This paper analyzes the effect of household non-farm income on agricultural productivity in rural Ghana using a
nationwide household survey data. Key features of the analysis are the problem of endogeneity and selection bias
which are addressed using the Heckman two-step procedure. The first stage probit regression revealed that the
coefficients of availability of telecommunication infrastructure, availability of banks, availability of motorable
roads, forest and savannah zones are significant in explaining non-farm income. The second stage OLS
regression showed a significantly negative relationship between non-farm income and per-capita farm income
given the other control variables.