forthcoming, U.C. Irvine L. Rev. Vol. 11
61 Pages Posted: 19 Jul 2019 Last revised: 25 Mar 2020
Date Written: July 18, 2019
Economic analyses of both crime and regulation writ large suggest that the subjective cost or value of incentives is critical to their effectiveness. But reliable information about subjective valuation is scarce, as those who are punished have little reason to report honestly. Modern “big data” techniques promise to overcome this information shortfall, but perhaps at the cost of individual privacy and the autonomy that privacy’s shield provides.
This Article argues that regulators can and should instead rely on methods that remain accurate even in the face of limited information. Building on a formal model we prove elsewhere, we show that variability in a defendant's subjective costs of punishment should be a key consideration in any incentive system, whether it be criminal law or otherwise. Our model suggests that this variability can be mitigated with some familiar and well-tested tools. For instance, in some situations ex ante taxes on behavior that creates a risk of harm can be preferable to ex post punitive regimes, such as the criminal law, that target primarily harms that actually arise.
Because of what we show to be the centrality of variation in subjective costs, we also argue that long-standing approaches to criminal theory and practice should be reconsidered. For instance, economic theory strongly prefers fines over other forms of punishment. We argue that this claim is typically right — indeed, it is understated — when applied to firms. But fines can be the wrong choice for incentivizing most humans, while ex ante taxes are a promising alternative. We also show that this same analysis counsels that, if prison is the most viable punishment available, it can be more efficient to make prisons safer and less alienating.
Keywords: criminal law, sentencing, optimal incentives, choice of instruments, fines, cost-benefit analysis, regulation
JEL Classification: H23, K14, K23, K42
Suggested Citation: Suggested Citation