Spatial autocorrelation occurs when either the value of an explanatory variable in one areal unit is correlated with the value of the variable in contiguous areal units, or when the regression error terms are correlated across contiguous areal units. There is reason to suspect that spatial autocorrelation exists in this study. Consider, for example, the station distance variable. Census area units situated relatively close to a certain rail or ferry station are clustered around that station, and thus around one another. As has been documented in the theoretical literature, e.g., by Anselin (1988), spatial autocorrelation violates certain central tenets of Gaussian linear regression models.