Peter Cuthbertson: Letwin's scepticism of "nudge" is entirely justified

I am pleased that Oliver Letwin has cheerfully admitted that the government's "nudge unit" may well have no impact. Whether or not it is right to have a unit in government to investigate the potential of behavioural economics, it is certainly right that ministers feel free to view the field with appropriate scepticism.

I find many of the claims of behavioural economics fascinating. How exciting, for example, to read that when you divide people in a laboratory into "employers" and "workers", these "employers" usually turn out to be willing to offer more of a salary than they really need to, and their "workers' in turn are more productive when paid above the odds. If there weren't already a field of inquiry into when and how people behave irrationally in economic matters, someone would have to invent it.

But I am very uneasy about reading too much into these findings. As Letwin implies, the field remains in its infancy - and this has its consequences.

Behavioural economics certainly appears better at making audacious claims than providing the rigorous evidence to back it up. The behavioural economists I read almost all seem to share the capacity for leaping from very simple laboratory experiments - in which people may well be more interested in impressing the experimenter than earning the trifling rewards on offer - to dramatic sweeping statements about human behaviour.

This has already led to telling embarrassments. Two academics, John List and Uri Gneezy, decided to repeat the experiment above that divided laboratory participants into 'workers' and 'employers'. But they did it for real, by hiring people to do work such as data entry and door-to-door charity collections. They paid some of them the advertised wage and others an unexpectedly high wage. Did the latter group work harder out of gratitude? Yes. But only for the first morning - or less. The laboratory experiments, and the conclusions initially drawn from them about the surprising benefits of paying people above the odds, had missed entirely how temporary this effect is. Employers who took seriously the initial laboratory experiments must be kicking themselves.

Then there is the problem of experiments that seem to contradict each other. Dan Ariely, a great writer and populariser of behavioural economics, conducted a now famous experiment with American chocolates. Hershy's Kisses were initially offered for one cent and Lindt truffle's were offered for 15 cents each. Although more expensive, the high-quality truffles were discounted by so much that unsurprisingly 73% opted to buy them rather than the Kisses. But when he offered the truffles for 14 cents and the Kisses for free, 69% now opted for Kisses. Conclusion: people really like getting things for free - irrationally so. Interesting.

But in his latest book, Ariely notes how much rats, given the choice between pressing a lever for their food and being given it for nothing, soon come to prefer the former. Conclusion: people* can be irrationally averse to getting stuff for free, and at least to some extent they want to work for what they get. Again, interesting - but which is it? Are people irrationally keen on "free" or irrationally averse to it? Or should we perhaps take all these experiments with a pinch of salt?

Finally, I get the sense that the best criticism behavioural economists make of conventional economists actually applies far better to them. Behavioural economists have a fair point when they say that we should be wary of simply assuming a great knowledge of how humans behave. But in so many conventional economic models, rational behaviour - people will be more likely to buy the same good if it is cheaper and less likely if it is dearer and so on - is a minor simplifying assumption.

Conventional economics, then, is very modest about the role it is willing to give in its models to an understanding of human behaviour. By contrast, behavioural economics by definition gives a huge role to irrationality and to the odd ways in which it makes people act. At first glance, behavioural economics seems to be the field that merely says "People behave in odder ways than economic models can account for - we must be cautious about making too many predictions based on our understanding of human behaviour". But blink for just a moment and it is saying: "Thanks to our experiments, we understand exactly the ways in which people behave irrationally and we can make accurate predictions based on this understanding!".

Behavioural economics will continue to attract fans. There is something enormously appealing about any field that appears to tell us we can get people to do all kinds of things they otherwise wouldn't without them minding, or that paying employees more is win-win because then they work so much harder, or that cutting bonuses for bankers won't reduce their performance. But I am reassured if some policy-makers are starting to wonder if this might all be a bit too good to be true.

* He argues that these results apply to humans, not just rats, by citing experiments that appear to prove only that people prefer work that is meaningful and recognised - which seems to me to be about as far as one can get from economically irrational behaviour