Diversity versus Equality

by on January 28, 2018 at 10:56 am in Economics, Law | Permalink

The Australian Behavioural Economics Team conducted a randomized trial of hiring in which applications for senior positions in the Australian Public Service were reviewed and ranked. By comparing outcomes in treatments in which gender, minority status and indigenous status could be inferred with outcomes using de-identifyed applications the researchers were able to test for bias and the effect of de-identification.

We found that the public servants engaged in positive (not negative) discrimination towards female and minority candidates:

Participants were 2.9%
more likely to shortlist female candidates and 3.2%
less likely to shortlist male applicants when they were identifiable, compared with when they were de-identified.

Minority males were 5.8%
more likely to be shortlisted and minority females were 8.6%
more likely to be
shortlisted when identifiable compared to when applications were de-identified.

The positive discrimination was strongest for Indigenous female candidates who were 22.2% more likely to be
shortlisted when identifiable compared to when the applications were de-identified.

Interestingly, male reviewers displayed markedly more positive discrimination in favour of minority candidates than
did female counterparts, and reviewers aged 40+ displayed much stronger affirmative action in favour for both
women and minorities than did younger ones.

The study was small and the participants knew they were in a study (although not what the study was studying).

This reminds me of the important Williams and Ceci paper which also found positive gender discrimination in academic hiring (with one notable exception of equal treatment):

The underrepresentation of women in academic science is typically attributed, both in scientific literature and in the media, to sexist hiring. Here we report five hiring experiments in which faculty evaluated hypothetical female and male applicants, using systematically varied profiles disguising identical scholarship, for assistant professorships in biology, engineering, economics, and psychology. Contrary to prevailing assumptions, men and women faculty members from all four fields preferred female applicants 2:1 over identically qualified males with matching lifestyles (single, married, divorced), with the exception of male economists, who showed no gender preference.

Hat tip: Phil Magness.

1 clockwork_prior January 28, 2018 at 11:04 am

‘with the exception of male economists, who showed no gender preference’

And one can likely further safely assume that white male economists are equally colorblind, too. One can check here concerning GMU full time econ faculty – https://economics.gmu.edu/people/full_time_faculty

Reply

2 Slocum January 28, 2018 at 11:14 am

“One can check here concerning GMU full time econ faculty.”

And what conclusion should we draw about gender ‘colorblindness’ here:

https://chhs.gmu.edu/nursing/faculty-and-staff.cfm

or here:

https://chhs.gmu.edu/socialwork/faculty-and-staff.cfm

Could it be — here’s a radical thought — that, on average, males and females have different interests?

And how, BTW, would you define fairness in hiring other than having the same ratings of candidates when the sexes are known as when they are not?

Reply

3 clockwork_prior January 28, 2018 at 11:45 am

Well, so much for closing tabs quickly.

‘Could it be — here’s a radical thought — that, on average, males and females have different interests? ‘

Which certainly explains why women did not become lawyers in the 1940s at the rate they do today, right?

‘And how, BTW, would you define fairness in hiring other than having the same ratings of candidates when the sexes are known as when they are not?’

Interesting question – but since absolutely no hiring decision is made on a purely objective standard, it would seem that this question is not especially relevant in a sense. Let us use an example – two candidates for a MIT econ dept. faculty position have precisely identical qualifications, except one earned their PhD at Harvard, the other at GMU. Venture to guess what the odds are for the GMU econ doctorate holder to be picked, based on only that single variable? And if they were not picked due to that single factor, would it be justifiable?

And this is not precisely hypothetical – it definitely applies to top tier law schools in terms of how their graduates have an advantage in the legal job market as compared being a candidate measured using a a set of (theoretically) objective measures.

Reply

4 aMichael January 28, 2018 at 1:04 pm

I think there are definitely structural biases that make it less likely that women (and men) pursue different careers, but 2017 econ departments are not 1940’s law schools and firms. Come on…

Reply

5 Alistair January 29, 2018 at 5:49 am

Slocums point was, certeris paribus, fairness required knowledge of gender not to change the prior rankings; that it added no _information_ to the output.

As usual, you didn’t answer the question and went off to talk to your strawman.

Reply

6 Phil Magness January 28, 2018 at 1:50 pm

If you’re going to keep a tally, it warrants mention that that the chair of the GMU econ department that built up its current program from scratch in the 1970s-80s was a female economist.

https://economics.gmu.edu/people/kvaughn

GMU has also tended to graduate a higher number of female econ PhDs than most other departments in the profession in recent years.

Reply

7 Philippe Lemoine January 28, 2018 at 11:11 am

Philosophy is heavily male-dominated, which made a lot of people claim that women were discriminated against, but as I discuss in this post, the data we have suggests that, at least in hiring, exactly the opposite is true.

I also use data from The Freshman Survey to argue that the proportion of women in various academic fields is mostly the result of the fact that, even before they have taken a single class in college, men and women have different preferences.

For instance, they are underrepresented in philosophy because, among incoming students, women are already less interested in majoring in philosophy, but they are overrepresented in psychology for the opposite reason.

Reply

8 anonprof January 28, 2018 at 3:22 pm

This certainly seems to be the case in Physics. About a fifth of students declaring a major in physics are women. This is what we see all the way through to tenure track positions.

Reply

9 Cathy Newwoman January 28, 2018 at 11:16 am

“men and women have different preferences”

. . . because of social brain-washing. Men and women who are treated the same show no difference in intellectual pursuits or hobbies.

Reply

10 Ivo January 28, 2018 at 11:33 am

It would be incredibly surprising if that were true, given the influence of nature on human development and the physical differences between an average male and an average female. You would expect all those differences to have at least *some* influence.

Reply

11 clockwork_prior January 28, 2018 at 11:51 am

‘It would be incredibly surprising if that were true’

Why?

‘given the influence of nature on human development and the physical differences between an average male and an average female’

Which is relevant to someone working in an office environment for what reasons? Or, in a slightly different context, working as a food safety inspector, for example. Or being a lawyer or politician – it seems as if women have all the necessary characteristics to handle such things as well as a man.

Reply

12 dearieme January 28, 2018 at 11:56 am

He was being sarcastic. If you find sarcasm hard to detect, simply observe his pseudonym.

Reply

13 Jeff R January 28, 2018 at 12:51 pm

The problem is people make this claim all the time, so it’s hard to distinguish sarcasm from earnest assertions of stupidity. See prior’s comment above.

Reply

14 dearieme January 28, 2018 at 6:29 pm

observe his pseudonym

15 Hazel Meade January 28, 2018 at 12:40 pm

It would be really hard to experimentally prove that ” Men and women who are treated the same show no difference in intellectual pursuits or hobbies”, since we can’t exactly raise human beings in a lab.
That said, there probably are some innate differences AND some social brainwashing/ acculturation which drive different interests. Obviously, women today pursue different occupations than they did 100 years ago, so it’s undeniable that the social environment plays a role. And of course these things are averages and individuals vary.

Reply

16 TMC January 28, 2018 at 1:04 pm

There were two articles in the past week that were interesting. One was written by a devout feminist whose daughter, despite the all the interventions by the mother, still preferred the color pink and playing with dolls. This vexed the mother to no end.

The second was how chims displayed the same sex based stereotypes as humans when young.

Reply

17 TMC January 28, 2018 at 3:25 pm

Sorry. ….how chimps displayed…

Reply

18 Aaron Luchko January 28, 2018 at 5:52 pm

There’s no intrinsic reason for girls to prefer pink, the association is only a recent phenomena.

That the daughter still preferred pink demonstrates that even a devoted parent may not be able to overcome societal conditioning.

Reply

19 TMC January 28, 2018 at 9:34 pm

From the article, it doesn’t seem like pink was very available, at least for the masses. Maybe like purple was for royalty because it was difficult to get.

20 Hazel Meade January 29, 2018 at 1:52 am

I hate pink, personally. Always have.

21 Hazel Meade January 29, 2018 at 1:59 am

Also, from the link:
Although it did not have any strong negative associations in these surveys, few respondents chose pink as their favorite color. Pink was the favorite color of only two-percent of respondents, compared with forty-five-percent who chose blue.[37] Pink was the least-favorite color of seventeen percent of respondents; the only color more disliked was brown, with twenty percent. There was a notable difference between men and women; three percent of women chose pink as their favorite color, compared with less than one percent of men.

I’m going to chalk down that 3% vs. 1% as social conditioning. No offense.

22 Alistair January 29, 2018 at 6:10 am

Yeah, Hazel, you’re got no chance against our Patriarchy. Tremble at its power! 🙂

Given Our Vast Social Conditioning Apparatus has successfully changed 2% of minds about favourite colour, it should be easy to persuade women to abandon political and economic autonomy, surrender to male authority and abandon any role outside childrearing and the home.

From this we can see there’s no end to what Social Conditioning can do.

23 Hazel Meade January 29, 2018 at 10:25 am

The Patriarchy hasn’t been very successful at convincing women of anything lately.
What’s more dismaying is is the feminists who run around wearing pink ribbons and insisting that women vote for Democrats because they care about children and health care – because women are so naturally suited to be teachers and nurses, right?

24 Alistair January 29, 2018 at 11:48 am

> because women are so naturally suited to be teachers and nurses, right?

Well…..there’s clearly some degree of selection-into those fields, just as there is evidence of selection-out from being engineers and offshore fisherfolk. Self-selection effects are clearly real and can be large; they should be controlled for before making claims of residual bias…

I’m happy enough if residual bias can be found, but I do want known confounding factors controlled for before the claim is made!

25 Aaron Luchko January 29, 2018 at 10:24 pm

Well…..there’s clearly some degree of selection-into those fields, just as there is evidence of selection-out from being engineers and offshore fisherfolk. Self-selection effects are clearly real and can be large; they should be controlled for before making claims of residual bias…

I’m happy enough if residual bias can be found, but I do want known confounding factors controlled for before the claim is made!

I’m sympathetic to the claims that there’s self-selection by females away from certain fields, and some of that selection is based on intrinsic sex characteristics.

I’m also sympathetic to the claim that hiring is often biased towards “diversity candidates”.

I also think there’s still a lot of bias again women and minorities in those same fields, both in stereotypes discouraging people from entering that field, but also in the education and work environments. Both because those stereotypes cause their contributions and competency to be underestimated, but also because they get excluded from helpful social groups at work.

26 carlospln January 28, 2018 at 6:05 pm

Thread winner

Reply

27 Hazel Meade January 29, 2018 at 1:54 am

LOL. I hate having to state obvious things to remind people of obvious things.
By hate , I mean kind of love, because it’s really easy.

Reply

28 Anon7 January 28, 2018 at 1:05 pm

The tired social construction claim. Men are naturally more likely to be abstract in their thinking and spirited, among other things, so expect more men to end up in positions like academic science (especially math and physics as opposed to biology) and politics where it’s an advantage to possess those qualities.

Reply

29 Anonymous January 28, 2018 at 2:17 pm

It was an obvious troll, but enjoy it if you will.

If it were real, I might link something like Male monkeys prefer boys’ toys

Reply

30 Anon7 January 28, 2018 at 2:36 pm

It’s a surprisingly common view, especially among academics, who complain the loudest about the War on Science. Steve Pinker has a good summary of the mountain of studies that refute the view that it’s nearly all social construction.

Reply

31 Anonymous January 28, 2018 at 3:15 pm

When I first read Pinker it felt odd because he was arguing against people and views I had never met. I still see these referred to far more often than I encounter them. Stock stories of feminist moms disappointed that daughters want to be Disney Princesses, etc.

http://www.pewsocialtrends.org/2017/12/05/on-gender-differences-no-consensus-on-nature-vs-nurture/

32 Alistair January 29, 2018 at 6:12 am

The monkeys have clearly been impacted by the Patriarchy. They need to be re-orientated and given gender sensitivity training.

Reply

33 Slocum January 28, 2018 at 2:12 pm

Wow — are you joking or are there really people who still believe this? If so, where can I find this culture where men and women display no differences in their interests?

Reply

34 Alistair January 29, 2018 at 6:30 am

Cathy,

It’s a wonderful tribute to human faith that, at your age, that you can continue to believe in such things, without the slight shred of statistical support. Just keep repeating the shibboleths!

Reply

35 Alistair January 29, 2018 at 6:31 am

Of course, the original post was a satire, and I’m an idiot 🙂

Reply

36 Words have meaning January 28, 2018 at 11:28 am

A strong can be made that the term “positive discrimination” should not be used.

It is inhumane to treat people as members of a group rather than as individuals.

Reply

37 clockwork_prior January 28, 2018 at 11:54 am

‘It is inhumane to treat people as members of a group’

Nothing more evil than citizenship, then? You know, where the person looking at your passport cares about your nationality more than you as an individual? Including deciding whether are allowed to legally pass a border to begin with.

Reply

38 Spock Talk January 28, 2018 at 12:03 pm

Intergalactic Studies shows that planets without nations have much lower crime rates and much higher incomes than planets that group people by citizenship and divide people by borders.

Reply

39 Kris January 29, 2018 at 5:14 am

Citizenship was a concept created for the purpose of mass military mobilization to fight enemies. If we all stopped fighting, we wouldn’t need citizenship.

Reply

40 Alistair January 29, 2018 at 6:15 am

Really, Kris? Does citizenship not impact on a dozen aspects of your life in terms of your taxes, benefits, and culture of your fellow citizens?

Reply

41 Careless January 28, 2018 at 12:55 pm

Assuming you meant to put “argument” in your post and forgot to, what does your second sentence have to do with your first? It seems completely irrelevant. Or are you unclear on what “positive” means?

Reply

42 Jack pq January 28, 2018 at 11:44 am

Could be a Hawthorne effect at play? Did the participants realize they were under scrutiny, and might they change their behaviour to please the experimenters?

Reply

43 Alistair January 29, 2018 at 6:18 am

Possibly, but they weren’t told purpose of the experiment. And a few questionnaire controls suggested that they were typical. Needs a random sample, but hey, its part of a wider cascade of negative evidence in the “bias” literature.

Reply

44 rayward January 28, 2018 at 11:47 am

There’s a very large not for profit hospital chain in the U.S. affiliated with a conservative religious organization that has a definite positive gender bias in hiring. It seemed strange to me given the religion’s very conservative view of the role of women. I discreetly asked why all of the women in key roles and the response was both encouraging and discouraging: Women are more loyal. One could interpret that in several ways, including that women know their place (i.e., subservience). Even it that’s true, should a woman complain because they hired her for the wrong reason? Life is complicated.

Reply

45 TMC January 28, 2018 at 1:08 pm

Loyal, in hiring terms, means they will stay with the organization for a long time. Part of men’s increased pay is that they are willing to job hop more often than women will. Men who stay at a company a long time are often under paid as well.

Reply

46 Guy Makiavelli January 28, 2018 at 12:17 pm

Here is the solution for REAL diversity:

Joonko helps your employees and managers to identify and overcome their hidden bias in real-time……Our Machine Learning and Artificial Intelligence algorithms will constantly evaluate existing data to identify events of Behavioral bugs or Cultural Typos…. Real-time actionable insights. When needed, Joonko engages executives, managers, and employees with insights and recommendations – so they can make real-time corrective actions.

http://www.joonko.co

Reply

47 Alistair January 29, 2018 at 6:22 am

Oh, this will be Golden.

I can’t wait for the next generation of AI recruitment algorithms to start producing the “wrong” answers. Then liberals will start shouting about how waycism and sexism got into the code….

Reply

48 Hazel Meade January 28, 2018 at 12:41 pm

Australian Public Service

What are the demographics of the Australian Public Service?

Reply

49 Hazel Meade January 28, 2018 at 12:50 pm
50 edgar January 28, 2018 at 2:24 pm

John Maynard Smith understood this behavior…

Reply

51 Jackson Monroe January 28, 2018 at 2:27 pm

The only things better than point estimates are point estimates with no standard errors!

Reply

52 Stefan January 29, 2018 at 7:29 am

+1

Reply

53 Roadrunner January 28, 2018 at 9:05 pm

What the hell is “positive discrimination?”

Otherwise known as discrimination against white males.

Reply

54 Careless January 29, 2018 at 3:39 pm

No, don’t confuse the appalling concept of “reverse discrimination” with “positive discrimination”

It just means discrimination favoring someone. Could be anyone, including a white male.

Reply

55 Sure January 28, 2018 at 10:58 pm

Being a minority in one of those heavily gate kept professions with lots of evaluations along the way, I have never had it work against me. In general every interviewer from med school to the present was thrilled to have a minority candidate to interview. Once I establish a few quality indicators (e.g. military service, class rank, credit score, lack of malpractice history), I basically need to kill someone to not be highly competitive.

Where I think things can hit against you is if you cannot escape stereotypes with some degree of statistical backing – then it I think you can get downgraded. For instance, banning the box makes it a lot harder for minority patients of mine to get quality work if you cannot check directly who is or is not a felon. Likewise, when I sat in for some residency meetings, foreign applicants who could demonstrate truly native level English fluency and a full understanding of expectations in the US medical system did vastly better than applicants where we just were not sure where they fell.

Frankly I suspect that over 90% of the anti-minority discrimination that does happens, happens only because people lack the ability to verify that the negative stereotype is not true of this applicant.

Reply

56 mpowell January 29, 2018 at 12:06 am

This is the problem with the ‘equivalent resume but black sounding name’ series of studies. But to a lot of people, allowing a decision to be influenced in any way by a stereotype, even a statistically accurate one, is racism.

Reply

57 Alistair January 29, 2018 at 6:25 am

Yes, obviously,

From a purely rational, statistical, perspective, [insert minority]-name is a negative but limited indicator that only applies in the absence of more detailed information. If you have the more detailed information, the decision value of the [insert minority] status falls to zero.

Reply

58 Anonymous January 29, 2018 at 9:21 am

It is also “yes, obviously” from the standpoint of this is prejudice:

Having a negative stereotype of a huge class of quite varied people, and then putting it on them to disprove it.

If you want “a purely rational, statistical, perspective” the in-group variance vastly exceeds the difference in means, between groups.

Reply

59 Hazel Meade January 29, 2018 at 10:29 am

Both excellent comments. I would say people have some obligation to seek more information before falling back on the [insert minority] stereotype. Especially in hiring decisions, where it should be pretty easy to acquire more detailed information.

Reply

60 Alistair January 29, 2018 at 11:41 am

Yes. The optimum effort to spend searching for more information can be determined too. Basic decision theory.

61 Alistair January 29, 2018 at 11:37 am

Firstly, I dispute your factual claim. In-group variance is DOES NOT vastly exceed cross-group variance in many real world contexts (e.g. factor of 10 or more).

Secondly, I dispute your claim of prejudice. From decision theory, ceteris paribus, it’s entirely the difference in means that matters. There are ways to wriggle about this, with a risk-averse utility function, strange distribution of variances etc. But let’s not pretend you’re really arguing the clever technical exceptions; you’re just plain wrong about basic operations analysis.

Would you like to have a betting game where we put our competing selection rules to the test?

Reply

62 Anonymous January 29, 2018 at 11:48 am

You have gone so far off the deep end now.

Negative stereotypes, the kind that stop you from selecting someone from “a group” imply a greater than 50-50 chance that “a member of the group is bad.”

Can you name any “maligned group” for which that is true?

Or are you making the very bigoted argument that a minority chance is sufficient for you to malign the whole group?

63 Alistair January 30, 2018 at 6:03 am

>> Negative stereotypes, the kind that stop you from selecting someone from “a group” imply a greater than 50-50 chance that “a member of the group is bad.” Can you name any “maligned group” for which that is true? Or are you making the very bigoted argument that a minority chance is sufficient for you to malign the whole group

It’s a selection Rule, Anon. You’ve just changed the question from “Is selecting A better than selecting B” to “Is B good?”. Are you aware you did that?

If groups A and B have means such that u(A) = u(B) + k, and std dev rk, for r<10 (ie. intergroup variance are non-negligible compared to group mean differences), then for normally distributed populations you would be stupid to not pick A all the time over B if you wanted to maximise the mean of a selected group. It is very unlikely, with a selected group of say, 30 persons, that the mean of selected group of B's would beat a similarly selected group of A (86% chance for r=10). Central limit theorem.

Of course, if you have additional information about individual members of A or B, ("B(i) is in the upper 50% of B") then some B may be selected preferentially to some A. But otherwise, it is rational (you WILL maximise your score) by just choosing A all the time.

This is probably novel to you. When you see something you don't understand you should take time to read it carefully rather than suffering an emotional response.

64 Alistair January 30, 2018 at 6:04 am

Typo: I transposed A and B in the outcome of the comparison. A group of A’s will beat a group of B’s 86% of the time in the example given.

65 Alistair January 29, 2018 at 6:26 am

It’s all basic Bayesian stuff.

Reply

66 Anonymous January 29, 2018 at 9:22 am

A real Bayesian would have the variance, uncertainty, preloaded.

Reply

67 Alistair January 29, 2018 at 11:39 am

Too many basic mistakes, anonymous. I don’t think you’re competent in statistics.

Reply

68 Anonymous January 29, 2018 at 11:52 am

I know the behavioral experiments with bags and colored balls.

You put 100 balls in the bag, 50-50 red and green. You let people draw 10 balls, and then ask them the ratio and their confidence. They are of course wrong and overconfident.

Just like bigots.

69 Alistair January 30, 2018 at 6:07 am

Crude Ad hominems aren’t effective against high IQs. High IQs know that. But mediocre IQs don’t.

70 blah January 29, 2018 at 12:17 am

For a moment I was shocked to find this article here. Then I found that it was written by Alex, and life became normal. 🙂

Reply

71 john January 29, 2018 at 9:20 am

Would be interesting to know the makeup of the participants and to look at the driver behind choice when the candidate’s characteristics were not know. Was it purely “skill set” or perhaps something more subtle in the communication and presentation styles.

Reply

Leave a Comment

Previous post:

Next post: