Spatial econometrics: Models, methods and applications.

摘要

Three essays comprise this dissertation. The first chapter considers the Two-Stage Least-Square (2SLS) and Generalized Method of Moments (GMM) estimators for spatial and temporal autoregressive panel data models under the fixed-effects framework. The proposed 2SLS and GMM estimators are consistent, asymptotically normal and robust against unknown initial conditions. The classical 2SLS estimators are linear with closed-form expressions. The GMM estimators make full use of moment conditions under the standard assumptions and are efficient relative to the 2SLS estimators.; The second chapter considers the decision of local school spending in a dynamic game-theoretical model. I formulate a dynamic game model with strategic interaction and show that the local spending follows an exact spatial and temporal autoregressive path in general equilibrium. Empirical work uses local school expenditure data published by the Ohio State Department of Education. Without time effects, the current spending among local school districts are strategically correlated, but including time dummy effects makes such strategic interaction statistically insignificant. The latter contradict the empirical results in recent public finance literature that has a positive reaction slope. Furthermore, I have found that the current spending of the school districts is negatively related with the temporally lagged spending of their neighbors. The third chapter considers specification test, identification and estimation of simultaneous system of spatially interrelated cross-sectional equations. The identification conditions of classical system are extended to the spatial system. Classical 2SLS and 3SLS estimators are derived for the spatial models. A simple Hausman specification test of exogeneity is provided.