学术活动

2025年12月15日（周一）

15:30-17:00

王玮宁教授

布里斯托大学

This paper analyzes spillover effects in spatial (network) models when measurement noise might contaminate the neighborhood (i.e. adjacency) matrix. We propose to adopt a low-rank and sparse structure to capture stylized network patterns in empirical datasets. We develop a flexible estimation framework via regularization techniques: a Least Absolute Shrinkage and Selection Operator (LASSO) penalty for the sparse component and a nuclear norm penalty for the low-rank component. We propose two estimators: (1) A two-stage procedure that first de-noises the adjacency matrix via regularization and subsequently integrates the purified network into a regression analysis, and (2) A single-step supervised Generalized Method of Moments (GMM) estimator that jointly estimates the regression parameters and refines the network structure. Simulation evidence underscores the superiority of our approach relative to conventional estimation protocols. In scenarios with noisy networks, our method reduces the root mean squared error (RMSE) of the estimate of spillover effects by 50–70% compared to conventional GMM. This advantage is more significant when measurement errors are correlated with the observed outcomes and network contamination is econometrically endogenous. We apply our framework to the dataset in Besley and Case (1995) and demonstrate its practical utility. The decomposition not only improves estimation reliability but also generates granular insights for policy design. These results highlight that we can bridge the gap between methodological rigor and policy relevance by explicit modeling of network structure heterogeneity.

JEL Classification: C21, C23, D57

Keywords: network analysis, policy effects, LASSO, nuclear norm penalty, penalized GMM

王玮宁博士现任英国布里斯托大学经济学教授。此前，她曾任荷兰格罗宁根大学经济与商学院计量经济学讲席教授、英国约克大学经济及相关研究系金融计量经济学讲席教授。王玮宁博士于德国柏林洪堡大学获得经济学博士学位。

她的研究领域主要涵盖非参数与半参数计量经济学、高维计量经济学、网络模型及时间序列分析，研究成果发表在Annals of Statistics、Journal of Business & Economic Statistics、Journal of Econometrics、Journal of the American Statistical Association、Econometric Theory 等经济学和统计学国际顶级期刊。

其研究课题集中于面板数据分析、高维时间序列模型及应用计量经济学方法，旨在解答经济与金融领域的关键问题，例如系统性风险模型分析、金融衍生品定价以及社会网络分析等。