My dissertation contains three chapters where I employ spatial hedonic modelling, reduced-form estimations and causal inference techniques to analyze urban demographics, migration and the housing market in the U.S. The first chapter employs generalized difference-in-difference to estimate the impact of COVID-19 regulations on the year-on-year of changes in housing sales and building permits approval. This chapter attempts to investigate if and how these interventions causally impacted the county-level housing sales and building permits approval in the U.S. The second chapter formulates a spatial hedonic equilibrium model to show inter-city impacts of the introduction of work-from-home, with multiple worker types. The third chapter explores how community characteristics at a city level correlate to fatal police encounters over the period 2000-2010.

The first chapter explores the effect of public policies like shelter-in-place and eviction moratoria on relative housing sales and building permits approved for construction. Shelter-in-place orders could potentially limit the ability of home-buyers and sellers to interact as well as they could pre-COVID, introducing frictions in the process of selling houses. Prolonged and over-lapping eviction moratoria could dampen the construction of multi-family units. This paper attempts to investigate if and how these interventions causally impacted the county-level housing sales and building permits approval in the U.S. The paper estimates the average treatment effect of these orders using a traditional generalized difference-in-difference estimator and a more recent variation of the estimator that is more suited multiple treatment groups with staggered treatment introductions and withdrawals. The results show that shelter-in-place is associated with significantly smaller year-on-year changes in sales of single-family houses, condominiums and the collection of all residences. Selective moratoria on eviction hearings and judgements are also found to be associated with smaller year-on-year changes in multi-family building permit approvals. Apart from the baseline model, the chapter also employs an additional estimator to add robustness to the results. SIP is found to be associated with significantly smaller relative changes in sales of single-family houses, condominiums and the collection of all residences. Selective moratoria on eviction hearings and judgements are also found to be associated with smaller relative changes in multi-family building permit approvals.

The second chapter formulates a spatial hedonic equilibrium model that shows inter-city impacts of the introduction of work-from-home. Following Brueckner, Kahn and Lin (2021), the paper attempts to theoretically study how the introduction of work-from-home, which allows workers to relocate across cities while keeping their original jobs, impacts housing prices, population and employment levels. Extending Brueckner et al. (2021), the current paper divides the workforce into two types of workers, remote and non-remote, to allow for a more realistic work-from-home model. Additionally, the model uses explicit functional forms for production and utility, resulting in closed-form equilibrium solutions conditional on the extent of productivity advantages, amenity advantages, and degree of complementarity between worker types. The current paper aims to examine whether the main results in Brueckner et al. (2021) still hold in the modified model.

The third chapter aims to answer if and how community characteristics at a city level correlate to fatal police encounters over the period 2000-2010. Panel and cumulative datasets on police-involved deaths from FatalEncounters.org, along with a relevant set of socio-demographic controls, are used in a negative binomial regression model to find factors that might affect the count of officer-involved deaths that occurred in a city. For the longitudinal panel data, the Black population and the Hispanic population in a city along with police employment, the level of crime and median income constitute significant determinants of officer-involved deaths. With panel and cumulative datasets used in negative binomial regression models, it is found that, for the longitudinal panel data, the Black population and the Hispanic population in a city along with police employment, the level of crime and median income constitute significant determinants of officer-involved deaths.