Non-farm rural enterprises (NFRE) are increasingly studied because of their role in poverty reduction. However, existing studies of the effects of infrastructure on NFRE may give incorrect inferences because they typically fail to account for spatial effects. Such effects could reflect either spatial errors due to excluded local effects or spatial lags due to excluded interactions, such as between households switching out of farm work. We use rural investment climate survey data from Indonesia that allow distances between each household to be measured so that spatial effects can be modeled to assess the bias from ignoring such effects.