Research

We provide systematic evidence on how retail real estate has been affected by the rise of e-commerce. Online shopping has shifted consumer spending away from brick-and-mortar stores, requiring shopping centers and malls to find new tenants and uses. We link establishment records to maps from OpenStreetMap and property operating income and debt from CMBS. We study how differences in baseline exposure to online shopping affect property use and performance, and trace out the effects of national chain store liquidations to understand amplification mechanisms. More-exposed properties lose tenants, employment, occupancy, and income beginning only in 2013. A 10-percentage-point increase in exposure lowers net operating income per square foot by about 2 percent by 2019. Malls have shifted away from exposed categories of retail towards tenants in health care and services, but the shift has been greatest in the malls with the least ex ante exposure. Malls that ultimately shut down do not have greater exposure to e-commerce or greater tenant turnover than others, but do a worse job of finding replacement tenants. Following a chain liquidation, occupancy falls by about 6 percentage points and income by 13 percent within a year. Losses are larger when properties are specialized, replacement markets are thin, or financing is tight. The results show how technological change can lead to the misallocation of durable capital after a shock.

Draft upon request

Older superseded draft here

Consumers prefer to obtain multiple financial products from the same bank. We provide causal evidence of this demand complementarity using a shift-share design that exploits local variation in Conforming Loan Limits, quantifying how shocks to mortgage lending spill over to banks’ deposit demand. We then use this spillover to estimate a dynamic model in which banks compete oligopolistically across multiple markets and strategically leverage customer complementarities. We show that these demand linkages represent an important source of economies of scope for multi-market banks.

Draft upon request

Financial technology has the potential to alter the transmission of monetary policy by lowering search costs and expanding banking markets. We study the reaction of online banks to changes in the federal funds rate. A 100-basis-point increase in the federal funds rate leads to a 30-basis-point larger increase in online-bank deposit rates relative to traditional banks. Consistent with these rate movements, online-bank deposits experience inflows while traditional banks experience outflows. Results are similar across markets with different levels of competition and different demographics, but vary with the stickiness of banking relationships.

This paper documents novel stylized facts linking the decline in new business formation to the rise of superstar firms using comprehensive French administrative data. Industries with larger increases in superstar firms’ market share experience more pronounced decreases in new business creation. Rising concentration discourages low-ability, but not high-ability, entrepreneurs from starting businesses. This results in higher average firm quality, measured by a higher fraction of entrepreneurs who are highly educated, former executives, or serial entrepreneurs. The findings help reconcile seemingly contradictory evidence and align with theories in which technological change benefits the most productive firms while raising barriers to entry.

Many fixed-rate mortgage borrowers in the United States voluntarily pay down low-rate debt even when safe assets pay more. We study this behavior after the 2022 rate hikes, when safe rates rose above many existing mortgage rates and extra principal payments became financially costly. More than one in five borrowers whose mortgage rate was below the safe rate prepaid in a typical month, generating $4.6 billion in present-value losses from 2022–2024. Prepayment is persistent and almost unresponsive to interest-rate spreads, but it responds to cash-flow shocks, including stimulus payments and increases in ARM payments. It is not concentrated among inattentive or unsophisticated households: prepayers have higher credit scores, are more likely to hold stocks and retirement accounts, and borrowers who refinanced when rates were low are just as likely to prepay. Prepayment nearly doubles toward the end of the mortgage as the balance approaches zero. The findings point to debt aversion—a preference for reducing debt even when returns are negative—as a first-order force in household borrowing.

Using detailed credit data and an empirical strategy based on the removal of Chapter 7 bankruptcy flags, we study the effects of credit access on internal migration and neighborhood choice. Increased credit access raises ZIP-code migration rates but has no positive effect on neighborhood quality. We interpret this as consistent with a setting in which unconstrained movers equalize utility across space through their willingness to pay for housing, so constrained households have no incentive to move to better-on-observable neighborhoods when their constraints are relaxed.

Advertising can both expand the market for a product and reallocate market share across brands. We quantify these effects in the mortgage refinancing market. Using sharp geographic discontinuities in television broadcast markets, we show that $1,000 in advertising per capita leads to 1.6 new mortgages, an ROI of about 3x. Nearly the entire increase is due to market stealing rather than new refinancing. Low-rate lenders do not advertise more than high-rate lenders, but their advertisements are more effective. As a result, we calculate that every marginal dollar in aggregate advertising per capita saves consumers 45 basis points on average.

Draft upon request

[slides]

Property tax limitations can create a transaction cost when a property’s tax changes upon sale. These limits typically work by capping reassessment between sales, so taxable value is held below market value while a property is owned and then resets when it changes hands. We quantify the transaction cost this creates and estimate its effects on sales and development in commercial real estate. For long-held, appreciated properties, the resulting “lock-in” can be large, reducing the likelihood of sale and keeping properties from moving to their highest and best use. We compare California to markets with conventional current-value taxes, where the same purchase histories carry no tax consequence. A 100-point increase in lock-in lowers the annual probability of sale by roughly 0.4 to 0.55 percentage points, with the clearest effects on conversion of vacant land and new construction. We then build and calibrate a dynamic model in which lock-in discourages the sales that would move a property to a better owner or use. A revenue-neutral shift to a current-value tax raises aggregate commercial output by roughly 5.6 percent.

Real rents in the U.S. Consumer Price Index rose 17.4 log points from 2000 to 2018. We develop a spatial-equilibrium framework to decompose the increase into several channels, including growing demand to live in cities with inelastic housing supply. Using parameterizations from the literature and a new rent index, we find that location demand accounts for 17 to 73 percent of the overall increase, and an even larger share in cities where the CPI is measured. We estimate the population elasticity to rents by comparing demand shocks across cities with different housing-supply elasticities. The results support a high population elasticity and imply that location demand explains more than half of the rent increase in our preferred specification.

New technology promises to expand financial services to small businesses poorly served by the banking system. We study FinTech’s response to demand created by the Paycheck Protection Program. FinTech is used disproportionately in ZIP codes with fewer bank branches, lower incomes, and larger minority population shares, and in industries with little preexisting small-business lending. Its role is also greater in counties where the economic effects of the COVID-19 pandemic were more severe. More PPP provision by traditional banks causes statistically significant but economically small substitution away from FinTechs, implying that FinTech mostly expands the overall supply of financial services rather than redistributing it.

We document a new fact: regional divergence—the rate at which rich states grow faster than poor states—explains most U.S. house-price movements since 1939, including the post-2000 boom-bust-boom cycle. An industry-share instrument provides evidence that the relationship is causal. We develop a model in which greater interstate inequality raises rents because relative demand for high-income states increases while housing supply in low-income states is elastic. Regional divergence also raises expected future inequality and rents, increasing current house prices. The model accurately predicts rents since 1929 and cross-sectional patterns in prices, rents, construction, and migration.

Remote work increased both the demand for housing and the demand for particular locations. Because housing supply varies across space and is more elastic in the long run, effects on rents and population differ over time. We use a spatial housing model with heterogeneous supply elasticities to identify changes in housing and location demand from 2020 to 2022. Although rents and prices rose substantially in the short run, we estimate that in the long run increased housing demand raises rents by only 1.8 percentage points, while changing location demand lowers rents by 0.3 percentage points, with larger declines in initially expensive cities and cities in which the CPI is measured.

This paper studies bank antitrust rules that discontinuously change the competitive impact of bank mergers. Mandatory divestiture becomes much more likely above a concentration threshold, increasing the number of banks in affected markets. Consistent with greater competition, intervention raises deposit rates and mortgage originations increase by 11 percent through both refinancing and purchases, while small-business loan quantities do not change. The effects do not dissipate over time, and nonbank lenders respond similarly to banks. Antitrust rules can therefore increase bank competition, but existing customer relationships protect banks from competitors.

Rising interest rates create mortgage-rate lock for homeowners with fixed-rate mortgages: they retain low rates if they stay but must take a new, higher-rate mortgage if they move. We show that mobility fell in 2022 and 2023 for homeowners with mortgages as market rates rose, both absolutely and relative to homeowners without mortgages. Changes in home values do not explain the decline. Our estimates imply that higher rates reduced mobility among mortgaged households by 16 percent in 2022 and 2023 and generated about $20 billion in deadweight loss.

We argue that rising regional inequality contributes to higher average housing prices and rents. Income growth has been faster in already high-income cities; this growth attracts population and raises relative housing demand; and housing supply in these cities is inelastic because of density and regulation. We illustrate the mechanism with a graphical spatial model in which rising regional inequality reallocates population toward high-income, supply-constrained areas, increasing national average housing prices and rents.

We estimate the fiscal externalities of the Los Angeles “Mansion Tax” on property-tax revenue when assessed values are closely tied to transactions. Because assessment growth between transactions lags market values, reductions in transaction frequency slow the growth of property-tax revenue. Assuming Measure ULA did not directly change transaction rates outside Los Angeles, our benchmark calibration implies that lost property-tax revenue offsets about four-fifths of the transfer-tax revenue; alternative calibrations range from 30 percent to more than 100 percent. Net revenue losses are larger for high-value and commercial properties.

Student-housing rents are a large component of college costs and high rents contribute to housing insecurity. Using data from U.S. student-housing markets from 2014 to 2022, we show that private-market student rents rose 15 percent more than national inflation and closely tracked local rents for general-purpose housing. On-campus rents rose only 7 percent and were largely uncorrelated with the broader local market. Enrollment growth raised on-campus but not off-campus housing costs. Universities therefore insulate students from local housing-market pressures, providing implicit housing subsidies especially in urban areas and at highly ranked institutions.

This paper provides a new explanation for regional variation in the 2000–2006 housing and consumption boom. Cities with larger shares of growing industries experienced larger housing-demand shocks, larger house-price increases from 2000 to 2006, and larger declines from 2007 to 2012. Price effects were stronger in cities with inelastic housing supply, and local demand was correlated with supply elasticity. Controlling for industry, the estimated durable-consumption elasticity to house prices is 0.08 during 2000–2006, about 40 percent below previous estimates. After 2006 the elasticity rises to 0.31, and house prices rather than local conditions explain consumption changes.

An instrumented difference-in-differences design scales a difference-in-differences effect on an outcome by a difference-in-differences effect on a mediating treatment. Despite widespread use, its identifying assumptions have received limited formal attention. We show that DDIV identifies a convex combination of average causal effects under exclusion, parallel-trends, and monotonicity assumptions familiar from both difference-in-differences and instrumental-variables designs. Clarifying these assumptions provides more precise justification for DDIV estimates and highlights potential pitfalls in causal interpretation.

CEO turnover is strongly procyclical, driven almost entirely by executives of retirement age. We show that executives time retirement to increase the value of their pensions: CEO pay is procyclical and pensions depend on pay in the final years of tenure, creating an incentive to retire when the economy is strong. Retirement cyclicality is especially pronounced at firms with stronger corporate governance, suggesting that firms use pension incentives and retirement timing to constrain executive behavior.