Wednesday 06 Oct 2021 | 11:26 | SYDNEY
Wednesday 06 Oct 2021 | 11:26 | SYDNEY

Reader riposte: Behavioural economics

17 August 2011 09:13

Richard Green responds to Sam's posts:

Behavioural Economics is rather less radical or revolutionary than is usually believed, something not helped by some of overwrought claims by some of its researchers. The important thing is to understand the real basis for the assumption of rationality. Here rationality means the maximisation of expected returns. No economist (or human being) of the most ordinary observational skills would say that a given person was rational, but it was assumed that departures from this irrationality were randomly distributed on either side.

A model that assumed rationality would not predict a single person's behaviour, but in a whole market, the random irrationalities of participants would cancel out. In aggregate it would look the same as if everyone was rational. Agents weren't necessarily rational, but the average representative agent was.

Where behavioural economics changes things is that it discovered that departures from rationality aren't entirely random, but there are systematic diversions discovered through empirical work and experiments. It's still not possible to predict a given person, but the representative agent could be expected to irrational in a predictable way — for instance, risk aversion with low stakes and risk seeking with high stakes. These diversions could even be assumed differently in different circumstances if social norms were not the same.

In this sense behavioural economics is an important progress, but it is utterly and completely rooted in orthodox neoclassical economics. It merely tweaks the assumptions whilst using the same tools.

By contrast there has been a long history — through early economists/sociologists and early institutionalists such as Veblen, game theorists and up to complexity scientists at the Santa Fe institute — to dispense with representative agents entirely. Instead, behaviour would be determined by agents' responses to other agents (whose responses were in turn responses to responses etc.). Neoclassicism, including behavioural economics, would assume a stable pattern of behaviour and then use a model to predict a response to variables, but the other approach would even attempt to predict how people would change the way they behaved. Instead of assuming a set of social norms, we could try and predict how the social norms developed and changed.

As yet this approach has proved fruitless and certainly has come nowhere near providing a guide to policy. Should it bear fruit it would be a genuine revolution, across every social science. Until then incremental modifications to existing tools will have to suffice.