Bias and a possible gender pay gap?

The media this morning is awash with breathless reports of a new study (and slides on the main results here), conducted for the Ministry for Women, on the differences between hourly earnings for men and women.  Not a hint of scepticism has been reported in what I’ve seen and heard, even from the Deputy Prime Minister in our ostensibly centre-right market-oriented government.  On principle, one should probably always be more sceptical of research results that confirm the preferences and priors of the agencies commissioning such research.

It isn’t my area of speciality at all, so this is just a brief note.  To the extent I have a dog in the fight, on the one hand I tend to believe that markets work pretty well most of the time, which makes me instinctively suspicious of the notion of “free lunches” or that somehow huge systematic effects result from conscious or unconscious bias or discrimination.  And on the other hand, I’m not in paid employment myself, while my wife is, and I have more daughters than sons.    If there was a material real gap, it wouldn’t be against our family interest to recognise it.

The authors use a fairly large sample of people, and look at quite a range of variables.  But, having read some summary articles on this issue from the US literature, I was quite struck by what appeared to be missing.  I couldn’t, for example, see any sign of a “time out of the workforce” variable.  I think there is huge value in one parent being at home, especially when children are young, but it isn’t necessarily experience that has a huge value back in the paid workforce (different skills, different roles).   Someone –  male or female, but most are women –  who takes five years out of the workforce to look after their young children is likely to set back their income-earning prospects back in the paid workforce, for any given set of qualifications, or even any particular type of job.   And, as far as I could see, the only proxy for input/effort (and thus, accumulated skill/expertise)  was a distinction between part-time work and full-time work, at a 30 hours a week cut-off.     For some jobs, that might be a perfectly reasonable marker,  but there is a big difference between, say, a lawyer working 32 hours a week and one working 60 hours a week.  More women are more likely to select for working relatively fewer hours, to prioritise family, and this study –  data-rich as it is –  doesn’t seem likely to be able to distinguish that point.

Are there pointers in the paper to these sorts of factors being part of the explanation?  Well, yes, there is, here in this chart.

gender pay gap

Note that there is no statistically significant difference in gender pay for the first three deciles, (and if anything the unexplained component goes the other way –  the purple line is above the “gap” line).  That appears to run contrary to, for example, one of the Ministry for Women’s other “causes” –  eg pay differences between rest home workers and other ostensibly similar occupations.    The big difference show up in the upper part of the income range, where the personal characteristics of the individual employee are likely to be both much more important (than for relatively homogeneous positions at the bottom of the income scale) and much harder for researchers in studies like this to observe.

If such large differences, for people with exactly the same characteristics, existed, it points in the direction of large “free lunches”.  Relatively “enlightened”, or unbiased employers, could profit hugely by replacing expensive male employees with cheap females ones, in relatively senior positions.  Perhaps it was to some extent true 100 years ago, when cultural expectations –  among men and women –  were quite different.  It defies belief now.

UPDATE: In some exchanges in the comments, not only did it became apparent that the AUT researchers had not even cited the leading work in this field, by Claudia Goldin at Harvard, but that the Ministry for Women policy staff, while aware of that work, do not refer to it.   Here is a link to an accessible discussion with Goldin  and an extract from that interview

DUBNER: Talk for a moment about potential categorical differences between men and women that have shown up in some research.  The different appetite for competition, as some have labeled it. Or, in another instance, the willingness to bargain on salary or flexibility.  How much might those contribute to the pay gap?

GOLDIN: I think there’s no doubt that they contribute to some degree. But let me tell you why I don’t think that they go the real distance. Some of the best studies that we have of the gender pay gap, following individuals longitudinally, show that when they show up right out of college, or out of law school, or after they get their M.B.A. — all the studies that we have indicate that wages are pretty similar then. So if men were better bargainers, they would have been better right then. And it doesn’t look as if they’re better bargainers to a degree that shows up as a very large number.  But further down the pike in their lives, by 10-15 years out, we see very large differences in their pay. But we also see large differences in where they are, in their job titles, and a lot of that occurs a year or two after a kid is born, and it occurs for women and not for men. If anything, men tend to work somewhat harder. And I know that there are many who have done many experiments on the fact that women don’t necessarily like competition as much as men do — they value temporal flexibility, men value income growth — that there are various differences. But in terms of bargaining and competition it doesn’t look like it’s showing up that much at the very beginning.

DUBNER: Let me ask you about one more contributory factor. The parent penalty, what’s often called the mommy tax. How significant is that as a contributory factor?

GOLDIN: Well, it seems as if it’s a very large factor.  That anything that leads you to want to have more time is going to be a large factor.


31 thoughts on “Bias and a possible gender pay gap?

  1. I totally agree with your last paragraph HOWEVER this does not occur. Surely this means rational economic man or homo economicus does not exist. Did he ever?


    • People will probably always display a slight preference for “people like them” (along whatever dimension – age, sex. race,religion,temperarment, old-school tie), and aren’t necessarily irrational to do so. So it wouldn’t surprise me if some ideal study (never likely to exist) found small male/female wage (and total remuneration) gaps, once all other characteristics were corrected for, but ones the size found in this study – and concentrated in the upper half of the income scale – still defy belief. It would surprise me more if such male/female gaps existed than if Maori/non-Maori ones did.

      Anecdotes mean almost nothing, but i spent 25 years sitting in salary-setting meetings for RB economists, and I’m pretty confident no such gap existed there (altho there were disconcerting questions at times as to why we didn’t attract/retain more women).


  2. A business operates on profit and the best person available to do a job in NZ. In my years of working experience there is minimal gender bias in terms of getting the best person to do that job to the degree that this ridiculous study is indicating.

    Over the years I have worked with both gender and personally I find it a lot tougher to work for and with women. There is a lot of emotional issues to deal with. The job is tough enough without having to deal with alot of gossiping, back stabbing, jealousy and self esteem issues.


    • Also with women in general there is alot of so called women’s intuition, in order words complete guesswork and a rather firm and stubborn belief that intuition replaces sensible research and logic. The number of arguments to get a decision over the line because of women’s intuition type arguments I have encountered over the years in management meetings suggests that women in general have still a lot of growing up to do.


  3. This is dangerous territory

    There is probably no correlation, but, while we have an immigration issue, allowing low-skilled bodies into the country to suppress wages, one could “command” those cheapskate employers to hire local female natives instead


    • Speaking of immigration, is there any data on gender splits, say by age bracket – given we are particularly interested in working age immigrants? Now that would be interesting… we could find our immigration settings have a gender bias.


      • I can’t readily spot a time series, but in 2015/16 of the residence approvals granted, 51 per cent were to females. Within that total 61% of “partners” (ie not the principal applicant) were female, and partners made up around 20% of the total approvals. Women were also 56% of the parent approvals – hot surprisingly I suppose given relative life expectancies.


      • oh, and by age, there is a dispropportion, in favour of woman, among the 20 to 29 year old age bracket , and in the age 50+, balanced by a slight preponderance of males in the other age brackets


      • Thanks Michael. How big is that disproportion in favour of women in the 20-29 yo age bracket? I wonder if that largely relates to the Aged Care Nurse category – as I seem to recall from one of your earlier posts that being second largest to Chef.


  4. on a five minute assessment the report seems to capture ‘effort’. They work on hourly pay rates which are based on actual hours of work


    • Not quite my point. someone working 32 hours a week as a lawyer is unlikely to build up the same accumulated expertise (10000 hours and all that) that someone working 60 hours a week will be – they are probably also less readily available to jump to meet all client demands/whims etc. So “effort” was a misnomer on my part.


  5. The report seems to refute the ‘glass celling hypotheses (at least as it is normally understood .) Fig one of the report show that the proportion of female workers in high earning jobs ($100K+ ) is the same as men. The differences are in higher female representation in low paid work and lower representation in middle income work

    – another five minute effort so I could have it wrong


  6. holding off until I get response to OIA request on whether MWA used any research by Claudia Goldin, one of the world’s top labour economist and best economist on gender

    google ‘We Are All Racists At Heart’ By AMY WAX and PHILIP E. TETLOCK WSJ 2015


      • There is no reference to the overseas literature apart from when a study supports their particular theme about unconscious bias. There is certainly no survey of the overseas findings

        There has been a radical transformation in the economics of gender in the last 15 years. When human capital stopped explaining the gender wage gap, clearly wages data ceased to be a useful measure of the gender gap. Fringe benefits, worklife balance, flexible hours, and occupational choice became paramount.

        The experimental literature based on sending fake CVs to jobs is not much good because it falls to James Heckman’s criticism that the marginal employer matters not the average one.

        If it is how the workplace is structured because of the rewards for working very long hours, particular hours or continuous availability to clients, that is a lot harder to change than boys picking boys.

        The unconscious bias hypothesis must explain why it is largely confined to female professionals at the peak of their career.

        I should add that the author is one of the best labour economists in New Zealand. Her work on the minimum wage is far better than anyone else in New Zealand.


  7. there is a reference in the paper to a UK paper 2007 that adjusts for degree type. Found that much of the difference between male and female pay disappears once you account for fact that males are heavily represented in engineering etc that pays better as opposed to teaching etc. Apparently they are going to adjust for qualification type in subsequent work.
    Would have been better to do the work first before making the big announcement.

    what could have been done too, is to present the stats for the male and female models separately rather then just some results of how much of the difference in earnings was explained by the model and how much was not. If the model could explain much of the variation in male rates then we would have more confidence that the difference in male/female rates was explained by discrimination. If the degree of explanation is relatively low then we would know that much was unexplained and and should be more cautious in drawing conclusions.


  8. Michael

    The authors say “the majority of the gender pay gap remains unexplained. The proportion of the average gap that is unexplained ranges from 64.4% to 83.4%.” So there is nothing here.

    I agree with you and with Jim Rose.

    Liked by 1 person

  9. Another confounding problem is that women more often than men end up working in fields they did not originally train for. Men tend to stick at the job or field they trained in; women are much more likely to retrain during their lifetimes for a new career, with all that implies for salary/wages in starting from the bottom again.


    • Agreed, I have 2 qualified young lady chartered accountants working for me. Both highly capable individuals. One is on maternity leave and having a child and as a result has been happy to work from home but not prepared to take on the more difficult jobs that require more attention time which would have broadened her experience and consequently limits her future self development potential. The other is now contemplating a career in human resources and is prepared to go down into a lower skilled payroll processing position in a larger company. She is our most highly paid individual compared to the rest of the team but does not keep updated with her day to day readings which equates to poor general knowledge even though processing and operationally technically capable but makes her very narrow in terms of meeting strategic type decision making.


  10. Just wanted to leave this quote from Goldin here, since there seems to be too much confirmation bias that based on Michael’s moral position and belief in efficient markets that the MWA report has missed something crucial that nullifies the debate on the gender pay gap.

    “The gender gap in pay would be considerably reduced and might even vanish if firms did not have an incentive to disproportionately reward individuals who worked long hours and who worked particular hours”

    The calls on a better measure of the value of ‘unpaid labour’ that women who take time off to support to their families and (hardworking) spouses. The pay-gap will exist for both genders, who may act as caretakers, as long as that care-taking labour (that requiers flexibility and time-off) is systematically unpriced in relation “productive” paid labour.


    • My only point in the initial post was that claims that pay differences results for bias and discrimation are unsupported, either in the AUT research, or in the results of people like Goldin. The differences exist for a variety of reasons, many of which probably have to do with preferences and opportunities – in some areas, for example, there are genuinely high returns available for people willing to work night and day. Is that socially desirable? I’m not sure – although in a family context, i see no particular problem if for example a couple agree that one will work long hours, and one will look after house and family.

      Liked by 1 person

  11. I forgot to ask where is figure 5 from. Cannot find it online through Google search?

    New Zealand data which I had a dompost op-ed on shows that the gender wage is about 2 or 3% except for the top 10% of workers


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s