There has been interest in the implications of learning by doing, and in particular in the possibility that learning by doing may be slower in less developed countries and in industries which use simpler technologies. This paper uses firm‐level data from Ghana to estimate learning‐by‐doing effects and generates three main findings. First, the learning curve, though present, is flatter in Ghana than in developed countries. Second, any industry‐wide spillovers are small and insignificant. Third, (contrary to the assumption of much theory) learning‐by‐doing effects are stronger at low levels of technology than at intermediate levels.