The positive political theory of cost-benefit analysis.

• Author: Adler, Matthew D.
• Position: Response to article by Jason Scott Johnston in this issue, p. 1343

The legal institution of cost-benefit analysis (CBA) may end up producing outcomes with lower social welfare (overall well-being), relative to the outcomes that would have been produced had the institution not been in place. This is easy to see, in a general way. What are much harder to understand are the various and potentially interacting mechanisms that generate slippage between the internal aim of CBA and its actual outcome, and to determine which institution--CBA, risk-risk analysis, benefits-only analysis, benefits-only analysis constrained by technological feasibility, or some other institution (1)--in fact maximizes social welfare within a given regulatory domain. Jason Johnston's article, A Game Theoretic Analysis of Alternative Institutions for Regulatory Cost-Benefit Analysis, (2) sheds much light on these issues. More generally, the article constitutes a substantial contribution to the positive political theory of regulation. My brief Commentary, however, will focus on what Professor Johnston's article says or implies about the welfare effect of CBA and its alternatives.

CBA might reduce welfare because: (1) CBA has direct procedural costs. Time and money need to be expended to perform a cost-benefit analysis. (3) (2) Agencies are epistemically imperfect. They might make frequent mistakes in performing CBA and fewer mistakes in applying some other test that is reasonably well correlated with overall well-being. (4) (3) Agencies may seek to maximize goals other than the goals specified by statute, and it may be harder for political principals--namely Congress and the President--to monitor agency compliance with CBA than with other, clearer goals that are reasonably well correlated with overall well-being. (5) (4) Even an epistemically perfect and faithful agency that could perform CBA with zero direct costs would not necessarily produce welfare-maximizing outcomes, given "wealth effects." CBA evaluates outcomes by aggregating the amounts that individuals are willing to pay or accept, in dollars, for the outcomes. One person, however, could be willing to pay more for an outcome than another would require to accept it, even though the outcome harms the second more than it benefits the first. (6)

Eric Posner and I describe these sources of slippage between CBA and social welfare in our article, Rethinking Cost-Benefit Analysis. (7) Johnston identifies a very different mechanism that bears on the welfare comparison of CBA and alternatives, namely firm lobbying and litigation, which is in turn driven by--and potentially revealing of--firms' compliance costs. At the heart of Johnston's article is a formal game-theoretic model with roughly the following structure. An agency decides whether or not to issue a firm-specific regulatory directive. If the agency issues the directive, the regulated firm can lobby for Congress or the President to impose political sanctions on the agency and, later, can litigate to overturn the directive. The agency can drop the regulation after the lobbying stage or finalize the regulation and defend it in court if the firm challenges it there. Both the firm and the agency can expend varying amounts on lobbying and litigation, which will in turn influence the size of the political sanction imposed on the agency and the probability of success in court. (8) Johnston analyzes firm and agency behavior under two scenarios: first, under a "benefits statute," where the agency is instructed to maximize some statutory benefit (such as health or environmental purity), and the regulated firm cannot judicially challenge the directive on cost-benefit grounds; (9) and second, under a substantive "cost-benefit statute," (10) which includes a judicially enforceable requirement that the costs of agency regulation not exceed the benefits, so that the regulated firm will prevail at the litigation stage for certain, or at least probabilistically, if the firm's costs of complying with the regulatory directive are greater than the directive's benefits.

Crucially, this model does not presuppose any of the sources of slippage between CBA and social welfare that Eric Posner and I described. The agency can costlessly calculate the benefits of the directive. In addition, at least under certain conditions, the agency is able to determine costlessly the firm's compliance costs. (11) Thus, at least under certain conditions, agency CBA has zero direct costs and the agency is epistemically perfect. Further, and interestingly, the agency is public spirited and does not "shirk" from statutory goals in a radical way. The agency genuinely seeks to maximize the statutory benefit, rather than seeking to maximize its budget, its power, or its own view of good policy.

Under the "benefits statute" scenario, the agency does not shirk at all: the statute instructs the agency to maximize some benefit type, which is exactly what the agency prefers to do. Under the "substantive cost-benefit statute," the agency shirks insofar as it fails to internalize the statutory cost constraint, but remains genuinely committed to the primary goal of the statute, namely to advance health, environmental purity, safety, or whatever other benefit type is specified by the statute. Finally, "wealth effects" have no essential role in Johnston's model. It is quite consistent with the model to assume that the numerical magnitudes that figure therein, and that drive agency and firm behavior--the benefit from the regulation (B), the political sanction (D), the lobbying costs ([e.sub.f] and [e.sub.a]), the litigation costs ([L.sub.f] and [L.sub.a]), and the compliance cost (c)--are measured on a scale of interpersonal utility rather than a dollar scale. (12) If so, the proposition that B is greater than c for a given regulatory directive entails that the issuance of the directive--leaving aside lobbying and litigation costs--increases social welfare.

Notwithstanding the absence of wealth effects and the existence of epistemically perfect, faithful agencies that can costlessly calculate the benefits and (under some conditions) the costs of regulation, CBA can produce outcomes that are no better than, and perhaps even worse than, the outcomes produced by alternative institutions. How? Let me describe and comment on the dynamics of Johnston's model relevant for my purposes.

(1) A benefits statute can "sort" between welfare-enhancing and welfare-reducing directives by inducing the agency to promulgate regulatory directives against low-compliance-cost firms but not against high-compliance-cost firms.

Consider the case of a benefits statute. In this scenario, the parameters of Johnston's model are: B, the benefit from regulating; D([e.sub.a], [e.sub.f]), a lobbying function that gives the political sanction (D) as a function of agency and firm lobbying effort ([e.sub.a] and [e.sub.f]); r([L.sub.a], [L.sub.f]), a litigation function that gives the probability of judicial reversal of the agency (r) as a function of agency and firm litigation effort ([L.sub.a] and [L.sub.f]); and I, the cost to the agency of learning the firm's compliance cost up front. If I is zero, then the agency is epistemically perfect because it already knows B and the model parameters and can (by assumption) costlessly learn the firm's costs. There are two kinds of firms: firms with low compliance costs ([c.sub.l]), where [c.sub.l] < B, and firms with high compliance costs ([c.sub.h]), where B < [c.sub.h]. (13)

Assume for now that I is in fact zero; the agency knows the firm's compliance costs up front. (14) Then, given certain values of B, D, and r, the agency will promulgate the directive regardless of whether the firm at hand is low cost or high cost. Given other values of B, D, and r, the agency will fail to promulgate the directive regardless of whether the firm at hand is low cost or high cost. But there are some values of B, D, and r such that the agency will promulgate the directive if and only if the firm at hand is a low-cost firm. In this last scenario, the agency is induced to "sort" between cost-benefit-justified and cost-benefit-unjustified directives--to promulgate the former but not the latter--notwithstanding the absence of a CBA requirement in the underlying statute. (15)

Why does the "sorting" scenario occur? It occurs because of the effect that firm lobbying and litigation have on the agency's expected net benefit from enacting the directive, and because high-cost firms have a greater incentive to lobby and litigate than do low-cost firms. Consider the final, litigation stage of the game. An incremental increase in the probability of reversal benefits the high-cost firm more than it benefits the low-cost firm because it generates an incremental reduction in expected cost (the change in probability times the cost of compliance) that is greater for the high-cost firm. As a result, the high-cost firm would spend more at the litigation stage than the low-cost firm. This produces a lower expected benefit for the agency from finalizing and litigating the directive, as Johnston explains:

[T]he higher is the firm's compliance cost, the higher will be the amount that the firm will spend challenging the regulation in court, and hence the higher will be the agency's optimal expenditure in defending the regulation. The higher is the agency's expenditure at the judicial review stage, the lower will be its net benefit from the regulation, since resources are diverted to legal defense that might have been used to pursue other regulations.... (16) Thus the agency, considering whether or not to finalize and litigate the regulation, might determine that the expected benefits of doing so in the case of a low-cost firm are positive, but the expected benefits of doing so in the case of a high-cost firm are negative. Of course, the agency's expectations about its choice at the finalization stage feed back to its choice at the initial, promulgation stage. Specifically, an...