Spatial Analysis of Food Inflation in Russian Regions

Spatial interactions among modelling economic variables observed in spatially distributed units (due to their economic and trade relations) may be considered as an additional explanatory variable in a regression model (which generally prevents from its misspecification). Usually, spatial interactions are included in a regression in the form of spatial lag. In this paper we conduct a spatial econometric analysis of consumer price indexes for foodstuffs (FCPIs) observed in Russian regions. There are 79 regions in our sample for the period of time from 2002 to 2015 (data structured in panel). Our research aims at testing hypotheses of 1) presence of spatial correlation, and 2) of its heterogeneity among regional FCPIs. We develop a spatial panel data model with two matrixes of spatial weights (which are inverse distances with the breakpoint distance of 5000 kilometers between administrative centers of regions measured by roads) to test research hypotheses. In our model, the first matrix serves to estimate spatial correlation among regions up to break point distance between them, while second matrix catches spatial interactions among regions farther than break point distance from one each other. We find strong empirical evidence that 1) there is statistically significant spatial correlation among Russian FCPIs, 2) estimated spatial correlation is heterogeneous and the degree of its heterogeneity depends on the distance. That is, spatial relation shrinks as distance between regions rise and vice versa, or alternatively the closer one region to another, the higher expected inflationary relations between them