The authors analyze five rounds of National Sample Survey data covering 1983, 1987/8, 1993/4, 1999/0, and 2004/5 to explore the relationship between rural diversification and poverty. Poverty in rural India declined at a modest rate during this period. The authors provide region-level estimates that illustrate considerable geographic heterogeneity in this progress. Poverty estimates correlate well with region-level data on changes in agricultural wage rates. Agricultural labor remains the preserve of the uneducated and also to a large extent of the scheduled castes and scheduled tribes. Although agricultural labor grew as a share of total economic activity over the first four rounds, it had fallen back to the levels observed at the beginning of the survey period by 2004. This all-India trajectory masks widely varying trends across states. During this period, the rural non-farm sector grew modestly, mainly between the last two survey rounds. Regular non-farm employment remains largely associated with education levels and social status that are rare among the poor. However, casual labor and self-employment in the non-farm sector reveal greater involvement by disadvantaged groups in 2004 than in the preceding rounds. The implication for poverty is not immediately clear - the poor may be pushed into low-return casual non-farm activities due to lack of opportunities in the agricultural sector rather than being pulled by high returns offered by the non-farm sector. Econometric estimates reveal that expansion of the non-farm sector is associated with falling poverty via two routes: a direct impact on poverty that is likely due to a pro-poor marginal incidence of non-farm employment expansion
and an indirect impact attributable to the positive effect of non-farm employment growth on agricultural wages. The analysis also confirms the important contribution to rural poverty reduction from agricultural productivity, availability of land, and consumption levels in proximate urban areas.