More rationality, and a million models

Here is an article on a conversation with Raquel Fernández that I found very interesting. The whole thing is worth reading, but I will quote at length this passage on rationality:

There is a beauty to the models in and of themselves. You assume, for example, that people are rational. I don’t think any really good economist thinks that people are perfectly rational, but, on the other hand, if you want to model people as not rational, all of a sudden it’s not clear what choice you should make. There are a million and one ways to be non-rational; there’s only one way to be rational within the confines of a model. Rationality means one thing: you’re maximizing your welfare subject to constraints. Now, if you say people don’t always maximize, and they’re beset by this and that, then all of a sudden you can have a million models. And that’s a little bit unsatisfactory too.

This is pretty close to my own view. My extra take is that any “irrational” behavior – and so anything that the “million models” generate – can be rationalized, either by revising what we assume that the decision-maker cares about, or by adding constraints on the decision-maker’s ability to choose: the question of rationality is a red herring.

Daniel Kahneman and the rational choice model

Daniel Kahneman has a new book, “Thinking Fast and Slow“, that is prompting a lot of excellent articles about his work. For example, Vanity Fair has a nice article by Michael Lewis. Back when I was an undergraduate, Kahneman and Amos Tversky’s Prospect Theory paper was one of the first academic papers I read that motivated me to become an economist.

The article quotes Kahneman as being astonished to learn early in his career that:

The agent of economic theory is rational, selfish, and his tastes do not change.

I just want to point out that this model of decision-makers in economic models is more innocuous than it seems. It means that when we put a decision-maker in an economic model, we assume that she

  •  has things that she cares about, and
  • will do the best she can to achieve the things she cares about. 

This is what we mean by rationality. The great thing about this assumption is that it is completely flexible – it can accommodate any preferences at all.

For example, the decision-maker might care about the wellbeing of her neighbor, and so give up some of her own material wealth to help her neighbor be better off. This is both rational and selfish!

For any decision we observe, some preferences will rationalize it. As economic theorists, the crucial assumption is therefore not rationality, but on what it is that the decision-maker cares about and what constraints there are on her making a good decision.

The great contribution of Kahneman and Tversky was to present evidence on the psychological constraints on our decision-maker’s ability to achieve the things she cares about. Making decisions is hard: even if I know what I want to achieve, “doing the best I can” is subject to how I process the options I have to choose from.

Rationality is not supposed to be a realistic model of the process by which people actually make decisions. Rather it attempts to capture the outcomes of decision-making in a plausible way, so that we can try bit by bit to analyze economic settings without simply saying “people are completely unpredictable, so let’s give up”. Kahneman and Tversky helped to show how to write models of rational choice that better reflect the decisions we observe people make.

Happiness and behavioral stuff again

Via Arts & Letters Daily (again) comes a wonderful (and long) article by Alan Wolfe for the New Republic: “Hedonic Man, The new economics and the pursuit of happiness.” There’s far, far too much for me to discuss in suitable depth here, but, beyond recommending the article, let me pick out a couple of choice bits.

I wholeheartedly agree with Wolfe when he says:

“The social sciences are not just empirical; they are normative, too. It was precisely the insistent normative preference for market-based social arrangements that turned me against Chicago School economics. Governmental regulation is always sub-optimal, they inevitably maintained. Individual freedom is worth more than social equality. If market logic works for firms, surely it can work for recruiting an army, fighting poverty, or providing kidneys. Non-Chicago economists were subtler about these matters, and at times questioned the reliance on markets; but for the many sons and daughters of Milton Friedman, we are hard-wired to be rational choosers, and any efforts we make to direct the course of our actions collectively are bound to fail. Myself, I do not believe that any of these propositions bring us closer to a good society. Other people feel differently. Democracy requires that we argue out our differences. But democratic debate is not well served by pretending that the empirical findings of a single controversial approach in a single academic discipline contain definitive answers to these questions.”

The presentation of economics as a discipline of study certainly does make it seem like we’ve magically cracked the nut of what is the ‘best’ way to organize a society, and, as I’ve argued before, that’s not just dangerous, it’s a misrepresentation of what good economic science is capable of. One reason why it’s a misrepresentation is the measurement issue at work again: to say ‘best’, we need a metric, and to do that we need to ask what people want, to speculate on their motivations and desires. Of course, as misrepresentations go, it’s a tempting one, because from day one of a Principles of Economics course it’s made again and again and again.

Wolfe’s article is built around the review of two books: Predictably Irrational by Dan Ariely (which I’ve talked about before, much to the author’s chagrin, so I won’t return to it now) and Happiness: A Revolution in Economics by Bruno Frey. ‘Happiness’, as discussed by Wolfe, is concerned with exactly that big question in normative social science, which is: what exactly constitutes a ‘good’ outcome? It’s exactly that question that’s the dangerous misunderstanding within economics, the dangerous belief that we know what’s ‘good’.

Wolfe talks at length about Daniel Kahneman and Amos Tversky, the pioneers of what has become ‘behavioral economics’, and I admit that I share Richard Thaler’s reaction, as reported by Wolfe:

“When I read this paper,” [Thaler] wrote of Kahneman and Tversky’s classic article “Judgment Under Uncertainty,” which appeared in 1974, “I could hardly contain myself.”

He talks further about the supposed ‘revolution’ in economics to account for the kind of behavior documented by Kahneman and Tversky and the whole slew of experiments run by economists since:

“One has to wonder why the revolution in economics failed so badly even before it really got off the ground. Neoclassical economics may in some ways be preferable to what the revolutionaries offer, but it remains a vulnerable approach, stuck in unrealistic assumptions about human behavior and all too complacent about the beneficial equilibria established by markets. Nor can one deny the ingeniousness of the early days of economic psychology, especially the inventive puzzles that Kahneman and Tversky devised. If ever a field were ripe for revolution, it is economics. Yet if these two books are any indication, supply and demand, marginal utility, rational choice, and cost-benefit analysis are not going away. At best, economists will tweak their models a bit to account for some of our odder calculations. More likely, they will simply reiterate their belief that we need not examine the internal mechanisms of utility satisfaction because the price someone is willing to pay for something is really all we need to know.”

Again, though, I offer this as the reason why the ‘revolution’ has ‘failed’: we simply can never say what motivates people, whether a person is “rational” or not. ‘Neoclassical economics’ does not restrict the range of assumptions one can make about human behavior. It is therefore completely resilient to any evidence on how people act in a given situation. That’s not a defense of the approach, it’s a fact. It’s distressing, because my prejudice is definitely to agree that economics is ‘complacent’ about the superiority of markets (again, because we ignore the variety of normative metrics of comparison), and that there’s too much arcana.

We need true normative debate. Look again at what Wolfe says about economics: “it remains a vulnerable approach, stuck in unrealistic assumptions about human behavior and all too complacent about the beneficial equilibria established by markets”. It’s a mistake to conflate this problem – the arrogant assumption that we know what’s ‘best’ – with the neoclassical assumptions on human behavior, because, as I’ve argued repeatedly, the method of scientific economics doesn’t actually assume anything about human behavior, as evidenced by the fact that the ‘behavioral revolution’ is comfortably within the confines of ‘neoclassical economics’.

The real revolution would be quieter, and would say something much more familiar: keep your science separate from your opinion.


As someone who laments misperceptions of what economists are and do, the barriers to communication with anti-capitalist groups make me very sad indeed. How did I get there? I was looking for something entirely different when I stopped to read an article by Roy Weintraub talking about neoclassical economics. To someone with my beliefs in what economics is, it’s a bit schizophrenic. This is nice:

“Neoclassical economics is what is called a metatheory. That is, it is a set of implicit rules or understandings for constructing satisfactory economic theories. It is a scientific research program that generates economic theories.”

This is pretty good news: the beast called “neoclassical economics” is merely a box inside which we concoct scientific theories: inside our box, this would lead to that. Weintraub continues to say that the assumptions of neoclassical economics

“include the following:

1. People have rational preferences among outcomes. 2. Individuals maximize utility and firms maximize profits. 3. People act independently on the basis of full and relevant information.”

Of these, 1 is redundant to me because I think rationality is not testable and is therefore irrelevant, especially since it’s probably implied by 2, and 3 is at best outdated (economists these days are very interested in the implications of imperfect or asymmetric information). If I was pressed to define neoclassical economics, I think perhaps the definition I would use is similar to 2. I’d say that neoclassical economics is the branch of economics that models entities (individuals, firms, governments, etc) as if they try to get the outcome they like best from the ones that are available.

I disagree more with the stance of the article when Weibtraub repeatedly invokes “the neoclassical vision”. The connotations of this phrase probably reinforce the misconception that economists think the box in which neoclassical economics works obeys the same rules as the real world. I doubt a physicist thinks that a vacuum is the same as the real world, just as I doubt that any economist thinks that the abstractions of economic modeling are the same as the real world.

It’s true that a positive economist who seeks to explore “what is” should not neglect to examine the differences between abstraction and reality, but again we must ask at what point the value of realism is eroded by its inability to draw any conclusions. I think the real choice we’re faced with is the application of the economic method that says “if this unrealistic simplification, then that” versus a shrug of the shoulders; if it were possible to achieve the ideal “if this, then that”, who would reject it? Should we stop trying because we can’t be perfect?

Perhaps partly because of such confusions, “neoclassical economics”, aside from having a silly name, seems to have become something of a lightning rod for the anti-capitalist set as much as it is for economists with different ideas. Google neoclassical economics and you get – on page one – a page from adbusters (an anti-consumerist publication – Wikipedia entry), and a less histrionic “critique of neoclassical economics” by Herb Thompson.

“Neoclassical economists normally treat economic instability as the effect of exogenous, stochastic factors even though nonlinear economics suggests that what may previously have been considered exogenous, or random, may more likely be endogenous to capitalist social formations.”

I confess I’m not sure what “nonlinear economics” means (the almighty Google was inconclusive): clearly I, too, have been indoctrinated to the neoclassical cabal. However, I actually think that the quotation touches on an interesting idea. Can we figure out if the primacy of money as a measurement of outcomes “caused” the rise of the capitalist method of organizing resources, or if the capitalist method “caused” the rise of the primacy of money?

A difficult one. For example, to take a typical example of an anti-capitalist complaint, do people buy sweatshop goods because they don’t know they’re sweatshop goods or because they care more about cheap goods than where they came from? I think the latter is more consistent with “money primacy leads to capitalism” and the former is more consistent with “capitalism leads to money primacy”, although I’m sure that could be debated.

It is possible to imagine that incorrect normatization of positive economics – by which I mean the mistaken assumption that some measurable positive economic variable is a measure of the quality of an outcome – actually causes problems within the economic system. People will do what they will, but if a policymaker chooses a policy based on the primacy of money as a measure of the quality of an outcome, there’s a real possibility that the system itself is influenced by its measurement.

The Thompson article also includes the following excellent paragraph:

“The ‘rational’ consumer of the mainstream economist is a working assumption that was meant to free economists from dependence on psychology…. The dilemma is that the assumption of rationality as intertemporally optimising is often confused with, and regularly presented as, real, purposive behaviour. In fact, the living consumer in historical time routinely makes decisions in undefined contexts. They muddle through, they adapt, they copy, they try what worked in the past, they gamble, they take uncalculated risks, they engage in costly altruistic activities, and regularly make unpredictable, even unexplainable, decisions.”

First of all, this is crucially wrong: “rationality” is not something that can ever be more than an assumption, unless you think you can test it. Further, assuming rationality does not exclude any of the motivations Thompson talks about. It would be trivial to write down a model of a rational person who “engaged in costly altruistic activities” – I simply have the person care about others and optimize rationally. The assumption that Thompson is really discussing here is the straw man of “rationality equals maximizes money”, which I have previously argued is absolutely not an assumption of any economic theory, neoclassical or otherwise.

Beyond that, this is really back to the same problem that the Weibtraub article was getting at: we’re doing the “if this unrealistic simplification, then that”. There’s a strong push in so-called “behavioral economics” to figure out if there’s a workable way to first make realistic generalizations on how people behave and second to incorporate them into the unrealistic simplification of neoclassical economics. While that goes on, the economist who seeks to defend his method must be clear on what his unrealistic simplification actually is and what it is used for.

As usual, no-one is fit to judge if the anti-capitalist model is “better” than the capitalist status quo, but I greatly hope that we would be able to talk about what each would mean. If somehow I were able to convince adbusters to sit down with me and I asked them what they wanted to do and what they wanted to achieve, what might they reply? I don’t know what they would say, but whatever their answer, I would like to figure out what it would take to achieve their goals, what the consequences of their chosen actions would be, what it would mean for people, not just them or me. I hope they would like to figure that out too. That’s positive economics.

Can you test rationality?

Is it possible to test if people are rational? I think the answer, practically, is a very short no: if a rational person tries to achieve his most preferred outcome of the ones that are available, we can’t distinguish the rationality or irrationality of his choice from his preferences. That is, if I don’t know what you like, I can’t tell if you did something because you liked it or because you’re “irrational”.

Yet rivers of ink have been spilled trying to “prove” or “disprove” models of rational choice. The most famous study of the type is the “Allais paradox”, discussed here. It says that when you pose different choices to people, their responses to pairs of choices are “inconsistent” with each other because the two choices really represented the same cash outcomes.

Whether you look at the question being asked by this type of work “are people rational?”, or “what do people care about?”, it’s pretty clear that any observation cannot answer either of these without knowledge of the other. In “The Methodology of Positive Economics” (pdf) Milton Friedman made the valid, general point that

“If there is one hypothesis that is consistent with the available evidence, there are always an infinite number that are.”

It just so happens that if we interpret some piece of evidence as being consistent with “people are irrational”, one of the “infinite number” towers above all others: “you guessed the preferences wrong”. There’s nothing wrong with trying to figure out how to better model the decisions of people, but claiming to have proved irrationality is nonsensical.