Profit rates differ across industries. Explanations have often relied on static models of imperfect competition. This paper develops a dynamic model of perfect competition to demonstrate that long-run average profit rates differ even across competitive industries when the effects of sunk costs on entry and exit are considered. The hypothesis that firms maximize their present expected values has few empirical implications for long-run average profit rates, but it does have implications for the behaviour of variables over time; for example, industries with high variability in the number of firms should exhibit low variability in firm values.

Optimal penal codes are constructed for a class of infinity-repeated games with discounting. These games can be interpreted as Bertrand oligopoly games with capacity constraints. No particular rationing rule is adopted; weak restrictions are imposed on the firms' sales functions instead. Models adopting the commonly used rationing rules are special cases of the general framework studied here. It is found that firms can be driven to their security levels by credible punishments.

An important question is how well competitive models approximate models of large finite economies. For a class of differentiated products models, static Nash equilibria, if they exist, always converge to competition as the number of firms increases. Dynamic Nash equilibria need not so converge. Easily checked conditions for convergence to competition do, however, exist.