Data-Driven Monopoly Regulation
Abstract
To regulate markets effectively, regulators need to anticipate market conditions that may be inherently unpredictable and constantly changing. In this paper, I propose regulations that leverage data to remain effective in unpredictable environments. These are relevant for natural monopolies and other settings where antitrust policy is either too costly or not viable.I introduce a new theoretical framework for robust dynamic mechanism design and apply it to a model of monopoly regulation (Laffont and Tirole 1986). There is a regulator, a firm, and a sequence of consumers that arrive over time. Regulations map the observed history (including prices, sales, and realized production costs) to payments for the firm and consumers.
I avoid many standard assumptions. The stochastic process that governs demand and costs need not be stationary, Markovian, etc. I do not assume that either the regulator or firm know this process (although the firm may). I do assume that if there exists a heuristic that guarantees the firm ?% of its optimal profits, the firm will not follow a strategy that obtains less than ?% of its optimal profits.
My results provide a robust foundation for customer-first regulation, reminiscent of Loeb and Magat (1979). Customer-first regulation involves a subsidy proportional to estimated con- sumer surplus and a tax on profits, where consumer surplus is estimated using an A/B test that always leaves participating consumers better off. It guarantees 81% of “first-best” welfare (if the marginal cost of public funds is 1.3). The optimal regulation guarantees 85% of first-best welfare and is a non-linear variant of customer-first regulation.
In an extension, I obtain similar results without relying on subsidies. Here, customer-first regulation simply caps profits at a small fraction of estimated total surplus.