Criminals are just like the rest of us. This is the finding of a fascinating new paper.

Researchers at the Norwegian School of Economics got inmates of a medium-security prison to play a dictator game, in which they were given 1000NKr (about ''175) and the option of giving 0, 20, 40, 60, 80 or 100% to a randomly-selected partner. On average, prisoners gave away 36.2%. This is statistically indistinguishable from the 34% given by non-prisoners of the same age and gender.

This similarity is not driven by honour among thieves. When the partner was a non-prisoner, prisoners gave an average of 34.2% whilst non-prisoners sent an average of 32.2%.

Much the same is true when subjects were asked to play a trust game. In this, the experimenter trebled the money the giver handed over to the recipient, and the recipient then chose to return some of the cash to the giver. Again, there were no differences in the behaviour of prisoners and non-prisoners, whether they were givers or recipients.

The message here is stark. Prisoners have almost identical pro-social preferences to the general population. Criminals are not unusually anti-social. There are three possible interpretations here:

1. This corroborates the economic view of crime, as inspired by Gary Becker (pdf). Criminals differ from the rest of us not by being anti-social or 'feral' but by facing different incentives. On this view, criminals are criminals because - for them - the expected benefit of crime exceeds the expected cost.

2. Maybe preferences are not the issue here. Some people might be criminals because they lack self-control, and can't go straight even though they'd like to. (The difference between this and point (1) might be slender for practical purposes, if criminals are easily tempted).

3. Social preferences are not consistent across individuals, but rather differ from context to context. Some people might be averagely pro-social in front of a computer screen and an academic, but not so when they are on the street.

Whatever the explanation, the fact seems to be that differences in behaviour cannot be easily explained by differences in observable preferences. It might be, then, that our attitudes to crime are contain an element of the fundamental attribution error - we over-rate the role of individual agency, and under-rate other forces.