Finite Optimal Penalties for False Advertising

I consider a setting in which firms have unverifiable private information about their type, which corresponds to their probable product quality; firms can expend a learning cost in order to observe their quality; and the regulator can enforce false advertising penalties contingent only on verifiable realized quality. I show that it may be socially optimal for high type firms to signal their type through ‘speculative claims,’ rather than to learn and signal their quality. This implies that socially optimal false advertising penalties are finite, in contrast to the literature's common assumption of arbitrarily high false advertising penalties, and that the regulator optimally tolerates the existence of some false claims in equilibrium.