한국은행은 이질적 공간 패널 모형의 공간 의존성 검증 기법을 제안하고 유효성을 살펴본 보고서를 발표하였다.
The widely used approach to testing spatial dependence formulates a hypothesis on a homogeneous spatial coefficient in spatial models. This paper proposes a novel test for spatial dependence in large panel data models with heterogeneous spatial autoregressive coefficients. Under weakly reciprocal interactions, the proposed test is asymptotically standard normal as both n and T tend to infinity jointly. We analyze its power under local alternatives. We show that the traditional test may lose power when spatial effects are heterogeneous, particularly in small samples. Monte Carlo simulations demonstrate that our proposed test has greater power than the traditional test in such settings. An empirical example illustrates that the proposed and traditional tests can lead to opposite conclusions regarding the presence of spatial dependence.