Group Differences Visualized

Last week, John Derbyshire wrote an article on race relations that led to him being removed from the list of contributing writers for National Review Online. Mr. Derbyshire made a series of factual claims, which were correct, and then proceeded to interpret these factual claims to give specific advise on how an individual should comport himself of members of other races.

Derbyshire’s column was widely criticized, but I have found only two examples of pieces which criticized the his factual claims, as opposed to his advise in the face of those claims. One was the short piece “By His Own Standards,” by Jason Lee Steorts that I previously discussed.

A much better criticism was written by Noah Millman in his article for The American Conservative. Noah states:

I’m all for frank talk, and not hiding behind politically-correct shibboleths. But I’m not for lazy talk, and I’m not for talk that implies a complete lack of concern for the social costs of one’s views.

Noah is getting at a point about racial differences that I previously mentioned

People talk of differences in “averages,” but this is a misleading way to talk. The difference between the 50th percentile and the 53rd percentile in extroversion, for example, is likely to be barely noticeable.

Rather, average differences matter in the extremes.

Here is an example of that, courtesy of Razib Khan. Using one measure of verbal intelligence, the average differences between blacks and whites are very small. But in the extremes this leads to a major under-representation of blacks among the very intelligent:

This is in contrast to sexual differences in intelligence, which basically do not exist:

and also in contrast to educational differences in intelligence. The average intelligence of the highly educated is greatly different from the average intelligence of the non-highly educated. This also shows up in the extremes.

And in contrast to Jews, who are highly intelligent compared to other groups. The average intelligence of Jews is greatly different from the average intelligence of non-Jews. This also shows up in the extremes.

These charts were created from data in the General Social Survey, using the WORDSUM variable as a a rough-and-ready measure of verbal intelligence. There are much better measures available, but GSS has the advantage of having asked the question to very large numbers of people and being easily accessible from a web browser.