Tuesday, September 11, 2012

Three things to remember about neoclassical utility theory

1. Utility and preferences are two different things: utility functions represent preferences.
2. Preferences are subjective.
3. Statements about welfare is really a sort of ethical or philosophical endeavor, not a scientific one.


I think these clear up some of the advantages that Jason Kuznicki attributes to "Austrian Utility Theory". I don't meant to denigrate Austrian views of utility. They are thoroughly neoclassical views, after all - and what I'd argue is that they are pretty comparable to mainstream neoclassical views.

Kuznicki begins: "To my mind, the Austrian school’s account of utility is (1) true (2) completely different from mainstream economists’, and (3) very interesting." I would say that the Austrian school's account of utility is (1.) true, (2.) almost exactly the same as mainstream economists, and (3.) very interesting.

He writes: "There’s no such thing as a hedometer, a device to measure utility, and without it we are thrown back on rank order. Each of us judges our own desires in a sort of footrace in our minds. Austrians strongly insist on ordinality, and it seems obvious to me that this is correct."

If you read an Austrian on utility you'll often get an answer like this and an initial claim that the mainstream thinks utility is cardinal and Austrians distinguish themselves by claiming it is ordinal. This is part falsehood and part improper comparison. It's true that the mainstream uses the word "utility" to describe a cardinal magnitude. But the important thing to remember is that if you stop there you're making an improper comparison. For the mainstream neoclassical, utility functions simply "represent" ordinal preference relations. So what Jason calls "utility" for the Austrians is what most of the rest of the profession calls "preference relations". This is often obscured because undergraduate textbooks gloss over the point - it's not a particularly important distinction unless you're neck deep in micro theory. Even easier graduate textbooks will only refer to it in passing. But higher level material maintains a very sharp distinction between ordinal preference relations (which is really what the Austrians have in mind) and cardinal utility functions. When talking about "the mainstream" Austrians often jumble this together. Utility functions are only acceptable if they maintain the critical characteristics that preference relations have to maintain. And it should be obvious that nobody thinks preferences or welfare is cardinal because an early exercise (even for undergraduates) is to show that you can monotonically transform utility functions.

Kuznicki says that the mainstream "grudgingly" came around to this. I don't know a whole lot of the history of thought on this, but I have my doubts about whether "grudgingly" is the right word. I read a little Jevons on this and see no signs that he ever say utility functions as doing anything other than representing preferences. It wouldn't surprise me if there were some early guys that actually thought "utils" had independent meaning and you could count them up, but if it ever characterized the "mainstream neoclassical view" that's news to me. You've got to remember that Menger, along with Jevons, is that father of the mainstream neoclassical view, so it's doubtful that Austrians ever thought anything different from the rest of us on this. Ragnar Frisch puled together modern formal treatments of preference relations in the 1920s, so certainly this was the established mainstream view by then. There's nothing grudging about it. We don't need to be grudging - it's just the way it is: preferences are ordinal.

Kuznicki then moves into utility-monster type stuff. Again, since mainstream economists view preferences as ordinal and magnitudes of utility as essentially meaningless and entirely incomparable to each other, none of this is as problematic for the mainstream as he suggests and the issues here are pretty familiar.

As I've said in the past, though, this does pretty much put welfare analysis outside of the realm of science and into the realm of ethical or policy judgement. That doesn't mean that welfare economics is wrong - it just means that we need to be aware that there's an implicit assumption that as far as mainstream economists are concerned, thou shall not suffer a utility monster to live.

Utility monsters do nothing to disrupt the mainstream neoclassical treatment of the scientific study of human exchange behavior.

Moving on, Kuznicki writes: "The comparison with the mainstream approach is sharp: Mainstream economists often build first-approximation models that assume perfect knowledge, then try to add the imperfections later."

Again I think this is a misconception. All preferences (and therefore all utility functions) are subjective. That means that it is not "perfect knowledge" that is being assumed, even if a mainstream economist tells you they're assuming perfect knowledge. Since preferences are subjective, what that implies is that agents facing a choice have some subjective view of the different trade-offs. "Perfect knowledge" is even brought up because "completeness" is one requirement for well-behaved preference relations. Complete preference relations imply some subjective ordinal assessment of every option available. It doesn't have to be accurate in an objective sense, and presumably heuristics could fill in the gaps for our frail minds. Completeness is not as objectionable an assumption as it is often treated, and certainly no one is assuming that agents are "omniscient" in an objective sense.

I think the discussion of transitivity also takes Austrian claims on the matter for granted. Kuznicki writes:

"The Austrian account of utility also handles intransitivity of preference both elegantly and well.

When declared preference is intransitive — that is, when actors say that preference A beats preference B, and preference B beats preference C, and preference C loops around and beats preference A — the problem is not, in fact, that their preferences are intransitive. It’s that to express an intransitive preference is easy, but to act upon it is impossible. One can always express intransitive preferences, just as one may always write intransitive inequalities using mathematical notation. But to act on an intransitive preference would require a time machine and/or some fairly weird assumptions about causality, identity, and multiple universes. Time doesn’t loop in the real world, so expressions of intransitive preference are always merely expressions."

He's right, but again this is pretty much how mainstream neoclasical economists talk about it. They would say that acyclicity holds if and only if a set of preference relations is maximizable. It's the exact same point. There's nothing particularly Austrian about this.

There is one thing in Kuznicki's post that I can't really speak to and that's this idea (allegedly Austrian - I believe him I've just never come across it myself) that we don't seek happiness, we seek to remove unease. That may indeed be a distinct viewpoint but I can't really figure out why that's an especially important distinction to make.

Generally speaking, in discussions of utility theory, a good rule of thumb is not to take everything an Austrian says about the difference between Austrian and mainstream approaches for granted.

22 comments:

  1. Good post. We've obviously been over this subject many times before, but the zombie... just... won't... die!

    I'd only add that the key point from analytical perspective is whether preferences and choices are rational (i.e. consistent). And, as I've tried to point out elsewhere, a lot of the experiments conducted in behavioural economics have actually validated the rationality conditions put forward by the likes of Samuelson (revealed preference theory, etc).

    This utility issue is a manufactured controversy and largely reflects an inability to understand or properly engage mainstream economics.

    ReplyDelete
  2. I don't know a whole lot of the history of thought on this, but I have my doubts about whether "grudgingly" is the right word. I read a little Jevons on this and see no signs that he ever say utility functions as doing anything other than representing preferences. It wouldn't surprise me if there were some early guys that actually thought "utils" had independent meaning and you could count them up, but if it ever characterized the "mainstream neoclassical view" that's news to me. You've got to remember that Menger, along with Jevons, is that father of the mainstream neoclassical view, so it's doubtful that Austrians ever thought anything different from the rest of us on this.

    Daniel, even the early Austrians wrote about utility as a cardinal thing. (For sure BB did all over the place; I am not sure if Menger did, I didn't read his as much, but I think he did.)

    In the mainstream there was a debate about it, with some saying it was dubious and others saying you needed it to do price theory. In Hicks' book (in 34?) he goes over the revolution inspired by Pareto of looking at indifference curves as showing bundles more or less preferred, rather than elevations on a utility mountain.

    You may be interested to know that in grad school I tsk-tsked the Austrians on this stuff too.

    Last thing: Just so you know, I think it sounds really weird for you to say a guy in 1871 never "disagreed with the rest of us."

    ReplyDelete
    Replies
    1. Bob, when you post comments like this, I have visions of us one day establishing a working relationship like this.

      In an economic sense. Obviously, and of course.

      Delete
    2. Wait stickman...which of us is insane and which is dumber?

      Delete
  3. It wouldn't surprise me if there were some early guys that actually thought "utils" had independent meaning and you could count them up, but if it ever characterized the "mainstream neoclassical view" that's news to me. You've got to remember that Menger, along with Jevons, is that father of the mainstream neoclassical view...

    The ideas of the marginal revolutionaries were actually preceded in large part by the likes of Gossen, Cournot and Dupuit. Of course, Jevons and Menger (and Walras) receive deserved credit because they were the chief popularizers of marginal utility and subjective valuation... But they weren't the originators. Indeed, Jevons graciously acknowledged as much during the second edition of Theory of Political Economy when he became fully acquainted with the writings of these forerunners.

    ReplyDelete
  4. Briefly put, I think you are mostly right. Some modern day Austrians wrongly ascribe Neoclassical microeconomic theory as utilizing cardinal utility. The accusation probably came from Rothbard, when he wrote that MU/P=MU/P is an admission of cardinal utility. Its no controversy though, as Professor Murphy explains, that older Neoclassical economics did work with a "cardinal utility." I've learned this in a History of Thought class, and Rothbard (right after making this accusation) does take a quote from Jevons on utility. But things did change in the 30s, and so its wrong to assert Neoclassical economics still holds this issue. In what I believe to be a useful analogy, holding that Neoclassical microeconomics requires cardinal utility is like holding that Austrians believe in multiple natural rates of interest. Both ideas were held by thinkers in the past, but both are outdated.

    Now what I always do take issue with is when welfare economics comes up (building off of things like consume surplus, the compensation principle, Kaldor Hicks, etc), the ironclad utility restrictions are "loosened" and now welfare economics becomes something of its own, even now requiring ethical judgements. Given the obvious restrictions ordinal-nointerpersonal utility comparisons places on welfare economics, it seems Neoclassical economics then works hard to readmit all those things through the back door to allow it to do cost-benefit analysis.

    ReplyDelete
  5. You could have mentioned the Ellsberg paradox as departures from utility, Daniel Kuehn. But otherwise, good post.

    ReplyDelete
  6. Kuznicki: "Austrians strongly insist on ordinality, and it seems obvious to me that this is correct."

    Gee, it seems obvious to me that preferences, hence utilities, are only partially ordered. Lack of transitivity is a big clue. :)

    ReplyDelete
  7. Kuznicki: "the problem is not, in fact, that their preferences are intransitive. It’s that to express an intransitive preference is easy, but to act upon it is impossible."

    Sigh. The lack of empiricism rears its head. Of course, people can and do act on intransitive preferences. Just look.

    ReplyDelete
  8. Don't forget that neoclassicals say increasing marginal utility is allowed, while Austrians claim it is impossible!

    ReplyDelete
  9. re:"Moving on, Kuznicki writes: "The comparison with the mainstream approach is sharp: Mainstream economists often build first-approximation models that assume perfect knowledge, then try to add the imperfections later."

    Actually, I think this is generally correct if you're talking about dynamic models. Rational expectations assumes perfect knowledge in a non-trivial way. You're not just assuming subjective preferences and knowledge, you're saying that, not only does the agent know everything relevant to the determination of future equilibrium variables, that agent also has the computational capacity to deduce what those variables will be. And this type of assumption is usually part of the baseline dynamic neoclassical model. So I wouldn't call this a complete misconception.

    ReplyDelete
  10. Thank you for the thoughtful and challenging reply. As I'm self-taught in economics, I am apt to make mistakes. I welcome corrections.

    I don't welcome all of them, however. Min is quite wrong to assert that a person can act in any literal sense on intransitive preferences. That would involve setting up some sort of time paradox. The best we can say is that people declare themselves to hold time-inconsistent preferences, not that they can act on them. They have stated inconsistent preferences, but revealed preference only ever reveals one order.

    Time is relentlessly linear, and the point I was making here was never to deny the existence of, say, regret in the absence of relevant new information. Of course it exists, and what it indicates is a defect of individual reflection or individual psychology, not some surprising new economic phenomenon.

    Daniel, as for your responses, I may have more to say later. My overall impression is that you inconsistently both affirm and deny my contention that, while mainstream economics has admitted utility to be ordinal, it keeps trying to bring cardinality back in anyway.

    Certainly that's the case with welfare economics, if nowhere else. I can hardly blame them for wanting it. A world of cardinal utility would be weird and surprising, but it would almost certainly be happier than our own.

    Finally, I am surprised that the idea of removing unease in Austrian economics is unfamiliar to you. It's a central theme early in Human Action, by no means incidental to the argument.

    ReplyDelete
    Replies
    1. "Min is quite wrong to assert that a person can act in any literal sense on intransitive preferences. That would involve setting up some sort of time paradox."

      Well, I would say, non-transitive, so maybe we mean different things. I don't think so, but maybe?

      Here is what I mean by non-transitive. Say that Jack, given a choice of two different fruits at a certain level of hunger, chooses apples over pears, pears over bananas, and bananas over apples. Those are actions. No time paradox is involved.

      Delete
    2. At any given time, he is only presented with two of the three. He chooses only one. There is no time paradox, but there is also no revelation of intransitive preference. All that is revealed is a changing preference over time.

      Delete
    3. There is no changing preference. The choices are clear and constant. To say that the preferences change over time is unparsimonious and unempirical.

      Delete
    4. Let me clarify a bit. Given the choice between two of the three fruits, Jack chooses apples over pears, pears over bananas, and bananas over apples. Over time, we see those three choices. Under the assumption that preferences change over time, we should see all six choices.

      Delete
  11. "Austrians strongly insist on ordinality, and it seems obvious to me that this is correct."

    Examining my own utility it is obvious that for those quantities that are divisible I do trade-offs that make it clear that utility is cardinal (and probably a function like sum( a[i]*ln[q[i]) where q is the measure of something I want and a[i] is a measure of the relative importance of q[i] to me. I try to construct an optimal portfolio - even if I am just choosing how much butter/bread/meat should go into a sandwich.

    Ordinal only arises when one has to make a once and for all either or choice and we make very few of those. Even with things that are inherently indivisible, like people, I might maximize utility by allocating the available time among the friends.

    ReplyDelete
    Replies
    1. Are your utility functions always (or ever) of a constant slope? I am sincerely puzzled. How do you decide when to start acting to attain a given good, and how do you decide when you're ready to stop?

      Delete
    2. Jason

      I do not think my utility function ever has a constant slope. It always has diminishing marginal returns. Fundamentally I allocate my time to activities and to earning money. Since I get paid by the hour worked there is a direct correlation between work effort and earned income (with a 30 to 90 day lag). I have a cash reserve equal to several years income so I can make inter-temporal choices.

      I try to maximize my utility but there is always going to be uncertainty in the "cost", the "success probability" and the "benefit" of the outcome. For example, if I pursue a pretty girl, I (generally) do not know in advance the costs or probability of having sex with her and I do not know how much I will enjoy sex with her if I achieve it.

      Most goods require an allocation of time and money. I have a budget constraint on my total time and a further long run budget constraint on how many hours a week I can work. I will start to act to attain the specific good when my best estimate of the expected utility (discounted appropriately for risk) net of the opportunity cost of forgoing other utility generating activities is positive. I stop when the expected utility of further effort, net of the opportunity cost, swings negative.

      Delete

All anonymous comments will be deleted. Consistent pseudonyms are fine.