Flood Insurance Premiums, Risk Beliefs, and Residential Mobility
Abstract: This paper evaluates the tradeoff between takeup of flood insurance and incentives to reside in areas exposed to higher flood risk, which poses a central tension in policymaking around insurance rate-setting and mitigation of natural disaster hazards. I leverage spatial and temporal variation in flood insurance premiums, due to discount changes and reverification cycles in the Community Rating System, to assess how flood insurance demand and housing prices respond to changes in the cost of insurance. I find that a 10% decrease in flood insurance premiums leads to a 3.1% increase in demand for insurance policies and a 2.7% increase in sale prices of houses in the 100-year floodplain. The magnitude of the housing price response is consistent with over-capitalization of insurance premium changes, and suggests that households may be adjusting their beliefs about flood risk based on observed insurance premiums. I then estimate a model of flood insurance demand and residential location choice with respect to flood risk, directly targeting the design-based estimates, and use the model to quantify welfare under alternative approaches to pricing flood insurance and incentivizing insurance coverage. I find that setting insurance costs at actuarially fair levels delivers a small welfare improvement, in part by improving risk communication.
Perception of Actual vs. Assigned Characteristics in Audit Studies of Housing Discrimination
Abstract: I provide an econometrically-specified framework to understand real estate agents' decisions of who they provide assistance to in the context of an audit study. I conceptualize latent quality of a prospective client as being a function of a number of actual, underlying characteristics of audit testers that are hidden from the agent but can be observed in the audit data. Although not all aspects of agents' observed decisions are consistent with them being fully informed about the relationships between observed demographic characteristics and true underlying characteristics, estimates of model parameters suggest that agents are able to incorporate information from both direct signals of latent quality and from predicted latent quality based on demographic characteristics into their decisions, and weight these two sources of information approximately equally. Removing the effects of both realized variation in testers' actual characteristics and agents' inferences of latent quality based on demographic characteristics reduces the predicted Black-white disparity in rate of receipt of assistance from 3.7 percentage points to 0.7 percentage points; this remaining disparity in treatment falls disproportionately on Black women.
The Impacts of Building Regulations on Disaster Damages and Recovery
With Swati Joshi