In my search for a new Dillow Towers in Oakham, I'm struck by the fact that several apparently attractive places are for sale very close to each other. This raises three possibilities:
1. The smart money - home-owners - know there's something wrong with the area. So I should avoid it.
2. This something might actually be something that doesn't trouble me, such as a desire to move near better schools.
3. It's just the sort of bunching that happens under random conditions.
Now, the mere existence of points 2 and 3 show that 1 is only a probability - and maybe a low one. But my problem isn't: which is right? It's: am I prepared to stake thousands of pounds on point 1 being wrong?
This bears on the question of statistical discrimination. Imagine two equally qualified 20-somethings, a man and woman, apply for a job with a small business. If the employer knows mere statistics - women are more likely to get pregnant than men, costing their bosses time and money - he'll be disinclined to hire the woman.
What's the solution? Bryan says the woman should use counter-sterotypical behaviour to stand out from the average woman. She could wear tweed and smoke a pipe to signal that she won't get up the duff.
The problem is, though, that such behaviour must send a signal strong enough to overcome two barriers. One is the employer's Bayesian prior that women are expensive employees - a prior that, in imperfect markets, might be held more strongly than rationality would warrant. The other is risk aversion. The boss might figure as I do in house-buying: "there's only a small chance this woman will get pregnant, but why bet thousands on this?"
Which raises the question. Is it really feasible for people to adopt sufficiently strong counter-stereotypical behaviour to overcome these barriers?
Bryan says it is, because black men who get an education enjoy big pay-offs. I'd add that lesbians do seem to earn more than heterosexual women.
But are these examples scalable? For example, some of the high pay-off to being black might come precisely because highly educated black men are so rare; a black man who made it from the ghetto to Harvard is signalling that he has remarkable qualities. If loads of black men made this journey, the signal would be greatly weakened. The same would be true if more women sent lesbertarian signals.
So, is statistical discrimination really so easily overcome?
... or, in the knowledge of litigation statistics, the employer will be more likely to hire the woman because (s)he knows that hiring the man has a non-zero chance of a discrimination charge being filed.
... or, in the knowledge that OTHER employers have an anti-woman bias, the smart employer will hire the woman on the basis that the imperfect market makes educated women undervalued, and hence there's an arbitrage option.
Both of which are my way of saying that I believe the COMPLEXITY threshold is higher than the bias threshold in many markets.
Posted by: Mark Harrison | February 14, 2008 at 04:18 PM
Is it really feasible for people to adopt sufficiently strong counter-stereotypical behaviour to overcome these barriers?
More importantly, is the signal reliable? Most people are capable of lying, after all.
Posted by: ad | February 14, 2008 at 06:59 PM
Sell your current place and then rent in Oakham until you've learnt the lie of the land.
Posted by: dearieme | February 14, 2008 at 10:09 PM
1. Do more research in Oakham before buying
2. Reject women of child-bearing age as employees, but make sure that you have the 'documentation' to back this up. (It's easy when you try, if a drag.)
Posted by: paul ilc | February 14, 2008 at 11:04 PM
I see some role for government here: the provision of childcare and leave, paid for by everyone's taxes, would be one way to remove that particular discrimination, which would otherwise be rational.
The black/white one I'd like to think will wash out, those with preferences beyond competence will have to overpay and thus fall behind.
Posted by: improbable | February 15, 2008 at 01:56 AM
CD is trying to explain something that doesn't happen.
Well, let me backtrack a bit. Men and women are different, physically and (or course) behaviourally. So with the same enviroment, we should expect different outcomes including salaries (just as if you control for genes and environment, men will be heavier than women, and taller).
That aside, of you control for relevant variables that affect salary (which is very seldom done) then this compares men and women who share the same employment-relevant characteristics (such as aptitudes and motivation) ... then men and women are on average paid the same
(or women are paid slightly more).
Those are the facts - but there is a heck of a lot of deliberate disinformation around. For example (as Tim Worstall frequently points out) official statistics that people comparing part time women with full time men, or comparing people with different levels of qualifications and so on.
(It's almost asif they *want* to have a gender gap, so they can sell their patent remedies...)
See Why Men Earn More, by Warren Farrell for a summary.
Posted by: BGC | February 15, 2008 at 06:22 AM
Where I live, there were two attractive houses offered DIRECTLY opposite one another. Then a month of so later, someone was nearly murdered in the same street. Maybe a case of #1. (i.e. Maybe the problem is very LOCAL like an impossible neighbour.)
Posted by: reason | February 15, 2008 at 09:41 AM
I don't suppose that it costs a lot these days to get rid of an impossible neighbour. So buy first and then get him bumped off.
Posted by: dearieme | February 15, 2008 at 11:12 AM
Getting back to your own dilemma, the answer is that you should sell-to-rent.
Posted by: Mark Wadsworth | February 15, 2008 at 12:13 PM
I think that overcoming social stereotypes is not statistical as much as it is psychological. Stereotypes are often wrong or misleading, and generalizations do not usually apply to specific circumstances. Generalizations and statistics usually give information only about group trends but tell very little about a specific case--assuming there is even a tiny amount of specific information about the specific case. But to overcome stereotypes you needn't convince the boss through mathematics, you must overcome his psychological assumptions--be they incorrect or correct.
Posted by: Scott Hughes | February 16, 2008 at 04:33 AM
Scott, ...overcoming the boss's correct psychological assumptions. Good luck with that.
What signal does it send for a man to say he needs time off to look after a sick child. I have done this with a boss who never hired women for this sort of reason. It wasn't pretty. And did my wife get any credit in her job for not taking time off? Unlikely.
Posted by: Joe Otten | February 19, 2008 at 06:01 PM
While the practice of billing by the hour has been long debated in the legal world, a group of lawyers has recently called for an end to the“ billable hour” for a new reason— because the practice disproportionately harms female lawyers. Women, they say, become“ discouraged” by the fact that they simply cannot work the long[…]
Posted by: employment discrimination senator email | March 20, 2008 at 04:47 PM
I also face the risk aversion & statistical discrimination I should avoid it. Thanks for your helpful tips. :)
Posted by: Green Palmer | November 30, 2010 at 04:35 PM