I’ve done a series of posts already on “utility” as defined in neoclassical welfare economics, including a post on social welfare functions in particular, but I’ve noticed whenever the word “utility” crops up in my various online conversations with other economists of both the professional and amateur varieties I invariably develop the uneasy feeling we’re not always on the same page as it were, not always using the word the same way. It’s comical to me because utility is probably the most fundamental concept in neoclassical welfare economics, which is ostensibly all about maximizing it. Maybe it makes sense to take yet another look at the ever so complicated albeit needlessly so heart of neoclassical welfare economics.
One element of normative neoclassical welfare economics I think creates enormous confusion and conflict is the simultaneous existence or discussion of both ordinal and cardinal conceptions of utility. They represent two very different versions or definitions of utility and the fact one is never entirely sure which version anyone else has in mind, or rather as soon as one starts talking about one whoever one happens to be talking to at that moment will invariably start talking about the other, adds a great deal of confusion to any conversation about normative economics.
Ordinal or ordinal only utility captures only the order or ranking of preferences for a given individual. It doesn’t really have any meaning beyond the preference rankings of individuals. In particular, it doesn’t refer to any underlying concept like “happiness” or “satisfaction.” One can see that easily enough when one realizes that, linguistically, one cannot discuss interpersonal comparisons of ordinal only preference rank utility because that type of utility is undefined in that context. Linguistically, grammatically, that would be nonsense, like talking about the square root of red, a fact which would be perfectly obvious were it not for the existence of other conflicting conceptions or definitions of “utility” in which that statement would not be grammatical nonsense and the ease with which one can equivocate on that term. Any concept of “happiness” or “satisfaction,” for example, that breaks no grammatical or linguistic rules, sounds sensible to the ear, when used in the phrase interpersonal comparisons of happiness or satisfaction cannot equate to ordinal only preference rank utility. Indeed, under preference rank utility it doesn’t matter if one’s preferences make one “happy” or more “satisfied” in any sense that adds content beyond the mere fact of the preference ranking itself. If one’s preference is to do something that makes one miserable that’s evidently what gives one higher utility. Under this definition of utility, the word “utility” could be removed entirely from economic theory and replaced with “preference ranking for individual X” with no loss of content. And, incidentally, that’s really what economists working with that sort of “utility” should do and would do if they were interested in intellectual clarity and honesty in the normative realm and not also in misleading rhetorical effect.
Cardinal utility, which adds the notion of an amount of “utility” so, for example, an individual can be said to get twice as much “utility” from fulfilling preference A than preference B, is an entirely different concept. In that case, one is obviously talking about something beyond merely a preference ranking. So what is the thing someone is doubling when we discuss someone getting twice the cardinal utility from A as from B? One likely suspect is we’re talking about some notion of happiness as with traditional philosophical utilitarianism. However, that doesn’t really fit with the notion of revealed preference and the idea it doesn’t really matter for utility if preferences lead to anything anyone would normally associate with happiness, per se, beyond the mere fact of the preference itself. The more relevant concept is really a sort of narrowly conceived satisfaction derived from preference fulfillment that applies even when one fulfills preferences that to an observer appear to make one miserable.
It’s important to note we must be talking about an inaccessible internal perception of satisfaction from preference fulfillment for this type of utility to work with the logic of neoclassical welfare economics because we need to preserve the result we cannot make interpersonal utility comparisons, which is central to the normative argument in neoclassical welfare economics under both formulations of utility. We can’t be talking about observable indications of satisfaction as we could and would with respect to happiness in traditional utilitarian ethical philosophy. As we discussed a moment ago, under ordinal only preference rank utility, talking about interpersonal utility comparisons is gibberish. In contrast, under cardinal perceptions of satisfaction utility, making interpersonal utility comparisons is impossible because we cannot as a matter of fact access or measure the relevant subjective perceptions. Same theoretical significance in terms of the math. Two very different concepts in terms of the normative or ethical content.
In that sense, revealed preference is the only objective indicator of both ordinal only preference rank utility and cardinal internal perception of satisfaction from preference fulfillment utility, but it would be confusing to call both “preference utility,” wouldn’t it? Confusing? That was my little attempt at a joke. Add it to the list, right? I’m sure people do it all the time. But let’s not do it here. Let’s reserve the term “preference utility” for ordinal only preference rank utility and use “perception utility” for cardinal internal perceptions of satisfaction from preference fulfillment, just to keep track of what we’re actually talking about.
In a normative context, preference utility is obviously an entirely different beast from perception utility. One can’t just switch between them willy-nilly. Doing so amounts to the logical error philosophers call terminological equivocation. The flippant attitude of purveyors of bad economics toward the distinction seems to me an example of mathematical formalism masking conceptual confusion in the normative realm. People used to working in a positive, scientific, predictive modeling mode may suppose all they’re doing is changing the mathematical specification in a model so what’s the big deal? And they’re right, in a positive context in which a model may be unreal in its entirety and still be highly evaluated because of its predictive ability, it really doesn’t matter. Use this, use that, use whatever you like. But in a normative context, of course, it does matter. In ethical philosophy and really philosophy in general, one must use language carefully, precisely, consistently, or confusion and error will be the inevitable result.
In a normative context, it’s actually much more philosophically useful to distinguish a version or conception or definition of “utility” where it doesn’t exist per se (exemplified by ordinal only preference rank utility) and a version where it does exist in some sense, if only as an internal perception (exemplified by cardinal perceptions of satisfaction from preference fulfillment utility).
In the case of the version of utility that exists in some sense, perception utility, the ethical proposition we should maximize total social utility can be shown to be weak, normatively or ethically implausible, and certainly very controversial, if one takes the time to analyze it properly and doesn’t just wave one’s arms about hoping other people lose track of what one is talking about. I’ve done it before a couple of times, but just to give the short version imagine having a dream where one can access or measure perception utility and hence make interpersonal utility comparisons. Think about the normative ramifications of accepting as one’s goal maximizing total social utility if one person’s capacity to generate perception utility were much higher than everyone else’s. Not necessarily a pretty picture, is it? Basically just whatever that person prefers; others be damned. It’s only a dream? A thought experiment? Sure. Of course it is. But it’s a revealing dream or thought experiment that shows no one who supports maximizing total social utility defined as perception utility in the context of neoclassical welfare economics does so out of any genuine interest in maximizing that type of utility. They support it because it’s inapplicable to real world conflict situations, because we don’t actually live in the postulated dream / nightmare world where we could really access it and maximize it, and hence it makes room for other ethical values in those conflict situations. Well, that and support for the trivial normative proposition we should let other people do what they want to do when what they want to do doesn’t conflict with what anyone else wants to do.
In the case of the version of utility that doesn’t actually exist per se, preference utility, the ethical proposition we should maximize total social utility just means we should generate an outcome consistent with certain propositions about how to treat other people expressing their preferences in non-conflict situations. That’s it. Under this version of utility, ethical or normative propositions about “maximizing utility” aren’t really about increasing the amount or level of something called “utility” but about generating outcomes consistent with certain ethical propositions about how to treat other people expressing their preferences in non-conflict situations. The propositions are just expressed in a funny way in an apparent attempt to deceive or confuse the unwary.
In previous posts, I discussed the distinctive form of neoclassical welfare economics as an ethical half theory. That applies to both preference utility and perception utility, but in the latter case only because of the unobservable or inaccessible quality of perception utility. Were it not for that problem one could, of course, base an entire ethical theory on the objective of maximizing perception utility, which would be somewhat similar to the program of traditional utilitarian ethical philosophy with respect to maximizing happiness variously defined. We already discussed it would be ethically implausible and certainly controversial to do that, but at least it obeys the laws of logic. However, if one does an end run around the inability to access or observe or measure internal perceptions of satisfaction directly by using a so-called social welfare function, where one introduces some weighting scheme to allow interpersonal comparisons of perception utility with the weights based on some criteria one finds ethically relevant, including for example criteria related to observable indicators or clues or beliefs relating to happiness or satisfaction (such as implicitly setting everyone’s capacity for utility or maximum utility to be at one level and then considering the implications of diminishing marginal utility), then one can mimic a more plausible ethical theory by expressing, using the type of utility used in economic theory, ethical concepts relevant to other definitions of utility from traditional utilitarian ethical philosophy or indeed other ethical principles entirely. It’s an awkward, unnecessarily confusing crazy quilt of an ethical theory, but it gets the job done in some respects. In contrast, trying to use social welfare functions with ordinal only preference rank utility doesn’t really make sense at all. Weighting something that doesn’t exist in the first place to facilitate interpersonal comparisons of a concept that is undefined in an interpersonal context is just gibberish. A confusing Frankenstein monster of unrelated ethical concepts and terms is one thing, but I think we really must draw the line at complete nonsense.
However, the mere availability of social welfare functions doesn’t convert neoclassical welfare economics using perception utility into a full ethical theory. The weights are not a part of the theory but exogenous to it, and are just as important or more important for the normative conclusions as the normative or ethical content and positive factual content that does appear in the theory. In addition, if one considers social welfare functions and the ethical or normative reasoning underlying the weights used in social welfare functions as applying only at the stage of deciding between perfectly competitive, Pareto Optimal, economic efficient outcomes, a great deal of the submerged normative content I’ve discussed in previous posts will remain. Honestly, it’s a bit of a philosophical mess, isn’t it? If we want people to be able to discuss these normative or ethical issues openly, honestly, transparently, then we really should remove the normative or ethical content from economics. Economists make bad ethical philosophers. They aren’t trained in it, aren’t interested in it, and produce risibly non-rigorous and amateurish normative theories when they set they hands to it. Ethical decisions relating to the evaluation of economic systems and outcomes belong with the people and should be addressed in the political arena by democratic decision making bodies, not in ivory towers by philosophically and in particular ethically clueless economists.
