Essays on the Dynamics of Regional Housing and Labor Markets Open Access
Downloadable ContentDownload PDF
In this dissertation, I explore a burgeoning area of interest in the field of regional economics: the dynamics of cities and regions. I show that regional differences in the state of housing markets and the specialization of industry serve to affect the dynamic responses to local demand shocks. In particular, urban decline lowers the elasticity of housing supply, causing demand shocks to affect house prices more in declining cities than growing or stable cities. Because the housing stock constrains the labor supply in cities, decline also serves as a predictor of the effects of demand shocks on employment and wages, causing the effects on employment to decrease and those on wages to increase.I also find that industrial specialization reduces the effects of demand shocks on the housing and labor markets. Specifically, cities with a high degree of diversification of regional exports experience stronger employment, wage, and house price effects of demand shocks compared to an equivalent shock in an industrially specialized city. When cities are highly diversified, intermediate inputs and final consumption goods are often produced locally, as opposed to specialized cities which must import these goods from other locations.Finally, I explore the ability of simple time series models to forecast regional house price dynamics. I find that theory-driven multivariate models were best able to forecast the declines in house prices experienced in California from 2007-2009. Univariate, atheoretical models, on the other hand, forecasted quite poorly and were unable to detect turning points in the housing market.In addition to the empirical results discussed above, this dissertation also makes methodological contributions. The essays on urban decline, industrial specialization, and regional dynamics use a new two-step estimation procedure that acts as a substitute for panel specifications. This procedure has only recently become possible due to the high data and computing power requirements. In the first step, individual time series models are estimated for each cross-sectional unit. In the second step, characteristics of the estimates in the first step are estimated as a function of cross-sectionally varying but time-invariant variables. Bootstrapping exercises are performed spanning both steps in order to establish the statistical properties of the estimates. This two-step procedure has broad applications for further research because it enables sophisticated time-series techniques to be applied to models where panel estimators are infeasible and panel restrictions are untenable.