In Defense of Legacy Admissions, Sex Discrimination and Racial Discrimination at Elite Universities

In a perfect world, people would undergo extensive cognitive, genetic, and neural testing at the age of about 15, and those results would become publicly available. But this world does not exist, so universities select their students for academic and nonacademic performance, and employers select their employees for educational attainment and a few other variables. There are also elite universities which attempt to select for the best students - these selection mechanisms are fairly effective, but they will never be perfect. Because of this, they should select their students based on other variables, namely:

Race

The base rate of eminence within each race, even after controlling for intelligence, is not the same. The relationship between national intelligence and billionaires per capita is strong (r = .69), but not perfect.

billionaire data from here, NIQs were an older version of this dataset

Hong Kong, Singapore, and Taiwan perform about as well as expected based on their national IQs, but these urban and extremely wealthy areas are likely to attract and create billionaires.

Larger and more representative samples of Asians such as Korea, Japan, and China perform at the level of Turkey, Romania, Brazil, and Uruguay, which score about 15 IQ points under them on tests of intelligence.

It is worth mentioning that the data on billionaires per capita does not come from country of birth, but country of residency. To test whether the amount of Asian billionaires was underestimated due to immigration, I consulted a table which reports the number of billionaires each race produces.

from here, 2013 data

Based on the population sizes of each racial group at this point in time (see appendix), the number of billionaires per million of each race is:

• Jews: 13.8 per million

• Europeans: 0.669 per million

• Arabs/Central Asians: 0.169 per million

• Asians: 0.140 per million

• Latinos: 0.114 per million

• South Asians: 0.0405 per million

• Blacks: 0.00594 per million

Close to the IQ rank order, but Asians stand out as being underperformers. Asian underachievement also holds up for cultural and scientific achievement as well. The distribution of significant figures in science and culture has been shifted towards Europe ever since the start of the 15th Century, as shown by Charles Murray.

from here

Some think that the fact Chinese/Asians study and cheat more on tests makes their academic test results less predictive — it’s possible to empirically test this using regression analysis; unfortunately I believe this has ever been tested properly. There are differential validity studies that test whether the correlation between SAT scores and outcomes differs by race, though that is not the proper way to do the analysis because races may differ in variance in these variables, which may inflate or deflate the variables.

In the College Board’s analysis of the racial bias in the SAT, they found that the correlation between outcome variables and SAT scores are roughly the same within Whites and Asians.

A = Asian, W = White, B = Black, H = Hispanic

Class

SAT scores and grades, while predictive of intelligence and competence, cannot perfectly measure it. If somebody’s parents were able to acquire status and resources, and parents and children relate in traits, then parental class will predict competence beyond SAT scores and grades.

Take, for example, legacy students. Based on a study of 11,837 students at 27 private and selective universities, legacies on average have higher GPAs than normal students in comparison to normal students, but after controlling for parental background, this advantage disappears, and possibly reverses.

from here - the difference is statistically significant based on a prior table

from here

This is a bit unintuitive, but the reversion is because legacy students tend to come from high class backgrounds, and class predicts GPA beyond whatever the admissions process is measuring. This is true even after controlling for SAT scores.

from here

Sex

Attracting the top talent is the most important feature of an elite university, but that’s not their only purpose. Another is to attract students who are wanted there, namely women. Women are unlikely to become extremely eminent - they make up only 2.2% of the significant figures in Charles Murray’s Human Accomplishment.

from here

But the fact remains that the male students want the women there - a lopsided sex ratio would disfavor them in the dating market. And I assume the women want to be there too, after all, they applied. Due to greater male variability, an unbiased admissions process at an elite university would let in more men (but not many more, as greater male variance in intelligence is not large). Instead, sex discrimination in favour of women should be practiced to force the ratio to 50:50.

Final notes

Many will object to having race or class based admissions in principle based on moral intuition. I ask: what distinguishes an SAT score or a grade point average from parental class or race? They are both, in the end, traits that are decided by genetic and environmental factors that predict competence. There is no real reason to select for the former and not the latter.

Some may be concerned that overt race or class based discrimination would harm the reputation of those universities; I agree that it would would. Universities optimize their admissions process for several outcomes: performance, public appeal, and political needs.

Appendix

Race populations (when possible, they were averaged):

from here

note: this table has the more accurate estimate of the Latin American population. The Hispanic population of the USA (~55M in 2013) was also added.

Europeans are overestimated in population size in the first two charts. To adjust for this, I subtracted the underestimated Latin American population (326000000) from the overestimate of the European population (1214375000) to calculate the true European population (888125000) and added the underestimated Latin American population to the original Latin American population to generate an estimate of 660000000. The Jewish population was estimated to be 21 million, a number that accounts for converts and mixed Jews.