New results on the economics of discrimination

There is a new and excellent paper by Uri Gneezy, John List, and Michael K. Price (gated version here, $5 charge for non-university readers), here is one good excerpt:

…in the automobile repair market, we find that the disabled receive offers that are 30 percent higher than the offers received by the abled.  One possible reason for this disparate treatment is search cost differences — one would expect search to be more costly for the disabled.  Under this scenario, such agents might search less and repairmen might capitalize on that fact.  To test this conjecture, and identify the underlying nature of discrimination, we employ a complementary field experiment where during the offer process all agents note that “I am getting a few price quotes.”  Upon having our agents make this simple statement, we find that offer distributions become isomorphic.

The paper has much more of interest.


I'd be interested in knowing the rural/urban distribution of the repair shops.

That is incredibly useful to know -- and surprising. I would have expected that auto repair would usually cost more for the disabled, purely because the use of a car is more inherently valuable to many disabled people than it is to the able (I hate using that word, but until a less judgemental-sounding synonym crops up, I'll have to stick with it).

I think car repair quotes are a bad test case. The variation in quotes between shops swamps their effect of interest.

e.g. For one of the cars (Table-1), the same disabled person was quoted a low quote of $3000 and a high of $4800. With respect to this large variation the disability penalty of $75 is peanuts.

Statistics, my boy. Or, as my statistics professor used to say "testa statsiticles"

They probably need all the statistical ammunition they can get. One of their surveys has a sample size of six.

Andrew's Corollary

There may not be a GOOD literature on everything.

I have to agree this is the sort of paper that is great for one's career but of very low value.

Now here's a real discrimination story:

Aren't you guys confusing discrimination with bigotry? You guys new to the language?

Isn't it funny that the discriminated can do something simple that makes the discriminators no longer discriminators?

Doesn't the result just imply that mechanics are less likely to cheat if they think their customers might be shopping around, regardless of the reasons they thought they could cheat to begin with? Maybe they assume the disabled are more reliant on their cars. Maybe they think their less likely to stand up (sorry) for themselves.
Also, "isomorphic"? Is there something wrong with "identical"?

"isomorphic" is simply wrong.

this is more interesting

"we find that when the discriminator believes the object of discrimination is controllable, any observed discrimination is motivated by animus. When the object of discrimination is not due to choice, the evidence suggests that statistical discrimination is the underlying reason for the disparate behavior"

the post is slightly misleading - they are considering a number of dimensions of discrimination.

i wonder what they mean by "controllable object of discrimination", since all the dimensions they consider (race, age, gender, sexual orientation and disability) are not really "controllable". unless you're faking it. but how do you know?

this isn't empty critique. for policy, statistical discrimination is a much harder issue to figure out than animus. or to put it another way, statistical discrimination could even be the right thing to do.

also, it isn't easy for many to do the field experiments the authors conduct; one needs significant amounts of funding to do this. given that most students who are writing dissertations are unlikely to get that level of funding, I'm confused as to how this can form a viable template for good research for the younger lot. \end{rant}

the only plausible controllable object of discrimination - they list race, age, gender, sexual orientation and disability - is sexual orientation. but that is objectionable. I can't see how any of these could be controlled.

Funny how the paper is itself a case of the phenomenon in question.

Comments for this post are closed