SES and the gap: It’s worse than that

Paul Kersey has a hilarious post on SES and the gap. Here’s a snippet:

Having compiled numbers that show the startling amount of money, time, and effort that the educators and political leaders in America have spent trying to close the racial gap in learning, this young member of Chinese intelligence – who studied at the University of California on a scholarship provided by American tax-payers – believes the US will bankrupt itself before it admits racial differences in intelligence, which could ultimately avert its decline and fall.

Addressing his audience of some of the most important people in China, he reads from a study:

For both blacks and whites, family income is one of the best predictors of a student’s SAT score. Students from families with high incomes tend to score higher. Students from low-income families on average have low SAT scores. Because the median black family income in the United States is about 60 percent of the median family income of whites, one would immediately seize upon this economic statistic to explain the average 200-point gap between blacks and whites on the standard SAT scoring curve.

But income differences explain only part of the racial gap in SAT scores. For black and white students from families with incomes of more than $200,000 in 2008, there still remains a huge 149-point gap in SAT scores. Even more startling is the fact that in 2008 black students from families with incomes of more than $200,000 scored lower on the SAT test than did students from white families with incomes between $20,000 and $40,000.

The room falls silent, until a small chuckle can be heard in the back. Someone clears their throat, but it’s too late; the entire room breaks out into laughter.

It’s actually worse than that. The gap doesn’t merely persist with SES, but — as with the heritability of cognitive ability, incidentally — it generally increases. The SAT SES data Kersey cited was somewhat of an anomaly. This phenomenon of increasing magnitude with SES puts the lie to the frequent claim that controlling for SES drastically reduces the gap. I’m sure you’ve heard it. The claim is accurate, but deceptive. Because Blacks have a disproportionately lower SES than Whites and as the gap itself is generally positively correlated with SES (See: Jensen 1973, 1998; and the graphs below), when scores are regressed on SES, one gets a larger reduction in the gap’s magnitude than one might otherwise anticipate. Raising the SES of Blacks relative to Whites will reduce the gap overall, of course, but it will also increase the gap relative to Whites of the same status.

Here is a graph from Murray and Herrnstein’s “RACE, GENES AND I.Q.-AN APOLOGIA“:

Anyways, I created some new RACE by SES graphs.

This entry was posted in Uncategorized. Bookmark the permalink.

11 Responses to SES and the gap: It’s worse than that

  1. JL says:

    One factor that inflates the b-w gaps by SES is that controlling for IQ blacks are more likely to go to college. If entry to colleges was race-neutral and always competitive we would expect the SES-controlled b-w gaps to be smaller.

  2. Steve Sailer says:

    Does that database you used for the “black net advantage” let you look at whether Hispanics are more or less likely than whites, all else being equal, to go to college? I suspect it may differ over the economic cycle: a lot of Mexican youths were dropping out during the Housing Bubble to work construction, but now they are trying their hand at junior college.

    • Chuck says:


      The Black Net advantage data base is the National Longitudinal Study of Freshman. It’s an elite college sample, so not nationally representative. Even if it were, it samples college Freshman, not High school seniors, which is what you would want. For your question, you could check — or have Inductivist check — as he’s quite familiar with it — the ADD health sample (Wave 1/Wave 4). (It takes a while to figure out the relevant variables.) I’m not sure what you mean by “all else equal,” but you would at least want to control for Language learner level, IQ, nativity status, and parental education.

  3. Matt says:

    I wonder why the API advantage on maths is so much higher by Grade 12 than at Grade 8 while the rest of their relative ability remains constant….

    It looks like all the ethnic minority graphs are flatter than the White graph – in the API case probably because many intelligent individuals did not get much education, while in the Black case probably because many unintelligent individuals did (are Hispanics more likely to follow the API or Black pattern?).

    • Chuck says:


      Good points. I don’t know why the slopes are different. It would be informative to look at Asians separate from Pacific Islanders. The new 2011 classifications separate them. But the results are only out for 4th and 8th graders. Perhaps you could check the data and let me know the effects of treating Asians separately.

      • Matt says:

        Sure. The Grade 4 data does not have the parental educational status data, but I’ll have a shot with the Grade 8 data.

        Maths –
        Reading –

        All the non-White groups, aside from the Asians, seem to retain their condition of being almost flat between the categories of parents who did not finish and did finish high school (no further than high school) and of the categories of getting some post-high school education and of graduating college (further than high school), with a step between no further than high school and further than high school.

        Seems like whether non-White, non-Asian (NAMs) parents graduated high school isn’t connected to their children’s ability, but their taking some form of post high school education is, but whether that is college or not isn’t too important.

        Asians seem to be like NAMS in that whether their parents graduated high school or not has no relation to their ability, but are like Whites in that college graduate parents have more able children than those that have only taken some education after high school.

        You could infer from that there isn’t really an IQ difference between high school drop outs and non-high school drop outs amongst non-White groups, and that possible other factors dominate whether or not these people will graduate or not. Though many of these parents are immigrants so that might not hold in the USA, particularly for Asians and Hispanics – for Blacks it seems like it holds true in the USA (since they aren’t very immigrant).

        It also looks like for NAMs, their intelligence has not much to do with whether they go to and graduate college or get some form of education after high school – once they’ve reached the same threshold it doesn’t matter. Although that might be a facet of an overlap between the categories of “some education after high school” and “graduated college” (apparently very few intelligent Blacks do not go to college – And may be due to some degree to stronger regression to a lower mean (my intuitive math sense is not good enough to tell me whether this is a bad explanation).

        NB, an Asian Maths gap with Whites, relative to your graphs, seems to come into existence once Pacific Islanders are broken out, but an Asian Reading gap does not emerge and Asians more or less stay where they are. Not sure what to make of this – i.e. if PI, who score about the same amount lower than Whites in both Maths and Reading, were depressing the Asian Maths score, surely they would also be doing the same in Reading? (Again my intuitive math sense may be failing me here).

  4. Chuck says:

    Matt, thanks. I’ll post the graphs when I get a chance.

  5. Chuck says:


    Which race variable did you use? Did you use school identified race?

    • Matt says:


      Yes, variables SRACE10 and PARED.

      I missed that there was a self assessed variable (DRACE10) present on the 8th Grade data when checking, so used school report by default.

      The variables seem to be in more or less rough agreement. There seem to be like 1 or 2 point differences between the two variables in each category in Math, which seem to leave Whites and Asians and Natives slightly smarter (by around 1 or 2 points, slightly more pronounced for low education Natives), Hispanics and Blacks largely with no change and Pacific Islanders slightly duller or unchanged. There is slightly larger difference between the two variables in Reading (but still largely less than 5 points either way in each parental status-race category in the cross tabulation).

  6. Pingback: The magnitude of the BW gap with parental education in the NLSY97 | Human Varieties

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s