主讲人：韩晓祎 教授（厦门大学）

主持人：谢聪 助理教授（辽宁大学中国经济研究院）

嘉宾介绍：马湘君 教授（辽宁大学中国经济研究院）

时间：2026年5月7日（周四） 14：00-15：30（北京时间）

地点：辽宁大学蒲河校区经济学部大楼547室

线上地址： #腾讯会议：690-510-966

语言：中文/英文

摘要：

In spatial and network panel data, time-varying (TV) dominant units exert disproportionate influence, invalidating standard estimation and inference methods. We study the generalized method of moments (GMM) estimation of a high-order spatial dynamic panel data (SDPD) model that features dominant units and unknown cross-sectional heteroskedasticity. We propose a novel classification of spatial weight matrices featuring dominant units and develop new general central limit theorems (CLTs) for linear-quadratic forms involving such matrices. We show that the GMM estimator (GMME) converges at a rate slower than the standard rate when the moment conditions include special matrices with dominant units that have unbounded row and column sums. To achieve efficient estimation, we propose a best GMME (BGMME) and a sequential root estimator (RTE). The RTE yields a closed-form solution, completely bypassing complex numerical optimization while preserving the asymptotic efficiency of the BGMME. We establish the consistency and asymptotic normality of these estimators. Monte Carlo simulations demonstrate that the proposed estimators have satisfactory finite sample performance, and the RTE has computational advantages over the BGMME. An empirical application examining the network peer effects of capital structure among Chinese listed firms illustrates the merits of our models and estimation methods.

主讲人简介：

韩晓祎， 2014年获美国俄亥俄州立大学经济学博士，现为计量经济学教育部重点实验室（厦门大学）副主任，厦门大学经济学科教授、博士生导师，入选国家级青年高层次人才。主要研究领域为计量经济学、应用计量经济学、区域经济学和劳动经济学。多篇论文发表于国内外经济学权威期刊。主持多项国家级和省部级科研项目。