The productivity of non-farm enterprises in rural Africa may be associated with the productivity of other spatially proximate farm and non-farm enterprises. To test for the presence and significance of such spatial autocorrelation we use data from the geo-referenced 2011 Ethiopian Rural Socioeconomic Survey (ERSS) and the 2010/2011 Nigeria General Household Survey (NGHS). We find evidence of significant spatial autocorrelation. Productivity of non-farm enterprises is widely dispersed across space in both countries. In Ethiopia rural non-farm enterprises are more productive in locations where farms are less productive. In Nigeria, we find evidence for spatial autocorrelation at the individual enterprise level but not at the community level, once we control for location variables. Hence, taking spatial autocorrelation into account using spatial lag and spatial error models, we find education, age, size of the household, religious affiliation and community infrastructure are significant determinants of the labour productivity of non-farm enterprises in Ethiopia and Nigeria. This is the first time, to the best of our knowledge, that the productivity of rural non-farm enterprises in Africa has been studied in this way.