect of employer concentration is at least four times higher for low outward mobility occupations than those with high outward mobility. Since the majority of U.S. workers are not in highly concentrated labor markets, the aggregate effects of concentration on wages do not appear large enough to have substantial explanatory power for income inequality or wage stagnation. Nonetheless, our estimates suggest that a material subset of workers experience meaningful negative wage effects from employer labor market power. Our findings imply that labor market regulatory agencies and antitrust authorities should take employer concentration seriously, but that measures of employer concentration – typically calculated for narrowly defined occupational labor markets – need to be adjusted to incorporate the availability and quality of job options outside the focal occupation.


Research in Progress


Automation, Career Values, and Political Preferences (w. Maria Petrova, Pinar Yildirim, Bledi Taska)

[Slides available on request]

Abstract: Recently, there has been much evidence linking economic shocks in the form of automation to employment and wage outcomes, as well as political outcomes. In this paper, we try to understand the mechanisms through which economic and political effects of automation are linked. In particular, we go beyond current worker outcomes by introducing a new measure of future career prospects. We show that automation does not only affect current wages, but that occupations also differ in the degree to which workers' career is affected by automation, as automation affects both wages in jobs that workers might aspire to move into, and the likelihood of different career moves. Moreover, the labor market effects of automation differ by demographic group and local area characteristics. We then demonstrate that these patterns of heterogeneity in the impact of automation align with shifts in voter preference towards Donald Trump in the 2016 election -- with negative impacts predicting a shift in preference towards Trump.


Career Assets and Mortgage Leverage

Abstract: The asset on workers’ implicit household balance sheet that best matches their mortgage liability in terms of location risk and duration is their “career asset” – the value of the expected future local trajectory of the worker’s income.  Locations with better career prospects in the form of higher income growth imply a larger career asset relative to current income – and a higher mortgage debt-to-income ratio if workers match mortgage liabilities to career assets rather than current income. To test this theory of household leverage choice empirically, I develop a new measure of the value of local career assets in different occupations using long-run worker mobility inferred from resumes. I combine this measure with data on the variation in debt-to-income ratios across cities from HMDA and Fannie Mae Single-Family Loan Performance Data to show the empirical effect of city differences in career prospects on mortgage leverage choices. 


Deep Learning Deep Skills: Firm Differences in Job Tasks and Stock Market Performance (with Gabriel Levin-Konigsberg)

Abstract: Are workers with the same job title performing different tasks at different firms – and what can we infer from that about future firm performance? In this paper, we study differences in social and interactive task content across firms within the same job.  We use a Recurrent Neural Network approach on a large data set of the text that firms use in job postings to infer linguistic features that predict more interactive processes requiring greater social skills. We then apply this classifier to individual firms’ postings to predict the degree to which firms are social workplaces with an interactive structure of internal processes. We document changes in the social task content of jobs over time and test the ability of the social organization index to predict wage differences across cities, as well as firm stock market performance and innovation.