Impact analysis for spatial autoregressive models: with application to air pollution in china

We investigate impact analysis and its asymptotic inference for spatial autoregressive models. We propose using the delta method, which enables us to obtain the dispersion in an explicit form. In addition, we provide an element-wise impact analysis. We first study the cross-sectional case, where various impacts are introduced to measure the interaction and feedback effects in a space dimension. We then study the spatial dynamic panel case, with simultaneous spatial and dynamic feedback in the effects. Monte Carlo results show that the proposed impact analysis has satisfactory finite-sample properties. Finally, we apply the impact analysis to investigate how meteorological factors and air pollutants affect PM2.5 in Chinese cities.