This paper reconsiders the well-known comparison of equilibrium entry levels into a Cournot industry under free entry, second best (control of entry but not production) and first best (control of entry and production). Allowing for the possibility of limited increasing returns to scale in production, this paper generalizes the conclusion of Mankiw and Whinston (1986) , that under business-stealing competition, free entry yields more firms than the second-best solution. We also show that under-entry always holds under business-enhancing competition. This confirms the general intuition given by Mankiw and Whinston, which does not rely on the convexity of the cost function. The same result is shown to extend (at a similar level of generality) to the comparison between free entry and the first best socially optimal solution, irrespective of business-stealing. Three illustrative examples are provided, one showing that the second-best and free entry solutions may actually coincide.