"Ugly criminals"?

"Being very attractive reduces a young adult’s propensity for criminal activity and being unattractive increases it. Being very attractive is also positively associated with wages and with adult vocabulary test scores, which implies that beauty may have an impact on human capital formation. The results suggest that a labour market penalty provides a direct incentive for unattractive individuals toward criminal activity. The level of beauty in high school is associated with criminal propensity seven to eight years later, which seems to be due to the impact of beauty in high school on human capital formation, although this avenue seems to be effective for females only" (Mocan and Tekin 2010).

I want to use this post to discuss a recent paper that appeared in The Review of Economics and Statistics (full version here if you can access it, otherwise the abstract above will have to do). Initially because I though the idea was so mad it was worth a few comments, but having re-read the paper more now because I think it throws up some very interesting points beyond its specific empirical content.

The article is based on analysis of the US National Longitudinal Study of Adolescent Health (n=15,338, which is relevant). It uses respondent’s ‘beauty’ (controlling for a very wide range of other variables) to predict self-reported criminal behaviour. It’s worth noting that the mechanisms the paper suggests for attractiveness affecting criminal activity are all social – it essentially claims that the less attractive are more likely to commit crime (and the more attractive less likely) because they are treated differently by those around them, in terms of access to the jobs market, treatment at school, and so forth (hence for example the lower vocabulary scores mentioned above - it's not that 'ugly' people are thick, its that they get treated differently at school and this inhibits their learning process).

Methods and stats

First, let’s get the methods/statistics out of the way. With two provisos they are actually quite convincing. It really does seem that the physical attractiveness of females (although to a lesser extent males – see below) is associated with self-reported criminal activity, in the ways suggested in the abstract above.

The two provisos are quite important, though. Firstly, how is physical attractiveness measured? It turns out by the interviewers in the survey. There is something immediately almost repellent about this, but the authors present some evidence that such ratings are fairly robust, and I don’t want to get bogged down in this point, so let’s just take it as read for now that they do have a somewhat accurate measure of the survey respondent’s physical attractiveness.

The second proviso is even more important. Namely, that the effect sizes they identify are very small. For example, “Being a very attractive female reduces the propensity to damage property by 1.1 percentage points” (controlling for personal and family characteristics). This effect is statistically significant, because of the very large sample size. But is it substantively meaningful? Does it make any sense to concentrate on such a small effect? The authors clearly think so, and up to a point I agree with them – the social world is a complicated place, and the unique association between any one factor (e.g. physical attractiveness) and another (e.g. propensity to vandalise things) is almost bound to be small. If there is a statistically significant association, robust to all the control variables that seem relevant, it probably does mean something.

But on the other hand, there must come a point where an association, although significant, is so small that it really can’t have much meaning – certainly if we want to change the world in some way. A one percentage point difference is interesting in the context of this paper, but would be so if, for example, some sort of policy intervention was being planned? I think we as social scientists do not pay enough attention to the pay off between statistical significance and effect size, and this paper butts right up against that point.

One other point on the specific empirical content – as noted the effect of ‘attractiveness’ on crime appear much more consistent for women than for men. I’m not too sure what to make of this, and neither I think are the authors. While it might be consistent with their theoretical framework (since it appears physical attractiveness as rated by others varies by more among women than among men, it should therefore have a bigger effect), it’s unclear how this fits with ‘explaining crime’, given the well known gender disproportionality in criminal activity.

Three more substantive points

There are (at least) three more substantive points I’d like to make about the paper, again because these seem to me to be relevant far beyond its specific content (there are more but these particularly strike me).

First is the rational choice theory model of criminal behaviour used in the paper. It essentially takes as a given the idea that people commit crimes when there is motive (reward) to do so, there is an opportunity, and the rewards outweigh the potential costs. Yet much recent work on why people commit crime would stress the importance of morals and normative values – to the extent that most people do not commit crime because they do not even see it as an option, even in situations where they would gain from doing so.

Now, this may not make much difference in terms of this specific paper, because we would expect normative orientations toward the law to be evenly distributed across ‘attractive’ and ‘unattractive’ people and thus not affect the results. But taking such a narrowly instrumentalist approach toward decisions to commit crime does I think in the wider context run the risk of over-simplifying what seems certain to be a very complex process.

More serious for the specific claims of the paper is its very one-sided view of what crime is. The types of crimes covered by the analysis are ‘damage’ (vandalism), burglary, robbery, theft, assault and selling drugs; essentially, ‘crimes of the poor’. What about white collar crime? Tax evasion? Fraud? Corporate crime? At the very least the omission of a wider range of criminal behaviour limits the claims that can be made by the paper. But worse, it may be that being physical attractive, and thus preferred in the job market to others, opens up greater opportunity to engage in white collar crime (by increasing both opportunity and level of potential gain). So it could be in these cases physical attractiveness is linked to higher rates of self-reported offending. Of course, such a impoverished of crime is not unique to this paper – it seems to bedevil many economic and similar approaches to criminological matters.

But to finish on a more positive note I think this is a very interesting paper not because of the specific claims it makes but because of the way it tries to integrate ‘innate’ characteristic (physical attractiveness), personal proclivities and propensities (to commit crime), and the social world which influences and interacts with them. That is, it takes seriously the idea that people are individuals (and individual), and make personal choices about how and when to act – but their individuality and their choices (and possibility of choices) are structured and limited by the actions of others around them and the social and economic structures in which all are embedded. It seems to me that this kind of empirically robust ‘cross-level’ analysis is exactly the kind of thing we should be trying do more of in British criminology.