Back to [Archive] Post-review discussions

[ODP]Putting Spearman’s Hypothesis to Work: Job IQ as a Predictor of Employee Racial
Hello Meng Hu,

Thanks for your review.

I can fix the missing bracket and add the following link as a footnote. It offers a detailed description of data, people, and things:

http://www.occupationalinfo.org/appendxb_1.html


ETA: What's next?

I like the article. That there is a prediction that the % of racial composition changes as IQ job (or complexity) increases is supportive of Spearman's hypothesis, also shows (or rather, suggests) that within-correlation of IQ*job complexity may also extend to between-group context (as Gottfredson compiled lot of research showing that this correlation holds within groups, white groups) as the hypothesis expected.

I have no problem with the consistency/stability of the correlations, and their magnitude and signs indeed are supportive of the studied hypothesis.

I would appreciate if you can describe a little bit more the variables of worker activity, namely, data, people, and things, because it may not be very clear to everyone (e.g., I have only a rough idea).

They score six (“speaking-signaling”) on people, and two (“operating-controlling) on things.


You forgot a bracket.

I was suggesting that you might qualify your statement. For example, for precision, I might say: Spearman's hypothesis would predict that group differences are larger on more g-loaded tests, assuming no countervailing psychometric bias. Likewise: Spearman's hypothesis would predict that employment differences are larger for more g-loaded fields, assuming no countervailing societal bias e.g., affirmative action or defacto quotas. If you think that the qualification is obvious, don't bother.


It has little to do with the core of this study but I don't remember that Jensen said anything like this. In fact, I think Jensen's reasoning is in accordance with Dolan, in that that if there is bias, any finding in support of SH would be unreliable. They however don't disagree about the method to detect bias (but that, you and I, we already know very well).
"Spearman’s hypothesis equates race differences with g differences, and predicts that the former will covary with how well mental tests measure the latter"

This is too strongly worded. Firstly, Spearman's hypothesis originally concerned only black-white differences even if it has later been applied (with less success) to other comparisons. Secondly, "equating" suggests a strong version of the hypothesis, which everybody agrees is unsupported. The hypothesis in its weak form suggests only that race and g differences are strongly collinear.

"We also found a very large correlation (.86) between job IQ and complexity"

I would clarify that this is an ecological correlation.

As to the discussion about whether the study really tests SH, I'd say it's an indirect test of SH that depends on the correctness of the theoretical assumptions you make. Specifically, you assume that job g can be equated with job IQ and that job complexity ratings can be equated with g loadings. Of these analyses the one with job IQ is somewhat circular (IQ varies between groups, so higher-IQ groups are better represented in higher-IQ occupations). The job complexity analysis is more interesting as it is, at least in principle, independent from job IQ data.

So, I think you can describe the study as a test of SH as long as you stress that you use theoretically plausible proxy variables for measures of g saturation.

Can you add a scatterplot (or some other kind of graph) with racial proportions in jobs on one axis and job complexity on the other (with the complexity/data variable reverse-scored)? Raw racial proportions may not be the perfect variable for your analysis, but a graph could make the relationships that exist more transparent.
Bryan Pesta, I am OK with the proposed changes. I'm giving you my approval. Although I concur with others that an analysis of the hispanic group can be done, I see no real problem with the analysis as shown.

Although it's not a condition, I would like that you add a third table for the regression analysis. I think the regression analysis is at least or more important than the correlation analysis, especially concerning the prediction about %asians. Without displaying results on a table, it may give the impression that it's of lesser importance while it's not.

It appears that values on People suppress the true correlation between Job IQ and percent Asian.


If you want, say explicitly that People variable act as a suppressor effect, as far as job IQ is concerned.

See below:
http://ericae.net/ft/tamu/supres.htm

Conger (1974) provided another definition of suppressor variables as, "...a variable which increases the predictive validity of another variable (or set of variables) by its inclusion in a regression equation" (pp. 36-37).
Hello,

I think I've addressed all concerns-- except one-- since version 2, dated 6/16. Please let me know if I missed anything.

1. De-hark. Done.
2. Change use of the word "correlation" to "relationship" regarding the true results for percent Asian.
3. I've included a footnote that links a detailed explanation of the DOT elements.
4. I fixed the missing bracket / quotation in the intro.
5. I toned down the definition (e.g., dropped "equates") of SH in the abstract.
6. I've mentioned the correlations are aggregate / ecological in the discussion.
7. I mentioned that IQ scores are only proxies for g in the limitations paragraph.
8. I took care of the sentence on supression and stated that this is a suppression effect.
9. I created Table 3, which includes Emil's suggested regressions.

I did not include a graph or figure showing how complexity and ethnicity relate. I spent some time on this with SPSS and the graphs looked stupid / non-informative. I think this is because I lack skill at making pretty graphs, but also because "Data" is binned into just seven possible values (0-6) on one axis, and on the other axis, things get smooshed, as % White hovers in the 70s, and the other race variables hover in the 10s.

Let me know if I missed something, and thanks to all for your input.

Bryan
Personally, I have nothing to add to what I already said before. Just waiting to have a look at your final version.
Can you post the newest version of the article? I accept it for publication, but I could take a look at the updated version.

I think a stacked bar chart with mean racial proportions on the y axis and job complexity on the x axis would work. Something like this (the other14 variable is 100-the other racial proportions, jobcomplexity is the data variable reverse coded):



Here's the SPSS code for that:

COMPUTE jobcomplexity=6-data.
EXECUTE.
COMPUTE other14=100-white14-black14-asian14.
EXECUTE.
GRAPH
/BAR(STACK)=MEAN(white14) MEAN(black14) MEAN(asian14) MEAN(other14) BY data
/MISSING=LISTWISE.
Wow, I could of swore I included the paper as an attachment. Here it is!

I'm ok with adding that figure, if you're ok with me stealing it!

Bryan

Can you post the newest version of the article? I accept it for publication, but I could take a look at the updated version.

I think a stacked bar chart with mean racial proportions on the y axis and job complexity on the x axis would work. Something like this (the other14 variable is 100-the other racial proportions, jobcomplexity is the data variable reverse coded):



Here's the SPSS code for that:

COMPUTE jobcomplexity=6-data.
EXECUTE.
COMPUTE other14=100-white14-black14-asian14.
EXECUTE.
GRAPH
/BAR(STACK)=MEAN(white14) MEAN(black14) MEAN(asian14) MEAN(other14) BY data
/MISSING=LISTWISE.
Admin
It is easier to keep track of files if they are put on OSF instead of as attachments here. Recall that I made a repository for the paper here: https://osf.io/pxmjc/

I will of course transfer the control of this repository to the authors if they make a user on the website. In the meanwhile, I have uploaded the (3rd) version of the paper posted above.
OK, I approve publication. Go ahead and use the figure, although I'd suggest you divide the y axis values by 100 and use more descriptive labels.
Admin
I will be attending the ISIR conference in Saint Petersburg and may not be able to publish this paper to the website until I get.

http://www.isironline.org/2016-st-petersburg-russia-july-15-17/
I will be attending the ISIR conference in Saint Petersburg and may not be able to publish this paper to the website until I get.

http://www.isironline.org/2016-st-petersburg-russia-july-15-17/


Enjoy ISIR.

I see LG is presenting a paper similar to mine there...

If you see Mike McDaniel, please buy him a drink on me. Ask him about the heathen in Ohio...
Admin
The paper has 3 approvals as required: Hu, Fuerst and Dalliard. Last reviewer (me/kirkegaard) disagrees due theoretical framing but not the actual analyses.

So, the paper can be published when the author submits a final version.

We are considering starting a service about creating prettier, more official looking papers for a fee. This is voluntary and will always be so. If wanted, Julius will set up the article in LaTeX, so that it looks like this paper. The price depends on the time it takes to set it up, but would presumably be around 30-50 USD, all of which goes to Julius.
The paper has 3 approvals as required: Hu, Fuerst and Dalliard. Last reviewer (me/kirkegaard) disagrees due theoretical framing but not the actual analyses.

So, the paper can be published when the author submits a final version.

We are considering starting a service about creating prettier, more official looking papers for a fee. This is voluntary and will always be so. If wanted, Julius will set up the article in LaTeX, so that it looks like this paper. The price depends on the time it takes to set it up, but would presumably be around 30-50 USD, all of which goes to Julius.


I'm happy to have Julius make the paper pretty. I've added the figure and made reference to it in the results section.
Admin
Because of this paper, I decided to read (or maybe re-read) the original Spearman's hypothesis paper.

Jensen, A. R. (1985). The nature of the black-white difference on various psychometric tests: Spearman's hypothesis. Behavioral and Brain Sciences, 8, 193—219.

It contains this quote:

The social context of g
The only commentator who brings Spearman’s hypothesis directly and specifically into apposition with its real-life social and economic consequences, is Cattell, who predicts that the percentage of blacks in different occupations should be inversely related to the mean intelligence levels of persons employed in the occupations. If shown to be true, this prediction would mean, of course, that disparities in the proportional representation of black and white workers in various occupational categories are not mainly attributable to prejudice and discrimination in hiring, but are due to differences in measurable g-loaded abilities, whatever the cause of the differences. I have not looked into data on this point myself, but quite precise data on a range of occupations (ranging from physician and engineer to truck driver and meat cutter), directly aimed at Cattell’s prediction, have been assembled by Linda Gottfredson (personal communication), a sociologist at the Johns Hopkins University. In light of Cattell’s query, it would be most valuable if Gottfredson submitted this analysis to Continuing Commentary. Gottfredson’s analysis, based on 1970 and 1980 statistics from the U. S. Department of Labor and the Bureau of the Census, strikingly bears out Cattell’s prediction, with a near perfect rank-order correlation between the theoretically expected and the observed ratios of black to white employees in different occupations.


Which is very close indeed to the idea with this paper.

---

At the ISIR 2016 conference, Linda Gottfredson (who also received the life-time achievement reward) presented the following talk (abstract):
http://www.isironline.org/2016-st-petersburg-russia-july-15-17/

Differences in the Distribution of G and Its Effect on Culture: The USA as a Case

Galton introduced a biometric approach to studying human populations. It was the foundation on which Spearman, Eysenck, and other staunch empiricists would erect the London School of Psychology’s biological approach to psychology. All were keenly interested in, not just the structure of intelligence, but also how its distribution within populations affects national well-being (e.g., Eysenck, 1973), including the mental distance between individuals, social cohesion, cultural level and social habits, diffusion of information («percolation range»), rate of innovation, proneness to extreme (simplistic) political positions, and economic divergence (Cattell, 1938). Drawing on these mechanisms, I look at how policies to achieve occupational parity, in spite of persistently large racial differences.

Data are publicly available in the United States for (a) number of black and white males employed in detailed occupations, (b) the ranges of IQ from which various occupations recruit their workers (recruitment ranges), and (c) the IQ distributions of blacks and whites. Thirty years ago (Gottfredson, 1986), I used these data to estimate whether black males were employed at the same rate as white males in 1970 and 1980 when they fall within the recruitment ranges. I examined 9 large occupations (physician, engineer, secondary teacher, real estate sales, fire fighter, police officer, electrician, truck driver, and butcher/meat cutter) falling in four IQ recruitment ranges (86–112, 91–117, 109–134, and 114+). Here I update those analyses with employment data for 1990 and 2005–2009.
Employment trends:
• If black and white males were recruited from the same IQ ranges, we would expect the B/W ratio of %-of-blacks to %-of-whites in an occupation to rise from .05 (physicians, engineers) to .72 (truck drivers, meat cutters).
• In 1970 and 1980, the B/W ratios for actual employment were more consistent with black males being recruited from .5 SD below that for white males (.22 to 1.07).
• The B/W ratios for actual employment rose in 1990 and 2005–2009; half were commensurate with black males being recruited from 1 SD below white males. Cultural trends: More blacks in good jobs and elite schools; working class white males have worse health and job prospects; institutions restructured; race relations worse; political discontent.

Human variation in intelligence is a biological fact that poses social challenges, especially when identifiable groups differ noticeably in mental ability. How a culture reacts to that variation also shapes its future. A nation that ignores or denies the practical and social significance of substantial intelligence differences risks its well-being. Constructive alternatives are available.

• Cattell, R. B. (1938). Some changes in social life in a community with a falling intelligence quotient. The British Journal of Psychology, 28, part 4, 430–450.
• Eysenck, H. J. (1973). The inequality of man. London: Temple Smith.
• Gottfredson, L.S. (1986). Societal consequences of the g factor in employment. Journal of Vocational Behavior, 29, 379-410.


So, the suggested study was in fact carried out using a method like I advocated above (demographic approach). First I used minimum IQs, and later realized that one should use recruitment ranges as LG did.

The value of the new study is three-fold. First, that it redid the older analyses with a much larger number of jobs. LG's original study used only 9 job types, the new study uses more than 100.

Second, the new study also examined Asian proportions, not just Black and White. It could have also examined Hispanic, but the authors chose not to do so.

Third, the new study incorporated the interest dimensions as well. These seem to be crucial to explain the unexpected Asian proportions.