Cuts both ways

Evidence from Australia suggests that discrimination towards male job applicants is just as likely to occur in female-dominated industries.

Studies of gender pay and employment differentials typically focus on survey-based data. Yet equilibrium outcomes reflect both productive traits and labour market discrimination.

An alternative approach – rarely implemented when considering gender discrimination – is to conduct audit experiments, sending matched CVs to employers in response to job advertisements. If only the names are changed, such an approach provides an unbiased estimate of the degree of labour market discrimination at the hiring stage. In their London-based field experiment, Riach and Rich (2006) found statistically significant discrimination against men in "mixed” occupations (trainee accountants, 31 per cent female; and computer analyst/programmers, 21 per cent female) and in "female” occupations (such as secretarial, 97 per cent female). They attributed this to gender stereotyping on the part of those making the decisions about who to call back.

New evidence from Australia

To further explore this unexpected finding, we chose for our field experiment four female-dominated occupations, ranging from 65 per cent to 85 per cent female (Booth and Leigh 2010). Where possible, we also obtained information about the gender of the person making the callback decision and whether the job was processed through a recruitment agency.

Our conjecture was that, in female-dominated occupations, we would find a pro-female bias that increases with the proportion of women in the occupation. In other words, our expectation was that the more skewed an occupation’s gender ratio, the greater the extent of gender stereotyping.

Why did we hold this conjecture? It is well documented that women are discriminated against in all sorts of ways in the labour market, and especially so in hiring for male-dominated jobs (Booth, 2009, provides a recent survey of various studies that have been put forward to explain these differences). An obvious question is therefore whether, given a pro-male bias in male-dominated occupations, this is symmetric across gender. In other words, will there be a pro-female bias in female-dominated occupations, ceteris paribus? A related rationale for our conjecture is that there may be gender stereotyping. If certain jobs are perceived as more appropriate for women, male applicants may be (implicitly or explicitly) evaluated less favourably because they do not fit society’s prescriptions about what is appropriate for men.

In 2007 we applied for several thousand jobs in the three largest Australian cities – Brisbane, Melbourne and Sydney. This was a relatively tight period for the labour market. In selecting appropriate occupations, we focused on female-dominated jobs that did not require post-school qualifications and had a relatively straightforward application process. These occupations were: waiting on tables, data entry, customer service, and sales. We created four sets of identical CVs, some with women’s names at the top and others with men’s names, and submitted them through a major job-finding website. (In a companion paper, Booth, Leigh and Varganova (2009), we report hiring discrimination across different ethnic minority groups in Australia.)

Extent of the pro-female bias

The responses indicated that the typical female applicant received a callback 32 per cent of the time, while the typical male candidate received a callback 25 per cent of the time. Consequently, an average male candidate would have had to submit 28 per cent more applications in order to receive the same number of callbacks.

But these averages disguise the fact that the gender differences in callback rates varied depending on the occupation. For table service and data entry jobs – which across the Australian economy are 80 per cent and 85 per cent female-held positions, respectively – the gap between women’s and men’s callback rates was very large. The differences were smaller for customer service and sales positions, which are 68 per cent and 69 per cent female-held jobs, and were statistically insignificant.

In summary, we find a pro-female bias in callbacks only in occupations in which the percentage of females is 80 per cent or more. For less female-dominated occupations, we find no significant bias towards either sex. This is in contrast to Riach and Rich (2006).

What explains this pro-female bias?

Just as males might prefer to be surrounded by men in jobs that have been traditionally male-dominated, so too might women in female-dominated occupations. The results from the field experiment confirmed our hypothesis that there is a pro-female bias in female-dominated occupations that increases with the proportion of women. Indeed, we can combine the results across four occupations in Booth and Leigh (2010), four occupations in Riach and Rich (2006), and five occupations in Riach and Rich (1987). When we plot these thirteen data points, there is a clear positive relationship between the share of women in the occupation and the pro-female bias.

The demand side: Gender of the recruiter

But can our data provide more demand-side information as to why this might be the case? To investigate this further, we used additional data about the gender of the contact person on the job advertisement, as well as whether or not recruitment was done by a professional recruitment company. Our goal was to test two more conjectures. The first was that firms with a female contact person on the job advertisement might be differentially prone to recruit women. The second was that different recruitment agencies might have a differential propensity to discriminate. (For instance, they might discriminate less if they are better trained in equal opportunity legislation, as one might hope, or discriminate more if they reflect their clients’ biases plus their own.)

To test these hypotheses, we included extra controls in some of our regressions of the probability of obtaining an interview. These additional controls were:

1. Interactions of female applicant and female contact person; and

2. An interaction between female applicant and an indicator for whether recruitment was done by a professional recruitment company.

Although positive, the estimated coefficients on both of these interactions were not statistically significant. Thus, neither the gender of the contact person nor the use of a professional recruitment agency explains the average pro-female bias in callbacks for our sample of female dominated occupations.

What might cause this pro-female bias in occupations that are heavily female? One explanation, mooted above, is gender stereotyping. If certain jobs are perceived as more appropriate for women, male applicants may be (implicitly or explicitly) evaluated less favourably because they do not fit society’s prescriptions about what is appropriate for men.

According to the gender identity hypothesis, individuals operate within society’s constraints and their utility may be powerfully affected by social custom and conditioning. Society's prescriptions about appropriate modes of behaviour for each gender might result in women and men experiencing a loss of identity should they deviate from the relevant code. If this is the case, employers might be happier employing women in particular posts, since both applicant and manager are then adopting modes of behaviour dictated by social custom. Much more research remains to be done in teasing out the workings of these demand-side mechanisms.

What is the relevance?

Does it matter that otherwise identical men are discriminated against in entering predominantly female jobs? And does it matter that otherwise identical women are excluded from predominantly male jobs? The answer to each of these questions has to be "yes”, both on equity and efficiency grounds.

On average, women are discriminated against more often than men (one only has to look at studies of pay differentials to see this). But it is nonetheless interesting and thought provoking to see that gender identity can cut both ways.

Alison Booth is professor of economics at the University of Essex and the Australian National University. She is also a CEPR research fellow.

Andrew Leigh is a professor in the economics program of the Research School of Social Sciences at the Australian National University.

Originally published on Reproduced with permission.

Related Articles