It’s the sort of research finding that seems perfectly designed for one of those free newspapers you can pick up in big cities: ‘women shy away from competitive workplaces’. And it’s partly true – a great new real-life experiment does show large differences in how men and women treat competition at work, and this may well contribute to gender inequalities.
But as ever, the stuff of real interest is only revealed when we dig deeper – into the details of the research, and into the limits of experiments.
Like most of my favourite naturalistic experiments at the moment, this one is reported in an NBER working paper – this time, conducted by Flory, Leibrandt and List. The design is simple: they placed job ads on real-life internet job-boards to get an initial sample of interested jobseekers. They then sent details back to the jobseekers, randomly varying the details of the job. For some they said they would pay a fixed $15/hour; for others they said they would pay $12 or $18, depending on whether they were more productive than a single co-worker. It’s this performance-related pay that can be interpreted as being about ‘competition’.
The results are striking. With the fixed wage, men were about 11% more likely to actually apply for the job, conditional on having shown an original interest. But when the job forces you to compete against your co-worker, men were about 18% more likely to apply – that is, the competition “widens the gender gap in propensity to apply by almost 70%.” You can see this in the figure below – T1 is the main original condition, and T4 is the condition with the large bonus that depends on competition.
But the figure also mentions a whole range of other ‘treatments’ (pay regimes) – and this is where the experiment starts to get really interesting. If the bonus for beating your co-worker becomes much weaker – 22% of the wage, rather than 50% – then this has almost no effect on whether men or women are more likely to apply. If the bonus is dependent on team performance rather than beating a co-worker, then the gender differences also disappear. It’s only when competition becomes particularly acute that women are deterred.
The researchers also used two different job ads – a ‘male-oriented’ job ad (on sports), and a ‘female-oriented’ job ad (on general admin; they justify these choices very reasonably in the paper!). And surprisingly the effects of competition are very different for these two jobs. For the female job, there is only a slight gender difference to begin with, and this is entirely unaffected by the different pay treatments. For the male job, though, there are much larger differences between the genders – and women are much less likely to apply to the most competitive pay structure.
So what does it all mean?
It’s not unreasonable to use this research to say women are deterred by competition at work – but it’s slightly misleading. In actual fact, both men and women are deterred by competition at work – it’s just that women are more deterred than men. For women, they are particularly deterred if there are other, equally good jobs around. And interestingly for men, its mainly those of ‘low abilities’ who are attracted by the competition-based pay structure. (Make of this what you will…).
What interests me more, though, is what exactly experiments like this can tell us. The paper’s authors rightly say that these results are much more revealing that the unreal lab experiments that have been common in this area. And as my repeated posting about natural experiments shows, I absolutely love this kind of research – not only are the results unusually revealing, but it involves researchers manipulating unsuspecting members of the public, which I’d like to do myself in future. (Incidentally, while I think about ethics – the study authors did actually hire people in this study!).
But there are limits to what experiments can tell us about wider structural inequalities and the societies in which they’re embedded. In a linked post in the new year, I’ll come back to this and explain why we have to go beyond randomised experiments – and why the associated rhetoric of ‘evidence-based policy’ can fatally undermine systemic attempts to reduce inequalities, which are likely to be the only policies that will actually have any effect.