African National IQs, redux

[See also: Waiting for Flynn & AfrIQ-Notes]

While searching around I came across this 2006 paper, which provided scores, based on international assessment tests, for 27 African countries, 11 of which L & V (2012) did not have int. test data for (Benin, Burkia Faso, Cameroon, Chad, Madagascar, Mali, Mauritania, Nambia, Niger, Sengal, and Togo), and 6 of which L & V (2012) had neither international data nor IQ data for (Benin, Burkia Faso, Chad, Mauritania, Niger, and Togo).

This allowed me to create the most comprehensive African IQ data set to date (excel file here). Listed are Malloy’s 2008 IQ estimates (N-weighted, international tests excluded),
L & V’s 2012 IQ estimates, Wichert et al.’s 2009 IQ estimates, Altinok & Murseli’s 2006 Int. test estimates, and L & V’s international test estimates. The last column gives the highest score out of the 5 estimates for each country. The average for 40 African nations comes out to 75. It should be noted that the correlations between some of the estimates were pretty low. Interpret that as you will.

While I was at it, I created a global National IQ excel file with Altinok & Murseli’s estimates and others. Here. If you find any mistakes, let me know.


I should add that most of Altinok & Murseli’s and L & V’s international African test data come from fairly recent studies. Here is some background info on the SACMEQ I, SACMEQ II, and PASEC, which supply most of the data points (based on international assessments). They’re from the mid ’90s to mid ’00s. So, it’s not as if we’re dealing with relics here.

About these ads

29 thoughts on “African National IQs, redux

  1. hi chuck, i have been researching this iq and race thing for a while, and have been looking at sites like yours that support the theory that iq differences amongs populations is genetically caused. One thing I have not seen your guys address was the research on stereotype threat when it comes to test scores, what do you think of this http://wicherts.socsci.uva.nl/wichertsST2005.pdf , http://wicherts.socsci.uva.nl/wicherts2005.pdf what about here. Do you think the whole thing about stereotype threat is bogus.

    • Hi Oval,

      We have, in fact, discussed this. See for example, here:
      http://occidentalascent.wordpress.com/2012/01/30/national-iqs-summarized/#comment-1227

      Basically, stereotype threat (ST) is a form of psychometric bias. If score differences are due to ST then they don’t represent true latent ability differences. As it is, one can test if differences do represent true latent differences and therefore test is differences are not due to ST. In the US, at least between Blacks and Whites, such tests have been conducted. The differences are real intelligence differences — whatever the cause — and therefore not due to ST. In other countries or between other groups the situation could be otherwise.

    • Those were pretty small Ns and the sample was unrepresentative. Compare this study to Dolan (2000); Dolan & Hamaker (2001), and Lubke, et al. (2003) which looked at standardization samples. See Lee’s discussion of this in section 2.1 http://lesacreduprintemps19.files.wordpress.com/2011/07/lee-2009.pdf

      See also the discussion here: http://statsquatch.blogspot.com/2011/01/stereotype-threat-and-measurement.html

      ST is a dead end for explaining an appreciable amount of the US gap.

        • basically, publication bias. I imagine that there are hundreds of baby ST studies using good measures of IQ showing a range of effects and that the only ones that get (published and) cited are the ones that confirm the ST worldview. The same happens with early childhood intervention studies. But then you see the meta-analytic results and you say .. oh.

    • You know, the whole idea of ST as an explanation is pretty ridiculous when you think about it. Here are typical findings:

      http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2678742/
      http://psycnet.apa.org/psycinfo/2008-11096-007

      The magnitude of the gap correlates strongly with the general intelligence loadings of a test which correlates strongly with the predictive validity of the test; and the gap can be found on job simulation “test” which merely test how well individuals can perform job related tasks.
      What ST proponents effectively have to argue is that cognitive tests are psychometrically biased against Blacks, yet Blacks just happen to underperformed — for some unidentified reason — as if the tests were not based. And they happen to underperformed on “tests” that are nothing like the typical cognitive ones. And that this strange synchronicity just happens to be ubiquitous:

      N=59 is not quite large enough to support this rather unparsimonious theory, which is, as it is, contradicted on numerous other grounds.

        • Not much.

          (1) The first part of the meta-analysis involved a heterogeneous group and doesn’t allow conclusions to be drawn about race differences in the US. (“The studies included a wide range of participant ages (kindergarten through college), stereotyped groups (e.g., African Americans, Hispanic Americans, Turkish Germans, women), manipulations of stereotype threat.”) It just tells us that test anxiety can — or does on average — influence scores.

          (2) The second part, which dealt with race differences, used heterogeneous measures (e.g., GPA, SAT, ACT) — and the authors didn’t bother to break the results down. The authors conclude:

          “The bias results from psychological threat. It causes measures of academic performance to underestimate the true intellectual ability and potential of ethnic minority students and of women in quantitative fields.

          The results quantify the bias. It is just under one fifth of a
          standard deviation: 0.17 ”

          So relative to the true latent ability gap, certain academic measures are psychometrically biased against some Blacks. But this doesn’t tell us much since we don’t know which ones. GPA, for example, is only modestly correlated with IQ (at 0.5 — i.e.., 25% of the variance is explained by IQ); it’s much more environmentally influenced. And the Black-White GPA gap is typically substantially larger than it “should” be given the IQ-GPA correlation and the magnitude of the IQ difference. Presumably this extra GPA gap is due to motivation or grade anxiety, since we know it’s not due to IQ. So showing that this gap can be reduced by changing levels of anxiety or motivation doesn’t tell us anything with regards to IQ. Since the SAT and ACT correlate with IQ much more highly (at about 0.8 — i.e., 64 % of the variance is explained by IQ) knowing the magnitude of the meta-analytic effect of ST (or motivation — I’m not sure that those have been disentangled) specifically on SAT/ACT would be more informative. But here again the SAT/ACT gap is higher than it “should” be.

          Generally, this study doesn’t conflict with anything I said and insofar as it shows a small effect size for tests for which IQ explains only one to two thirds of the variance, It accords well with my point. Thanks for re-bringing it to my attention though. .

    • That is Walton’s paper not WIcherts. Right? in I read that a long time ago. Did they look for publication bias? Even if ST has an effect on test scores they have not proved it some how decrease the actual latent ability. If it was a major cause of the gap then you would expect minorities to outperform their scores while they actually under perform.

      I think Wicherts found an example where there was a lack of measurement invariance in some dutch minority data. Not sure he has a position on the BW stereotype threat. There was a rumor he had a meta analysis that showed publication bias but that study has not been published.

      • It was more than a rumor. It’s been cited a number of times, for example, here.
        Apparently it’s a rather reclusive paper.

        Anyways, as noted, one problem with Walton was that they were working with GPAs and SATs and the effect sizes were not decomposed by type of measure. GPAs are not the best indices of g. But maybe they have more info in the supp.

    • “Prof Wicherts seems to think that ST does have a serious effect on test scores, have you read his PHD theses.”

      Ya, I read it — and you’re incorrectly characterizing him. His point was that ST affects scores rather than latent ability. As such, it’s a form a psychometric bias which can be tested for by using MGCFA. Whether or not a difference is, in part or full. induced by ST can be determined by factor analysis because ST will alter the relation between scores and latent factors. Right? So the argument is not that “ST does have a serious effect on test scores” but that ST doesn’t have a serious effect on latent ability.

    • I read that a while ago. sounds plausible and agrees with my own experience. (I actually have grammar bias — induced by my 3rd grade english teacher. Hence why this site is replete with errors.) It would still be a form of psychometric bias, though, if that’s what you’re getting at.

  2. Pingback: linkfest – 08/26/12 « hbd* chick

  3. These scores are all messed up. For example many sub-saharan states have average IQ’s way below 80 on most studies while others, like zimbabwe for example, have quite reasonable IQ scores above 80 on one study. I still don’t take these findings seriously, and the only way i would is if a mixed team of people consisting whites, blacks and so on, contributed to the study.

    • Africa is exceedingly culturally, genetically, and environmentally diverse. Given that, it would be shocking if there was little variance between countries in latent cognitive ability, no? Moreover, the scores above are based on a mix of assessments: IQ tests, international school assessments, and regional school assessments — which were given to different cohorts at different times. Again, it would be shocking, given this, if there was little variance in scores.

      Generally, the individual scores should be taken as rough estimates.

      As for your comment, “the only way i would is if a mixed team of people consisting whites, blacks and so on,” who do you think is conducted the regional African assessments?

      http://www.iiep.unesco.org/en/news/single-view.html?tx_ttnews%5Btt_news%5D=778

      Ditto IQ.

      Don’t comment on this blog if your going to make imbecilic statements as the above.

  4. Pingback: IQ Estimates for Various Races | Robert Lindsay

  5. There is enormous room for environmental improvement in SSA and wont that inevitably lead to IQ increases? If the average IQ for SSA now is 75 with Africa in the state it is in, then wouldn’t you expect it to at least be able to reach 90 as Africa develops and conditions improve?

    90 wouldn’t be too bad. That is like the IQ in the west 3 or 4 decades ago, right? Plus that would mean Africa would probably have the potential to develop to the level of the west in 1970, wouldn’t it? And who would care about race mixing for the sake of 5 iq points? It wouldn’t be a disaster.

    • “There is enormous room for environmental improvement in SSA and wont that inevitably lead to IQ increases”

      You assume that Africans are passive actors who merely experience their environment. Yet, if a strong hereditarian hypothesis is correct, the African environment represents an extension of the population geneotype; it’s the result of, not the cause of, the population IQ. If so, the environment could be improved, but this wouldn’t improve the genotype. And so constant intervention would be required..

      As for your second point, this issue has been discussed over and over again. The secular differences are in large part due to psychometric bias — meaning: they don’t have the same meaning as do differences between random persons at a given time within a given population. It’s not clear if the international differences are more like the secular ones (and due to psychometric bias) or if they are more like the differences between random individuals.

      As for your last point, people who do oppose miscegenation do so for different reasons. I don’t imagine that out-breeding depression is an important one. I imagine that the issue has more to do with identity — the same reason why catholic parents would want their kids to marry in the faith. Doing so increases the probability that the grand kids will share the same belief.

      • Thanks for your reply Chuck.

        “As for your second point, this issue has been discussed over and over again. The secular differences are in large part due to psychometric bias”

        By the secular differences are you referring to the differences between generations ie the Flynn effect? I thought you might say something like this because I’ve heard it before but I haven’t seen it discussed and I don’t understand it. Could you please either explain or refer me to a fairly simple an explanation or discussion of this? I’m sure that will be easy if its been discussed over and over!

        thanks…

        • Okay I just looked up psychometric bias and I can see what you’re getting at but still if you could refer me to something that discusses this in regard to the Flynn effect that would be great…

        • Read the two papers below. A finding of measure invariance implies that the difference between groups is of the same psychometric nature as that within. You have an apples to apples comparison. MI has generally been found to hold for the US B-W gap; as such, it’s like a W-W or B-B gap. The generational gap is something else. A substantial component of it is due to “psychometric/cultural bias” in the sense that the tests don’t have the same meaning across groups as within. The generational gap is more like a test-retest gap then a within group gap.

          Flynn Effect : Wicherts, J. M., Dolan, C. V., Hessen, D. J., Oosterveld, P., Van Baal, G. C. M., Boomsma, D. I., & Span, M. M. (2004). Are intelligence tests measurement invariant over time? Investigating the nature of the Flynn effect. Intelligence, 32(5), 509-537.

          BW gap: Lubke, G. H., Dolan, C. V., Kelderman, H., & Mellenbergh, G. J. (2003). On the relationship between sources of within-and between-group differences and measurement invariance in the common factor model. Intelligence, 31(6), 543-566.

comments do not require an email

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

WordPress.com Logo

You are commenting using your WordPress.com 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