This is based on a similar earlier MIT study which sent out the same resume with black and white names and found that there was a clear and statistically significant difference in response rates (Employers were uninterested in Tawana, but they were interested in Terri).

It was a very nicely structured experiment that managed to control for many of the confounding factors.

This is based on a similar earlier MIT study which sent out the same resume with black and white names and found that there was a clear and statistically significant difference in response rates

Yup, Marianne Bertrand of Chicago's Graduate School of Business and Sendhil Mullainathan of MIT. The paper can be downloaded here. Anyway, while discrimination is awful, I have a hunch that S.Asian/E.Asians/Blacks/Arabs/Latinos are far more likely to discriminate against Whites, as well as other minorities, than Whites are likely to discriminate against them.

The "Berkeley Discrimination Center", huh? Seems like the institute is out to affirm the consequent...

Guess this horrific discrimination explains why East Asians, South Asians, and Arabs have incomes significantly higher than the national average.

oh, wait a sec.... :)

by the way, the bertrand and mullainathan study lacked some rudimentary controls in their rush to blame whites...truly pathetic coming from MIT.

missing controls:

1) checking what black employers did to white applicants. I'd bet black employers practiced discrimination as well, probably in favor of blacks (though it's an empirical question).

2) including east asians in the applicant pool. They didn't control for the affirmative action effect, which makes a black & a white with equal paper credentials (say, a CS degree from State U) unequal in the real world b/c the black had to pass a lower threshold (e.g. in terms of SAT scores) to get that credential.

That discounting effect wasn't included...but would have been controlled for had East Asians been included.

Gotta read this study in full, but while they seem to have done the 2nd control, they haven't done the first.

Guess this horrific discrimination explains why East Asians, South Asians, and Arabs have incomes significantly higher than the national average.

Yeah, but that by itself doesn't imply an absence of discrimination. You also need to control for education levels for it to be meaningful. Consider the UK where Indians have higher education levels than Whites, yet rake in less than their White counterparts. And we aren't talking about unassimilated brown hordes here, but mainly second/third generation desis.

GC:

Looking at what black employers did to white applicants is a different question. They sent the resumes to large corporate employers, but didn't try to spin them to ones with black CEOs or white CEOs, so your first point is off base.

Your second, I grant a bit, but not entirely. Much of that should be controlled away by the pattern of achievement on the resume. AA gets you in the door, but it doesn't do the work for you.

Also, given the numbers involved, which were large but not huge, the statistics were clearer with a 2-way split.

They were hardly in a rush to blame whites -- they had a neat experiment designed to ask a question. I heard them on the radio -- they were surprised by the results, and were agnostic about what they meant.

AA gets you in the door, but it doesn't do the work for you.

AA/racial preferences means 200 points on the SAT, as both the released scores at UMich + the data in Bowen and Bok's book reveals.

In fact, racial preferences are more than that...because they also impact admissions to *grad school* as well as job hires (b/c of constant EEOC lawsuit threats). It is not as if the only racial preference is at the gate to undergrad. Instead, racial preferences are active at every level of academia.

In other words, racial preferences are equivalent in effect to race-norming the grade distribution in undergrad, as they impact grad admissions as well as all hires out of undergrad and beyond. Ever read "Diversity Inc."?

It couldn't be otherwise if you wanted to maintain the rickety illusion of h-bd denial - the first *real* merit threshold would cause step-function drop off in "diversity" (= non-underrepresented minority presence).

The only reason AA can survive is because all the quantitative metrics are *intentionally* kept secret. For ex. on the Berkeley website, you can get all kinds of breakdowns of the undergrad populatoin on geographic distro, ethnicity, gender, SAT scores, etc....including SAT scores by gender.

But they've kept SAT scores by ethnicity off the website or buried it somewhere...though it would clearly be trivial to tabulate given the relational db that they surely have set up behind the scenes.

Bottom line: you are fooling yourself if you think Harvard or MIT would be even 10% "underrepresented minority" without racial preferences. You can do the calculation using SAT score distro as a proxy. If the admitted blacks and Hispanics were forced to have SATs comparable to the rest of the class, their combined proportion would be around 1%. See here. Do you seriously deny this? I would suggest checking the ETS score distros first, and then comparing it to the US News reported median SAT scores before challenging this, b/c it is pretty easy to show...

They were hardly in a rush to blame whites -- they had a neat experiment designed to ask a question.

Their experiment wasn't "neat"...it lacked basic controls. And they set up the study with the following tacit assumptions:

1. racism = irrational hatred, rather than rational statistical discrimination on a hidden variable (e.g. SAT scores) using race as an imperfect proxy (thereby giving the false impression that "racism" means employer reactions to blacks & East Asian resumes will be the same).

2. Only whites can be racist.

And let's just say that peer review is just a TAD more lax for the latest bogus discrimination study (whether it be stereotype threat, post-traumatic slavery disorder, or name-bias) than it is for any kind of study on IQ.

Suddenly MRI measurements of brains and correlations of .7 become controversial when you talk about IQ...yet the sort of uncontrolled rubbish that Bertrand and Mullanaithan published gets a credulous rave in the WSJ. Oh well...the days of h-bd denial are numbered anyway...you just can't deny base pairs and brain scans.

Consider the UK where Indians have higher education levels than Whites, yet rake in less than their White counterparts.

That is not true:

http://www.jrf.org.uk/knowledge/findings/socialpolicy/sprN48.asp

The largest group analysed by the research consisted of families below pensionable age with at least one worker. Chinese and Indian working families averaged slightly higher earnings than white people. Overall, Caribbean and African earnings were significantly lower than whites', though this was not true for black women. Pakistani and Bangladeshi families' earnings were much lower than those of any other ethnic group - partly because of low wages, but also because relatively few married women in these groups had a job.

That is not true

Err, not really. My point still stands. Go through this paper if you want to get a clearer idea of what the British labor market is like. Average earnings of Male full-timers per week: Whites - 331 pounds, Indians - 317 pounds. Reason : In a nutshell, Indians have larger families with a higher proportion of working-age members. That's why "Chinese and Indian working families averaged slightly higher earnings than white people" .

Also, look up Chart 1.3: Average weekly earnings of full-time men and women employees. Both Indian males as well as Indian females earn less than their White counterparts.

GC:

You astound me sometimes. In science you don't make up an experiment to study "evolution" or "relativity", you make up a study to examine one of the observable implications of that theory.

They weren't studying racism (which is a large topic), they were looking at whether there was discimination in hiring, particularly at the early phase. For this reason, they sent resumes to the large companies. They never denied that there might be racism by blacks. There might be racism by Inuit also, but it doesn't have much of an affect on the job market outside of small villages in Alaska.

Secondly, you're probably the only person who would hire based solely on SAT scores. Most employers hire based on experience. If that's true, then the thing to control for would be experience in the resume. It would have been easy enough for them to add an SAT score into the resume, but that would have looked strange, since most employers don't ask for it, and it would have introduced a difference between experimental resumes and the rest of the resumes being received.

Even if somebody was admitted under AA, as I said, that gets them in the door, that doesn't account for their performance. I haven't seen the sample resumes, but most of the accomplishments on an average resume are things like exceeded sales quota 3 years running. That has nothing to do with AA.

(You're also presuming that there are no quotas for white folks either. While usually smaller, some universities reserve up to 25% of their pool for legacy admits. Similarly, geographical diversity clearly benefits white folks. A student from Idaho or Montana can have a much lower SAT than a student in San Fransisco, and still get into MIT or Stanford. If you control of income as well, the gap isn't as large as you're making it seem. Personally, I'd be suspicious of people with waspy names with numbers after them. Trustafarians are very lazy.)

I know that you think that SATs proxy for intelligence, and that intelligence determines job prospects and economic growth on a national level.

However, I've tought SAT test prep. The company I worked for guaranteed a minimum 100 point improvement (this was on the old test), and I personally saw a reasonable number of students go up 200-300 points just by learning how to take the test and some very targetted studying. That would make it a poor proxy for intelligence, even assuming that all intelligence can be compressed into a single dimension.