[Presentation Schedule]
13:00-14:30 Nov. 27th, Thu, @ED-107, East, Kyushu University
Moriya, Koichiro. (Keio University), with Noda, Akihiko. (Meiji University) Finite-Sample Properties of Model Specification Tests for Multivariate Dynamic Regression Models
Abstract: This paper proposes a new multivariate model specification test that generalizes Durbin regression to the seemingly unrelated regression (SUR) framework. The proposed Durbin-SUR estimator explicitly models the joint second-order dynamics of regressors and disturbances and accommodates cross-equation dependence. It remains consistent even in environments where OLS and GLS estimators are inconsistent and is asymptotically efficient under certain conditions. Monte Carlo experiments show that the Wald test based on the Durbin-SUR estimator performs well in finite samples. An empirical application further demonstrates its reliability and practical usefulness.
Workshop on Mathematical and Data-Driven Social Science（数理・データ駆動型社会科学ワークショプ）is supported by JSPS KAKENHI 24K16368（科研費若手研究）and Kyushu University SCS research grant.
Contact: kurita.kenichi.564@m.kyushu-u.ac.jp