Rationality
A very good article from John Kay in the FT throws doubt on the notion of rationality as a fixed, unchanging phenomenon that can be pre-defined by economists. Instead, he says that:
Irrationality lies not in failing to conform to some preconceived notion of how we should behave, but in persisting with a course of action that does not work.
This view conflicts with the generally-accepted view among neoclassical economists that rationality is simply what we ‘should do’, given particular motivations and constraints, and that economic outcomes can therefore be modelled with some certainty using appropriate mathematical techniques.
In reality, rationality is moulded by particular social influences and varies according to circumstance. Rationality is what people do in response to real world situations, and it may vary according to time and place.
When I was working in Vanuatu, traders on the southernmost island of Tanna, often preceived to be the least developed of the chain, were often observed not to take price into account when selling meat at the market. An entire cow’s leg sold for the same price as a large steak. The standard economist’s view would be that the supply curve slopes upward from left to right, with price equal to marginal cost. The traders should be selling the leg for much more than the steak. Yet these people weren’t silly or irrational. They had plenty of cattle; the volcanic soil is very fertile and it rains a lot; the roads were bad and it was difficult to get to market; and most people lived subsistence lifestyles where cash was needed only periodically to buy kerosene or to pay school fees. They simply wanted to go to market, get some cash, and head home.
In the last decade or so economic models, and even central tenets of neoclassical economics, that imagined that future outcomes could be predicted with a high degree of certainty based on a fixed view of rationality have repeatedly failed. General equilibrium analysis has been criticised for its reliance on assumptions that are so unrealistic as to make it inapplicable in reality. The over-certainty of the Black-Scholes model helped lead to the near-collapse of Long Term Capital Management in 1998. Soros has thrown doubt on the relevance of some sort of ‘neutral’ notion of rationality in financial markets, showing that reflexive, herd behaviour can lead markets away from equilbrium. On this view, the efficient markets hypothesis is reduced to an academic plaything rather than a predictor of behaviour.
Kay’s definition of irrationality also undermines the current fashion for experimental economics.
If you ask people whether they would rather have $100 now or $110 a week from now, many people will plump for the $100. But if you ask if they would rather have $100 in 52 weeks’ time or $110 in 53 weeks, almost everyone prefers the larger sum. Yet 52 weeks from now, all those who postpone gratification could have had $100 in cash rather than $110 in a week’s time. Faced with just this situation a year ago, many chose the immediate $100…
The experimenter’s trick is to construct an artificial situation in which normally sensible behaviour gives what he thinks is the wrong result. The “mistake” is detected in a meaningless problem designed solely to elicit the “mistake”.
It is not that people are stupid. It is that in this experiment they make a particular response in an unrealistic context which has little correspondance with the actual world.
Good economics involves examining choices made in various real situations, taking into account historical evidence, where people are responding to the kind of genuine influences and constraints that they might encounter in real life. The standard neoclassical view of rationality and the findings of experimental economists at best reveal only a limited amount about rational behaviour in the real world.
Emergent Economics 
Are you saying that only by observing behaviour, taking in to consideration the widest possible range of human responses no matter how weird they are, that it is possible to build realistic economic models? Surely there has to be some sort of rationally negotiated benchmark that smoothes over aberrant reactions like those quoted by Kay.
I’m not sure Kay was suggesting that the reaction to the experiment was abberant. He was simply defining rationality in the negative, saying that irrationality is repeatedly doing something that doesn’t work. His point is that it can be hard to pin rationality down, to pre-define it in an experimental or modelling situation. That rationality is hard to define in the positive doesn’t make modelling impossible; rather, as economists, we should use historical and empirical information from real situations to achieve answers that are as useful as possible. The abstract techniques of the mathematical economist, based on an empty definition of rationality, can sometimes be so unrealistic as to make the results inapplicable to the real world. The contrived experiments of the behavioural economist are often similarly removed from reality.
In my view, the best political economists have used historical and empirical information from real situations rather than solely basing their arguments on the deduction of results from abstract models. Adam Smith famously derived his notion of the division of labour from the workings of the pin factory. Hume had strong views on the importance of history and made detailed notes on historiography. Marx made history central to his political economy. Veblen, using an evolutionary approach, criticised the neoclassicists for aping their physicist contemporaries. Keynes worked as much on policy as on economic theory, developing the idea of effective demand partly in response to observation of the great depression. Keynes, incidentally, viewed Jan Tinbergen’s econometrics as ‘a mess of unintelligable scribblings’. Hirschman opposed the exclusive use of mathematical models, changing his views according to what he observed in different situations.
Finally, few would say that models should be entirely realistic. In ordinary everyday reasoning we all regularly use unrealistic assumptions to draw out a feature of behaviour to which we want to draw attention. The problems start when the assumptions of a model become so unrealistic as to make the answers irrelevant, when the focus is on complexity of technique rather than usefulness, and when apparent ‘laws of behaviour’ are deduced on the basis of these techniques. The real world is complex, messy and changeable; and answers can vary according to context.