My articles sometimes require a bit of reaching for the cold towel round the head. This one will push that envelope a bit - it's almost into Graeme's league of technicality, I'm afraid. You have been warned...
As Sam notes, Daniel Finkelstein has offered a list of "The five sexiest ideas in politics". One was as follows:
Prospect Theory: Thirty five years ago, innocent A level economics students would ask their teachers "but what if the consumer isn't rational?". It would patiently be explained to them that rationality was a modelling assumption, that could very easily be relaxed. It turns out that relaxing the assumption has produced the most interesting work in economics over the last three decades, winning a Nobel Prize for Daniel Kahneman and Amos Tversky. Prospect theory considers how people actually make decisions. It looks at, for instance, the ways people weigh up risks. It leads to a richer understanding of the way consumers respond to the design of policy.
In my office, when (as we increasingly do) we encounter debates about how to account for consumer rationality when considering the basis for and impact of regulatory changes, there is an important divide of opinion:
- One view (advocated by some of the smartest people in my team, and by most of the young ones) is that Rationality is an assumption of much of modern economics, somewhat akin to the assumption of perfectly elastic collisions in Newtonian gas theory, or of frictionless movement in classical mechanics. On this (Baconian) view, it is for experimental evidence to tell us whether people are in fact rational, and then if experiments suggest that people are not perfectly rational but operate in some other way, then we should incorporate their actual behaviour into our models. I describe this research tradition as that of behavioural economics. Finkelstein calls it "prospect theory" (or perhaps he focuses on the particular branch of it called "prospect theory").
- The other view (of which I am the key proponent, and to which I have converted/misled most of the main high-powered academics in my team) I term Kantian. According to this Kantian approach, Rationality is not an assumption. It is an axiom (or set of axioms), and it is an axiom that conditions both how we construct well-formed theories and also how we interpret data. According to this view, economics and game theory just are the application of the Rationality axiom to modelling and experimental evidence. There is no possibility of an experiment to prove that agents are not Rational, any more than of an experiment to prove that 2 + 2 = 5. Of course, just as we could (as some people - the "intuitionists" - have done) form a new subject a bit like mathematics in which, say, the law of excluded middle does not apply (i.e. in which it is not so that every proposition is either true or not true), likewise we could form a new subject a bit like economics or game theory in which there was some other axiom than Rationality. There is an ongoing research programme of attempting to come up with such alternative axioms. I call this the bounded rationality research tradition.
I like to illustrate why I think the second approach is far superior with the following cautionary tale. In the late nineteenth century there was active research interest into an apparently strange anomaly. It appeared that, under certain circumstances, when the price of bread rose people bought more of it instead of less. This research led people to distinguish between "substitution effects" whereby people switch from one good to another, and "income effects" whereby how much people purchase is influenced by how affluent they are. This distinction led to much of modern consumer theory. Now, I submit that if the behavioural economics advocates had been around then they would have said (as they say of many analogous hard-to-explain puzzles today): "Well, of course, if people were rational then they would buy less of goods as the price rose. But they aren't always rational, as can be illustrated by the example of bread!"
In this way, the behavioural economics research tradition explains everything - like the grand conspiracy theory of economics. But, of course, since everything can be explained so simply and so negatively, the research programme lacks precisely those constraints that make theorising fruitful. It is hard to explain why certain behaviours are, in fact, to be interpreted as Rational. But we struggle until we come up with an explanation, and our results are exciting and fruitful and have changed the world. We should not surrender our Rationality axioms lightly - we should only do so if someone shows that some other set of axioms leads to an approach even more fruitful and informative and useful than economics and game theory have been. Perhaps one day soon there will be some new such axioms - I have worked on some myself. But the bar to overcome is set high (for example: wisely, no-one liked any of my alternative axioms).
Finally, another thing I don't like about the behavioural economics tradition. If you have the stomach, you can read here and here my argument that non-mathematical intuitive thinking about what happens when agents are irrational will often lead us astray in our policy recommendations - the thing that seems obvious to do turns out to be worse, and the situations in which we expect to have the most problems if people are irrational turn out, actually, to be the situations least affected. Economists are trained to think in rational paradigms, which typically guide us well when forming policy, with models only being crucial (a) to help form new intuitions; and (b) to fine-tune policy decisions at the end. But if agents are not rational, intuitive policy formation will very often fail, and only experts in bounded rationality modelling will be competent to make policy decisions. That does not seem a happy outcome to me.
