The Measured Proficiency of Somali Americans

The discussion of the performance of African immigrants led by Chanda Chisala has been of unusually poor quality. As such, I thought that I might write a brief tutorial post on how to locate data and estimate differences in hopes that this will inspire better research practices and more rigorous debate. I will also elaborate on the Jensenist position and its predictions, as Chanda, and apparently many others, do not seem to have a good grasp of it at least in its quantified form.

I. Locating data and computing differences

We start first by picking a subgroup of interests. Given Chanda’s intended argument, Somalis, as a Somali American commenters noted, are theoretically a good choice since most are nominally fairly unselected refugees. I say nominally for the reasons discussed in section III. Having decided on a group, we first check national level databases and then do a Google Scholar search. If we can locate little to nothing there, we look up the group’s dominant areas of residence, so that we can check city and/or state educational reports. Wikipedia informs us that:

In terms of cities, the heaviest concentrations of Somalis in the US are found in the Twin Cities (Minneapolis and Saint Paul), followed by the Columbus, Ohio, Seattle, San Diego, Washington, D.C., New York City, Portland, Maine and San Francisco metro areas.

A quick Google search turns up proficiency rates for Seattle, Portland, Minneapolis, and St. Paul Somalis. We can check the American Community Survey (ACS) to get a rough sense of the proportion of all Somali’s living in these cities so to decide which to focus on. In 2013, the proportions would have been:

2013 Somali Population Per City      
    Number % of Total
Seattle CCD   10,133  
Seattle East   1,223  
Seattle all   11,356 10%
 
Portland East   852  
Portland West   239  
Portland all   1091 0.01%
 
Minneapolis   12,702 12%
 
Saint Paul   5,558 5%

Since this is a tutorial we will not go through every data source or conduct a thorough search. Instead, we initially will focus on Minneapolis scores as this city appears to have the most Somalis of any U.S. city. Having decided on our region, we head over to the Minneapolis Public school website and look for a data explorer and/ or technical reports which might provides some type of scores for the group in question. We find the former.

MinAf

MinSom

We see that Minneapolis Somalis had a 2014 MCA-11 math proficiency rate of 22% and a reading rate of 21%. This is compared to an African American math rate of 21 % (d= -0.03) and a reading rate of 21% (d= 0.00). Here, we calculate our d-values using Excel’s normal distribution function; we assume, and are explicit about this, a normal distribution and equal group variances, assumptions which often fail to hold, but ones which we are forced to make, given the data available. Next we go through a series of steps to norm our values on a national reference, which will allow us to better evaluate them. First, we want to compare the performance of Minneapolis African Americans to Minnesota ones. To do this we check the state level results for the same test for the same year. We see that Minnesota African American’s math and reading proficiency rates were 33.5% and 33.2% respectively. Granting the previously mentioned assumptions, this represents an advantages of d= 0.38 (math) and 0.37 (reading).

StateBlack

Second, we want to compare the performance of Minnesota African Americans to African Americans nationally. Since the African American Minnesota population is small, we use NAEP results for two years to increase reliability. We go to NAEP explorer. We select:

1. Select criteria: Math/reading composite National/Minnesota
2. Select variable: Race/ethnicity using 2011 guidelines, school-reported
3. Edit reports: statistical options: Average Scale score and standard deviations
4. Build reports

We then take the results and compute standardized differences:

Year Group Math Reading Average
Mean SD % Black %All Mean SD % Black %All
2013 Black, National 263 34 15 250 33 15
2013 Black, Minnesota 260 39 10 ? 248 36 10 ?
2015 Black, National 260 34 15 248 34 15
2015 Black, Minnesota 262 36 8 ? 247 36 8 ?
d-value 2013 -0.09 -0.06
d-value 2015 0.06 -0.03
d-value average -0.01 -0.05 -0.03

We do not need to compute state B/W differences, because this has already been done for us. Now, we put all the numbers together. We also compute composite scores a la Sackett and Shen (2010) since these are more comparable to full scale IQ ones.

Groups Math d Reading d Average d Composite d
Minneapolis Somali/ Minneapolis Black -0.03 0.00 -0.02 -0.02
Minneapolis Black/ Minnesota Black 0.38 0.37 0.38 0.41
Minnesota Black/ US Black 0.01 0.05 0.03 0.03
US Black/US White (2011, 2013) 0.90 0.98
Minneapolis Somali/US White 1.29 1.40
Average ACHQ Composite ACHQ
Minneapolis Somali ACHQ (White US reference) 80.65 78.96

After we have computed our initial scores, we seek validation. We can do this using the Seattle proficiency rates provided by Chanda. Following the procedure outlined above, we compute the magnitude of the differences between Seattle Somalis and Seattle Blacks for all students. We then compute the difference between Seattle Blacks and Washington Blacks using Washington’s 2011-2012 Annual Measurable Objective Summary scores. Finally, we calculate the difference between Washington Blacks and African Americans nationally using 2011 and 2013 NAEP scores. The latter are shown below:

Year Group Math Reading Average
Mean SD % Black %All Mean SD % Black %All
2011 Black, National 262 33 15 249 32 15
2011 Black, Washington 265 34 5 ? 254 34 5 ?
2013 Black, National 263 34 15 250 33 15
2013 Black, Washington 269 39 4 ? 258 30 4 ?
d-value 2013 -0.09 -0.16
d-value 2015 -0.18 -0.24
d-value average -0.13 -0.20 -0.17

We are then in position to add the differences up. Our results look as follows:

Groups Math d Reading d Average d Composite d
Seattle Somali/ Seattle Black 0.32 0.22 0.27 0.29
Seattle Black/ Washington Black 0.08 0.18 0.13 0.14
Washington Black/ US Black -0.13 -0.20 -0.17 -0.18
US Black/ US White (2011, 2013) 0.90 0.98
Seattle Somali/ US White 1.13 1.23
Average ACHQ Composite ACHQ
Seattle Somali ACHQ (White US reference) 82.99 81.55

We see that our Seattle and Minneapolis estimates are remarkably similar. The average of the Somali/ White (national) composite gaps comes out to d= 1.32. Since we are not lazy, we also check the data for Portland and St. Paul to verify that the differences are at least as large. We find that they are. So we have four samples from regions which cover more than 25% of the Somali populations which show convergent results. We are, of course, not satisfied as we are always data hungry; but we are not completely dissatisfied.

Now, one could skip a few steps and simply try to directly transform Somali/ state White proficiency difference using the method of thresholds. We did not for sound statistical reasons. Standardized differences are linearly dependent on the relevant groups’ standard deviations. Using larger standard deviations will deflate group differences. Standard deviations will be larger when populations are heterogeneous and total sample standard deviations are employed. Proficiency percents are based on total sample standard deviations. Thus gaps are deflated when populations are heterogeneous. This can be shown by comparing Black/White state differences computed using percents (the method of thresholds) and by using means and pooled standard deviations. The results for Washington and Minnesota are shown below. The B/W gaps are smaller by about 0.20 SD.

Comparison of B/W gaps by method
Method Groups Math d Reading d Average d Composite d
Thresholds Minnesota Black/ Minnesota White 0.92 0.87 0.89 0.97
Means and SD Minnesota Black/ Minnesota White 1.17 1.01 1.09 1.18
Washington Direct transformation from White
Thresholds Washington Blacks/ Washington White 0.68 0.60 0.64 0.69
Means and SD Washington Blacks/ Washington White 0.80 0.82 0.81 0.88
Average d-value -0.19 -0.18 -0.18 -0.20

Whatever the case, the results based on both methods correspond well enough. Those generated using a direct transformation are shown below. Subtract about 0.2 standard deviations from the final results and you get something closer to the “true” gap:

Direct transformation from White        
Groups Math d Reading d Average d Composite d
Minneapolis Somali/ Minnesota White 1.29 1.24 1.27 1.37
Minnesota White/ US White (2011, 2013) -0.24 0.00 -0.12 -0.13
         
Minneapolis Somali/ US Black (2014)     1.15 1.24
         
Groups Math d Reading d Average d Composite d
Seattle Somali/ Washington White 1.08 1.00 1.04 1.13
Washington White/ US White (2011, 2013) -0.05 -0.05 -0.05 -0.05
         
Seattle Somali/ US White (2010)     1.00 1.08
         
Groups Math d Reading d Average d Composite d
St. Paul Somali/ Minnesota 1.04 1.24 1.14 1.24
Minnesota White/ US White -0.24 0.00 -0.12 -0.13
         
St. Paul/ US White (2003-2005)     1.02 1.11

I will reference my original results in subsequent discussions as I believe that they are more statistically sound.

II. Evaluation and critical assessment

Now, Chanda, not without reason, has suggested that the proper scores to use are the English Language Learner (ELL) excluded ones. So, we can compute the d-values for those too. The results for Seattle are shown below:

Minnesota Somali IQ (White US reference)
Groups Math d Reading d Average d Composite d
Washington no-ELL Somali/Black 0.04 -0.70 -0.33 -0.36
Seattle Black/ Washington Black 0.08 0.18 0.13 0.14
Washington Black/ US Black -0.13 -0.20 -0.17 -0.18
US Black/US White (2011, 2013) 0.90 0.98
no-ELL Seattle Somali/US Black -0.37 -0.40
no-ELL Seattle Somali/US White 0.53 0.58
ACHQ Average ACHQ Composite
no-ELL Washington Somali IQ (White U.S. reference) 91.99 91.32

The non-ELL Seattle Somali/White (national) gap comes out to 53% of the total Seattle Somali/ White national one. Assuming that this % is roughly the same for Minneapolis, our Minneapolis ELL corrected composite gap would come out to 0.74 SD and the average of the two would be 0.66 SD. These numbers assume that ELL status, which is assigned based on English language proficiency test scores, indexes, in full, English language familiarity and in no part cognitive ability. JimC, commenting on Chanda’s argument, suggested that this may not be the case. If so, this would fit with my own findings concerning international differences. A while back, I used path analysis to show that National IQ scores predicted national English Proficiency Index and TOEFL test scores — purported measures of English language familiarity — at least as well as did the % of the national populations who spoke English!

EnglishProfNIQ

There is then ecological support for JimC’s proposition. We can check the NAEP Main results by ELL status & home language usage to get a sense of the plausibility of this supposition as applied on the individual level. Some results are shown below:

ELLnation

We see that ELL students do worse that non-ELL ones, as expected. And, surprisingly, that this holds for Whites, Hispanics, and Asians who reportedly never speak a language other than English in their homes! Confirming this observation, regression analysis shows that other-than-English language use in a home has practically no effect on scores independent of ELL status. Yet, presumably, typical language spoken at home is at least as much of an index of English language familiarity as is test-measured English language proficiency. We notice also, curiously, that non-ELL non-Hispanic Blacks who speak a home language other than English all/most of the time or half of the time do little to no better than those who only speak English. This is odd since a good number of these African Americans should be of recent African origin, Spanish speakers presumably being largely excluded from the non-Hispanic Black sample. All of this suggests that low scoring, foreign appearing students might be shuttled into ELL programs, regardless of their true level of English familiarity. News reports offer anecdotal support in favor of this conjecture:

When 10-year old Khadar Yassin came to his first class in the U.S., he spoke no English. He sat in a classroom full of mostly white students and handful of Somali kids. For the next year, he would only speak with other Somali kids, his cultural liaison and his accommodative English as Second Language (ESL) teacher.. Hassin, now 18 and a senior, communicates in good English with everybody but is obligated to take ESL classes. He profoundly detests that.

“I can read, write and speak in English perfectly,” says Yassin in a slightly accented intonation “but they still keep me in ESL courses. I don’t get it.”

Yassin’s frustration with ESL courses echoes a growing dissent among Somali students in Public schools and colleges. Many of them who were born and raised in the U.S. are obligated to take ESL courses by the school system, because English is not their “at-home” language. That haunts them the most.

The evidence is more than anecdotal. Educational Realist provides some background which helps to make sense of our NAEP by home language usages results:

Immigrants aren’t even half of the ELL population.

ELLbygeneration

Citizens comprise from just over half to eighty percent of the ELL population, depending on who’s giving the numbers, but while the estimates vary, the tone doesn’t: no one writing about English language instruction seems to find this fact shocking.

Twenty percent of elementary school kids and thirty percent of middle and high school ELL students have citizen parents. Their grandparents were immigrants…

Read any study on long term ELLs, the bulk of whom are citizens classified LEP since kindergarten, and it’s clear that most are fluent in oral English—that English is, in fact, their preferred language, the one they use at home with friends and family. They just don’t read or write English very well. And then comes the fact, expressed almost as an afterthought in all the research, that long-term ELLs don’t read or write any language very well…

So it seems that we have a genuine confound here. Are our ELL students just not good in math and reading or are their scores depressed by language unfamiliarity?

To get a rough sense of the typical effect of language bias on test scores, we might compare the performance of first and second generation Hispanics. The generational difference comes out to about 0.3 SD or 30% of first generation Hispanic/ third generation White difference. A plausible estimate might then be that no more than 50% of ELL/ non-ELL differences typically found owe to language unfamiliarity and the rest owes to cognitive variation. With a 50% correction, in the case of the Somali/ White difference, we get a composite d-value of 0.99, which is relative to a B/W national level difference of 0.98.

Of course, Chanda would argue that this is an outrageous over-correction. And he might be right. First generation Somali immigrants really are quite English unfamiliar. And about 70 percent of all Somali Americans are of the first generation (computed based on the numbers provided in this report). Presumably the number would be lower for k-12 students. But perhaps not. So this issue deserves a little more investigation. First we can look at the non-ELL/ all advantage for all Seattle ethnic groups. These are shown below. We see that the ELL effect is greater for Hispanics and African Americans:

Group N (all) N non-ELL N ELL Math Reading Average
Asians
Chinese 1516 1233 283 -0.21 -0.51 -0.36
Vietnamese 1436 1064 372 -0.24 -0.50 -0.37
Filipino 1061 881 180 -0.14 -0.29 -0.22
Other Asian 716 570 146 -0.18 -0.36 -0.27
Japanese 317 308 9 -0.05 -0.14 -0.10
Korean 158 149 9 0.05 -0.20 -0.08
Asian Indian 158 132 26 -0.23 -0.47 -0.35
Samoan 144 126 18 -0.13 -0.13 -0.13
Cambodian 80 60 20 -0.13 -0.35 -0.24
Average -0.14 -0.33 -0.23
n-Weight -0.29
Hispanics
Spanish Speaking 1706 930 776 -0.38 -0.49 -0.44
African
Somali 978 447 531 -0.51 -0.69 -0.60
Amharic 238 143 95 -0.28 -0.57 -0.42
Tiginya 165 106 59 -0.30 -0.54 -0.42
Oroma 166 94 72 -0.35 -0.70 -0.52
Average -0.49
n-Weight -0.55

The systematically larger reading gaps suggests language bias. On another hand, culturally loaded language tests have generally been found to be more g-loaded than less culturally loaded tests, so a general intelligence model could also explain some of the excess reading deficiencies. And on a third hand, since the ELL students are screened based on English language proficiency tests they could also just have larger broad ability language deficiencies unrelated to language bias. It is worth noting that attempts have been made to minimize the effect language on NAEP-esque math tests. For example, Abedi and Lord (2001) gave English simplified and regular NAEP math questions to 1,174 8th grade ELL and non-ELL students. Interestingly, they found virtually no interaction between ELL status and linguistic modification of the tests. So we have yet another hint suggesting that the ELL/ non-ELL difference is not simply due to psychometric (language) bias. Whatever the case, we can estimate the score differences between non-ELL Seattle students and Washington Whites to see how excluding ELL students effects the ethnic score differentials across all groups. For reference, we include Californian scores for the same ethnic groups (from: Fuerst, 2014) and a 50% correction value, which assumes that 50% of a typical non-ELL/all gap indexes cognitive ability:

Asians Math Reading Average If 50%[a] California d (2003-2008)
Chinese -0.84 -0.53 -0.69 -0.51 -0.46
Vietnamese -0.42 -0.33 -0.37 -0.19 -0.13
Filipino -0.08 -0.09 -0.09 0.02 0.13
Other Asian 0.14 0.11 0.12 0.26 0.01
Japanese -0.73 -0.43 -0.58 -0.53 -0.40
Korean -0.63 -0.13 -0.38 -0.34 -0.41
Samoan 0.65 0.55 0.60 0.66 0.76
Cambodian 0.30 0.22 0.26 0.38 0.51
Hispanics
Spanish speaking 0.40 0.28 0.34 0.56 0.71

What we notice is that excluding ELL scores reduces the Hispanic and Cambodian/ White gaps substantially and it increases the Chinese, Vietnamese, Filipino, and Japanese/ White ones. What catches our eye is the Hispanic/ White difference. We know that second generation Hispanics score 0.71 SD below Whites. (The difference is smaller for 3rd+ generation Hispanics, but these generations are highly “assimilated” via intermarriage into the White population; when correcting for this assimilation effect, the 3+ generation difference appears to be no different than the 2nd generation one; for computations, see Fuerst (2014) pg. 19-20). As second generation Hispanic students were born and raised in the U.S., we have reason to believe that they are not unfamiliar with the English language. This suspicion is confirmed by psychometric analyses which find little to no language bias for this group (discussed elsewhere on this blog). Thus, it seems that removing ELL students overcorrects for language bias, at least in the case of Seattle Hispanics. It possibly does the same for our Seattle Asians since the n-weighted ELL excluded differences turn out to be larger than Fuerst’s (2014) second and third generation Asian/third generation White ones (-0.30 versus -0.18 and -0.01).

In light of what has been said above, a 50% correction seems plausible. Our final assessment is that the English language familiar Somali/ White composite gap is between about 0.66 (lower bounds) and 0.99 (based on a 50% correction) standard deviations. And we tentatively endorse the latter figure. Of course, all of this residual gap could be do to environmental factors x, y, and z.

III. Contextualization

Since Chanda takes the Somali American scores to represent conclusive evidence against a Jensenist model, we should take a few minutes to evaluate his position. Let us first put on our Jensenist hats and determine a priori what we would expect. According to Jensen and Rushton (2005) 50-80% of the Legacy Black/ White gap owes to hereditary effects:

The hereditarian model of Black–White IQ differences proposed in Section 2 (50% genetic and 50% environmental), far from precluding environmental factors, requires they be found. Although evidence in Sections 3 to 11 provided strong support for the genetic component of the model, evidence from Section 12 was unable to identify the environmental component. On the basis of the present evidence, perhaps the genetic component must be given greater weight and the environmental component correspondingly reduced. In fact, Jensen’s (1998b, p. 443) latest statement of the hereditarian model, termed the default hypothesis, is that genetic and cultural factors carry the exact same weight in causing the mean Black–White difference in IQ as they do in causing individual differences in IQ, about 80% genetic–20% environmental by adulthood.

While according to a recent international survey of Intelligence researchers the 50% figure is the mainstream one, we will consider both estimates.

As I have noted elsewhere, expressing expected differences this way is confusing because expected test gaps owing to genetic g differences should vary in line with both the heritability and g-loading of the tests in question. And test heritabilities systematically vary by age and other factors. It makes more sense to express differences in terms of allelic risk score d-values and then extrapolate an expected test score difference from this based on an assumed h^2 and g-loading of a given measure. Assuming a B/W full scale IQ difference of 1.00 on a test which has a 0.90 g-loading for which both the within (WGH) and between group heritabilities (BGH) are the same (Jensen’s default model) at 50% to 80%, the predicted allelic scores difference would be 1.00/0.90*SQRT(0.5) = 0.79 to 1.00/0.90*SQRT(0.8) = 0.99 SD. (This, of course, translates back to our expected phenotypic gap — given the same measures and hertabilities — of 0.79*0.90*SQRT(0.5) = 0.5 to 0.99*0.90*SQRT(0.8) = 0.80.) This would be the Jensenist position expressed in quantitative genetic terms.

Importantly, the cause of the genetic component of the Legacy Black/ White difference has been attributed by Jensenists to slave/African subpopulation selection (Jensen, 1973; Eysenck 1971), differential dysgenic reproductive patterns in the U.S. (Jensen, 1998; Shockley, 1974 1:40 on), and evolved global difference (Rushton, 1995; Jensen, 1998; Lynn, 2008). Of these, only the latter has import when it comes to discussions of difference between major geographic races. It is impossible to determine the effect of differential dysgenics and of selection, but we can guesstimate a plausible upper bounds value. As for dysgenics, Jensen (1998) estimated that owing to this the U.S. B/W IQ gap increased 0.6 SDbetween the 1970s and 1990s. He noted:

The effect thus increases the W-B IQ difference from 15 IQ points in the parent generation to 15.6 IQ points in the offspring generation — an increase in the W-B difference of 0.6 IQ points in a single generation. Provided that IQ has substantial heritability within each population, this difference must be partly genetic. So if blacks have had a greater relative increase in environmental advantages that enhance IQ across the generations than whites have had, the decline of the genetic component of the black mean would be greater than the decline of the white genetic mean, because of environmental masking, as previously explained. We do not know just how many generations this differential dysgenic trend has been in effect, but extrapolated over three or four generations it would have worsening consequences for the comparative proportions in each population that fall above or below 100 IQ.

Assuming that one half this apparent dysgenic effect has a hereditarian basis and that the magnitude of the effect was the same for each of the 6 generations (on average) which Legacy African Americans were present and free in the U.S. gives us a reasonable upper bounds estimate for differential dysgenics: 0.18 SD. Using other estimates (for example: Vining, 1982; Meisenberg, 2010; Woodley and Meisenberg, 2013) gives somewhat different values, typically higher, but all are positive.

As for sub-population selection, this could not have been large given the number of Africans enslaved and exported. Nunn (2007: table 1) estimates that 6 million slaves were exported to the New World between 1700 and 1799. In, “African Population, 1650 – 1950: Methods for New Estimates by Region”, Manning (2013) provides a graph of the West African population. Eyeballing it, the per generation population seems to have been 65 million during the 18th Century. Assuming 20 year generations, this gives us (65 x 5) a total of 325 million West African who lived from 1700 to 1799, meaning that 2% were exported to the New World. So at very most if the genotypically dullest of the dull were sold off then the average IQ of those sent could only be 2 SD below the African mean. But there is no possible way the selection could be this efficient. If we arbitrary assume that the effect of slave selection could be 5% of this at most, we get a possible upper bounds negative selection effect of 0.1 SD in allelic scores (and about 0.07 in phenotypic IQ).

Thus a best guess estimate is that only about 50 to 30% of our proposed Legacy Black/ U.S. White gap could be accounted for by non- global evolutionary genetic factors. For the sake of argument we will assume that this value is 0.

Legacy African Americans were until recently 80% West African. In the most recent cohorts they are more ancestrally European owing to intermarriage, but the products of intermarriage are now typically classified as “mixed race” by the Department of Education (DE) and so largely excluded from “African American” educational samples. Thus DE classified African American students tend to be around 80% African. Granting the 80% estimate, for hypothetical pure African Legacy African Americans, the B/W allelic scores gap should be around 0.79/0.80 = 0.99 to 0.99/0.80 = 1.24 SD. And the advantage, owing to genes, that typical Legacy Blacks should have over hypothetically pure-blooded African Legacy ones would be a measly 0.15 to 0.18 SD in IQ (assuming the previously discussed test characteristics and a WGH of 0.65). This is typically quite a bit less than what one derives when extrapolating from the regression line for indexes of legacy Africaness and cognitive ability. For studies on the association between IQ and indexes of African American ancestry in this population see, for example, here, here, here, here, and section O & P here. Also see here. An appreciation of this point led Jensen (1998) to attribute much as the association between e.g., skin color and IQ in the African American population to cross-assortative mating for the two traits. (E .B. Reuters, 81 years prior, proposed a similar idea, for the same reason, in his paper The Superiority of the Mulatto.)

Thus,

(1) Ignoring possible (global) non-evolutionary genetic differences…

(2) If Somalian Americans were perfectly representative of their home populations…

[Note: Back in 2008 a US state department investigation found a high degree of East African refugee application fraud. Specifically DNA tests of 5,000 family refugee applicants showed that around 80% of purported nuclear family members were unrelated to one another. Some of the fraudsters could have been non-refugees who bought their or their families member’s way in to the program and some could have been more advantaged refuges who did the same. Moreover, while the Refugee Processing Center reports that between 1984 and 2013 109,000 Somali refugees were admitted to the U.S., a number which closely matches the Community Survey reported Somali population size of 115,000 for 2013, the actual number of Somali Americans is unknown. Back in 2003, Lehman and Omar estimated it to be 150,000. They noted that the “total number of Somalis living in the United States is estimated at 150,000, of whom about 40,000 are Somali refugees from the dominant clans.” Somali advocate groups themselves claim that there are 150,000 Somalis in Minnesota alone. Generally, the census data can be unreliable because population estimates are based on domicile numbers, yet often multiple Somali families live in one domicile (Source.) Knowing the precise ratio of refugees to economic immigrants is important since Somali emigrants, in general, are highly unrepresentative of their home populations. According to the IAB brain-drain data bank’s Emigration rates file, for example, in 1990, 2000 and 2010, respectively, 1.08%, 3.52%, and 4.43% of all Somalis emigrated somewhere, but at the same time 20.03%, 37.72%, and 44.86% of highly educated ones did. Thus educated Somali’s are seemingly massively over-represented in the emigrant pool. So, to reiterate, it’s not clear what precise portion of all Somali Americans were nominal refugees or the descendants thereof. And knowing this is of non-trivial importance. Nonetheless, a large portion of the Somali American population must be actual (refugee camp) refugees. And presumably such refugees, fraudulent or not, are not particularly cognitively advantaged relative to average members of the home populations.]

(3) If Somali populations are perfectly representative of West African Black ones in terms of cognitive ability…

[Note: Horn-of-African populations are 30 to 50% Out-of-African (OOA). Thus geneticists tell us that “East African groups, such as Ethiopians and Somalis, have great genetic resemblance to Caucasians and are clearly intermediate between sub-Saharan Africans and Caucasians” (De Stefano et al., 2002) and “On the basis of autosomal polymorphic loci, it has been estimated that 60% of the Ethiopian gene pool has an African origin, whereas ~40% is of Caucasoid derivation.” (Passarino et al. 1998). And it so happens that OOA ancestry among African admixed populations is associated with differences in the frequency of “genes with known ontologies [that] are involved in neurological processes or brain development”.]

(4) If, over the last 200 years, there was no differential dysgenic effect between Western populations — including Afrodescent populations in the West — and African ones…

[Note: Hereditarians including Michael Woodley, Richard Lynn, and Helmuth Nyborg maintain that the genotypic IQ throughout the West has been deteriorating over the same time period.]

(5) If achievement tests index general intelligence just as well as IQ ones do …

[Note: As Gottfredson (2006) points out, the magnitude of the expected achievement gap owing to (general) intelligence differences would be no more than the magnitude of the g-gap times the g x achievement correlation: in the U.S., then, the typical B/W ACHQ gaps (e.g., NAEP MAIN test gaps) are significantly larger than any plausible Jensenist model would predict, assuming only inter-individual and no population-level effects.]

Then the Somali-American allelic risk scores should be 0.79/0.80 = 0.99 (lower) to 0.99/0.80 = 1.24 (upper) SD below the U.S. White scores. And consequently the Somali-American/ White measured aptitude score gaps should be no less, assuming no environmental advantage, and assuming WGH = 0.65 (a split between Jensen and Rushton’s 50-80%) than d= 0.70 to 0.88. This should be compared to my estimated (English familiar) difference (from section I) of 0.66 to 0.99.

Now Chanda would argue that the Somali American/ White gap is inevitably inflated given such and such considerations. And he would try to dismiss concerns 1 though 5. But this is something to be demonstrated — which I am not saying that he can not do, just that he hasn’t yet done it.

To note, he also mentions Seattle Ethiopian (Amharic, Tigrinya, and Oromo) scores (n = 569 with ELL, 343 without). Using the same method as above, we get an Ethiopian/ White composite gaps of 0.73 (relative to a B/W one of 0.98). The English familiar Ethiopian/White d-values would be 0.24 (lower bounds, ELL students removed) to 0.48 (with a 50% ELL correction). However the sample sizes are tiny, we have as yet no other samples to validate the results with, a much smaller fraction of all Ethiopians live in Seattle (~2% as of 2013 according to the ACS) and thus those is Seattle are more likely to be unrepresentative of all, many more Ethiopians were economic immigrants (about 80%, computed based on Migration Policy figures in “Ethiopian Diaspora in the United States” — but this is based on a conservative estimate: see Terrazas (2007) for an alternative — 2/3rds of the under age 18 Ethiopian population is second generation (based on the MPI figures, page 6) and thus many more (than Somalians and first generation Hispanics) should be English language familiar, making more questionable the removal of all ELL students. And since Somalis in Ethiopia perform on par with the mentioned Ethiopian ethnic groups, there is no theoretical justification for disaggregating the groups except in an attempt to control for emigrant selection. So either we should look at Somalis alone or all East Africans together in which case we get d (weighted) = 0.77 with the 50% correction (and 0.48 without).

Later, I can reanalyze my National Achievement Semifinalist data to provide another data point — but without a robust estimate of the Ethiopian American population — or indeed the SubSaharan African American population in general– the method is pretty worthless.

IV. Other considerations

Now some will note that my estimates of the Negroid-Europid gap are relatively small. Indeed, I suggest, and have for the past several years, a 1 SD allelic score gap. Which would show up as a 0.73 SD IQ gap on tests with a h^2 of 0.65 and a g-loading of 0.9, given equal environments and representative populations. (My feelings about differential dysgenics are mixed. It’s possible that the genetic quality of Europeans with respect to complex cognitive ability has deteriorated such since the time of Galton that the gap is significantly smaller! And it’s possible that, owing to negative selection and dysgenics, the Legacy African American/White allelic gap is larger.) Now, this is less than half of Rindermann’s (2013) estimated African/ UK regional ACHQ/IQ scores gap. And it is much less than the e.g., Ethiopian/ UK national IQ score gap given by Lynn (2016, forthcoming):

Ethiopian National IQ Scores
Ethiopia 5/14 162 CPM 64 Aboud et al., 1991
Ethiopia 15 250 SPM 68 Kaniel & Fisherman, 1991
Ethiopia 15 250 SPM 65 Lynn,1994b
Ethiopia 14/16 46 SPM 69 Kozulin, 1998
Ethiopia 6/7 29 CPM 86 Tzuriel & Kaufman, 1999
Ethiopia 7/11 108 CPM 70 Ayalew, 2005
Ethiopia 7/13 2,000 SPM 66 Poppe, 2012
Ethiopia 8 1,000 CPM 68 Dendir, 2013
Ethiopia: median 68

Nonetheless, I am an HBD radical because I conjecture that the other ~ 50% is due largely to a population level effects. As I noted in my NofR paper:

The considerations above lend themselves to the conclusion that there are at least some interesting behavioral genetic differences between certain races. On the individual level, these differences do not typically seem to be large, at least by social scientific standards. This, though, does not imply that some of the known large cultural, national, and regional differences are not explainable in terms of them. It has been shown that in theory small individual level behavioral differences can amplify such to produce large population level ones (Dickens and Flynn, 2001); more specifically, population level differences involve not just the direct effects of individual level ones but they also involve the indirect effects of these, thus allowing for a population level amplification (Gordon, 1997). As such, there is no inconsistency between small individual level differences and large population level ones. This point was made by the prominent early 20th century psychologist William McDougall. In 1920, he noted, “The principle is that, though differences of racial mental qualities are relatively small, so small as to be indistinguishable with certainty in individuals, they are yet of great importance for the life of nations, because they exert throughout many generations a constant bias upon the development of their culture and their institutions” (Lamb, 1999). Ninety-four years later, after discussing a number of societal and civilizational differences in light of human biodiversity, Wade (2014) concluded much the same:“[T]hese minor differences, for the most part invisible in an individual, have major consequences at the level of a society.

Thus I can have my cake and eat the environmentalists’ one too by invoking something akin to population level active gene-environment co-variance! This population effect, which largely acts by way of the social polices designed by a country’s smart fraction, leads to increased broad and narrow cognitive differences. When you have a Robert Mugabe running a country, the measured national IQ is bound to be depressed by terrible socioeconomic decisions. This depressive effect, which loads on narrow abilities, is why international cognitive differences have repeatedly been found to be measure non-invariant, and thus not comparable to inter-individual within nation ones. Nicaraguan Nicaraguans, who are 57% Europid, really are depressed 2 SD (!) in “cognitive ability” broadly understood as Lynn (2016, forthcoming) finds. But only part of that difference represents a general intelligence one and only part of that (possibly) owes to genes. And this is why after 1 or 2 generations in e.g., the U.S. “IQ” scores or Central Americans miraculously rise. This all should be pretty obvious.

V. Conclusion

Some African immigrants to the U.S. do quite well. Based on national achievement semifinalist qualifications, I recently estimated that Nigerian Americans are no less apt than White Americans. However a cursory analysis of the vast “brain drain” research on emigrant selection shows that most African emigrants to the U.S. and the OECD in general are highly unrepresentative in education. (It’s not clear if they are “educationally selected” for per se or if they are selected for in some other trait(s) that correlate with education.) This situation opens up the possibility that they are also highly unrepresentative in cognitive ability. To get around this confound, Chanda pointed to Somali and Ethiopian proficiency rates in Seattle. Since most Ethiopian Americans were or are the descendants of highly selected economic emigrants, Seattle Ethiopian scores, even were they representative of those of the broader Ethiopian American population, could not provide dispositive evidence in support of his argument. Since Somali immigrants, mostly being nominal refugees, are much less selected, their scores could disconfirm a global Jensenist hypothesis (with respect to SubSaharan Africans and Europids) — maybe this should be called a Lynnian or Rushtonian hypothesis instead, since Jensen concerned himself mostly with U.S. differences — so long as we are willing to discount the fact and any implications drawn from this that Horn-of-African populations are substantially non-Negroid. Unfortunately, for his argument, though, the Somali performance in the U.S. — as yet — seems to be not inconsistent with a Jensenist position. More convincing data could be found. In the future it would be nice if any such was presented in a rigorous manner.

Edited 10/10/2015

7 Comments

  1. Chuck

    November 6, 2015 at 6:23 pm

    Comments are now enabled.

    • There’s no selection for intelligence among African immigrants. They’re generally brought over by their relatives, who couldn’t give a crap about their level of intelligence. Some of the most brilliant Africans remain in Africa, while some of the dumbest immigrate. You’re just deluding yourself.

      • Chuck

        June 16, 2016 at 3:14 pm

        As noted in the discussion, African immigrants from the 80s to 2000s tended to be highly educationally selected: i.e., African Brain Drain. I linked to databases showing this. Whether they are presently or are with respect to IQ is another matter.

        • You guys are always linking to “studies.” But they’re usually fake or fabricated. You can’t tell me about my own background. I have dozens of relatives here from Nigeria. No one gave any of them an intelligence test before arrival. In fact a dumb Nigerian is just as likely to immigrate as a smart one. But overall, it’s a cross-section, not some elite. The immigration process has no way of selecting for intelligence. Most Africans that come here end up doing odd jobs and driving taxis while going to school. They’re not the upper echelons or the wealthy or highly educated. They acquire their college degrees in the United States.

          I understand that this myth is of critical importance to you. You want to believe that the only Africans that can compete with and even out-compete whites is if their the very best and brightest Africans. But that myth has no basis in reality. Need I remind you that the vast majority of whites are nothing to write home about. They certainly don’t constitute any sort of intellectual elite. So one doesn’t have to be a genius or some super selected person in order to out-compete them.

          • Chuck

            June 16, 2016 at 8:43 pm

            If you are Nigerian, you might instead comment under this post, where I show, based on PSAT scores, that Nigerian-Americans are approximately as apt as White Americans. I concluded:

            African immigrants seem to have been more selected in the 1990s, so this could be a factor (7). Also, there could be a generational effect, with second generation Africans being more acculturated, etc. The relatively large number of high performing African born students is nonetheless curious given claims that African Black populations in Sub-Saharan Africa are disadvantaged, for example by parasite load or radiation exposure, in biological based intelligence. Why, if so, are so many just-off-the-boat African Blacks performing so well? Either they must not be representative of their region of origin population or their home populations must not be biologically disadvantaged, either environmentally or otherwise. The latter possibility has some interesting implications for models of global cognitive differences, many of which are predicated on the assumption that differences are not just psychometrically but also physiologically real.

            I would say that that was a rather nuanced conclusion. The Ghanese researcher, Chanda Chisala, with whom I discussed and shared the results, seemed to agree. Nonetheless, you accuse me of presenting “fake or fabricated” studies, when my footnote to my rather tepid statement, “African immigrants seem to have been more selected in the 1990s, so this could be a factor” links to none other than the often employed “IAB brain-drain data bank’s Emigration rates file”.

            I would be interested in conversing with you further, but, to be honest, it’s difficult when you deride and insult, accuse me of passing off fraudulent data, and then engage in egregious bulverism: “You want to believe that the only Africans that can compete with and even out-compete whites is if their the very best and brightest Africans.” But let me address that. Primarily, I would like to uncover the facts. Which is why I looked into the matter in the Using Surnames to Assess Ethnic Aptitude analysis, which took a couple of months. Of course, there is an ulterior motive: I would like to believe that “racism” is not suppressing contemporaneous legacy (slave descendant) Blacks performance. And one contra-explanation happens to be a hereditarian model. And one version of that is an evolutionary-heredity model. One which the performance of recent African Black immigrants complicates, owing, of course, to their racial affinity. So there is a degree of truth to what you say, but it is very slight and contingent.

            Whatever the case, I simply can’t dialogue with someone who feels the need barrage me with insults — which is unfortunate since otherwise I would have liked to hear your opinion.

  2. Congratulations on your extremely detailed analysis. Had I read your work first, I would not have bothered to comment on the Seattle schools data and the arguments put forwards by Chisala. In mild self defense, I note we are agreed that requiring English Language Learning probably scoops up children who have a more generalized difficulty with learning.

  3. There are several majority Somali high schools (upwards of 98% Somali) in Minnesota with published state standardized testing results.
    This is a thread discussing these results on a Somali website:
    https://www.somalispot.com/threads/mismeasurement-of-the-somali.31553/

    Data can be found at http://rc.education.state.mn.us/#

Leave a Reply

Your email address will not be published.

*

© 2017

Theme by Anders NorenUp ↑