Research
Research
"Does Homeownership Matter? The Long-Term Consequences of Losing a House during the Great Recession" with Patrick Bayer (Duke), Fernando Ferreira (Wharton), and Stephen Ross (UConn)
Abstract: We examine the long-term impact of owning versus renting a home in the United States. Our research design compares two groups of homeowners who experienced similar negative income shocks during the Great Recession but had different probabilities of retaining their homes due to mortgage modifications. While the two groups exhibit nearly identical pre-trends, they diverge by 36 percentage points in home retention. More than half of this disparity persists nearly a decade later, translating into average housing capital gains of $83,000. However, homeownership and housing wealth did not affect creditworthiness, consumption, or neighborhood quality, partly due to tightening credit constraints. NBER working paper 33692.
"The Push of Big City Prices and the Pull of Small Town Amenities" with Jeffrey Brinkman (Philadelphia Fed) and Svyatoslav Karnasevych (Princeton) (Revise and resubmit, Journal of Urban Economics)
Abstract: As house prices rise in large, supply-constrained cities, what are the implications for other places that have room to grow? We explore how location amenities have differentially driven population and price dynamics in small towns versus big cities. We provide theory and evidence that demand for high-amenity locations has increased. High-amenity counties in large metropolitan areas have experienced price increases, while high-amenity counties in small metros and rural areas have absorbed increased demand through population growth. High-income and college-educated workers sort into large and high-amenity places, while retirees and other households detached from local productivity gravitate to high-amenity small towns. Draft version.
"College Alumni Networks and Mobility Across Local Labor Markets" with Richard Jin (Berkeley)
Abstract: We quantify the impact of alumni networks on the geographic mobility of job seekers for nearly 1,400 US colleges and universities. We use detailed employment and education information on LinkedIn users to isolate college-educated workers who faced an exogenous job separation in a mass layoff or firm closure. Using a nested logit model of location choice, we compare the migration decisions of job seekers who were displaced in the same city and who attended different but similar and geographically proximate universities. We find that a 1% increase in the number of co-alumni in the city of displacement increases a job seeker's odds of staying there by 0.4%. Conditional on moving, a 1% increase in a potential destination's number of co-alumni increases the odds of choosing that city over another by 0.9%. Co-alumni may both impact job search and provide local amenities. Using data on the presence or absence of co-alumni at new jobs, we conclude that the job search channel is particularly important. Co-alumni from the same or neighboring graduating class have much larger impacts on location choice, indicating true network effects rather than idiosyncratic matches between alumni of certain colleges and jobs in certain cities. We also find strong impacts of having more local co-alumni who work in the same industry. Draft version.
"Work from Home and the Geography of House Prices"
Abstract: I quantify the impact of remote work on the geography of house prices. I overcome challenges in measurement and endogeneity using an instrumental variables strategy paired with pre- and post-COVID data on remote work. I find that a one-percentage-point increase in metro remote work results in $10,500 of additional price growth for a median-distance suburb over a central neighborhood of average initial price, which represents $126,000 at average remote work growth. This profound shift in the distribution of home value stands to increase wealth inequality. Draft version.
"The principal problem with principal components regression" with Gary Smith (Pomona), Cogent Mathematics and Statistics, May 2019, DOI:10.1080/25742558.2019.1622190
Abstract: Principal components regression (PCR) reduces a large number of explanatory variables in a regression model down to a number of principal components. PCR is thought to be more useful, the more numerous the potential explanatory variables. The reality is that a large number of candidate explanatory variables does not make PCR more valuable; instead, it magnifies the failings of PCR. Link here.
"Another Look at Dollar Cost Averaging" with Gary Smith (Pomona), Journal of Investing, May 2018, DOI:10.3905/joi.2018.27.2.066
Abstract: Dollar cost averaging—spreading an investor’s stock purchases evenly over time—is widely touted in the popular press because of the mathematical fact that the average cost per share is less than the average price. The academic press has generally been skeptical, and attributes dollar cost averaging’s popularity to investor naiveté and cognitive errors. Yet, dollar cost averaging continues to be recommended by knowledgeable investors as a sensible way to avoid ill-timed purchases. We argue that dollar cost averaging is, in fact, an imperfect, but helpful strategy for diversifying investment decisions across time. Link here.
"Labor Demand and Neighborhood Choice: Spatial Sorting in U.S. Metros"