Research

Using LinkedIn career histories of more than 52 million workers between 2005 and 2023, we construct annual job-loss risk indices at the occupation-by-area and industry-by-area levels. The indices combine the frequency of involuntary job separation with the severity of the wage loss following separation. Linking these measures to mortgage performance, we find that borrowers in riskier occupations and industries are more likely to default within the first two years after origination. A one-standard-deviation increase in job-loss risk is associated with a 0.27 percentage point increase in the default rate, about 12 percent of the sample mean. Yet lenders only partially recognize this risk: borrowers in riskier industries are less likely to be approved, and their loans are more likely to be sold, but interest rates reflect neither occupation nor industry risk. Job-loss risk is therefore underpriced at origination, and the unpriced exposure is transferred to the secondary market.

with Jon Wiley

We investigate inconsistencies in property tax assessment regressivity across property classes using public record data from Cook County, Illinois during 2006-2022. Across a battery of tests, we document that assessment regressivity is severe within Class 3 (large residential), Class 5A (commercial), and Class 5B (industrial), whereas assessment ratios are only marginally regressive within Class 2 (small residential). Property Classes 3, 5A, and 5B are significantly more likely to appeal, have successful appeals, and have greater value reductions following a successful appeal, compared to Class 2. The appeal process acts to further exacerbate regressivity within Classes 5A and 5B. We calculate the counterfactual tax rate that could have been applied under the hypothetical scenario of zero regressivity within each property class and assuming that assessment ratios equal official assessment rates. The weighted average tax rate across jurisdictions in Cook County during 2006-2022 is 7.8 percent, and the counterfactual tax rate that produces the same property tax revenue under the above conditions would have been 6.1 percent.

with Yongqiang Chu, Thao Le, and Lily Shen

We study the impact of buyers’ political identity on home purchasing using North Carolina voter records linked to more than 250,000 housing transactions and over six million comparable MLS listings. Using a Bartik-type shift-share IV framework, we find that buyers are significantly more likely to purchase homes in politically aligned neighborhoods, with alignment increasing purchase probability by about 1.58 percentage points relative to a 3 percent baseline. The effect is strongest among Republican buyers and declines sharply from 2010 to 2019 despite rising national polarization.

Presented at: AREUEA-ASSA Conference (2027)*; ARES 42nd Annual Meeting, Destin, FL (2026); Homer Hoyt Institute, West Palm Beach, FL (2026)*

* presented by co-author

with Lily Shen

We study whether actuarial flood insurance repricing is transmitted to real estate markets through mortgage credit. FEMA’s Risk Rating 2.0 replaced the National Flood Insurance Program’s zone-based pricing system with property-level actuarial rates. Using NFIP policy microdata, ZTRAX housing transactions, and HMDA mortgage applications for Florida, we identify tracts exposed to premium increases and examine both housing-price capitalization and mortgage-credit responses. Treated tracts experience an average 2 percent sale-price decline and a 1.4 percentage-point drop in mortgage approval rates. Application volume does not fall detectably, but approval declines are concentrated among low-income applicants, while high-income applicants experience little detectable change. Price effects do not vary systematically with tract-level educational attainment, providing limited support for a pure household-learning explanation. Banks and nonbanks reduce approvals at similar rates, but nonbanks additionally raise interest rates and securitize more originated loans in treated tracts. The results identify a lender-mediated affordability channel: flood insurance repricing raises current and expected total housing costs, and mortgage lenders transmit this shock by tightening credit for borrowers with limited income capacity.

Presented at: AREUEA-ASSA Conference (2027); Cambridge Real Estate Finance and Investment Symposium (2026)*; JRER/UCF Current Issues in Real Estate Symposium, Orlando, FL (2026)*; ARES 42nd Annual Meeting, Destin, FL (2026)

* presented by co-author. Previously circulated as “Who Prices Flood Risk and When: Evidence from Housing Markets.”