Research Papers

Several studies suggest that “hot spot” and “proactive” policing – i.e., increased deployment of police or private citizens in areas with high crime rates – are effective at reducing crime. However, most of these studies analyze short term exogenous changes in the deployment of police following a terror attack, short term “crackdowns”, or randomized experiments. In this paper, we analyze the effect of adding guards near schools on crime rates over a sustained period of time. The focus of the study is the Safe Passage program, which was implemented in Chicago to increase the safety of children traveling to and from school. We use difference-in-difference and matching techniques to examine the effectiveness of the Safe Passage program in reducing crime. Using crime data from January 2006 to August 2016 and the location of Safe Passage routes, we analyze crime within small geographic units of one eighth by one eighth of a mile. We find strong evidence of the effectiveness of the program, with violent crimes declining by around 14%. However, we do not find a significant decline in property crime. Additionally, our results indicate that the program does not lead to significant spatial spillovers to neighboring unguarded areas. We also find that the program has a positive effect on attendance rates: the presence of a safe passage route is estimated to reduce the rate of absenteeism by about 2.5 percentage points.​​

​

In 1972, the Alquist-Priolo Zoning Act provided for the publication of earthquake fault maps in California. I exploit revisions in these official maps over time to estimate the rate of capitalization of seismic risk into property values using a difference-in-differences framework. Using geographically consistent data from 1970 to 2010 at the census tract level, I find that on average property values decline by 6.6 percent after the delineation of the fault zone, while rents decline by around 3.3 percent. I also examine the risk gradient and heterogeneity in willingness to pay using individual sales transactions, assessors' records, and publicly available mortgage data for the City of Los Angeles. The analysis of micro transactions data from 1997 to 2016 reveals that, on average, house prices increase by 1.8 percent for a one-mile increase in distance from the fault zone. The evidence also indicates that holding income and gender constant, Blacks and Hispanics are less willing than Whites to trade other forms of consumption to avoid earthquake risk. ​​

A unique data source is used to analyze the effect of the fair market rent (FMR) on the distribution of apartment rents in Los Angeles for 2007-2014. The data set includes the results of semi-annual surveys of rents in Los Angeles since 2005. The surveys include information on location, characteristics of the property, and a rating of the overall quality of the building exterior. Histograms suggest there is a tendency for rents to cluster at values around the FMR, particularly for small (studio and 1 bedroom) units. We also estimate a model explaining the probability that a unit’s rent lies in a set of intervals, including regions just below and just above the FMR. The estimates suggest that the probability that the rent is in the region just above the FMR is significantly higher than would be expected given the underlying characteristics of the properties. An analysis of changes in rents over time suggests that this tendency for rents to cluster in the region just above the FMR is persistent for individual properties over time. ​​

Homestead exemptions typically produce a moderately progressive statutory incidence for the property tax in the US when progressivity is measured using the ratio of a household’s tax to the sale price of a home. However, the tax is levied on the assessed value of a home rather than the sale. A stylized fact from the assessment literature is that assessment rates tend be lower for higher-priced home. In this paper, we first document the pattern of assessment regressivity. Next, we show that this pattern of regressivity is capable of reversing the statutory progressivity incidence of the property tax, particularly for low-priced homes. We then use HMDA data to show that this pattern of assessment regressivity also leads to regressivity when measured as the ratio of taxes to household income.

This paper has explored the stringency of land-use regulation in US cities, focusing on building heights. Substantial stringency is present when regulated heights are far below free-market heights, while stringency is lower when the two values are closer. Using FAR as a height index, theory shows that the elasticity of the land price with respect to FAR is a proper stringency measure. This elasticity is estimated for five US cities by combining CoStar land-sales data with FAR values from local zoning maps, and the results show that New York and Washington, D.C., have stringent height regulations, while Chicago’s and San Francisco’s regulations are less stringent (Boston represents an intermediate case).

Using data from the National Survey of Mortgage Originations (NSMO), we document that borrowers’ financial sophistication (measured as self-reported understanding of mortgages and the mortgage process) and their exposure to competition (measured as the number of lenders they seriously considered) are both associated with lower mortgage rate spreads. However, competition is not a substitute for sophistication: the benefits of competition are at least as strong for sophisticates as they are for naifs. Our results complement those from the literature on financial sophistication which detail the limits of advice (e.g., Guiso et al. (2018)) and education (e.g., Fernandes et al. (2014)), and collectively paint a pessimistic view about the prospects for simple interventions to close the mortgage rate gap between the informed and the naive.

We analyze the causal effect of air pollution on Iran’s housing market by exploiting increases in air pollution due to sanctions that targeted gasoline imports. In response to the sanctions, Iran converted petrochemical plants into refineries that produced low-quality gasoline, which led to a drastic and heterogeneous increase in air pollution across neighborhoods in Tehran. Using unique administrative data on Tehran's housing market, we estimate a fixed effects model that exploits this variation in air pollution and compares housing values within neighborhoods over time. We find that a 10% increase in the outdoor concentration of nitrogen dioxide leads to a decrease in housing prices of around 0.6% to 0.8%.

Teardowns provide direct information on land values in fully developed urban areas because such properties are valued only for their land and location rather than for the characteristics of the structure. We use two approaches to estimate land values. The first approach is a Stein-like procedure that uses teardown properties and makes efficient use of limited data when a group of variables – in this case, the structural characteristics – is expected beforehand to provide little explanatory power. The second approach is based on an unconditional expectation for the pooled data set of teardown and non-teardown sales. We use data from Chicago and Maricopa County to demonstrate the two approaches.

  •  Do School Shooting Erode Property Values?  (with Juan S. Muñoz-Morales),  Regional Science and Urban Economics, Accepted

We examine how residential property values are impacted by school shootings, which are crimes that have a very low probability of being repeated in the same location. We exploit the exogenous timing of eleven mass shootings that took place between 1998 and 2014, and use a difference-in-differences framework to estimate the causal effect of these school shootings on property values in the affected school attendance area. We find that house prices decline by an average of around 2.4 percent in a four-year period after a mass school shooting. We also find that enrollment falls in the affected school district, suggesting that families subsequently avoid schools in areas that have experienced such events.

  • ​Measures of Vertical Inequality in Assessments (with Daniel McMillen), Revise and Resubmit at Journal of Housing Economics

Standard measures of vertical inequity suggest that assessments are regressive in the sense that high-priced properties are often assessed at lower rates than low-properties. Conventional measures of measuring vertical inequality include a simple descriptive statistic – the price-related differential – and measures based on regressions. We show that regression-based procedures are seriously flawed, with a bias that tends to imply regressivity even when it is not present. To supplement the price-related differential, we propose two approaches that focus on the entire distribution of assessments rather than attempting to provide a single measure to characterize the entire assessment process. The first is to compare Gini coefficients and Theil Indices for sales prices and assessments. These statistics directly measure vertical inequality by determining whether the distribution of assessments is not as skewed toward low-value properties as are sales prices. We also measure inequality using the Theil index and get similar results. The second approach is to test formally whether the distribution of log sales prices is statistically different overall from the distribution of log assessed values. We compute all of the measures using data on sales prices and assessments for 48 large central city counties.

  • Long Term Impact of Rwanda Genocide: A Quantile Regression analysis

This paper uses Rwanda's Demographic Health Survey along with data on war intensity to examine the long term impacts of 1994 Rwanda genocide. The identification strategy is to exploit variation in timing of birth and geographic variation in war intensity using ordinary least squares and quantile regression. On average, there isn’t much impact on education and child mortality but the quantile regressions show that the left tail of the survival distribution, i.e. children who are weaker are adversely impacted. However, for adult height the impact of the genocide is felt across the distribution, i.e. even individuals who are in the upper tail of the conditional height distribution are impacted.


​

Work in Progress

  • The Effect of Riots on the Property Tax Base in Los Angeles
  • A note on the Optimal Lambda for Minimum Power Divergence Estimator (with Anil Bera)​