Abstract
A common assumption in the Schumpeterian growth literature is that the innovation size is constant and identical across industries. This is in contrast with the empirical evidence which shows that: (1) innovation size is not identical across industries and (2) the size distribution of profit returns from innovation is highly skewed toward the low value side, with a long tail on the high value side. In the present paper, we develop a Schumpeterian growth model that is consistent with this evidence. In particular, we assume that when a firm innovates, the size of its quality improvement is the result of a random draw from a Pareto distribution. This enables us to extend the class of quality-ladder growth models to encompass firm heterogeneity. We study the policy implications of this new setup numerically and find that it is optimal to heavily subsidize RXX1D for plausible parameter values. Although it is optimal to tax RXX1D for some parameter values, this case only occurs when the steady-state rate of economic growth is very low.