Testing initial conditions in dynamic panel data models
提出一种新框架检验动态面板数据模型中的“均值平稳性”假设，该假设是系统GMM估计量一致性的前提。新方法通过扩展矩条件集将假设转化为参数约束，并利用LM检验验证其有效性，相比传统Sargan/Hansen检验具有更高检验功效。
We propose a new framework for testing the “mean stationarity” assumption in dynamic panel data models, required for the consistency of the system GMM estimator. In our set up the assumption is obtained as a parametric restriction in an extended set of moment conditions, allowing the use of a LM test to check its validity. Our framework provides a ranking in terms of power of the analyzed test statistics, in which our approach exhibits better power than the difference-in-Sargan/Hansen test that compares system GMM and difference GMM, that is, on its turn, more powerful than the Sargan/Hansen test based on the system GMM moment conditions.