This study explores how human capital affects farm household earnings using two tools to refine measurement of human capital effects. First, it employs a two-sector model to allow the allocation of family labor between farm and non-farm activities. Second, it accounts for village fixed effects to evaluate whether results from panel data differ meaningfully from a cross-sectional data analysis with local binary variables. The results show that education has a negligible effect on farm earnings; instead, experience appears to be the principal channel by which human capital affects agricultural performance in a traditional rural setting. Our results also suggest that prior models that fail to separate non-farm activities spuriously exaggerated the effect of education to the farm sector. In addition, typical cross-sectional analyses that ignore fixed effects may cause the effects of education on rural household earnings to be significantly overstated. The fact that panel data regressions accounting for village-level fixed effects found only one instance of education raising earnings the effect of literacy on non-farm income suggests that considerable heterogeneity may have been ignored in cross-sectional data analyses, especially ones that omitted village binary variables.