Tim Worstall

Interesting Statistical Nonsense of the Day

Over at the Fawcett Society they claim the following (in their submission to the Select Committee, bottom right hand of the page).

The
Women and Equality Unit’s own research concludes that the biggest
single cause of the gender pay gap is discrimination rather than
occupational segregation1.

1
Walby, S. & Olsen, W. (2002) The
impact of women’s position in the labour market on pay and
implications for UK productivity,
Women and Equality Unit, DTI

Really? Here’s that paper. On page 11 we get a table. "Discrimination and other factors associated with being female" amounts to 29% pf the gap. So it isn’t discrimination alone, it’s with other things as well.

Then, the lower level of full time employment experience provides 26% of the gap. The next line is very interesting as well: interruptions due to family care are a further 15% of the gap. Plus the next line, part-time employment experience leads to yet another 12% of the gap.

Now, the lower levels of full time employment experience, the levels of part time and the interruptions for family care are all really the same thing. Whether it’s a good thing or not, whether it’s a moral outcome, whether we should be trying to change this, all are very different from the fact that all three are part of exactly the same thing: women do more of the family care in our society and thus spend fewer years in full time employment.

And that explains 52% of the gender pay gap, even by the figures of researchers who really don’t want to let us see that, which is why they’ve split the figures out.

So well done all those at the Fawcett Society. You’ve proven Disraeli’s dictum yet again. By far the largest cause, the majority of it, even by the figures you yourselves quote, of the gender pay gap is because women spend fewer years in the full time workforce than men.

It would be just wonderful to see you campaigning on that issue, rather than ignoring it.

Antonia, you’re a trustee there. Over to you.

In

2 responses

  1. Bruce G Charlton Avatar
    Bruce G Charlton

    Residual (uncontrolled) confounding probably accounts for most/ all of the rest of the supposed gender gap in pay.
    Here comes some elementary – but too often ignored – stats:
    If, every time you introduce a control variable (eg. employment experience, family care, hours of work), you reduce the supposed gender gap; then it is reasonable to assume that more and higher quality (more precise) control variables would continue to reduce the supposed gender gap.
    In other words, the assumption should be that perfect control of confounding would eliminate the gender gap; and that gender gap is probably due to residual uncontrolled confounding (and not to gender discrimination).
    By contrast – with genuine explanations, the more controls you introduce, the bigger the difference gets.
    For example, men are taller than women on average in most studies – but this difference may be obscured by the fact that height is correlated with educational level. So that if you compare highly educated women with poorly educated men in the same society the average height difference between men and women will be smaller or may disappear.
    But if you control for education level, so that highly educated women are compared with highly educated men (ie you control for the confounding effect of education) then the average height difference between men and women gets bigger.
    And if you could control for everything perfectly, all men would be taller than all women (because, in our species, as in most primates, males are biologically bigger – in humans by about 15 percent).
    So – if adding controls makes the difference smaller – then the observed difference is probably an artifact due to poor control; but if adding controls makes the difference bigger, then the difference is probably real.
    Therefore: the gender gap in pay is probably an artifact, due to inadequate control of confounding variables.
    However, the fact that the variation in ability is greater in men than women men means that you get more extremes in men (more geniuses, more dunces) – so this may be a real cause of gender differences in pay – even without any sex difference in average ability.

    Reply
  2. Mark Wadsworth Avatar
    Mark Wadsworth

    That’s why under my Citizen’s Income reforms, child benefit will be £34 per child payable to the mother, so £68 a week for an median woman.
    Let’s say she earns £100 a week less than average man, so what? He’s paying £33 flat tax, so he’s £67 ahead on net salary and she gets £68 tax free child benefit.
    That’s that fixed.

    Reply

Leave a Reply to Bruce G CharltonCancel reply

Discover more from Tim Worstall

Subscribe now to keep reading and get access to the full archive.

Continue reading