A0700
Title: GMM and root estimations of spatial dynamic panel data models with unknown heteroskedasticity and dominant units
Authors: Chen Yahui - Xiamen University (China) [presenting]
Han Xiaoyi - Xiamen University (China)
Zhang Jiajun - Shanghai Jiao Tong University (China)
Abstract: The spatial dynamic panel data model is considered in the presence of dominant units and unknown heteroskedasticity. The dominant units can vary over time, and the number of dominant units can be finite or infinite are allowed. Since the quasi-maximum likelihood estimator (QMLE) is inconsistent under the heteroskedasticity, the generalized method of moments estimator (GMME) and the root estimator (RTE) RE proposed, and the consistency and asymptotic normality of these estimators are established under both scenarios:(1) large $n$ and large $T$, (2) large $n$ with small $T$. Our RTE is asymptotically as efficient as the GMME under the heteroskedastic case. Mento Carlo simulations demonstrate that the estimators have satisfactory finite sample performances even if the strength of the dominant units is equal to 1. Finally, an empirical application is presented on the peer effects of firm finance decisions across Chinese listed corporates.