The puzzle of "ex ante efficiency": does rational approvability have moral weight?

• Author: Adler, Matthew D.
• Position: Preferences and Rational Choice: New Perspectives and Legal Implications

INTRODUCTION

What is the moral status of ex ante efficiency? Many within the law and economics tradition seem to think that the ex ante efficiency of legal institutions has moral weight--that legal officials have a moral reason to create ex ante efficient institutions. (1) I think the view is questionable.

In this Article, I seek to undermine the view by attacking a more general claim; the putative moral weight of ex ante efficiency is a particular variant of this generic claim, one that marries the claim with a particular view about the nature of rationality. The generic claim is that rational approvability has moral weight--that legal officials have moral reason to choose actions that citizens would be rational to approve.

Part I fleshes out the concept of rational approvability and describes the link between rational approvability and ex ante efficiency. Part II asks whether rational approvability has moral force given the particular account of rationality dominant within economics: subjective expected utility (SEU). I consider various arguments for an affirmative answer--grounded in welfare, agency, stability, and the legal official's own uncertainty--and reject them all. Part III assesses the moral force of rational approvability on the assumption that SEU has been replaced with alternative accounts of rationality. What is rational for someone to do or approve might be a matter of maximizing objective expected utility, realized utility, subjective expected value, objective expected value, or realized value. I suggest that these variations on SEU either fail to be plausible accounts of rationality or, if plausible, generally fail to support the claim that rational approvability has moral weight.

1. EX ANTE EFFICIENCY AND RATIONALITY

   SEU is the standard account of rationality within economics, (2) and stipulates the following, given some actor in a choice situation at time T, choosing among actions [A.sub.1] ... [A.sub.n]: (1) each choice, [A.sub.i], can be represented as a lottery over possible worlds, where the probabilities in the lottery are numerical measures of the actor's subjective degrees-of-belief that the choice, if made, will result in the various possible outcomes; (2) the actor's preferences over possible worlds can be represented by a numerical "utility" function; (3) for each choice, an expected utility can be calculated by using the subjective probability numbers in the matching lottery to discount the utilities of the various possible outcomes, and then aggregating; and (4) the rational choice is the choice with the largest expected utility.

   It is common in law and economics scholarship to describe some law, doctrine, rule, statute, or other legal institution as maximizing everyone's expected utility and therefore "ex ante efficient." Note the problems in this formulation, at least within an SEU account of expected utility. First, subjective expected utility, strictly construed, is a feature of each person's choices. What does it mean to speak of the expected utility, for a person, of some other kind of thing--here, a legal institution? Second, subjective expected utility varies over time. A person's subjective expected utility for her own choices (and presumably for derivative things, such as institutions) depends upon her subjective probability assessments, which are not temporally fixed. A person's probability assessment at some time, [T.sub.1], that some action will produce some possible world may be different from her assessment at another time, [T.sub.2], that the action will produce that world. (3) Legal institutions, however, endure over time. It is not as if a legal institution exists only at one moment and we can (somehow) use everyone's subjective probability numbers at that time to determine whether the institution is ex ante efficient. So what time is relevant in determining ex ante efficiency?

   Here is an attempt to resolve the problems and provide a clearer definition of ex ante efficiency. As just mentioned, rational choice theory is, centrally, prescriptive for an actor with regard to the actor's own choices. That is, the theory identifies the action, from the various possible actions open to person Pin a choice situation, which P would be rational to choose. P rationally orders the possible actions of P. By extension, however, one can describe the rational ordering, by some other person, of P's possible actions. Imagine an Agent and a Patient. (4) The Agent is in an actual choice situation. At a particular point in time, T, he is choosing among actions {[A.sub.1] ... [A.sub.n]}. Since these are the Agent's choices, not the Patient's, how does one construct a rational ordering of the choices by the Patient? I suggest that one might ask which choice by the Agent it would be rational for the Patient to approve. More precisely, we can imagine the Patient in a hypothetical choice situation at T, faced with the options {[A.sub.1]' ... [A.sub.n]'}, where each [A.sub.i]' is a determinative act of approval by the Patient: if the Patient were to choose [A.sub.1]', the Agent would do [A.sub.1] ; if the Patient were to choose [A.sub.2], the Agent would do [A.sub.2]; and so on. Since, at a minimum, any theory of rational choice can rank choices, any theory will be able to rank the Patient's hypothetical choices of approving one or another of the various choices open to the Agent.

   With this concept of rational approvability in hand, we can provide the following definition of ex ante efficiency, one that treats ex ante efficiency (as well as ex post efficiency) as a feature of particular, institution-affecting choices by legal officials:

   Ex Ante and Ex Post Efficiency: Imagine that a legal official, the "Agent," is choosing among {[A.sub.1] ... [A.sub.n]} at some time T. The Agent's choices may affect legal institutions in various ways; they may create, support, destroy, or undermine such institutions. Some action A* by the Agent is "ex ante efficient" if, for each individual in the population (each "Patient"), it maximizes the Patient's subjective expected utility at T to approve A* rather than the other actions in {[A.sub.1] ... [A.sub.n]}. Some action A+ by the Agent is "ex post efficient" if the actual world that would result, were A+ to be chosen, is preferred by all Patients at T to the actual worlds that would result, were the Agent to choose other actions from {[A.sub.1] ... [A.sub.n]}. (5) This definition resolves the problems mentioned earlier. Given some institution-affecting choice by an Agent at time T and some group of "Patients" affected by the choice, for each individual Patient one can coherently assign subjective expected utilities to the Agent's choices and determine which choice it would maximize subjective expected utility for that Patient, at T, to approve. If the maximizing choice by the Agent, for all Patients, is the same choice, that choice is "ex ante efficient."

   The problem of determining ex ante efficiency, given the temporal variability in any person's subjective probability assignments, is also solved by thinking of ex ante efficiency as a feature of particular institution-affecting choices by legal Agents. Choice situations presented to Agents, unlike whole legal institutions, can be thought of as occurring at particular times; it is the affected persons subjective probabilities at the time of the Agent's choice that, I suggest, might be used to calculate the expected utilities that are determinative of ex ante efficiency.

   Finally, the definition offered here draws a clean distinction between ex ante and ex post efficiency. It is common for legal scholars to draw this distinction. "Ex ante" and "ex post" efficiency are seen as logically distinct properties of legal institutions: some institution might be ex ante efficient but not ex post efficient, or vice versa. My definition warrants this sort of talk. Given some Agent placed in a choice situation at time T, choosing among {[A.sub.1] ... [A.sub.n]}, the choice that some Patient "ex ante" prefers, versus the choice that she "ex post" prefers--the choice that it maximizes subjective expected utility for her to approve, versus the choice whose outcome she most prefers--clearly can be different. Thus, too, it might clearly occur that the ex ante efficient choice by some Agent is not ex post efficient.

   Ex ante and ex post efficiency, as I have clarified these concepts, cannot be ascribed to institutions simpliciter. Rather, they are ascribable to particular choices by legal officials: to possible actions by them, in particular choice situations at particular times. Is this a flaw in my conceptualization? I don't think so. If we are to preserve the standard link between ex ante efficiency and the SEU account of rationality, I see no way to license talk of the ex ante efficiency of an institution simpliciter, detached from a particular time. Again, subjective probability assessments vary over time. And since SEU is fundamentally an account of rational choice, defining ex ante efficiency as, first and foremost, a feature of choices--and derivatively, of legal institutions, relative to particular choices that affect the institutions--seems the natural way to proceed. At a minimum, the choice-based account is a plausible conceptualization of ex ante efficiency and is the one I shall rely upon in this Article. (6)

   Given this definition of ex ante efficiency, the claim that ex ante efficiency has moral force emerges as a particular version of the generic claim that rational approvability has moral force. (7) Is the claim true? Imagine that an Agent is choosing among {[A.sub.1] ... [A.sub.n] }. For some Patient, it would be rational to approve A* Does the Agent therefore have moral reason to perform A*? (8) If so, and if SEU in particular is the correct account of rationality, it follows that ex ante efficiency has moral force. The premise that rational approvability possesses moral weight is...