Autocorrelated residuals are common, although not universal, in standard regression analyses of dependent variables that are themselves autocorrelated. Figure 1 strongly suggests that county-level child poverty is one such variable. Broad regions exhibit high child poverty, where any selected county and its neighbors likely share high levels of poverty. Similarly, other regions exhibit almost uniformly low levels of child poverty, where a selected county is likely to have a low rate of child poverty similar to that of its neighbors. (For the moment, we set aside the matter of how the term ‘‘neighbor’’ is defined.) Consider some of the mechanisms that can cause spatial autocorrelation (Wrigley, Holt, Steel, & Tranmer, 1996):