Hispanic performance by generation

In light of Ron Unz’s assertion that the Hispanic-White (non-hispanic) differential has greatly narrowed in recent years, I decided to examine the change in the magnitude by generation. Numerous samples over the past three generations have consistently point to a gap of between 0.5 and 0.8 standard deviations. The largest analysis to date was conducted by Roth et al. A meta-analytic general intelligence gap of 0.72 SD was found for the 70s through 90s:

It should be noted that Spearman’s hypothesis has generally been found to hold for the the White-Hispanic gap. According to this hypothesis, the magnitude of the differential is correlated with the general intelligence-loadings of the tests. When evaluating the evidence this should be kept in mind.

Now, as noted, the evidence to data points to a difference of between 0.5 and 0.8 SD. However, there is a continual flow of Hispanic immigrants to the US; as such it’s possible that a narrowing of the difference is being masked by this continual inflow. Perhaps second and third generation Hispanics, against whom there should be little cultural bias, have greatly narrowed or even eliminated the difference, but this remarkable fact has gone unnoticed due to the under-performance of recent Hispanic immigrants, who are handicapped by a limited familiarity with English. This is a plausible scenario. To evaluate this possibility I looked at the cognitive ability differences by generation in The National Longitudinal Study of Adolescent Health (1994-1994), Program for International Student Assessment (PISA) (2000,2009), Trends in International Mathematics and Science Study (TIMSS) (2007), Progress in International Reading Literacy Study (PIRLS) (2001, 2006), and the General Social Survey (GSS) (2002-2010). The Add health data contained a picture vocabulary test; PISA, TIMSS, and PIRLS contained math, reading, and science tests which are highly correlated with IQ (Lehrl, 2005); and the GSS contain a vocabulary test which is also highly correlated with IQ. Below is a summary of the standardized differences between 3rd+ generation Hispanics and 3rd+ generation Whites:

The differences ranges from 0.35 SD (GSS) to 0.77 SD (TIMSS grade 8 science). The average of the differences comes out to 0.61 SD or, when averaging PISA, TIMSS, and PIRLS tests scores per year, per grade (e.g., PISA 2009 MAth + Reading), 0.59 SD. This is not largely different from the general intelligence difference reported by Roth et al. (2001), which is notable given what was said about Spearman’s hypothesis. On re-analysis, Ron Unz’s claim concerning the difference in the GSS sample was upheld; this claim, though, was contradicted by all other samples. As for the GSS sample, it’s notable that the Black-White difference is also greatly reduced. Noted this, Murray (2007) has pointed out:

The GSS vocabulary test. GSS data are now available through the 2004 survey, 6 years longer than the observation period available to Huang and Hauser (2001), and they show a continuing decline in the B–W difference for persons born into the early 1980s (author’s analysis of the GSS). But if the question is whether black performance on the vocabulary test has improved, there is no inconsistency with the Woodcock–Johnson results. The GSS has an absolute scale of correct answers, from 0 to 10, and the vocabulary items have remained unchanged since the advent of the GSS. The highest black mean score, whether measured in a single birth year or in five-year aggregations, occurred among blacks born in 1945–1949. The decline in the B–W difference in the GSS vocabulary test for persons born since mid-century is entirely attributable to a decline in white performance, not improvement in black performance

This sample, then, is probably best to be treated with caution.

Two of the studies, the GSS and Add health, allowed Hispanic scores to be decomposed by nationality. For both studies I compared the Hispanics (aggregated) and Mexican scores. For third generation individuals no significant differences was found; this wasn’t surprising as Mexicans constitute over 2/3rds of third+ generation Hispanics. This implies that Hispanic scores can be used as a proxy for Mexican scores. Tables below:

For interested readers, the tables below contain the generational scores for PISA, TIMSS, and PIRLS.


Lehrl, 2005. PISA – ein weltweiter Intelligenz-Test. Geistig fit.

About these ads

8 thoughts on “Hispanic performance by generation

  1. Chuck,

    I just always naturally assumed you were the same Chuck from Gucci Little Piggy. But he believes differently, and who am I to argue? Could you give me an email address or write me at jmalloy (at) nyc (dot) com?

  2. These are quite interesting results. I regret my insulting tone on the TAC comments page, but I disliked having my claims called “egregiously dishonest” and “laughable commentary” when it was very obvious that the error was not mine.

    It is certainly correct that the PISA and TIMSS results you cite do seem to indicate a much larger white/Hispanic gap than was found in the most recent decade under Wordsum, with the PIRLS and Add Health results being intermediate.

    Obviously, Wordsum is an imperfect proxy for IQ, let alone the Math/Science/Reading results of those other tests, but the gap is still much larger than I would expect. Also, the pattern of rapidly rising Wordsum scores seems too strong to be coincidence. One obvious difference is that the Wordsum results are based on adults, while nearly all the others being tests for children in school. It’s not obvious to me why the adult/child difference would skew the results in a particular direction, but it certainly seems possible, given the American-born Hispanic age-distribution curve. I assume there are no older tests of these types, which would allow longer term trends to be detected. I’m also not too familiar with the details of these other tests, and how well they are found to correlate with IQ (though I assume it’s at least as good as Wordsum).

    As I’d mentioned in my original article, the overall Wordsum-IQ gap between farm-raised whites and urban/suburban whites was almost as large the black/white gap, and when I went back and confirmed the Inductivist Mex-Am calculations last night, I decided to look more closely at that issue as well. It turns out that during the 1970s the white rural-urban/suburban gap was almost exactly the same size as the black/white gap. The trendlines were that the urban/suburban Wordsum-IQ remained almost exactly constant between the 1970s and 2000s, while the white farm Wordsum-IQ rose fairly rapidly, reducing the gap by about 1/3 between the 1970s and 2000s.

    From a genetics analysis, I’d think a contrary result would be the more expected, as the smarter and more ambitious farm whites moved off to the cities, leaving the dumber ones behind. Since the result is exactly the opposite, I’d suggest that the most plausible explanation would be the huge increase in the “cognitive development environment” in farming communities, as TV and other complex modern technology increasingly penetrated over the last few decades.

    Anyway, the innateness of the Wordsum-IQ results for white farmers during the 1970s seems unlikely. I find it rather implausible that almost 10% of all farm-raised whites during the 1970s would have suffered from clinical mental retardation.

    • Ron,

      It’s late and I don’t have time for a detailed reply. So I will briefly note the following:

      (a) The genetic case for the Black-White difference is exceedingly complex. I outlined it here:

      (b) One important aspect of that case is that many of the psychometric properties of the difference are known. For example, measurement invariance holds for the gap and the gap is g-load. It’s clearly mot merely a score difference.

      (c) When it comes to other gaps such as that between the North and the South, Rural areas and Urban ones, or Nations, little is known. My own guess would be that the Rural-Urban gap is not a general intelligence gap, whether in the statistical or biological sense. Whether this is the case can be tested easily but oddly it has not been. (I noted elsewhere, though, that the Protestant-Catholic gap has turned out to be negatively correlated with g, as with the Flynn effect.)

      (d) The point above is that not all score gaps have the same meaning, Refer to the discussion in the comment section here.

      (e) There is some support for the claim that National IQ gaps represent g-gap, but this is tentative.

      (f) Each National IQ data point is not particularly meaningful since the proper statistical studies have not been done to determine the meaning of the scores relative to other National scores or international scores. The aggregate of the data has more meaning as aggregating minimizes error. Aggregation can’t, of course, establish that the differences are true latent ability differences, but, at least, it’s plausibly that the aggregated IQ differences do represent true intelligence differences, give their predictive validities; this can not be said for intra-national IQ differences for which no formal studies have been conducted. In your own sample, there were as many countries for which intra-national IQ scores correlated with GDP as not; the relation seemed to be random.

      (g) Now, the above are pretty basic points– which is why I considered your discussion, which took none of these points into consideration, to be disingenuous. Its seems that I misunderstood your intent. Anyways, if you’re interested, I could point you to other samples. Here were some results from the nationally representative NELS as given in “Parental influences on the education outcomes of immigrant youth.” The difference comes out to 0.55 SD for third generation Hispanics (achievement tests were used). From what I recall similar results can be found in the High School Longitudinal Study of 2009. I posted a bunch of immigrant articles here which contain cognitive ability scores, if you’re willing to dig. As for adult IQ, I’m not aware of any nationally representative studies that are broken down by generation. I’ve come across some normative data which supports your position of a secular change, for example: http://occidentalascent.wordpress.com/?attachment_id=6017 (see: WAIS-IV Clinical Use and Interpretation: Scientist-Practitioner Perspectives page 122), but I’m not really in the mood to argue against myself.

      (e) All of that said, you can find some support for your rural-urban hypothesis here: http://sgdp.iop.kcl.ac.uk/davis/teds/geocoding/

      You will notice that IQ heritability is much higher in Urban than Rural regions; this implies more environmental depressors in rural regions.

  3. Pingback: Unz on Race/IQ – The Rural/Urban Divide | The American Conservative

  4. Pingback: Randoms « Foseti

  5. Pingback: Unz on Race/IQ: The Entire Series and Debate | Ron Unz – Writings and Perspectives

  6. Pingback: Race/IQ: The Entire Series and Debate | The American Conservative

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