Answers are grounded in this paper and cite their sources.
Previous articleNext article FreeComment on "Explaining the Standard of Proof in Criminal Law: A New Insight"Abraham L. WickelgrenAbraham L. Wickelgren*University of Texas at Austin. Search for more articles by this author PDFPDF PLUSFull Text Add to favoritesDownload CitationTrack CitationsPermissionsReprints Share onFacebookTwitterLinked InRedditEmailQR Code SectionsMore"Explaining the Standard of Proof in Criminal Law: A New Insight" by Nuno Garoupa (2017) is one of the first articles to apply the insights from the principal-agent literature in economics to standard of proof issues.1 Given that our justice system relies on the actions of many different agents, all of whom almost certainly have objective functions different from that of the social planner, Garoupa's approach is clearly a valuable one. As a first step in applying an agency model, the article naturally has some limitations. Because a comment is more interesting and useful when it focuses on the limitations of an article, most of my remarks will be about why there might be advantages in applying the principal-agent approach to model burden of proof in a somewhat different way. First, I will discuss the advantages of taking a more comprehensive, from first principles, approach to finding the socially optimal burden of proof and how that might affect the results. Then, I will discuss how a more explicit integration of the agency issues in law enforcement might affect the results. But, at the outset, I want to make clear that Garoupa's approach represents a clear step forward for this literature. As such, I hope it will inspire future authors to build on and improve upon the path that this article has begun.1. Deriving the Welfare Function from First PrinciplesThe central agency problem in the article is that enforcers have stronger relative preferences for not acquitting the guilty than for not convicting the innocent (punitive preferences) than does the court. The enforcer, however, has the information that the court uses to decide whether to convict a suspect or not. That is, the enforcer knows p, the probability that the suspect is guilty. In the model, the court moves first, setting a threshold level for p, which the article calls p′. If the enforcer reports that p > p′, the court convicts, otherwise it acquits. Because lying is costly for the enforcer (and more costly, the larger the lie), the court can induce the enforcer to only recommend conviction (reporting a p that exceeds the threshold) when the true value of p is above the court's ideal threshold simply by setting the nominal threshold far enough above the ideal threshold.This strategy works, and it achieves the court's first best outcome given the specified objective function. That said, it does not achieve the first best outcome in terms of social welfare because the enforcers have to incur lying costs (which are not included in the court's objective function in the model). If one includes lying costs in the social welfare function, then the optimal standard of proof is probably lower (so as to induce less lying) than the standard that re-creates the same set of convictions and acquittals that a perfectly informed court would prefer.Because the revelation principle (Myerson 1979) does not always hold with reporting costs, the optimal solution might still involve some deception on the part of the enforcer. The general solution to this problem would involve the enforcer announcing some p′ and having the court commit to a function that maps this announcement to a probability of conviction. If the court is maximizing social welfare, including reporting costs, my guess is that the solution to this second best optimal mapping is probably not a step function like the outcome in the article. That said, the small literature on mechanism design with reporting costs (see, e.g., Crocker and Morgan 1998; Lacker and Weinberg 1989; Maggi and Rodriguez-Clare 1995) does suggest that it will likely involve some misreporting and some biasing of the mapping away from the first best in the direction of higher standards of proof.However, even if the court/social planner brings lying costs into its objective function, the planner's objective function and the article's analysis seem too ex post focused. That is, the planner is taking the crime rate, the number of suspects brought before it, and the social costs of mistakes as given despite the fact that the legal rules surrounding standard of proof likely affect all of these. Why not model the decision to commit a crime as a function of the standard of proof and other parameters (e.g., the probability of detection and the punishment if detected and convicted)? This generates a crime rate that can be put into a social welfare function—thus, generating costs for mistaken acquittals endogenously and possibly wrongful convictions as well, depending on whether you think of criminal opportunities as exclusive or not—that might also include social costs from imprisonment and maybe other costs from wrongful convictions.2Based on this social welfare function, one could then incorporate the agency problem in the article. This would involve allowing the enforcer to learn p and lie about it and then also incorporating lying costs into the social welfare functions as well. One way to view the Garoupa (2017) article is that it takes a reduced-form approach to all of this. It is imagining that all of this optimization from first principles is going on in the background to generate an optimal standard of proof, and then its problem is how to implement that given that enforcers have an incentive to lie about p. If we ignore lying costs as part of the social planner's objective function, this works well because the article finds that we can implement the first best real threshold for p by creating a formal threshold that exceeds p. If we view the lying costs as social costs, however, then we not only need to know the socially optimal p, we also need the social loss function for when the actual threshold differs from the socially optimal one.Again, one could take a reduced-form approach and just specify an arbitrary loss function, L(p), that is minimized at zero, the socially optimal threshold. This article, however, is implicitly assuming a particular functional form for this function when it assumes a constant social cost for each wrongful conviction and mistaken acquittal. This is not the loss function that a first-principles analysis of social welfare would generate. The advantage of deriving a social welfare function that depends on p from first principles is that it would enable us to get much more insight into how the agency function affects the optimal formal p than simply specifying an arbitrary loss function.2. Explicitly Incorporating Agency ProblemsBecause I really like the incorporation of the agency problem into a model of optimal legal rules, I would have also liked to see a fuller integration of that problem. The motivation for the assumption that law enforcers have stronger punishment preferences is from Dharmapala, Garoupa, and McAdams (2016). In that model, it is optimal to hire punitive enforcers because they are willing to work for lower wages as a result of the utility they get from the act of punishing. I find this plausible, but it means that our social welfare function should take into account the cost of hiring the enforcer. If we do not, then we might as well hire nonpunitive enforcers. If we do, then this may interact with our optimal evidence threshold. For example, if the cost of hiring an enforcer depends on the utility that enforcer receives from doing the job, then establishing a legal rule that either (a) requires the enforcer to bear lying costs to obtain the convictions that increase its utility and/or (b) leads to fewer convictions that reduce the utility from the job and will increase the wages required to hire enforcers and will tend to lead us to choosing a nominal standard of proof that is less than the unconstrained optimum. It would be nice to have explicit conditions for the interaction of the cost of hiring effect and the need to control the different preferences of the agent. It is also possible that in such a world, it would be less costly to simply try to screen out punitive enforcers.The downside of doing so, however, is its effect on another important agency problem, the canonical problem of inducing effort. An enforcer's job is to exert costly effort to gather evidence that can both help identify the perpetrator of a crime and lead to that individual's conviction. It is certainly easier to induce the enforcer to exert any given level of effort if the enforcer gets utility from the outcome of the effort. This is a reason to prefer punitive enforcers despite the costs associated with controlling their divergent preferences. But, if we raise the standard of proof and impose lying costs on the enforcer, then the net utility from gathering evidence that leads to a conviction is lower because it requires incurring lying costs. This suggests another reason why we might want punitive enforcers but not want to bias our standard of proof very much to compensate for their divergent preferences.Garoupa's (2017) article represents a clear step forward in its integration of agency considerations into models of optimal standards of proof. I hope that this comment provides some suggestions for ways that Garoupa and others could build on this first step in future work in this area. Notes 1 Articles such as Wickelgren (2003) have discussed related issues with respect to punishment.2 If criminal opportunities are not exclusive, then if I don't commit a particular crime, someone else might. The probability that I might be convicted for that crime if I don't do it would give me an incentive to commit it. If criminal opportunities are exclusive, then if I don't commit a given crime, it won't happen. The fact that I might be wrongfully convicted for a different crime will not affect my incentive to commit this crime if the punishments are additive (see Schrag and Scotchmer 1994).ReferencesCrocker, Keith J., and John Morgan. 1998. "Is Honesty the Best Policy? Curtailing Insurance Fraud through Optimal Incentive Contracts." Journal of Political Economy 106:355–75.First citation in articleLinkGoogle ScholarDharmapala, Dhammika, Nuno Garoupa, and Richard McAdams. 2016. "Punitive Police? Agency Costs, Law Enforcement, and Criminal Procedure." Journal of Legal Studies 45:105–41.First citation in articleLinkGoogle ScholarGaroupa, Nuno. 2017. "Explaining the Standard of Proof in Criminal Law: A New Insight." Supreme Court Economic Review 25:111–122.First citation in articleAbstractGoogle ScholarLacker, Jeffrey M., and John A. Weinberg. 1989. "Optimal Contracts under Costly State Verification." Journal of Political Economy 97:1345–63.First citation in articleLinkGoogle ScholarMaggi, Giovanni, and Andres Rodriguez-Clare. 1995. "Costly Distortion of Information in Agency Problems." Rand Journal of Economics 26:675–89.First citation in articleCrossrefGoogle ScholarMyerson, Roger. 1979. "Incentive Compatibility and the Bargaining Problem." Econometrica 47:61–73.First citation in articleCrossrefGoogle ScholarSchrag, Joel, and Suzanne Scotchmer. 1994. "Crime and Prejudice: The Use of Character Evidence in Criminal Trials." Journal of Law, Economics, and Organization 10:319–42.First citation in articleGoogle ScholarWickelgren, Abraham L. 2003. "Justifying Imprisonment: On the Optimality of Excessively Costly Punishment." American Law and Economics Review 5:377–411.First citation in articleCrossrefGoogle Scholar Previous articleNext article DetailsFiguresReferencesCited by Supreme Court Economic Review Volume 252017 Sponsored by the Antonin Scalia Law School, George Mason University Article DOIhttps://doi.org/10.1086/699730HistoryPublished online December 06, 2018 © 2018 by the University of Chicago. All rights reserved.PDF downloadCrossref reports the following articles citing this article:Alice Guerra, Barbara Luppi, and Francesco Parisi Accuracy of Verdicts under Different Jury Sizes and Voting Rules, Supreme Court Economic Review 28 (Sep 2020): 221–236.https://doi.org/10.1086/709734Alice Guerra, Barbara Luppi, Francesco Parisi Jury Size and Voting Requirements for Criminal Convictions: Assessing the Tradeoffs, SSRN Electronic Journal 258 (Jan 2019).https://doi.org/10.2139/ssrn.3432392
No takes yet. Share an insight, caveat, or question.
Abraham L. Wickelgren (2017) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: