Small sample properties of maximum likelihood versus generalized method of moments based tests for spatially autocorrelated errors |
| |
Authors: | Peter Egger Mario Larch Michael Pfaffermayr Janette Walde |
| |
Affiliation: | aIfo Institute for Economic Research, University of Munich, Poschingerstrasse 5, D-81679 Munich, Germany;bUniversity of Innsbruck, Faculty of Economics and Statistics, Universitätsstraße 15, A-6020 Innsbruck, Austria |
| |
Abstract: | Many applied researchers have to deal with spatially autocorrelated residuals (SAR). Available tests that identify spatial spillovers as captured by a significant SAR parameter, are either based on maximum likelihood (MLE) or generalized method of moments (GMM) estimates. This paper illustrates the properties of various tests for the null hypothesis of a zero SAR parameter in a comprehensive Monte Carlo study. The main finding is that Wald tests generally perform well regarding both size and power even in small samples. The GMM-based Wald test is correctly sized even for non-normally distributed disturbances and small samples, and it exhibits a similar power as its MLE-based counterpart. Hence, for the applied researcher the GMM Wald test can be recommended, because it is easy to implement. |
| |
Keywords: | Spatial autocorrelation Hypothesis tests Monte Carlo studies Maximum likelihood estimation Generalized method of moments |
本文献已被 ScienceDirect 等数据库收录! |
|