To improve the welfare of the rural poor and keep them in the countryside, the government has been spending 40% of the value of agricultural GDP on agricultural support services. But can investment make smallholder agriculture prosperous in such adverse conditions? This paper derives an answer by applying a two-output six-input stochastic translog distance function, with inefficiency effects and biased technical change to panel data for the 18 districts and the commercial sector, from 1979 to 1996. This model demonstrates that herds are the most important input, followed by draft power, land and seeds. Multilateral indices for technical change, technical efficiency and total factor productivity (TFP) show that the technology level of the commercial sector is more than six times that of traditional agriculture and that the gap has been increasing, due to technological regression in traditional agriculture and modest progress in the commercial sector. Since the levels of efficiency are similar, the same pattern is repeated by the TFP indices. This result highlights the policy dilemma of the trade-off between efficiency and equity objectives.