General and specific spatial autocorrelation: Insights from country-level analysis of species imperilment

We explore the consequences of including both general and specific controls for spatial autocorrelation in country-level models of species imperilment. A general spatial autocorrelation term constructed as a binary contiguity matrix aggregates an unknown number of possible influences, some of which exert a negative impact on species imperilment in the referent country, others of which exert a positive impact. Adding spatial dependency measures that are based on specific cross-border effects can substantially change not only the size and statistical significance of the general spatial dependency term but also the size, sign, and/or statistical significance of the explanatory variables.