Clustering of Auto Supplier Plants in the United States
提出一种线性化logit版本的空间GMM估计方法，能处理大数据集，并通过蒙特卡洛实验验证其准确性。应用于美国汽车供应商工厂选址，发现新工厂的集聚程度并未超过现有工厂。
A linearized logit version of Pinkse and Slade's spatial GMM estimator reduces estimation to two steps—standard logit followed by two-stage least squares. Linearization produces a model that can be estimated using large datasets. Monte Carlo experiments suggest that the linearized model accurately identifies the presence of spatial effects and is capable of producing accurate estimates of marginal effects. In an application to the location of supplier plants in the U.S. auto industry, the results imply no additional clustering of new plants beyond the level of clustering of existing plant locations.