The Ghana cocoa market has been extensively liberalised over the period since the mid 1980s. Two issues have been prominent in microeconomic research on agricultural supply response to liberalisation. The first has been the response to reduced subsidies on inputs, the second whether innovation has occurred. In this paper we investigate these two issues by estimating a production function for cocoa in Ghana drawing on two household surveys covering the period from 1991 to 1998. The estimated production function allows identifying the factors underlying the change in output. It is shown that the use of non-labour inputs increased substantially while we find no evidence for any rise in total factor productivity. We show that labour inputs declined substantially so that the ratio of both land and non-labour inputs to labour rose implying a rise in labour productivity of 39 per cent. Thus the analysis of the micro data shows that, contrary to much of the discussion of the effects of trade reform, the contribution of non-labour inputs to cocoa production has increased both relative to land and, very substantially, relative to labour. Over the 1990s labour productivity has increased and land productivity has not. Reforms have not led to innovation in techniques which raise total factor productivity.