Research
Working Papers
The Lock-In Effect of Rising Mortgage Rates, with Ross Batzer, Will Doerner, and Mike Seiler
FHFA Working Paper Public Use Data Media Coverage: New York Times, Business Insider ( 1 ) ( 2 ) ( 3 ), Newsweek, Fast Company, Realtor.com, Investopedia, Barron's, Calculated Risk Blog, Zillow, Top of Mind Podcast
People can be “locked-in” or constrained in their ability to make appropriate financial changes, such as being unable to move homes, change jobs, sell stocks, rebalance portfolios, shift financial accounts, adjust insurance policies, transfer investment profits, or inherit wealth. These frictions—whether institutional, legislative, personal, or market-driven—are often overlooked. Residential real estate exemplifies this challenge with its physical immobility, high transaction costs, and concentrated wealth. In the United States, nearly all 50 million active mortgages have fixed rates, and most have interest rates far below prevailing market rates, creating a disincentive to sell. This paper finds that for every percentage point that market mortgage rates exceed the origination interest rate, the probability of sale is decreased by 18.1%. This mortgage rate lock-in led to a 57% reduction in home sales with fixed-rate mortgages in 2023Q4 and prevented 1.33 million sales between 2022Q2 and 2023Q4. The supply reduction increased home prices by 5.7%, outweighing the direct impact of elevated rates, which decreased prices by 3.3%. These findings underscore how mortgage rate lock-in restricts mobility, results in people not living in homes they would prefer, inflates prices, and worsens affordability. Certain borrower groups with lower wealth accumulation are less able to strategically time their sales, worsening inequality.
Capitalization of Property Tax Incentives: Evidence From Philadelphia
Abstract: In 2000, Philadelphia enacted an abatement policy that exempted new development from property taxes for 10 years. This policy provides an ideal natural experiment to test property tax capitalization because it creates contemporaneous intra-jurisdiction tax variation within a finite and known duration. Consistent with theory, the tax benefits are initially capitalized fully into home prices. However, as abatements near expiration, the benefits become overcapitalized in home prices. This paper also finds that escrow payment shocks cause delinquencies for owners of homes with expiring abatements.
Works in Progress
The Effects of Philadelphia’s 10-Year Tax Abatement
In 2000, the City of Philadelphia implemented a generous tax abatement policy that exempts the value of new development from 10 years of property taxes. This paper uses a dynamic version of the DiPasquale-Wheaton four-quadrant model to show that, in theory, this should lead to a large (~50%) initial spike in development but a much smaller (<5%) long-term increase. Philadelphia did indeed experience a development boom in the early 2000s. However, this boom was not unique to Philadelphia. A synthetic control analysis cannot reject the null hypothesis that the policy did not affect development. This may be because, prior to the abatement policy, construction costs in Philadelphia were well above the price of new development, and therefore, supply was relatively inelastic.
Endogenous Depreciation in Housing
Housing depreciation is typically modeled as an exogenous parameter. In models where endogenous depreciation results from owner decisions about maintenance. This theoretical paper models that builders exercise some degree of control over the depreciation of the housing they build. This means that deprecation should vary predictably depending on things like real interest rates and property tax rates when a structure was built. A 100 basis point increase in real interest rates can raise optimal depreciation rates by as much as 50 basis points.
Alternative RDD Methodology
Regression discontinuity designs (RDD) are often used to estimate causal effects when the probability of treatment is discontinuous in some running variable. A popular RDD methodology estimates the relationship between the dependent and running variables on each side of the discontinuity separately. This traditional estimator is described by Hahn, Todd, and Van der Klaauw (2001). An alternative is to use a semiparametric estimator inspired by Robinson (1988) that leverages all the data simultaneously. The properties of this estimator in RDD settings are explored by Porter (2003). I build on Porter's theoretical results using Monte Carlo simulation to demonstrate a variety of scenarios where this estimator will outperform the Hahn et al. estimator. Such scenarios include typical sharp and fuzzy discontinuities as long as the dependent variable is continuously differentiable in the running variable. It also includes scenarios where the discontinuity point itself is unknown or fuzzy. These scenarios can arise if treatment is unobserved or measurement error in the running variable.
Mortgage Delinquency After Expected and Unexpected Payment Shocks
Are Credit Card Revolvers Subsidizing Transactors?