The View From the Cello Section

I do not think every difference in occupational choice is created by discrimination. People have different abilities, interests and priorities, and equality does not require everyone to make identical choices. I have no ideological objection to the possibility that average interests differ.

But neither do I believe that millions of choices occur in a cultural vacuum.

If women disproportionately enter lower-paying occupations, then yes, occupation explains part of the pay gap. But if the occupational distribution is itself shaped partly by gendered expectations or barriers, then controlling for occupation may mathematically remove part of the very structural effect we are trying to measure.

That does not make every occupational difference discrimination. It means we cannot automatically file occupation under “innocent variable.”

Part-Time Work, or: Down the Research Hole

Around 42% of working women aged 16–64 were working part-time in the second quarter of 2013, compared with 12% of working men. Full-time men also worked an average of 44 hours a week, against 40 for full-time women. Those figures come from the same ONS report.

This is one reason the gap across all workers is so much larger than the full-time gap, and it is perfectly legitimate for the opposition to point out that the headline all-employee figure compares people working in different circumstances.

What is not legitimate is to say:

The gap is explained by women working part-time, therefore it has nothing to do with structural discrimination.

That skips the most important question. Why are women more than three times as likely to work part-time?

Sometimes the answer is preference. Sometimes it is study, health or the availability of work. Very often it is caring responsibility.

Then there is the pay. Part-time work pays less per hour on average, and I went looking for why. Alan Manning and Barbara Petrongolo looked at this in Britain. Their figures showed women working part-time earning around a quarter less per hour than women working full-time, but when they controlled for occupation, most of that unexplained penalty disappeared.

One possible reaction is: Excellent. Nothing to see here. Part-time women simply work in lower-paying occupations.

Except that immediately raises another question. Why does working part-time so often mean working in a different, lower-paying occupation?

Sara Connolly and Mary Gregory looked at the earnings paths of British women moving into part-time work, and found, among other things, occupational downgrading among women moving from higher-skilled jobs. In other words, a woman may not simply take her existing career and perform it for three days a week instead of five. She may discover that the part-time version of her career barely exists. So she changes job, possibly changes occupation, possibly accepts work below her qualification level.

Now imagine someone saying: “The pay gap isn’t discrimination. She earns less because she chose a lower-paying occupation.”

That is factually true at one level. But if she entered the lower-paying occupation because higher-skilled work was unavailable on hours compatible with caring responsibilities, then calling the result “choice” conceals quite a lot.

Again, this is not proof of discrimination.

It is evidence of structure.

Children

We cannot discuss any of this seriously without discussing children.

Women have children. Men also have children. This biological revelation has apparently surprised parts of the labour market.

Pregnancy creates some unavoidable differences. Only women become pregnant and only women give birth. Nobody designed that. But the economic consequences after birth are not dictated entirely by biology. Who takes time away from work? Who returns full-time, and who returns part-time? Who becomes the default parent when the nursery phones? Whose career is treated as more important when a couple has to decide whose hours change?

Those are household choices. I can already hear someone saying so, and they are. But household choices are influenced by wages, available childcare, working practices, social expectations and previous choices, and they can become circular.

Imagine a heterosexual couple. He earns more. They have a child. Childcare is expensive and one of them needs to reduce their hours. The economically rational option is for the lower earner to reduce theirs.

She is the lower earner. She goes part-time. Part-time work limits the roles available to her. She gains less experience, receives fewer opportunities and progresses more slowly. Five years later he earns substantially more than she does. When another child arrives, the economic case for her to compromise her career is even stronger.

At no point does anyone have to announce that a woman’s career matters less. Each individual decision can make perfect economic sense. The combined outcome can still reinforce a gendered structure, because the initial difference is strengthened by every decision that follows it. The outcome becomes both cause and consequence.

Rational individual choices can collectively reproduce inequality.

Economics is annoying like that.

The ONS found that women with dependent children were less likely to be employed than women without them. For men the relationship ran the other way: fathers were more likely to be employed than men without children. Parenthood appears to weaken women’s attachment to paid employment while strengthening men’s.

That does not look like two sexes independently expressing unrelated preferences. It looks like a social division of responsibility.

And it is not costless for men either. The expectation that fathers should be uninterrupted earners restricts their choices too. Structural discrimination can advantage a group economically while still imposing costs on individual members of it.

“But the Gap Shrinks When You Control for Things”

Yes. It does.

This is probably the strongest argument against the lazy version of the claim, and any honest case has to concede it. Men and women do not, on average, work identical hours, do identical jobs, work in identical industries, accumulate identical amounts of continuous full-time experience, take identical career breaks, perform identical amounts of childcare or reach identical levels of seniority. Account for all of that and the unexplained gap becomes smaller.

That matters.

But I think the debate usually stops one question too early.

“The gap is explained by women working part-time.” Fine. Why do women disproportionately work part-time?

“It’s explained by occupation.” Fine. Why are men and women distributed so differently across occupations?

“It’s partly career breaks.” Fine. Who takes career breaks, and why?

“It’s seniority.” Fine. Why are men disproportionately represented at higher levels of it?

An explanatory variable can itself require explanation. If women have less continuous experience because they undertake more unpaid care, removing experience from the calculation does not establish that the resulting inequality is harmless. It may simply remove one of the mechanisms through which the structure operates. If women are channelled towards lower-paid occupations, controlling for occupation describes the result while concealing the process that produced it. If promotion favours uninterrupted full-time service, controlling for seniority does not tell us whether access to seniority was equal.

A variable does not acquire moral innocence merely because it fits inside a regression equation.

Statistical adjustment can answer: are apparently comparable women and men paid differently? That is an important question. It cannot answer: why did women and men become apparently incomparable?

That is a different question, and it is the one this motion asks us to consider.

The Gap Grows With Age

One paper I found particularly interesting was Alan Manning and Joanna Swaffield’s work on early-career wage growth. They found that the gender pay gap at entry into the labour market was approximately zero but widened substantially over the first decade of employment.

That matters to me. If young men and young women start on broadly similar wages and a significant gap develops afterwards, something is happening during their careers.

Part of it can be explained by differences in human capital, employment history, job mobility and various behavioural factors. But the researchers still found an unexplained gap even among women with continuous full-time employment, no children and no expressed intention to have them.

The ONS pattern points the same way. In the 2012 figures, the full-time median gap was 9.6%. At the bottom tenth of the earnings distribution it was smaller. At the top tenth it was substantially larger, around 19%. If gender were irrelevant to career progression, I would not automatically expect the difference to widen at the upper end of earnings.

The ONS report also found that men and women were more evenly represented among the highest-paid tenth in the youngest age groups. Women’s representation then fell most sharply between ages 25–29 and 30–34, around the age at which many begin having children. That is not conclusive proof of a motherhood penalty, and the ONS itself offers possible explanations rather than claiming causation. But it is evidence of a life-course divergence.

Men and women begin closer together. Something happens as they age.

The possible explanations include different occupations, industries, career interruption, working hours, promotion, bonus structures, seniority, selection, negotiation and discrimination. Probably several at once. But a gap that is especially large among higher earners supports the idea that progression matters, not merely starting pay. Which returns us to structure. Who gets promoted? Who gets the assignments that lead to promotion? Who gets mentored? Who can work the hours associated with senior positions? Who is assumed to be committed, and who is assumed to be leaving soon to have children? Who asks for more money, and how do employers react when they do?

None of these questions is answered by the headline statistic. They are questions the headline statistic gives us reason to ask.

“Unexplained” Does Not Mean “Discrimination”

Please read that sentence again.

An econometric residual does not acquire a tiny badge saying SEXISM. Think of a wrong note in a performance. It is evidence that something went wrong, but it doesn’t tell you whether the cause was the player’s left hand, the bow, the instrument, a lapse in concentration or the person beside her turning the page at an unfortunate moment. You investigate.

If someone on my side of the house says, “We controlled for the variables, 8% remains, therefore that 8% is discrimination,” I think the opposition should attack them. I would.

But it cuts both ways. You cannot say, “We explained 60% with occupation, hours and experience, therefore discrimination accounts for none of it.” That does not follow either. Some of the explained variables may themselves reflect structural inequality, and some of the unexplained difference may be discrimination. Some may not.

Unexplained means unexplained. The correct answer is irritating: we have to investigate the mechanisms.

“Women Make Different Choices”

Yes. They do.

So do men.

This is probably the most contentious part of my argument, so let me be careful. I do not think saying that women have agency is anti-feminist. Women choose occupations. Women choose whether to have children, whether to work full-time or part-time, which employers to work for. Women negotiate. Good.

But human choices are not made in a vacuum.

I chose Law. It was my choice. It was also shaped by my education, my family, my teachers, which subjects I enjoyed, which universities I could realistically attend, my finances, which careers I knew existed, what I thought I would be good at and probably a hundred influences I could never identify. Acknowledging all that does not mean I did not choose Law. It means that explaining a choice requires more than observing that somebody made one.

The same applies to employment. If a woman chooses part-time work because childcare costs more than the additional income from full-time work, she has made a choice. If a man works longer hours because his partner is covering childcare, he has made one too. If both are responding rationally to the incentives facing their household, both retain agency. But those choices can still produce a predictable, gendered outcome.

“Women choose family over career.” Some do. Some men do. More women do. But describing a patterned outcome as a choice does not tell us how that choice was produced, what alternatives were realistically available or who bore its cost. Childcare is expensive. Flexible work is not equally available across occupations or seniority levels. Maternity has physical consequences that paternity does not, but it does not follow that mothers must carry most of the care for the next eighteen years.

We should be able to acknowledge that the choices are real and the structure within which they are made is also real, without treating women as either helpless victims or perfectly independent economic units untouched by family, money or social expectation.

Choice and structure are not opposites.

That may end up being the central sentence of my speech.

The Inconvenient Evidence

Among part-time employees, women’s median hourly pay in 2012 was slightly higher than men’s. An honest proposition case should say so.

Does that disprove structural discrimination? I don’t think so. Men and women who work part-time are not necessarily comparable populations. They may differ in age, occupation, industry, career stage and reason for working part-time. The median favours women while the mean favours men, which tells us the earnings distributions differ considerably.