Incomparability and Precise Equality

Options can be neither better nor worse (NBNW) with respect to overall well-being.213 _{One option is NBNW than another with respect} to overall well-being if (a) the two options are equally good, or (b) the two options are incomparably good. If two options (S1 and S2) are equally good with respect to overall well-being, then any third option is better than S1 just in case it is better than S2 and, conversely, worse than S1 just in case it is worse than S2. Incomparability is less demanding: if S1 and S2 are incomparably good with respect to over- all well-being, then a third option can be better than S1 without be- ing better than S2, or worse than S1 without being worse than S2. In practice (I would guess) agencies are not often faced with a choice between options that are equally good with respect to overall well-being. Why? Where NBNW options are good for overall welfare in different ways—where, for example, the extent of physical pleas- ure is greater in S1, while the extent of friendship is greater in S2, or longevity is greater in S1, but convenience is greater in S2—then there will virtually always be a hypothetical option which is a small improvement over one option but still NBNW than the other. If so, the two options by definition are not precisely equal.214

On the other hand, it could well be the case that regulatory agen- cies do frequently confront incomparable options. Recall my account of interpersonal comparisons: S1 is better than S2 with respect to overall well-being, or worse than S2 with respect to overall well- being, if and only if all idealized observers would have convergent re- stricted preferences as between S1 and S2. Will this occur if S1 is somewhat, but not dramatically, better than S2 with respect to one welfare good while S2 is somewhat, but not dramatically, better than

S1 with respect to another? (For example, what if there are 1000 more annual deaths in S1, but the unemployment rate in S2 is a half percentage point higher? What if consumer surplus in S1 is $100 mil- lion greater than in S2, but twenty endangered species go extinct? What if workplaces in S1 are moderately safer, but the quality of education is moderately worse?) When we think hard about the wel- fare tradeoff between such options, we may conclude that the options are welfare-comparable—that one option does emerge as better, and the other as worse. But it is at least plausible to think that a signifi- cant fraction of agency decisions involve incomparable, and thereby NBNW, alternatives.

213. See Ruth Chang, Introduction to INCOMMENSURABILITY, INCOMPARABILITY AND

PRACTICAL REASON, supra note 163, at 1-34 (defining and discussing incomparability); sources cited supra note 163 (same).

214. See Ruth Chang, Introduction to INCOMMENSURABILITY, INCOMPARABILITY AND

PRACTICAL REASON, supra note 163, at 23-27 (concluding that the “small improvement” ar- gument is a plausible argument for incomparability).

Would this create a divergence with neoclassicism? Note that two options can be NBNW with respect to efficiency. Clearly, two options can be equally efficient; it also turns out that two options can be in- comparably efficient.215 _{On the other hand, the proportion of regula-} tory agency choices that are either equally efficient or incomparably efficient seems to be low.216 _{So it is at least plausible that many} agency choices are NBNW with respect to overall welfare, but few agency choices are NBNW with respect to efficiency.

Imagine that two options are NBNW with respect to a particular moral criterion X* (where X* is the criterion of overall well-being, for the welfarist; and the criterion of efficiency, for the neoclassicist). How should the agency choose between the options? There may be some other criteria, X1, X2 . . . Xn, that possess moral force, and that the agency is charged with implementing. (That is, there may be other criteria not covered by a “partition” that effectively places them within the jurisdiction of legislatures, courts, or the taxing authori- ties, and outside the jurisdiction of regulatory agencies). If so, the agency should choose the option that is better with respect to the ap- plicable X1, X2 . . . Xn.

But it may be that no such X1, X2 . . . Xn exist; or (more likely), they may exist, but the option may also be NBNW with respect to the

X1, X2 . . . Xn. In such a case, I suggest, the agency is free to choose at random between the options. It possesses moral discretion. Where an agency’s alternatives are NBNW with respect to the applicable moral criteria X*, X1 . . . Xn, the agency might as well decide between the alternatives by flipping a coin. This is clearly true when the options are equally good with respect to the X*, X1 . . . Xn; and it is also true (I have argued elsewhere) when the options are incomparably good with respect to the X*, X1 . . . Xn.217 _{If, for example, our beneficial me-} 215. This is a result of the Scitovsky paradox. Take two outcomes O1 and O2 that in-

volve the paradox, that is, there is some redistribution from Winners to Losers in O1 that

makes it Pareto -superior to O2, but there is also some redistribution from Winners to Lo s-

ers in O2 that makes it Pareto-superior to O1. Then, O1 and O2 are NBNW with respect to

efficiency. However, they are not precisely equal with respect to efficiency, since there will typically be at least some O* that is efficient relative to O1, but not O2 (or vice versa). For

example, create O* by taking each person’s holding of each good in O1 and increasing it

slightly. That guarantees that O* is efficient relative to O1, but there could still be some

redistribution in O2 that makes it Pareto -superior to O*.

216. Why do I say this? Cost-benefit analysis is closely related, if not equivalent, to Kaldor-Hicks efficiency, and it is highly unusual for an agency performing cost-benefit analysis to conclude that one option has neither positive nor negative net benefits relative to another.

217. See Adler, supra note 63, at 1401-08. Amartya Sen makes this very claim: Some see completeness as a necessary requirement of consequential evaluation, but it is, of course, nothing of the sort. A consequentialist approach does involve the use of maximizing logic . . . [but] [m]aximization only requires that we do not choose an alternative that is worse than another that can be chosen in- stead. If we cannot compare and rank two alternatives, then choosing either from that pair will fully satisfy the requirement of maximization.

teor can be used to help Phil’s or Pat’s emphysema, and the choice of one or the other would not affect the fair distribution of welfare (which would be true if, for example, poverty-line egalitarianism ob- tained and both Phil and Pat were well above the poverty line) or perfectionist values, then the meteor-possessing agency is morally free to give the rock to either Phil or Pat. Either choice is morally permitted; neither is morally required or morally prohibited.

The upshot of this analysis is that welfarism may confer greater moral discretion upon agencies than neoclassicism. If, in practice, agency choices are rarely NBNW with respect to efficiency, then agencies rarely have moral discretion (within the framework of neo- classicism).218 _{If, in practice, agency choices are frequently NBNW} with respect to welfare, then agencies may frequently have moral discretion (within the framework of welfarism), depending upon the applicability and content of perfectionist, distributive, deontological, and other such supplementary moral criteria.