Skip to content

Economists don’t think like accountants—but maybe they should

Joseph Delaney quotes Frances Woolley:

In other words, the reason we care about inequality is that it reduces the happiness achievable from a given amount of income. How much depends upon the happiness/income relationship. Does the marginal utility of income fall rapidly? Or is the happiness from the 100,000th dollar almost as great as the happiness from the 100th?

Delaney goes on to discuss the marginal utility of income, but I have a different point to make.

Woolley is a professor of economics.

Economists seem to rely heavily on a sort of folk psychology, a relic of the 1920s-1950s in which people calculate utilities (or act as if they are doing so) in order to make decisions. A central tenet of economics is that inference or policy recommendation be derived from first principles from this folk-psychology model.

This just seems silly to me, as if astronomers justified all their calculations with an underlying appeal to Aristotle’s mechanics. Or maybe the better analogy is the Stalinist era in which everything had to be connected to Marxist principles (followed, perhaps, by an equationful explanation of how the world can be interpreted as if Marxism were valid).

I can understand this from a historical perspective—this sort of insistence on microfoundations is a tradition in economics—but from the outside it looks a bit silly.

OK, now back to the Woolley quote, which reminds me a bit of another economics professor, Greg Mankiw, when he wrote about taxation as a moral issue, implying that that the state only has a right to tax things that are “unjustly wrestled from someone else.” This didn’t make much sense to me–whether it’s the sales tax, the income tax, or whatever, I see taxes as a way to raise money, not as a form of punishment.

Similarly, Woolley connects beliefs about inequality to attitudes about a utility function. There’s a difference, though. Inequality is macro, and it’s real (in the sense of being directly measurable), and utility is micro and doesn’t exist.

To return to the physics analogy, suppose someone said (correctly) that to understand solid structures, you need to understand electrons. Solid structures exist and can be observed; electrons are more hypothetical. The difference is that electrons are not merely hypothetical, they are also a fundamental part of a model that has been tested and applied a million times. In contrast, “utility,” like “the life force,” is a sometimes-useful construct but it’s based on an intuitive notion of psychology that we have no particular reason to trust (any more than we trust the various principles of folk physics, folk biology, and so forth).

Economists, though, are so used to requiring their microfoundations that I don’t think they reflect enough on how silly this can sometimes be. I have no problem with utility analysis in its place (we have a whole chapter on utility-based decision making in Bayesian Data Analysis, and I’ve published some papers on the topic too), but when it comes to taxation, there are some more direct principles at hand, such as the need to fund the government and ability to pay. I have a feeling that right now that various bankrupt governments have more pressing concerns than the (hypothetical) marginal utility of someone’s 100,000th dollar!

I do think there can be good reasons for opposing tax increases (starting with the feeling that the money, if taken, will be spent in some wasteful or distasteful manner), but I think here the irritation is coming more from the sense of powerlessness—these bozos are taking my money!—than a loss of marginal utility.

To put it another way, I think economists might benefit from acting a bit more like the stereotype—as accountants who carefully keep track of the dollars and cents—and consider the psychological modeling as an interesting sidebar rather than a foundation to their analysis.

P.S. More along similar lines from John Quiggin. He is making slightly a different point than I am, though. He is discussing the unfalsifiability of the rational actor model (as in Popper’s classic criticisms of Marxism and Freudianism), which is a fine point, but I’m making more of a sociological point here, which is that even those economists who do not follow any sort of “party line” still seem to see the utility model as a necessary microfoundation of any serious social-science analysis.

And that’s what’s particularly bugging me right now, the idea that we should be using 75-year-old psychology as a foundation for modern economics. Rather than making the usual analogy of a house with a foundation of sand, I prefer to think of it as a house that could stand on its own, with an unnecessary sand foundation introduced below, a foundation that makes the house itself that much less stable!

To get back to the title of this post, I’d rather see some really good accounting than some really clever but misinformed amateur psychology.

P.P.S. Further discussion in comments below. You can judge for myself if I have convincingly replied to the doubters.


  1. Ian Fellows says:

    At the very least there should be (and may be as I don’t know the lit at all) a clear distinction between the accountancy and philosophy. Physicists are in the nice situation of only needing to describe nature, where as economists often both describe the economy and proscribe solutions. What you call micro-foundation could also be described as philosophical underpinnings. The certainly have no place in the science of economics, but they most certainly have a place in figuring out how to order our society in a way that is consistant with economic realities and philosophical morality

  2. Florian says:

    I agree with you that using “utility functions” as foundation of moral arguments are kind of sketchy.

    The use of utility functions by itself is OK though – it just depends on how carefully you specify them. In particular, as long as a subject’s behavior satisfies some consistency requirement (“Weak axiom of revealed preferences”), you can always find a function such that it looks exactly *as if* the subject were maximizing ‘utility’ as expressed by that function.
    The next step, to put some further assumptions on such a function, is also still OK with me – for example, choosing a concave function to capture some sort of saturation/smoothing behavior often seems plausible.
    What is usually not OK, and I’m with you on that, is to do interpersonal comparisons of utility. Especially the utilitarian version of just adding up everybody’s ‘utility’ is unfortunately something you see quite a bit. I bet Professor Woolley is not a theoretical microeconomist.

  3. Jon Baron says:

    I don’t get it. You say “ability to pay” is OK, but you don’t want to use utility as part of the definition of ability. What is ability to pay? It is the opposite of the loss that you suffer from paying. The loss not in money, but in what matters to you. Ability to pay is the relative lack of suffering from paying. That is, the disutility of paying. That is what utility is about.

    People with high incomes suffer (on the average) less, lose less utility, from paying a given amount. That is the utilitarian justification of progressive taxation.

    Even better, people who consume a lot will surely benefit less (in utility) per dollar, the more they spend. A graduated consumption tax would be better still than a graduated income tax, in utilitarian terms. When you are buying your 7th house, you are not even going to have much time to enjoy it.

  4. Anderson says:

    Just use a linear utility function, problem solved. :-)

  5. Jonathan says:

    Economics can of course be written entirely in terms of preference relations instead of utility, but it’s considerably more complicated and it’s unclear (to this economist) whether it’s worth the effort. But the test of utility theory isn’t neurology — it’s results. To continue your physics analogy, light “chooses” the least time path. If we endow light with desires to travel with minimum time paths we describe the universe, we carry out silly folk notions of intentionality, but we also say interesting things about the structure of the universe. No economist I know believes literally in utility as an actual description of the neurology. But the notion that people seem to consistently prefer more to less (in many contexts) and will trade off good A with good B in ways that often represent the equivalent of diminishing marginal utility both explains factually what people do in many circumstances and creates intersting insight when it’s factually wrong. None of this requires “utils” or perfectly consistent rationality, but solving the problem that homo economicus faces helps us understand what real people do, however unrealistic the factual underpinning.

  6. Andrew says:


    I agree that the concept of a utility function can be useful; where I disagree is with the idea that it is necessary or that it is somehow more foundational than a direct analysis. The point of “ability to pay” is that people making more money will be more able to actually come up with the money to pay the taxes. “Ability to pay” comes up in all sorts of situations, from bill collecting to child support, and it can be discussed without reference to utility.

    I have no problem with your utility-based justification for a graduated tax rate; I just don’t think that has to be the reason we care about inequality (as the economist wrote).

    • Jonathan says:

      Oh, OK. Then it seems to me your problem isn’t with utility… it’s with utilitarianism, which lots of people, economists and noneconomists, have lots of problems with.

      • Andrew says:


        My problem is with the assumption, prevalent among economists and some political scientists, that every analysis needs a utility-based microfoundation. I take a more statistical, or phenomenological, view, and believe that it can be fine to analyze data as they are. For example, consider our paper on police stops. We just presented the numbers with no utility model of the behavior of police or of suspects. Similarly, I think it’s just fine to analyze proposed changes in the tax code without needing a model for the marginal utility of dollars. My problem with the utility model is partly that it can be completely wrong (see my paper from 1998 for an example) but, more generally, that it’s silly. It’s folk psychology. It’s a reasonable-seeming model of human cognition and behavior, just as folk physics is a reasonable-seeming model of light and motion. But psychology has advanced from there.

        It can be fine to use utility as a convenient model (see chapter 6 of my book with Jennifer for our discussion of choice models) but I don’t see it as foundational, and I’m bothered by those who do, especially considering that this view is often implicit, as if a social science model must have this sort of microfoundation or is otherwise a failure. I say, No! Often the foundation is weaker than the house that it purportedly supports.

        • Utility theory provides the justification and the reasoning for normative conclusions, though. I think that’s why so many economists want to keep it.

        • Zach says:

          Ok well I think there are several problems with your argument (I am a political scientist). Yes it is folk psychology, people don’t really act as though they are expected utility maximizers. I would also agree that RCT theory is basically non-falsifiable (but it can generate falsifiable predictions). Human behavior is obviously enormously complex, and since cognitive neuroscience doesn’t have a real handle on how people make decisions, it is safe to say the social scientists don’t either. Given that constraint, we still want to try to understand how groups of people interact in political situations. One of the things that makes political science interesting to me is the competition between groups: strategic interaction. Here, RCT is a useful abstraction. We assume away all the real cognitive processes going on in individuals (also because we don’t have a good grip on how individuals interact to form group preferences) and just look at how strategic interaction works. We’ve been working on this since the 60’s really, and have just barely scratched the surface. RCT provides analytical leverage to study strategic interaction that would otherwise not be possible. Also bear in mind that we are dealing (especially in IR conflict stuff) with data that is sparse, censored, etc. and if we want to get at a causal mechanism we think exists, the currently best way to do that (since you can’t actually get reliable information from decision-makers) is to create a formal model and then test empirical implications you derive from that.

    • Why shouldn’t tax policy try to minimize the reduction of psychological well-being that comes with increasing taxes? It’s necessary to be practical, it’s great to have morals on your side.

      Utility != experienced satisfaction in general. But there is plenty of evidence that subjective well-being goes up as the log of money, or at least concavely in it. So in the case of taxes relatively less psychic harm would be done by raising $1,000 from a rich person than from a poor one.

      If rank effects and work satisfaction dominate in the upper incomes, as I think they do, then even less satisfaction will be robbed from the upper earners with a progressive tax structure.

  7. numeric says:

    better analogy is the Stalinist era in which everything had to be connected to Marxist principles

    Actually, everything had to be connected to Stalin. Solzhenitsyn describes in the Gulag Archipelago how professors who cited Marx and or Lenin but not Stalin were invariably arrested.

  8. Merijn Knibbe says:

    Well, let’s take a look at what neurologists write about this. Oops. The modular structure of the human brain itself disables the very possibility of a ‘well behaved’ (i.e. concave, stable) utility function. Interestingly, it can be proved that part of our brain is specialized in making us believe that our behavior is consistent and planned, even when it isn’t… Seems that the first principles are void and inconsistent with scientific investigation.

    Read Victor Lamme (of Nature/Science fame), ‘De vrije wil bestaat niet’ (‘There ain’t no such thing as a free will’ ). Or research the resources on:

    • Acilius says:

      Good point. As I’m sure you know, the philosophical school known as “Virtue Ethics” tends to start with the neurological work that made unidimensional concepts of “happiness” untenable, and to loop back through Aristotle and other premodern thinkers before rejoining contemporary, post-Utilitarian thought.

  9. Two good concepts regarding taxation I’ve been exposed to, are:

    via Prof. DeLong: “Taxes are like plucking a hen, you want to gather a maximum of feathers with a minimum of fuss” -I think this points us towards high rates but relatively lax enforcement of harsh penalties.

    and via Prof. Mankiw: “Pigovian taxes is the concept of using taxation in a regulatory way, to reduce activity society doesn’t want, to subsidize activity society does want”. -I think this points us towards taxing stuff like wasteful consumption, consumption with high negative externalities.

    • It also assumes that “what society wants” is well-defined. In the case of the US it’s a map from 300 million dimensions onto one.

      • human mathematics -there’s always death by a million niggles.
        But I think the concept of pigovian taxes + cost benefit analysis points usefully towards things like pollution taxes and away from things like vaccine taxes (taxing people for getting the annual flu vaccine).

        • You’re right, that critique is a bit off-topic, as are others I could make of those statements. They do both point toward taxing inelastic things. I think it was a good juxtaposition.

  10. Frank says:

    I think something is being overlooked here. Unless you use lump-sum taxation, which is rare in the US, the change in incentives resulting from a new tax changes what and how much is produced and consumed in the economy, eventually feeding back to the amount of revenue the tax collects.

    To take an extremely simple example, it would be mistake to assume that because Y gallons of gasoline were consumed last year a $10/gallon gasoline would raise $10 * Y in revenue. To understand this kind of behavioral response, economists need models of individual behavior, indeed including those based on the idea of utility.

    If we were to raise the top income tax bracket to 90%, people at the high end of the income distribution would face two effects: first, the payoff to another hour spend working has dramatically fallen; second they no longer earn enough to afford the lifestyle to which they have become accustomed. The first effect would induce them to work less, the second would induce them to work more. The amount of revenue raised by such a tax depends sensitively on individual’s preferences and in this case involves the marginal utility of income. If you don’t like that term, just put it another way: whether the rich person facing such a tax works more or less than before depends on how she values the stuff she can buy with an amount of money equal to the difference between her income before and after the tax. Obviously “value” is relative, and in this case it is relative to how much she values leisure, time spent with family and friends and so on.

    Another important lesson from economics is that the person on whom a tax is levied is not necessarily the person who “pays” the tax. If you’re interested in inequality and related issues, this is clearly important. Employers seem relatively well-off, they have a high ability to pay, right? Let’s charge them a payroll tax to pay for social programs that we think are important. The question is, how does the tax get paid? Do employers end up earning smaller profits or simply paying workers less while earning the same profits? The answer depends on market factors that, once again, boil down to models of individual choice, and again are represented by utility.

    In public policy debates, these kinds of questions seem to me to be receive less attention than they merit from non-economists. Without models of individual behavior based on micro-foundations, it’s not clear how we could be able to address these important macro-phenomena. No economist I know believes in utility as anything more than a ranking and modeling device, but it does prove useful in examining these issues.

    • Andrew says:


      It makes a lot of sense to model how individuals might react to a change in the law—in particular, the idea that if higher-earning, more productive people work less, that this could reduce the nation’s overall level of wealth. I just don’t see “the marginal utility of income” as a useful concept here. I’d rather model the behavior directly than through a “utility function” which is a product of mid-twentieth-century folk psychology.

  11. James says:

    1) Jonathan mentioned this but may not have been sufficiently clear: the primary microfoundations of economics have not required the idea of utility since Arrow-Debreu in the early 1950’s. We still talk about utility because it is easier and because the results are usually equivalent to the more rigorous Arrow-Debreu way of thinking.

    2) There is a bunch of actual supporting evidence on the matter, for instance Justin Wolfers using survey data to find that happiness seems to rise with the natural log of income.

    3) If you have this problem with economics, you should have a much bigger problem with philosophy, where utilitarianism is one of the most common ethical philosophies, and as far as I know they rarely use evidence besides introspection and folk psychology.

    • Andrew says:


      I view philosophy in a “cafeteria Catholic” sort of way: I take what I like out of Popper, Kuhn, Lakatos, Mayo, etc., and I don’t worry about the rest. Economics is different because it is more closely linked to policy, hence I’m more bothered by its reliance on folk psychology as a purported foundation.

    • Carsten Bern says:

      “ou should have a much bigger problem with philosophy, where utilitarianism is one of the most common ethical philosophies,”

      And philosophers now there are problems with utilitarianism.

      ” and as far as I know they rarely use evidence besides introspection and folk psychology.”

      When you do, e.g., meta-physics (what comes before physics, cf. Aristotle), it’s tough to rely on evidence! Furthermore, if you deal with the very conditions of human existence (cf. Sartre), a survey really isn’t going to help much.

      It’s frustrating, how little people(including the blogger) know about philosophy.

  12. Cyrus says:

    With respect to the broader point, biologist David McFarland has argued (persuasively, to me) that in developing theories of behavior, there’s no escaping utility functions. Here is a pithy statement:

    D McFarland, “Defining motivation and cognition in animals”

    And here for more elaboration:

    D McFarland, T Bosser, Intelligent Behavior in Animals and Robots.

    McFarland has noted that utility maximization should not be confused with reasoning, as processes other than reasoning can be crucial — less in humans than, say, insects, but important nonetheless. The point is to characterize the mechanisms that govern regularities in behavior. Look at it like this: if you were to program a robot to act like a person (or an insect, or rabbit, whatever), how would you do it? Specifying a utility function would be necessary, no?

    On the more specific point about whether models like this should be applied to policy debates like taxation, I guess the problem is that the models have not been calibrated sufficiently to attend to the question at hand. We need experiments (e.g., tax rebate experiments) that allow that allow us to calibrate the model so that we can make reasonable approximations.

    • Andrew says:


      I agree with you that the concepts or rationality and optimality are important and fundamental. But Woolley was not making a serious model of cognition or behavior. Rather, she was relying on a folk-psychology concept of the utility function.

      Now, you might say that “utility” is just a shorthand for a bunch of assumptions about behavior and decision making. But that’s my point: that these assumptions are an outdated bit of folk psychology.

      You might also respond to me by saying that I should be able to evaluate Woolley’s arguments on an economic level and not worry about the validity of her psychology model. After all, she’s an economist, not a psychologist: It’s not her job to plausibly model human behavior and decision making, she just needs a framework that will allow her to do economics.

      And that’s my point: I think Woolley (and others) can do their economics directly without thinking that utility theory is some sort of microfoundation. Similarly, I think Mankiw could analyze the economic impact of taxes without assuming that the state only has a right to tax things that are “unjustly wrestled from someone else.” In both cases, the non-economic arguments (related to psychology and morality) can be evaluated on their own, but I don’t think it helps to treat them as foundational.

    • Corey says:

      > Look at it like this: if you were to program a robot to act like a person (or an insect, or rabbit, whatever), how would you do it? Specifying a utility function would be necessary, no?

      If you wanted to program a robot to act like a person, specifying a utility function would be the last thing you’d want to do. Most people’s choices do not satisfy the consistency axioms that lead to the existence of a utility function; any robot that was a utility maximizer couldn’t duplicate these sorts of inconsistencies. See this nice paper by Matthew Rabin for an example involving risk attitudes over modest stakes.

  13. gappy3000 says:

    One should comment too much on this issue, since Prof.Gelman has been very consistent on the issue. I find his stance unworthy of him, because of the superficiality of the arguments he keeps using, but hey, it’s his blog. Just two very quick notes. First, from a normative viewpoint, it’s legitimate to assume utilities or preferences and take it from there. It’s a deductive approach with very few assumptions. Don’t like the consequences? Revisit the assumptions, don’t just dismiss this as “folk psychology”. From a descriptive viewpoint, very non-trivial predictions (e.g., Rational Expectations) of this “microfoundation” approach have been empirically tested over and over and over. In nearly all tests I have seen, they perform better than available alternatives. Not that I am wedded to utilities, but just think they should be criticized on a different ground, as done by some commenters.
    Second: “This didn’t make much sense to me–whether it’s the sales tax, the income tax, or whatever, I see taxes as a way to raise money, not as a form of punishment.”. OK, aside from utilities, we know Prof.Gelman is a consequentialist by default, no questions asked. We need the money, now let’s raise them. A hospital *needs* two kidneys, a heart and a liver to save 4 people. and a guy checks in with stroke. We might as well let him die and harvest his organs. Less dramatically, we can take one kidney only.

    • Andrew says:


      I appreciate your patience in commenting. (I mean this sincerely and non-ironically.)

      Regarding utilities: I agree that if the science of psychology had stood still for the past 60 years or so, it could make sense to assume utilities and go from there. Just as, if biology had made no progress in the past century or whatever, medical researchers could go with various early models of human circulation. At this point, though, it all seems a bit silly.

      That said, I have no problem with utility analysis. I think it’s great. I devoted a whole chapter of Bayesian Data Analysis to the topic! I also like rational choice theory, and I even published a paper in the journal Rationality and Society on how it can be rational to vote.

      My problem is with the idea of utility theory as a necessary microfoundation to an economic analysis. I sometimes hear the attitude that social science without a utility microfoundation is somehow incomplete. I disagree, I think that in many cases we can study social science problems directly without feeling the need to derive our logistic regressions etc. from some underlying model which is itself bogus. Hence my analogy to the foundation that’s weaker than the house.

      Regarding taxes: I accept that Mankiw and I have different perspectives on taxation. It’s a topic Mankiw knows a lot more about than I do; nonetheless I believe that he has occasionally been led astray by his own expertise (as I discussed in the context of his distress over his projected 93% marginal tax rate). In any case, I don’t object to his inclusion of a moral dimension in his tax analysis. I just object to (what I see as) his implicit assumption that this dimension is necessary. That’s the connection to my critique of utility analysis when it is treated as a necessary foundation.

      • gappy3000 says:

        Yes, I think that of all people a bayesian statistician should like rational choice theory! I also beleive that completeness or (alleged) consistency is not the main goal of micro foundations. The goal is to clarify the assumptions underlying otherwise arbitrary phenomenological, ad hoc assumptions. For example, rational expectations can be interpreted as a limit version of adaptive expectations. It also allows for more empirical testing. Without utilities, most of the taxation literature becomes moot, including Mankiw’s contributions (or at least the “taxing height” paper). In fact, without this utility-based approach, I am afraid we’d resort to shaky moral arguments a la Mankiw.

        For the time being I’d go with mainstream Economics, and try not to think like an accountant, a moral philospher or a practical statistician.

  14. Jason Eisner says:

    Hi Andrew:

    I’m confused. Are the economists in question really assuming that individuals act to maximize their individual utilities? That’s what you seem to be objecting to, with your comments about “folk psychology,” “psychological modeling,” and “micro” models. But I don’t see it in the quotation above. In fact, I had the impression that (many) economists had moved past the assumption of rational actors.

    But one would still like to rationally choose among social policies. For that, one needs an objective function. “Societal utility” just refers to the function whose expectation one decides to maximize, right?

    Utilitarians take this societal utility to be a sum of individual utilities. A first-year econ student might think that this stance implies indifference to inequality, but it doesn’t — Woolley makes the standard rebuttal that utilitarians “care about inequality” because a sum of utilities, unlike a sum of dollar amounts, might be improved by a different distribution of the same total wealth.

    Woolley then goes on to observe that how one should change this distribution depends on the particular distribution and the actual utilities involved. One shouldn’t simply follow the gradient of the Gini index — that’s what her post was criticizing. She could have added that by placing everything in the common framework of utility, you can discuss tradeoffs between a policy’s redistributive effects and its many other costs and benefits (some of which you allude to).

    I agree that we can’t point to “utility” anywhere in the bubbling soup of the human psyche. Still, one does have to operationalize the benefit of a policy somehow in order to make policy choices. (What is the value of a life, etc.)

    I’d also agree if you said that it’s tough to come up with a specific utility function, and that my happiness at having $X presumably depends on more than just X as in the above cartoon (it also depends on local prices, my neighbors’ wealth, my emotions around earnings and taxes, etc.). But I think Woolley would agree with these points too. She’s picking on the traditional Gini index as being an arbitrary and insufficient statistic, not recommending a specific substitute.

    • Andrew says:


      I agree that Woolley’s post was interesting and informative. That’s one reason I picked on her: even in the midst of an interesting and informative discussion, she couldn’t help formulating her ideas in terms of the microfoundations of folk psychology.

      To respond to your first paragraph above: I think utility theory is folk psychology. It’s a crude and outdated model of human cognition and behavior. I agree that such a crude model can be useful (let me again point to the relevant chapter in Bayesian Data Analysis), but it’s still folk psychology and I think it’s still not a good foundation. My problem with (many) economists and political scientists is that they seem to often feel that an analysis is incomplete without a utility-based microfoundation. To me, the analysis comes first and the microfoundation might be useful in some settings but is hardly required.

      • Jonathan says:

        At the risk of violating some law of blogospheric debate, I’m going to agree with you, Andrew, for the most part, now that I understand your argument somewhat better. There are very few circumstances in which a utility foundation is required to actually do microeconomics, and I agree that there is a bias in academia to take arguments more seriously that start with utility (or at least with preferences, Arrow-Debreu style, as James and I both pointed out above.) Logit analysis is probably the most egregious example: McFadden demonstrated that logit analysis make sense if individuals have random utility corresponding to Type II Extreme Value Distributions, and he got a Nobel Prize for that singularly unenlightening result. Practicing economists use logit analysis because, y’know, it usually works. Logit and probit analysis give almost the same results with entirely different underlying utility microfoundations, and it’s unclear why anybody should care.

        What practicing economists do all the time, of course, is estimate and use, market demand curves, downward sloping things that allow us to make all sorts of predictions and calculations. it is only slightly embarassing to economists that individual utility theory doesn’t actually constrain market demand curves in any meaningful way, a result associated, among others, with Afriat and Hal Varian (of Google.) It is only slightly embarassing because theorems about the pathological cases don’t really concern us, and besides, we are trying to explain something practical, and downward sloping demand curves make more sense about 99 percent of the time.

        So I agree, Andrew, that utility foundations are often a waste of time. Fortunately, they aren’t needed most of the time, except by Journal Editors. That said, I don’t think economists care, for the most part, whether they are right or wrong as an actual desciption of neurology…. except for neuroeconomists, of course.

        • Andrew says:


          I agree. I think the microfoundations are often added in at the end, but the rule is that they’re supposed to be treated as foundational. The result seems to be an unfortunate undervaluing of the “accounting” or logistic-regression world of practical economics, and an overvaluing of the twists and turns needed to trace a path from practicality to microfoundations.

      • Jason Eisner says:

        Andrew: I’m still confused! I don’t think Woolley is trying to model behavior. So why are you talking about “modeling human cognition and behavior”? Who are you criticizing when you say earlier, “I’d rather model the behavior directly than through a ‘utility function'”?

        Do you think that Woolley is employing utility as a quantity that individual economic actors are trying to maximize? That would indeed be a theory of behavior. However, I think she is just talking about the quantity that social planners ought to optimize.[*]

        In other words, she’s not talking about how people will behave, only how happy they would be under a given arrangement. Maybe you are saying that social planners should try to optimize for people’s behavior rather than for their feelings?


        [*]Once we agree on this objective, we could try to identify a policy that induce people to behave in a way that increases the objective. This would indeed require a theory of behavior — but not necessarily a utility-based theory.

        • Andrew says:


          I’m objecting to Woolley’s statement that “the reason we care about inequality is that it reduces the happiness achievable from a given amount of income.” I think there are a lot of reasons to care about inequality, irrespective of the marginal happiness associated with someone’s 100,000th dollar.

          Similarly, I disagreed with Mankiw’s implicit claim that the state only has a right to tax things that are “unjustly wrestled from someone else.”

          In both cases, I find the objectionable statements to be symptomatic of a general view among economists that it’s not enough to simply do economics—to study dollars-and-cents issues such as economic inequality and tax rates—and that such studies must be placed in a foundational context of utility and justice. I think this unnecessary belief in these foundations causes confusion (as in Woolley’s statement above).

          • Jason Eisner says:

            Thanks, Andrew. Leaving aside what was meant by these particular quotes, I’m trying to understand the more general complaint in your last paragraph:

            I’d think that economists (just like statisticians) ought to be thoughtful about basic issues of what they’re doing and why. You might want to say that they’re not being thoughtful enough, or propose entirely different foundations, but it seems odd to say that they should know their place and “simply do economics” with no guiding philosophy.

            Suppose economists stick to “simply” studying the data as you suggest. Following Woolley’s question, how do you think they should quantify inequality when they study it? Are they allowed to summarize the wealth distribution as a single number? If so, should it be the Gini index or some other number? On what grounds should they choose?

            You might answer: Depends on what they’re studying about the data. But what constitutes an interesting or useful question about the data? What ought economists to tell policymakers about inequality, even if the economists themselves abstain from policy recommendations and notions of justice?

          • Andrew says:


            I’m happy with economists being thoughtful and I’m happy with the use of utility theory as a way to understand some decision making. But I’m not happy with the idea that utility-based microfoundations are necessary for serious social science work, nor do I think it makes much sense for economists (or others) to take their folk psychology models so seriously.

      • Carsten Bern says:

        I am amazed that there aren’t more comments picking up on Gelmans “bag of tricks”. The “argument” he is making by 1) attempting to critizing utility theory, and then 2) suddenly call it folk psychology is fascinating. There is hardly any link inbetween, i.e. backing for these statements. He does say a little bit more, e.g. stuff like crude and outdated, but also somehow acknowledges that it can work and is better than other stuff – but making that analogy makes sure that it is completely dismissed.
        My experience is that people who need to make such flawed analogies, are most often people who don’t have more subtle and detailed arguments to rely on.

  15. […] a piece by andrew gelmanon economists (see quote below as well). andrew makes a very good point, and one that’s […]

  16. […] Causal Inference, and Social Science Skip to contentHomeBooksBlogrollSponsorsAuthorsFeed« Economists don’t think like accountants—but maybe they should Ethnicity and Population Structure in Personal Naming Networks Posted by Andrew on 26 […]

  17. when it comes to taxation, there are some more direct principles at hand, such as the need to fund the government and ability to pay.

    Both of those concepts are slippery. Many would prefer a smaller government, so “need to fund” is moveable. And “ability to pay” is one of the most malleable concepts of all time.

    Fun fact on ability to pay (also relevant to Quiggin’s linked post). My wife works as a mortgage servicer. Her company tries to figure out when people receive their paycheck and get them to set up bill pay on or right after that date. Delinquency is much rarer from the same person, with the same income, if their mental accounting is done on a smaller budget because the mortgage payment is “already gone”.

    Back to taxes, didn’t Peter the Great institute a soul tax on peasants who weren’t, essentially, able to pay? Anyone can argue that they’re not able to pay. Surely you’ve heard a rich person explain why they’re poor. But in the USA there’s 99.999% a possible reduction in consumption that would allow someone to pay more tax.

  18. […] because philosophy has been based on “failed models of the brain.”This reminds me of my recent remarks on the use of crude folk-psychology models as microfoundations for social sciences.The article also […]

  19. mclaren says:

    Your physics analogy proves excellent, but in fact better than you realize.

    “To return to the physics analogy, suppose someone said (correctly) that to understand solid structures, you need to understand electrons.”

    The claim that in order to understand solid structure we need to understand electrons is false. The science of mechanics became well-developed from the 17th through the 19th centuries without any knowledge of electrons. Indeed, continuum mechanics, gives superbly accurate results without any knowledge of the existence of smaller particles: statistical mechanics as developed by Boltzmann and Maxwell, proves entirely superfluous to an accurate understanding and meticulously precise prediction of the behavior of solid materials.

    Such bulk quantities as specific density, mass, tensile strength, and resultants like Young’s Modulus, prove entirely sufficient to do all the predictions that need be made about any material structure. Given the essential bulk properties, architects can accurately predict exactly how an arbitrary structure will behave under loads, tensions, tortions, and other kinds of shear and compressional and tensional and torsional forces, without needing to know that electrons even exist.

    The issue at the heart of this point boils down to the fact that in physics, entirely different laws often govern different levels of organization. Electrons operate according to wildly different laws than bulk materials in the macroscopic world. For example, electrons can diffract through solid matter (Bragg diffraction) because electrons also behave as waves. This never ever ever occurs in the macroscopic world. It is impossible to throw a baseball at a wall and get it to diffract through the wall. But in the microscopic world, this happens all the time.

    Likewise, electrons can tunnel through potential barriers that they don’t have the energy to penetrate. Once again, this never happens in the macroscopic world. You never ever see a baseball sitting on the ground, then jumping up to sail through a window — but electrons do this sort of thing all the time. In fact, if quantum tunneling didn’t happen, the CMOS junctions in your computer wouldn’t work.

    Economists appear to be too ignorant or too incompetent or too self-deluded to recognize that entirely different laws often govern different levels of organization in a system. It seems highly likely that this is true in economics. If so, attempting to use marginal utility at the micro-level to explain crowd behavior at the macro-level is not merely silly or superfluous, but clearly and foolishly and obviously wrong, and a gross mistake.

  20. David B. Benson says:

    My economics class, from Samuelson’s book, was about 50 years ago and I’m trying to ‘brush up’ on the foundations of taxation policy. (If I studied that then I’ve forgotten it all.) In finding various papers starting with a web serch on ‘economic utility tax’ I’ve so far read (after warm ups in Wikipedia)
    On the Edge: Declining Marginal Utility
    and Tax Policy
    Sarah B. Lawsky
    Minnesota Law Review, 2011.

    Optimal Taxation in Theory and Practice
    N. Gregory Mankiw, Matthew Weinzierl, and Danny Yagan

    with a view to offering some utility functions of income which (so far) don’t seem to be popular with economists, to wit the cumulative probabity functions of stable distributions. This is the maximum Util is one and that only obtained at the limit of infinite income. Stating another way, Frederick Pohl’s Midas Plague applies; no single individual can consume more than one Util’s worth of goods in one year.

    For the purposes of describing function from $ to (fractional) Utils the only stable distributions of interest have beta = -1, i.e., support only for positive $. The stable distirbution with alpha = 2 has finite moments; the moments go away as alpha descreases so by the time the Levy distribution is reached at alpha = 1/2 no moments exist.

    Nonetheless, the stable distributions for a fixed alpha have the important property of clsure under convolution (and many other beautiful properties; see James Nolan’s introductory chapter avaialble on the web). Nonetheless, I haven’t found a single economics paper making use of any of the stable distributions. Is that because there just aren’t any (yet) or am I overlooking some important intuition that eleimates consideration of the stable distributions as suitable utility functions?

    I suppose I am primarily interested in how much individuals should (both want and be forced to) contribute to the functioning of their government (via taxation), the Hobbesian alternative being too unpleasant to contemplate.

    • David B. Benson says:

      H.P. Young
      Progressive Taxation and the Equal Sacrifice Principle
      J. Public Econ. 32 (1987) 203–214.

      H. Peyton Young
      Progressive Taxation and Equal Sacrifice
      The American Economic Review Vol. 80, No. 1, Mar. 1990

      appear to provide enough of an answer to my questions.

  21. […] physical inferences in general and also why they show several systematic biases.Cool. I love models of folk science.P.S. Here’s a short conference paper by Hamrick, Battaglia, and Tenenbaum […]

  22. […] economics paper.Not including microfoundations—that’s actually just fine with me! My impression is that microfoundations in economics and political science are typically a version of […]

  23. […] the perspective of an empirical modeler such as myself) is that they add unnecessary layers of implausible psychology on to empirical models that might very well stand up just fine on their own.In particular, I […]

  24. […] of economic models—and things are just as bad in other social sciences. (I’ve criticized economists and political scientists for taking a crude, 80-year-old model of psychology as […]