[E]ntrepreneurship in a developing country consists of discovering the underlying cost structure–what can and cannot be produced profitably. Initial investors in a new line of economic activity face a great amount of uncertainty, since foreign technology always needs some local adaptation. Plus, their cost discovery soon becomes public knowledge–everyone can observe whether their projects are successful or not–so the social value they generate exceeds their private costs. If they succeed, much of the gains are socialized through entry and emulation, whereas if they fail, they bear the full costs.
I find this fairly compelling. Of course this sort of argument has to compete with public choicy arguments that governments are bad at picking winners and that such policies will create more opportunities for rent seeking.
I think Dr. Rodrik has more arguments for industrial policy in this paper (which I have not read yet).