Monday, September 17, 2012

Normative Rationality and the Euvoluntary Spectrum

NYU economics and law professor Mario Rizzo spoke in Arlington last week on the topic of rationality and irrationality. Like him, I am fascinated by paternalism. His project these days is to challenge the new paternalism (unfortunately dubbed by its proponents "libertarian" paternalism), where most of the examples we reference here at EE invoke the old-fashioned variety.

That's us, like a slow ride down the Mississippi, sipping on a cool sasparilla, we're just a little old fashioned here.

Rizzo makes the very interesting point that there's a big difference between what economists think of as "rationality" and the way everybody else uses the word. In the econ profession, we use the Arrow-Debreu criteria, which boils down to complete and non-transitive preferences. Much of what behavioral economics has done is to take the A-D definition of rationality and poke holes in it. These researchers then find that real live people violate assumptions of rationality so are therefore irrational. Irrationality has policy implications, which helps explain why Cass Sunstein can end up as the administrator of the White House's Office of Information and Regulatory Affairs.

Let me give you one of my favorite classroom examples of irrationality-by-economics*. Lady Moffet's beloved nephew calls and asks if she'd like to come over for a nice cup of tea and some scones. Delighted, she readily agrees and sets a date for a fortnight hence. Three days later, her nephew calls back with a slight change to the invitation: instead of just tea and scones, he will have tea, coffee and scones. She's not a coffee drinker, so the addition of the extra beverage is what we'd call an irrelevant alternative. Accordingly, and consistent with strict economic rationality, she confirms the appointment, well aware she has no intention of drinking coffee. Three days after that call, he phones yet again. Now, instead of coffee, tea and scones, he will have coffee, tea, scones and cocaine. She abruptly cancels the plans and slams down the receiver as hard as she is able with her dainty, birdlike hands. What gives? She doesn't consume candy of the nose variety, so why should the blow be any different than the coffee? It's just another irrelevant alternative, right?

Here, according to Mario is where the strict neoclassical definition of rationality falls flat on its face. Humans are contextual creatures and they make decisions as such. The behavioralists have turned the definition of rationality as used by the econ profession into a totemic standard, labeling deviations as irrational. Contrast this with the way most regular folks would look at Lady Moffet's outrage: her decision was entirely rational, since she didn't really care much about the scones or the tea, but instead the pleasure of her nephew's company which may have been tainted by coke-fueled rants about the corrupting influence of Hollywood on the national dialog. When people use the term "rationality", they use a normative definition that comports with "what a level-headed person would do in this situation" as opposed to the positive definition assumed by the econ profession.

The rest of Mario's talk explored this idea in greater detail, but I couldn't shake the suspicion that there's some overlap with what he's doing and the EE project. Both projects take seriously the idea that moral intuition drives attitudes towards permissible exchange. Both projects take aim at paternalism. Most interestingly, both projects worry about deviations from morally acceptable trade. Here at EE, we concern ourselves mostly with exchanges made on unequal footing (BATNA disparity), whereas the Nudge folks fret over regret (I've posted before on regret, and I still think that post has some interesting ideas worth pursuing). The questions that occurred to me following Rizzo's talk were something like:

  • Does euvoluntary exchange presume rationality? Are there euvoluntary exchanges that violate either normative or positive rationality conditions?
  • Does the rationality spectrum map to the euvoluntary spectrum? Is it rational to agree to a coercive exchange, for example?
  • If there is a map of exchange in rational-euvoluntary space, what does this imply for policy, especially once we incorporate some of my musings on the nature of regret?
  • Most importantly from a purely selfish perspective, are these questions interesting enough to become a research project? 
  • And finally, what about Andrea's Question? The new paternalists already have their hands on the levers of policy. How do we excite popular condemnation of a kinder, gentler paternalism? What's a euvoluntarist to do?
I'll try to address some of these questions in a forthcoming post. I'll expand my 2x2 regret matrix a little bit to accommodate normative rationality and see what happens when we vary coercion. Stay tuned!

*I wish I could remember the attribution for this. I think I heard it first from Jerry Brito, but it's been borrowed so much, I'm not sure who originally wrote it. Heck, for all I know, it might have been Munger himself.


  1. I don't understand how the example is "irrational" from an economist's standpoint.

    You are assuming the economist looks like the consumption goods offered at the meeting are the only things that she's making a decision about.

    What about time with the nephew?

    Haven't economists always recognized time use and conditions that introduce compensating differentials as part of a person's choice?

    Clearly the nature of the time with the nephew was devalued by the introduction of drugs.

    This seems like an explanation that is unfortunately too common in arguments about "mainstream" or "neoclassical" arguments: dumbing down what economists say in order to dismiss them. If you had asked me "would an economist consider that a rational response?" without giving your answer, I would have said that of course it's rational - who would argue otherwise?

  2. I'm not the one making the assumptions about consumption goods only. Laboratory violations of (eg) the Allais Paradox are held up as evidence of irrationality. Kahneman and Tversky's work on prospect theory and Dan Ellsberg work on uncertainty was underpinned by seeming violations of IIR. These studies are the foundations of new paternalism. If interpretations (especially popular interpretations) of this early literature were as nuanced as you suggest, new paternalism wouldn't enjoy the popularity it now boasts.

    1. I appreciate that calling these things "irrationality" and "rationality" may not reflect the way those words are used in common conversation. I just don't see what providing a bad application of what economists would say about IIR has to do with claims about the use of heuristics, etc.

      You seem like you're arguing that being irrational by an Arrow-Debreu definition isn't automatically grounds for intervention. I agree with that. I would think most economists would.

      What I don't understand is what a bad application of Arrow-Debreu to a nephew offering cocaine to his grandmother tells us about the invalidity of the libertarian paternalist arguments.

  3. In that case, I encourage you to read what Rizzo has already written on the subject, available here:

    I'm less troubled by the Chinese telephone problem as economics moves from the lab to the legislature and more interested in how to include the intuitions of the median voter into a model of commonly-held values and beliefs.

    You make a good point about communication in economics. What we teach undergrads is very different from what we teach graduate students, which is in turn somewhat different from what we discuss amongst ourselves (though perhaps not too much the latter). There's a lot of normative stuff left unsaid in academic papers, so it's sort of easy to see how miscommunication happens.

    At any rate, the arguments made in Ellsberg's '61 paper are as devoid of context as my example above (available here: To the extent that economists have limited resources that prevent them from addressing nuanced arguments in survey-type undergraduate courses, the level of discussion will perpetually be 30-40 years behind the state of the art.


Do you have suggestions on where we could find more examples of this phenomenon?