Pilot project key to regulatory compliance.

Rigorous policy evaluation often involves randomization, and both federal and state governments have used randomization for a variety of purposes for approximately two centuries. But in the particular context of random government audits, transparency of process is crucial—especially when non-compliance can have reputational effects.

One of the first recorded government randomizations was by states in the early 1800s. It involved the shameful removal of indigenous people from their land followed by distribution of that same land to white people via land lottery. Other randomizations over time have included diversity visa lotteries, the draft lottery for selective military service, and education lotteries to place students in public schools. Panels of judges are often randomly assigned to review cases, and administrators on some government boards are also randomly assigned to make decisions.

Randomizations for policy evaluation are now so common that the U.S. General Services Administration has an Office of Evaluation Sciences that has conducted over 70 randomized policy experiments. Governments also regularly conduct audits, some random and some not. Examples include audits of financial records by the U.S. Securities and Exchange Commission, random audits of lobbying disclosures by the U.S. Government Accountability Office (GAO), audits of federal taxpayers to assess compliance with the Internal Revenue Code, and randomized review of Board of Veterans’ Appeals opinions for quality control. Those audited include regular citizens, firms, agencies, policies, lobbyists, and even elected officials.

As we explain in a recent article published in the journal Public Choice, government regulators may want to conduct random audits for several reasons.

First, random audits are a fair method of allocating resources. Assuming that regulators are not able to audit all entities of interest, random audits allow every potential person or firm subject to audit to have a similar probability of being audited.

Second, random audits by government agencies are effective policy tools to encourage regulatory compliance. In addition, ongoing random audits may allow for longer-term equilibrium effects to emerge, as people and entities select into or out of the regulatory regime because the audits raise the costs of both compliance and non-compliance.

Third, randomization by government agencies allows for policy evaluation: Are regulatory policies working as intended? Do the incentives embedded within policies change behavior? Audits can also allow regulators to test the outcome of policies that may have unintended effects. Before a full launch of a new regulatory policy, piloting the policy on a randomly selected subset of intended program members allows policymakers to refine the policy—or abort it—before the final launch.