Literacy and Numeracy in England by Race

Below are results from the 2003 and 2011 English skills for life survey. Literacy and Numeracy were measured for the age 16 to 65 population. (Here.) Annoyingly, only %s by skill level were given. Moreover, the distributions was highly unnormal. This was unfortunate since the Literacy and Numeracy tests seemed to have been good measures of ability. To extract a d-vale, I compared the percent of Whites who fell above the skill level where approximately 50% of Whites fell to the % of Black Caribbeans who fell at and below this level. If you can think of a better method, feel free to offer it. I only compared Whites with BC because these were the only groups for which the English as first language rates were above 90%. The d-value came out to 0.77. I estimated a composite d-value of 0.86.


This entry was posted in Uncategorized. Bookmark the permalink.

10 Responses to Literacy and Numeracy in England by Race

  1. I wonder why the master race always has to prove their the master race. It makes me wonder if indeed the master race is truly master. Probably not. Or this could be just a race in it’s death throes, in which case go dig a hole and jump in, lest we have drag your carcass to the hole.

  2. Julian says:

    Hey, I thought this piece in Nature may be of interest.

    • Chuck says:

      As for the paper under discussion, my colleagues and I investigated it somewhat. See here and here and here.

      Here were some of our conclusions:

      [Kan et al’s cultural Load] variable behaves as if it was some sort of heritable-g index rather than a generalized cultural factor as it is supposed to approximate. The absence of any relationship with environmental indices such as FE gains, adoption gains, e2 (and perhaps c2) substantiates my point. That culture load correlates both with Gc and Gf could mean either than culture load variable is not well constructed or, rather, that Gc and Gf are indistinguishable. If that is the case, the inescapable conclusion would be that Kan’s separation of g-crystallized from g-fluid, a dichotomy that Flynn has considered before, is highly misleading. And thus not justified.


      The correlations between cultural-loadings and the BW gaps and between cultural-loadings and heritability estimates were both fully mediated by g-loadings. This implies that g is a mediating factor in both instances. What one needs to explain, then, is why more culturally-loaded tests are better measures of g.

      Generally, Kan et al. found an interesting relationship between cultural loading (defined so and so) and g-loading. It’s not clear why this exists. There are many ways to account for it. The findings are consistent with a largely additive genetic model of individual and race differences.

    • Chuck says:

      I re-read the paper.

      The conclusion is rubbish, though the found associations are not uninteresting. The basic problem is that the authors failed to demonstrate the construct validity of their measure of “cultural loading” as a measure of ‘dependency on prerequisite information for which the population is heterogeneous in its exposure to’ (from here on: heterogeneous load). They basically show that test dependency on cultural information correlates with g. (So tests that ask one to define “travesty” are more g-loaded than tests that ask one to say “7,5,3,8,1” backwards.) But they fail to establish the crucial link — needed for their conclusion — between cultural load and heterogeneous load. They merely assert that this link is likely. But their own findings can be read in support of the contrary conclusion. For example, the findings show that the magnitudes of differences between MZ twins reared in the same environment negatively correlate with cultural loadings. Since MZ twins reared together share the same rearing environment, the same genes, and the same prenatal environment, one would normally think that they only differ due to exposure to information. That is, one would normally take the differences in the correlations between MZ twins reared together as a perfect index of cultural load as they mean it (i.e., as heterogeneous load)!

      The evidence they present, then, can only provide evidence for their model if they assume their model in the first place to explain why, contrary to what one would expect, the magnitude of differences between MZ twins reared together negatively correlates with their supposed index of heterogeneity load. Since their cultural loadings negatively correlate with many presumably culturally induced effects e..g, the Flynn Effect, we suspect that it’s really not indexing what they contend that it is.

      • KJK says:

        “The best examples of scientific rubbish is what you find here…”

        • Chuck says:

          Sorry, I have been busy. No need to get hostile.

          As I noted, I liked your analysis — but felt that the conclusion was rubbish. Nothing you have said has changed that view. Now, onto specifics:

          (1) As for g versus g-load, I don’t recall saying that these were the same. When I discussed this issue in my HV post, I said:

          “g-loadings fully mediate the association between cultural loadings and the two other variables noted and therefore that what is in need of explanation is only the association between cultural-loadings and g-loadings. I will then proceed to offer an account for this….The correlations between cultural-loadings and the BW gaps and between cultural-loadings and heritability estimates were both fully mediated by g-loadings. This implies that g is a mediating factor in both instances…What emerges robustly from the above analysis is that g-loadings mediate the association between cultural-loadings and the magnitudes of the Black-White gaps. I have argued elsewhere that such mediation more strongly evidences g-differences than do bivariate correlations between g-loadings and the magnitude of the subtest differences…I take the above as strong evidence for the veracity of Spearman’s hypothesis. Taken together with numerous other lines of evidence, it must be accepted that Spearman’s hypothesis has graduated to the level of scientific theory.”

          You might quibble with my use of the term “implies” but in English “implies” can mean “suggests”. I don’t see anything else potentially problematic here. Can you point to the exact sentence with which you disagree.

          Now, more generally put, IF the correlation between two vectors, X and Z, is interesting because it’s “something that needs to be explained”, THEN the mediation of this correlation by another vector Y should also be interesting, because in this scenario that vector, Y, does some explaining. In the case being discussed, the correlation between cultural load and heritability is explained by g-load. That is, there is no (or seems to be no) correlation between cultural load and heritability independent of the correlation between cultural load and g-load.

          Now, this is what you say:

          “Whereas ‘mediated by g’ denotes something meaningful (and testable with Structural Equation Modeling) ….‘mediated by g-loading’ is meaningless. g-loading is not an interindivual variable but a test characteristic….The point of our paper is that we found an effect that was not predicted, and is not explained by, g theory. It thus deserves an explanation.”

          I disagree with (2) as explained just above. I explained your r ( cultural load x h^2). Now, I just have to explain your r (cultural load x g-loading).

          If you want, we can use a larger collection of subtests to check if, in fact, there is no correlation between cultural load and h^2 after taking into account g (edit: g-loadings). I have a large collection of batteries with g-loadings and ACE estimates. If you are willing to code the subtests for cultural load, we can test this point further. If a significant correlation is found, controlling for g-loading, I will grant your point, as a g theory could not possibly explain a correlation between cultural load and non-g genetic variance (as indexed by a correlation between cultural load and h^2 taking into account g-loadings). .

          My argument, though is that a correlation between your cultural load and g-loading poses no challenge for a “biological g theory”. I don’t know why you are having trouble with this point.

          You say: “Most importantly, g theory might explain why vocabulary is highly heritable, but not why is is the HIGHEST heritable.”

          IQ tests measure information obtained. When information is roughly equally available, these tests measure the ability to learn information.

          You assume:
          (a) that “cultural load” indexes the amount of prerequisite information needed for the successful completion of a subtest (that is, prerequisite information load)
          (b) that “prerequisite information load” indexes the “heterogeneity of information” or how unequal the prerequisite information is spread in a population
          You show that:
          (c) “cultural load” correlates with heritability
          You argue that:
          (d) The above correlation is not easily explained by biological g theories but is by COV(GE) g theories, because …
          (d1) biological g theories would not predict a correlation between cultural load and heritability
          (Maybe you didn’t make this argument.)
          (d2) biological g theories would not predict a correlation between cultural load and g-load, given investment theory

          (If I substantially misunderstood your argument, let me know.)

          As for (a), it’s not clear that your more “culturally loaded” tests require more prerequisite information, within a given population, than do your less “culturally loaded” tests. This problem is illustrated by the strong positive correlation between the magnitude of the secular changes and the inverse of your “cultural loadings”. .As noted prior, one might just as well as argue that the magnitude of the Flynn effect better indexes “true” cultural load (in the sense of subtest dependency on prerequisite information) and after draw the opposite conclusion as you do.

          As for (b), it’s not clear that within a given population “prerequisite information load” measures “heterogeneity of information load”. Can you make some further predictions that I could test? Alternatively, can you think of a more direct way of testing if “heterogeneity of information load” positively correlates with h^2. (If you want, for example, I could look at — or, better, give you the data so that you can — the heritability of NLSY79 ASVAB subtests for mixed language Hispanics. The Hispanic sample showed zero heritability for AFQT using SEM. Would you predict a positive subtest heritability x cultural load correlation in this linguistically heterogeneous sample?) Whatever the case, you didn’t demonstrate (b).

          As for (c), I would like to see the finding replicated on a larger selection of tests. We a number of batteries here:

          As for (d1), I showed — at least based on the sample which you used — that the correlation between the vector of cultural load and the vector of heritability can be explained by a correlation between the vector of cultural load and the vector of g-loading. Again, if you want, we can look at this using a larger sample.

          So what are you left with? Not much… d(2).

          That is, all you have is your continual assertion that a “biological g theory” can’t explain the correlation between g-loading and cultural-loading or can’t explain the high g-loading of vocabulary. (You have not shown that the heritability of vocabulary is high independent of its g-loading; rather, you just evade the point and repeat: “but it is clear that vocabulary is highest heritable…. g theory might explain why vocabulary is highly heritable, but not why is is the HIGHEST heritable). You make it sound as if no “biological g theorist” has noticed the g-loading x cultural-loading association. But Jensen noted it in his magnum opus, in his classic 1985 spearman’s hypothesis paper, and in his 2001 commentary on vocabulary, which I thought I sent you.

          I don’t see why this is difficult to understand. Both learning new words and correctly retrieving learned words — searching one’s mind for the correct associations — is cognitively complex. Try running a search program on an old beat up computer.

          But you say:

          You: “Most importantly, g theory might explain why vocabulary is highly heritable, but not why is is the HIGHEST heritable.


          Note that if there is no gene-environment covariance heritability of educational attainment will be 0. ”

          My comment: This is a really silly point. According to this conceptualization of COV(GE) in absence of COV(GE) the heritability of IQ will be 0 — since all IQ tests test some form of learned information, some implied rules, or some understandings. If you disagree, please show me a purely “culturally free” IQ test. Generally, you are conflating here what you previously called intra-individual COV(GE) with inter-individual COV(GE). When I speak of COV(GE) I mean, properly, the inter-individual type, as the context of discussion is population variance as in P = G + E + COV(GE), variance which is inter-individual. No?

          You: “Note that if education is assumed homogeneous, variance in Edu equals/approaches 0 (and covariance between fluid intelligence and education will be absent anyway)..”

          My comment: Come on, stop with the games. If “education” in the sense of “exposure to information” is homogenous then environmental variance will approach 0 and genetic variance will approach 1, since most differences will be due the biological ability to acquire, understand, integrate, process, and store that information, an ability which is largely genetically conditioned. .Now the issue here is the relative homogeneity of “exposure to information”. So the question is: Is the exposure to the prerequisite information needed to do well more heterogeneous for “cultural loaded” tests than for “culturally reduced” tests. Again, you have provided no evidence that the answer is “Yes”. Recall, I said:

          “Now, I would tend to think that “Western society creates a homogeneous learning environment” more so with regards to the information that is measured by more cl(a)-loaded tests than with regards to the information that is measured by less cl(a)-loaded tests for the simple reason that the “schools and school systems” etc., homogenize environments somewhat and also happen to teach more cl(a)-loaded information than not”

          Does the common core in the Netherlands focus more on Raven’s matrices like problems or more on vocabulary like problems. You tell me.

          You continue: “Heritability of crystallized intelligence is thus a weighted sum of the heritability of education (which is 0 under the assumptions above), the heritability of the residual and the heritability of fluid intelligence and…”

          Yes, I understand the argument form gC/gF. And I noted the issue with the Fluid-Crystal model. In reply, though, you said: “I devoted an entire chapter to the interpretation of crystallized inteligence. I don’t confuse factor models with developmental models, for example, as you do here.”

          But I referenced Johnson and Bouchard. To quote them:

          “This hypothetical causal model of the Gf–Gc distinction has led to the prediction that Gf should be under greater genetic influence than Gc. As acknowledged by Horn (1998), however, this prediction has been repeatedly disconfirmed. While there are other ways to posit relative genetic and environmental influences on Gf and Gc, this finding is a rather strong refutation of one of the major and novel predictions based on the theory as conceived by its conceptualizers, calling into question at least theirdevelopmental framework”.

          I made the exact same point as them. Did they also confuse “factor models with developmental models” in their discussion?

          So all you have is a very week argument from Fluid/Crystal intelligence and h^2 which reduces down — unless you can show otherwise — to an argument from
          Fluid/Crystal intelligence and g-loading. That is: (d2).

          And I did note: ” That said, there is something to the Gf-Gc model, so there is something of interest here.”

          The issue then seems to boil down to: Why are the g-loadings larger on crystal than fluid intelligence tests? (And why is the Flynn effect larger on fluid than crystal intelligence tests.)

          As for the first question, MH dd seem to find a positive correlation between “fluid-loading” (in the sense of the magnitude of the association with Raven’s, see Flynn 2000 for a discussion of method). I will assume then that Fluid intelligence loadings, understood this way, correlate with g loadings. So, to be clear, the issue involves the relative association between g-loadings, fluid loadings, and crystal loadings. Again, why are crystal tests MORE g-loaded?

          As has been noted prior, the situation is ambiguous for several reasons. Recall JL’s comment here:

          It’s not clear if this stronger association is an artifact of battery composition/ subtest design or if it is “real”. If the association is real, then there is, in fact, something of interest. That is, you would have established (a). But your conclusions would still not follow for the various reasons discussed above.

    • Chuck says:

      Excuse the numerous redundancies in the reply below.

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