Awarded the C. Lowell Harriss Dissertation Fellowship
This paper uses a natural experiment in Japan to provide evidence of the feedback loop between corporate borrowing and commercial real estate investment emphasized in macro-finance models with collateral constraints. Japan enacted a series of reforms in the early 1980s which relaxed national regulatory constraints on the height and size of buildings. Combining originally-constructed local commercial land price indices for over 400 localities with geocoded firm balance sheets, I show that these land use deregulations generated a boom-bust cycle in corporate real estate values, borrowing, and real estate investment. Firms located in more ex ante land use constrained areas both issued more debt and invested more heavily in real estate, thus amplifying the initial positive shock to commercial real estate prices. I develop a multi-city spatial sorting model with production externalities and real estate collateral which uses the estimated reduced form effects of my local regulatory instruments on firm outcomes to assess aggregate effects of the reform. I find that the deregulatory shock to commercial real estate markets and corporate borrowing environment amplified the real estate cycle in the 1980s and led to an increased incidence of zombie lending in the 1990s.
with Takashi Unayama
This paper documents heterogeneity in consumption responses to a large stimulus tax rebate based on household exposure to a housing price cycle. Linking geocoded household expenditure and financial transactions data to local housing price indices in Japan, we estimate a U-shaped pattern in the marginal propensity to consume with respect to housing price growth. Recipients living in areas with the smallest housing price gains during the 1980s spent 44% of the 1994 rebate within three months of payment, compared to 23% among recipients in areas which experienced the largest housing price gains. While we find limited heterogeneity in marginal propensities to consume among households in less-affected areas, MPCs are higher for younger, renter households with no debt residing in more-affected areas. These findings are consistent with near-rational households for which the pricing shock was small relative to permanent income spending a larger fraction of the tax rebate. Our analysis suggests fiscal stimulus payments primarily induce spending among "winner" households who face minimal exposure to housing price cycles.
The question of how governments should choose the frequency of payments has received little attention in the literature on the optimal design of public benefits programs. We propose a simple model in which the government chooses the length of the interval between payments, subject to a tradeoff between the administrative cost of providing more frequent benefits and the welfare gain from reducing deviations from full consumption smoothing. In our empirical application, we examine consumer and retailer responses to bimonthly payments from the Japanese National Pension System. We exploit variation in the duration of payment cycles using a unique retail dataset that links consumers to their purchase history. Our difference-in-differences style approach shows a clear spike in spending on payment dates for customers who are of retirement age relative to those who are not. While within-store average prices increase by 1.6% on payday, this effect is almost entirely due to consumers substituting towards higher quality goods rather than a retailer response. We use these reduced-form estimates to parameterize the model and conclude that the optimal frequency of Japanese public pension payments is less than one month, implying the government could improve welfare by increasing payment frequency.
Works in Progress
This paper analyzes the effects of property transfer taxes on real estate prices and taxpayers’ investment decisions using a tax reform in Taiwan which required sellers of non-owner-occupied real estate to pay large percentages of the full selling price for properties resold within one year (15%) or resold after one year but within two years (10%) from the original purchase date. We link the universe of personal tax returns to transaction records to show bunching at the 12-month and 24-month holding period thresholds under this tax, but no such bunching prior to implementation, suggesting that owners were highly attentive to the tax. We apply a sharp RD to examine the effect on prices, exploiting the fact that the tax applied retroactively to properties purchased prior to the implementation date. While average transaction prices do fall immediately after implementation, this is due to a decline in the number of high-end property sales that would have been subject to the tax. Our initial results suggest limited effects of the transfer tax towards curbing the rapid rise in real estate prices in recent years.
Microbubbles and Local Property Tax Regimes
with Takashi Unayama
Delaying the Honeymoon: The Great Decline in American Vacation
Revisiting the Incidence of Mandated Benefits: Evidence from U.S. Sick Pay Mandates
Recent city and state-level mandates impose minimum requirements on employers for paid sick leave provision. Yet, it is unclear who bears the costs of these mandates. I use geocoded establishment-level microdata from the BLS National Compensation Survey to document that employers respond to sick pay mandates by creating combined paid time off (PTO) plans. While PTO plans offer employees more flexibility by granting a fixed amount of paid leave days that may be used for any purpose, they also enable employers to provide fewer total days of paid leave while complying with sick pay laws. In contrast to other labor market incidence studies, my difference-in-differences (DD) and triple differences (DDD) designs exploit staggered law implementation across both local jurisdictions and across groups of employers for which the mandates are binding or non-binding. I show that employers’ self-reported costs for providing vacation, holiday, and sick leave fall by 2% while costs for “other leave” plans increase within four quarters after law enactment. My findings suggest that identifying changes in total compensation, rather than changes in wages, is crucial to determine the labor market incidence of mandated benefits.