Notions of fairness versus the Pareto principle: on the role of logical consistency.

• Author: Kaplow, Louis
• Position: Response to article by Howard Chang in this issue, p. 251

In other writing, we advance the thesis that legal policies should be evaluated solely on the basis of their effects on individuals' well-being, meaning that no independent evaluative weight should be accorded to notions of fairness.(1) In that work, we consider a variety of principles of fairness, justice, and corollary concepts that are conventionally employed in the assessment of legal rules.(2) In the course of our research, we discovered that each of the leading notions of fairness that we examined could be shown to conflict with the Pareto principle; that is, consistent adherence to any of the notions of fairness would, in some circumstances, make everyone worse off. This observation led us to inquire about the generality of the conflict, and we explored it in two articles. In the first, we demonstrated that, in symmetric settings (in which every person is similarly situated), every individual will necessarily be made worse off whenever a welfare-independent notion of fairness is decisive.(3) In the second, a short, technical article intended for economists, we presented a formal proof of the proposition that, in all cases (symmetric or not), if a welfare-independent notion of fairness is given any weight in making social decisions, there will exist circumstances in which everyone is made worse off.(4)

Howard Chang has written an article in this journal that addresses aspects of our second article.(5) He accepts the validity of our formal proof, but he challenges the appeal of our assumptions. Furthermore, he suggests that certain notions of fairness that do not satisfy these assumptions, including his own conception of liberal welfarism, could be sufficiently modified so that they would not conflict with the Pareto principle. (Chang's precise claim is unclear, however. The overall thrust of his article may give the reader the impression that many familiar notions of fairness might be susceptible to modification so as to avoid conflict with the Pareto principle, yet the analysis itself suggests only the logical possibility that the conflict can be circumvented with respect to modifications of certain types of notions of fairness, which are not formally specified.(6))

We begin this Reply by summarizing the demonstration in our first paper of the conflict between notions of fairness and the Pareto principle. This demonstration, which Chang does not contest, is easier to understand than our second, and it is independently sufficient to establish our conclusion. Then we consider our second, more technical demonstration of our conclusion. We emphasize that the two assumptions that Chang challenges are really minimal in character; in essence, they amount to requirements that normative theories be logically consistent. (Indeed, the relevance of centuries of moral philosophers' normative discourse depends on one of these assumptions.) We also explain how, even when one does not require logical consistency in the respects to be articulated, Chang's effort to show that certain notions of fairness can be altered so as to avoid conflicts with the Pareto principle is unsuccessful. Finally, we offer brief remarks on Chang's liberal welfarism, drawing upon arguments from our larger work.

Before proceeding, we should note that we are well aware that many readers will be reluctant to accept our claims. First, the idea that all plausible notions of fairness conflict with the Pareto principle may seem surprising; indeed, it came as a surprise to us during the course of our research. Yet, analysis has revealed it to be true to a general extent. Second, some of our analysis--particularly the second demonstration with which Chang takes issue--is of a technical nature. We endeavor to explain the relevant points in accessible terms so the reader can see what is really involved in considering them. Third, the fact that notions of fairness have broad intuitive appeal to everyone--including to us--seems in tension with our critique. In our conclusion, however, we briefly draw upon our other writing to sketch some of the ways that this appeal can be reconciled with our overall thesis that notions of fairness should not be employed as independent evaluative principles in the assessment of legal policy.

1. OUR FIRST DEMONSTRATION: NOTIONS OF FAIRNESS ALWAYS MAKE EVERYONE WORSE OFF IN SYMMETRIC CASES

   A basic, natural setting to consider is the symmetric case: that in which all persons are similarly situated with respect to the policies under consideration. For example, in analyzing tort rules and the principle of corrective justice in the context of automobile accidents, one might examine cases in which every person is equally likely to be an injurer or victim and faces the same costs of accident avoidance, the same harm if an accident occurs, and so forth.

   In such symmetric settings, we have demonstrated that, whenever a notion of fairness leads one to choose a policy different from that which would be chosen were the social goal concerned exclusively with effects on individuals' well-being, everyone will necessarily be made worse off.(7) The reasoning is straightforward. Suppose that a notion of fairness leads to the choice of legal rule A when overall well-being would be greater under rule B. Since overall well-being is greater under rule B, and since each person is identically situated, it must be that each person's well-being is greater under rule B (and by the same amount). Hence, everyone is made worse off by choosing rule A.

   This simple demonstration, which Chang does not question, is by itself sufficient to establish that there is a conflict between any notion of fairness and the Pareto principle. For if one endorses a notion of fairness, consistency requires that one cannot ignore symmetric cases, in which the conflict arises.(8)

   Moreover, we suggest in our writing that the symmetric case has special significance for a range of systems of morality, including those associated with the Golden Rule, the categorical imperative, and choice behind a "veil of ignorance."(9) The reason is that such frameworks, which are designed to capture a notion of impartiality among individuals, in essence require that proposed moral principles pass muster in symmetric settings. (Consider, for example, the categorical imperative. If one did not demand that individuals be viewed as if they were symmetrically situated, a person who is mighty could advance the principle "might makes right," for such a person would happily generalize that principle, as that would simply enlarge his or her personal benefit. Only if one is required to assume symmetry among individuals--that no person is more mighty than another, or that a person is not more mighty than others any more often than anyone else is--would the test of the categorical imperative rule out such a self-interested principle.) We suspect that many readers, as well as most commentators upon whom Chang relies for various normative principles, do endorse the fundamental requirement of impartiality that is embodied in these frameworks. Thus, our demonstration that notions of fairness can only make individuals worse off in the symmetric case should be viewed as particularly important.

2. OUR SECOND DEMONSTRATION: IN GENERAL, NOTIONS OF FAIRNESS SOMETIMES MAKE EVERYONE WORSE OFF

   In our second demonstration that any notion of fairness conflicts with the Pareto principle, we do not restrict analysis to symmetric settings. We believe it will be helpful to sketch our proof briefly. (The proof is somewhat abstract, and we ask the reader to bear with us.) Let us begin by considering any notion of fairness. (For example, consider the principle of corrective justice, under which wrongdoers are required to compensate their victims.) Now, as a theoretical matter, if this notion of fairness is given any weight (that is, if, other things equal, it sometimes affects the social decision), we must be able to imagine two regimes--call them Fair and Unfair--that have the following characteristics: First, every individual is equally well off in Fair and in Unfair; and second, one regime--Fair--is definitely more fair than the other, Unfair, and hence it is deemed normatively superior.(10) (To be concrete, suppose that Unfair, unlike Fair, does not follow the requirement of corrective justice that wrongdoers compensate victims in a class of cases; nevertheless, injurers are not better off than in Fair because they pay higher fines in Unfair, and victims are not worse off in Unfair because social insurance is provided to them.)

   Next, consider a slightly modified unfair regime, Unfair-II, that is identical to Unfair except in one respect: There is a tiny savings of administrative costs in Unfair-II, which is distributed uniformly per capita. Now, if fairness has any real weight, it must be true that Fair is deemed superior overall to Unfair-II: After all, Fair was definitely superior to Unfair, Unfair-II is every bit as unfair as Unfair, and the cost advantage of Unfair-II over Unfair was stated to be tiny. (For example, compared to Fair, suppose that Unfair-II involves a major sacrifice of corrective justice and a trivial administrative cost savings, perhaps a penny per person.) However, observe that everyone is worse off in Fair than in Unfair-II. (This is because everyone is equally well off in Fair and in Unfair, while everyone is worse off in Unfair than in Unfair-II.) Hence, the notion of fairness has been shown to...