ESSAYS ON SPATIAL ECONOMETRIC MODELS

<p dir="ltr">This study investigates several different specifications of high-dimensional spatial econometric models with various empirical applications to urban economics, COVID-19 vaccination rate, and the real estate market. The first chapter proposes two models that incorporate both heterogeneity and multiple sources of spatial correlation for dynamic panels. One uses convex combinations of them to form a single weight matrix. The second one includes explicitly different spatial weight matrices to form a higher-order model. We use a Bayesian scheme for model estimation by deriving the full conditional distributions of heterogeneous parameters. Our Monte Carlo experiments demonstrate their finite-sample performance relative to a baseline model. In our empirical study, we find the importance of including both geographic and non-geographic information in capturing correlations in real house price growth in the US.</p><p><br></p><p dir="ltr">The second chapter proposes a Bayesian approach to estimating heterogeneous spatial dynamic panel models, subject to possible shrinkage on spatial dependence parameters. This amounts to heterogeneous selection of candidate spatial weight matrices that represent different spillover channels. The shrinkage methods include both the traditional and more flexible ones that allow the shrinkage strength to vary across spatial parameters. Monte Carlo results indicate that when the true model has a relatively low proportion of nonzero spatial parameters, flexible shrinkage in general leads to lower average root mean squared errors in estimating these parameters. An empirical study using this approach shows that there exists substantial heterogeneity in spillover channels across counties that determine the correlation patterns of county COVID-19 vaccination rates in four states in the United States. The first and second chapters are based on the joint work with Dr. Yong Bao.</p><p><br></p><p dir="ltr">In the third chapter, I investigate the heterogeneous selection of house price spillover channels among metropolitan areas in the United States. I propose a Bayesian approach to incorporating shrinkage on groups of spatial dependence parameters in a heterogeneous higher-order spatial dynamic panel model. Results from Monte Carlo experiments demonstrate that including group-level shrinkage would lead to better finite-sample performance in estimating the spatial dependence parameters when there exists group-structured sparsity in the true model. Empirical results using the Bayesian approach show that the geographic proximity, followed by the migration channel, makes the greatest contributions to the correlation pattern in regional house price growth.</p><p><br></p><p dir="ltr">The fourth chapter provides a test for measurement of spatial competition in residential real estate markets. Several alternative spatial competition measures are tested. We employ a Bertrand oligopoly model with differentiated products and adopt a Spatial Autoregressive model using a two stage least squares estimator. Our results show that commonly used count-based measures using the number of competitors in specific geographic radii are outperformed by price-based measures using prices of nearest competing neighbors. The main reason is that the latter measure accounts for heterogeneous neighborhood density of competitors. The measure captures the decaying pattern of spatial price competition over distance. The measure also stands out in capturing heterogeneous spatial price competition effects. We find that spatial price competition is more intense among high-value homes within the five nearest competing houses. The fourth chapter is based on the joint work with Dr. Ralph Siebert.</p>