【主讲】邓莹 助教授 (对外经济贸易大学国际经济贸易学院)
【主题】Estimation and Inference for the Spatial Threshold Model
【时间】2014年5月23日 (周五) 15：30-17：00
【地点】上海财经大学经济学院楼801室
【语言】英文
【摘要】The spatial autoregressive (SAR) model suggested by Cliff and Ord (1973) has received a lot of attention in various areas of economics including urban, labor, trade and development as well as environmental economics. It allows one to model interactions among regions, counties, neighborhoods, firms to mention a few. However, there’s an implicit assumption assumed by the existing literature is that the magnitude of the spatial dependence is the same over the entire population. This may not be the case in some economic examples popular in this area of research. For example, rich countries' growth may affect poor countries' growth differently from the way poor countries' growth affect that of rich countries. Similarly, housing prices in neighborhoods with high income may affect their poor income neighbors’ prices differently from the way poor income neighbors’ prices affect their rich income neighbors’ housing prices. The typical SAR model does not allow one to investigate this asymmetry in effects as it assumes the same effect for all neighbors. This paper extends the conventional threshold model to the spatial cross-section framework, and we allow different spatial autoregressive coefficients for different subsample groupings of countries, regions, states, firms, etc. We will refer to it as the spatial threshold model. We suggest a spatial two-stage least squares (S2SLS) estimator for the threshold and slope parameters based on Kelejian and Prucha (1998) type instruments. We also prove the consistency of the threshold parameter estimator. A Monte Carlo study is provided to examine the finite sample properties. Our results show that the performance of our estimator improves as the sample size increases and as the difference between the spatial parameters of the two subsamples increases.
