Pollution at work

While much progress has been made, the gender wage gap remains evident in even the most progressive and equal societies. For example, the latest data for Denmark show that women’s earnings are 88 per cent of men’s — and lower still when comparing the earnings of female managers and professionals relative to their male peers.

Part of the explanation for the gender wage gap relates to gender-based differences in occupational profiles. Some jobs are disproportionately filled by male workers, while others are dominated by female workers. At one level, this can be understood as a form of path dependence: a profession which has a high share of women is perceived as ‘women’s work’, and therefore continues to attract more women than men.

Yet the strength of path dependence in this context is questionable, given that changes in gender profiles within occupations have taken place over time. In some cases, the changes have been profound. For example, across the developed world, teaching was historically a male-dominated field, but switched to a female-dominated one during the nineteenth and twentieth centuries. (In recent decades, the pendulum has begun to swing back, though to differing degrees for primary and secondary levels and across subjects.)

This phenomenon — and its implications for differences in male and female earnings — is central to a model proposed by Goldin. Her pollution theory of discrimination offers a novel framework, which boils down to an interaction between job status and technical change. In the absence of information, gender stereotypes influence how jobs are perceived.

Perception is reality

In crude terms, Goldin’s pollution theory starts from the premise that men — at least in a historical context — view women as a threat to the status of their jobs. In the absence of evidence to the contrary, women are perceived as inferior to men with respect to labour output. Thus, if it turns out that women can do a given job, it must be a job requiring relatively less skill than one that only men can do. When women enter male-dominated professions, it ‘pollutes’ or diminishes the reputation of that profession.

That this logic builds on a faulty assumption — that of inherent female inferiority — does not matter for our purposes. The model considers how individuals (men) behave given that assumption.

The kernel of truth in this stylised narrative is that job status and skill are linked. All else being equal, the higher the skill level of a job, the more prestigious that job is. But the skill level of any job is not fixed over time: technical advances continuously change the expectations for and demands of workers. Agricultural production — and the role of farmers — has been repeatedly transformed over the centuries with the introduction of new tools and equipment. Large swathes of manufacturing, which used to be labour intensive, are now heavily automated. Office typing pools have been made redundant by the rise of personal computing.

Technological shocks, and their effect on skill demands, can diminish the status of a particular job. But we do not necessarily know when a shock is hitting — rather, its true impact first becomes apparent when looking backwards. Pollution theory suggests that the entry of women in a previously male-dominated profession may be interpreted as evidence of technical change reducing skill requirements.

Man (bottom right) surrounded by pollution. Source: FJWvanBlokland / Wikimedia Commons.

In terms of basic mechanics, it helps to think of a simple two-period model. In the first period, the labour force consists only of men. In the second period, women start to work.

There is some distribution of skills across the population: that is, workers are heterogeneous. Each job requires a level of skill. Men enter jobs in the first period commensurate with their skill level.

Individuals only know their own skill level; that is, skill is not directly observable. But the matching of individuals (men) to jobs in the first period allows for the skill level of men to be inferred.

At the start of the second period, when women start to work, the distribution of skills among men is known, but nothing is known about the distribution of skills among women. Men assume that, in general, women are less skilled than them. (Whether the distribution of skills is actually different between men and women is immaterial here. What matters is men’s perceptions of women’s skills.)

When women enter a relatively high-skilled profession, the status of that profession is depressed. That is, the perception (among men) is that the skill requirements for that profession have weakened. That the women who enter the given profession might be just as — or more — skilled than the men already working in it is immaterial to the ‘pollution’ effect.

Generalising from the simple model, job status is contingent on a job’s required skills. If women are perceived as lower skilled than men, then the entry of women into a previously male-dominated professions adversely affects the prestige of that profession. And with a decline in prestige, jobs that might previously have commanded higher wages may no longer be quite so lucrative.

Evidence from the US

Viewed through a contemporary lens, the story I’ve sketched out is hard to recognise. In today’s world, many professions involve a mix of men and women. The point is that this has not always been the case. Pollution theory gives us a framework for understanding how we got to where we are.

Goldin demonstrates how well pollution theory holds in reality by considering historical evidence on office workers in the United States. Using labour market survey data from 1940, Goldin examines earnings for men and women, controlling for (among other things) the shares of men and women by occupation. Office work is a broad category, encompassing multiple occupations — in Goldin’s sample, 75. Whereas stenographic functions (typing) were typically performed by women, accounting functions were mostly the domain of men. Other occupations, including a range of general clerical roles, involved a mix of men and women.

The table below includes two sets of information. To the left of the variables are sample averages, which give an overview of the profile of male and female workers in the dataset. The right hand side of the table reports regression results, estimating the contribution of the model variables on workers’ earnings.

Gender in the office

Average:
Female
Average: MaleLeft: Summary statistics
Right: Estimated effect on earnings
(1)
Female
(2)
Female
(3)
Male
(4)
Male
6.957.34Annualised full-time salary (log)
10.3512.77Years of office experience0.0320
(0.00220)
0.0315
(0.00368)
0.0507
(0.00240)
0.0483
(0.00608)
Years of office experience squared × 10–2– 0.0588
(0.00613)
–0.0577
(0.0097)
– 0.0804
(0.00507)
– 0.0766
(0.0109)
7.5810.20Years with current employer0.0154
(0.00155)
0.0151
(0.00199)
0.0131
(0.00155)
0.0129
(0.00288)
11.511.9Years of schooling0.0393
(0.00364)
0.0368
(0.00555)
0.0605
(0.00343)
0.0551
(0.00632)
0.1970.484Married– 0.00760
(0.0147)
– 0.00774
(0.0186)
0.140
(0.0169)
0.131
(0.0197)
0.7110.733Sex ratio: share of (respectively) female/male workers in current occupation– 0.818
(0.350)
– 0.588
(0.233)
Sex ratio squared0.720
(0.305)
0.659
(0.182)
Constant6.151
(0.0472)
6.339
(0.115)
5.978
(0.0470)
6.184
(0.139)
Observations1393139314911491
R20.4840.5040.6260.648
Columns 2 and 4 are clustered by occupation. In these columns, the sex ratio reflects the fraction of female (column 2) and male (column 4) workers in each included occupation. Standard errors in parentheses.
Source: Goldin 2014, table 9.1

On average, male and female office workers had broadly the same education level — while Goldin notes that more men than women had university degrees, most office work at the time required only high school graduates. Nevertheless, the returns to a year of schooling for male workers were significantly higher than for female workers. Likewise, the returns to work experience are greater for men than women — at least at earlier stages of one’s career. The returns to experience with the same employer, however, are reasonably consistent.

Furthermore, while married men enjoyed higher wages on average than their unmarried peers, the same is not true for women (perhaps even the reverse, though the results are not significant). Two factors are relevant here: one, that women often left the labour force once they married; two, those married women who continued to work after marriage might nevertheless be expected to leave to raise children.

The most interesting results in the context of pollution theory are highlighted in the sex ratio rows. (Note that the sex ratio used in column 2 is the share of women employed in a given occupation; for column 4, it is the share of men employed.) The quadratic relationship implies that, all else being equal, female-dominated and male-dominated professions deliver higher earnings than gender-mixed professions. This finding might seem curious, but is consistent with pollution theory.

Pollution theory does not predict men will seek to block women from working at all. Rather, the intent is to defend the prestige of certain (relatively high-skilled) professions. Two options are available: one is explicit barriers to women entering certain roles, which Goldin also provides several examples of. The other is to provide incentives for women to find other jobs. The results on sex ratio are indicative of the latter.

Both men and women could be found working in offices in 1940’s United States, but in different roles with distinct career paths. This was not only limited to more senior roles; even entry-level clerical positions were typically divided into male and female jobs. (Being the office errand boy was not a prestigious position, but it was virtually always a young man in the position.) That is, similar to the ‘glass ceiling’ locking women out of positions of power, efforts (by men) to preserve job status also encouraged ‘glass walls’ to separate the sexes on every level.

Cleaning up

There are still examples of glass walls to be found, but most of those once erected no longer stand today (at least in the developed world). Goldin suggests that technical progress — the advances in production processes that can threaten job status — is a key driver of this.

High-pressure treatment of pollutants. Source: CEphoto, Uwe Aranas / Wikimedia Commons.

The changing nature of production and work, especially since industrialisation, has shifted the emphasis from physical labour to intellectual labour — from ‘brawn’ to ‘brain’, as Goldin puts it. If men on average are stronger than women, then men have a skill advantage in jobs requiring physical strength. But as the share of jobs dependent on physical strength declines — with technical advances allowing machines to substitute for labour’s brawn, while opening a range of new occupations that increase demand for brains — then men’s skill advantage evaporates.

Furthermore, as the pace of technical change increases, the returns to brain rather than brawn also increase. The advantages of trying to lock women out of the new ‘brain’ occupations are low in an environment where demand and productivity growth push wages up. The share of female workers becomes a far weaker signal of job status.

Goldin acknowledges that there are still pockets of gender discrimination — what she terms ‘frog ponds’. For example, in many US communities, there remain significant barriers to women becoming firefighters. Such barriers can include both discriminatory entry requirements, as well as harrassment by male colleagues of any women who do get through. But such examples are the exception rather than the norm. The long-term trend shows a strong reduction in pollution.

Pollution theory provides a useful perspective on long-term trends in male and female labour force participation, with relevant lessons for other forms of discrimination. (Goldin also discusses racial differences in the US context.) The real ‘pollution’ has not so much to do with the role of women in the workforce. Rather, it is the pollution of perceptions — misguided and ill-informed as they can be.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.