Hello There, Guest!  
1 2 3 4 5 6 Next 

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

#1
Dear Editors,

We attach a manuscript entitled, "Putting Spearman’s Hypothesis to Work: Job IQ as a Predictor of Employee Racial Composition". We ask that it be peer-reviewed for potential publication in ODP.

However, we cannot completely comply with your submission requirements, as one column in our data set (Wonderlic IQs by job) is copyrighted / proprietary.

Please advise,

Sincerely,

Bryan


Attached Files
.docx   Spearman_odp.docx (Size: 31.88 KB / Downloads: 442)
 Reply
#2
In general, I don't see any large problems with this study. An idea was tested with the relevant datasets in an appropriate way.

All quotes are from the authors/the paper unless otherwise stated.

Quote:However, we cannot completely comply with your submission requirements, as one column in our data set (Wonderlic IQs by job) is copyrighted / proprietary.

One cannot copyright datasets (https://en.wikipedia.org/wiki/Sui_generi...ted_States ). This column is not copyrighted. Note that this information has already been published elsewhere, e.g. by Gottfredson (1997). This was however the old version (1992 as I recall). If you don't dare publish it, please ask Wonderlic for permission to share this dataset.

Speaking of which, did you cross-check the IQ by job with the old Wonderlic data? It may not use exactly the same job types, however. Jobs change over time, of course.
There is also a large collection of IQ by job types here: http://emilkirkegaard.dk/en/?p=5644 Most of them are not as extensive as yours.

Aside from that one column, I note that the remaining dataset is not published attached either. All data that can be shared, must be shared for publication in this journal, so I'd like the authors to please provide the dataset. The best solution is to use OSF and make a public repository: http://osf.io/
However, I will do this for the authors, if they do not have the time or expertise.

Quote:Scores on diverse sets of mental abilities tests correlate positively.

I would cite Carroll 1997 and/or Dalliard 2013 for this.
http://www.cambridge.org/us/academic/sub...ic-studies
http://humanvarieties.org/2013/04/03/is-...-g-a-myth/
The first being the conservative choice.

Quote:Biologically, g likely reflects the speed and efficiency with which brains process information (Jensen, 1998, 2011; Pesta & Poznanski, 2008).

Perhaps take a look at: http://www.nature.com/nrn/journal/v11/n3...n2793.html
There are more studies too. There is a brief summary of these findings in http://emilkirkegaard.dk/en/?p=5034 (unfinished research project)

Quote:Galton (1892) first documented Black / White (e.g.) differences on cognitive tests, and these differences have persisted in the literature since then (Roth et al., 2001).

I find that hard to believe. Galton did not have any IQ tests since the first was invented by Binet some years later. He did make some rough estimates of the race groups. Perhaps that's what you have in mind?

Quote:This century, Roth et al., (2001) conducted a massive, meta-analytic summary of race / IQ differences (N = 6,246,729 people). They reported an overall Black / White effect size of 1.10.

Perhaps the most comprehensive and recent analysis is that unpublished study by Fuerst at
http://humanvarieties.org/2013/01/15/sec...ility-gap/
I told him to get it published. Let's see what happens! :)

Quote:Other racial and ethnic groups show different mean IQs. The Asian IQ mean is estimated at 106 (Rushton & Jensen, 2005).

This is ”Asian” in the US. Asian is a broad category that includes e.g. Indians. Indians in the US do very well (https://en.wikipedia.org/wiki/Indian_Ame...ioeconomic ), and may be the top scoring group, but in India, they do very poorly (mean IQ around low 80s). You should clarify.

Quote:The digit span task is a classic example illustrating this effect. Here, the tester vocalizes a random number string (e.g., 5, 4, 7, 1, 9) at the rate of one number per second. The respondent then immediately attempts to recall each number in either forward (i.e., 5, 4, 7, 1, 9) or backward (i.e., 9, 1, 7, 4, 5) serial order. The backward version is roughly twice as g-loaded as the forward version, and it produces roughly twice as large a difference across Black / White groups (Jensen & Figueroa, 1976).

This claim was also made last time and it is not right. See:
http://openpsych.net/forum/showthread.php?tid=239

Quote:Lower scores on a DOT element indicate higher complexity and / or stature.

Think you mean status? Stature is height. Perhaps you mean it in the extended meaning of “social standing” or so.

Quote:Tying together IQ, race, and job complexity, we return to the method of correlated vectors. We argue that a job is analogous to a mental test. Different jobs require different IQs (i.e., levels of g) to perform successfully. A novel test of Spearman’s hypothesis is to use the IQ of various jobs as predictors of the racial composition of their employees. If Spearman’s hypothesis is correct, as job IQ increases, the percentage of Black workers holding the job should decrease (but the percentage of White and Asian workers should increase). The opposite is predicted as job IQ decreases.

I would definitely recommend reading:
Gordon, R. A. (1997). Everyday life as an intelligence test: Effects of intelligence and intelligence context. Intelligence, 24(1), 203-320.
http://citeseerx.ist.psu.edu/viewdoc/dow...1&type=pdf

Great paper. You are doing much the same. You treat the different jobs at items in a test, with their own g loadings and then do a test of Spearman's hypothesis on these.

By the way, this kind of study has been done for Academic majors (students):
http://www.sciencedirect.com/science/art...6902002155

It has also been done for gender, although this is more dubious due to the lack of agreement on whether there actually is a mean IQ difference or not.

Quote:We coded all but percent-Hispanic, as the BLS considers this group to be an ethnicity versus a race. That is, people in the BLS’ Hispanic category could be either White, Black or neither race. We also coded the identical data set for the year 2012 (BLS, 2013). We did this to cross-validate results found with the 2014 data, and to test whether the demographic content of different jobs is stable across years.

You say, “we”. Did you have two coders (there are two authors)? If so, where there any disagreements about the coding? How were these resolved?

Quote:Although the effect is not large, the mean IQ of all jobs, 104.5, is higher than the population IQ mean of 100 for individuals. We speculate this is likely because there are not many jobs with considerably low IQ requirements. That is, few jobs exist where the minimum IQ is in the 10th or 15th percentile. Consistent with this is the fact that the lower one’s IQ, the less likely he or she is to be in the labor force (see, e.g., Gottfredson, 2003; Herrnstein & Murray, 1994). At any rate, the mean value of 104.5 represents the population parameter for all jobs listed in the Wonderlic (2002) test manual.

This corresponds to the low correlation found between being out of a job and IQ at the individual level. You could back-estimate this correlation using this mean. Just a minor check.

Quote:As a cross check on validity, we correlated these same variables, but used the 2012 BLS data. The correlations were .46, -.57, and .20 respectively. Also, the 2012 and 2014 BLS data showed very strong stability in the demographic makeup of jobs over a two-year period. The correlations were .88, .87 and .87 for percent White, Black, and Asian (paired as 2012 and 2014 values), respectfully.

It would be better to just average the BLS values across years, no?

I also recommend reading the following paper, which is similar to yours:
McDaniel, M. A., & Kepes, S. (2014). An Evaluation of Spearman's Hypothesis by Manipulating g Saturation. International Journal of Selection and Assessment, 22(4), 333-342.
http://emilkirkegaard.dk/en/wp-content/u...ration.pdf

This study checked SH by creating different IQ tests out of a collection of subtests. As expected, the group difference goes up the higher the g-loading of the IQ test created.

Quote:We found consistent support for this hypothesis. As job IQ increased, the percent of Black workers decreased. Conversely, the percent of White and Asian workers increased. Although the correlation was small for Asians, it was non-trivial for Whites, and large for Blacks. Spearman’s hypothesis is therefore supported in a novel way.

I'd like the authors to add some ideas for why the Asian% X IQ is so low. Theory-wise, it is predicted to be larger due to Asians' mean IQ being further away from the mean IQ, I think.

My ideas: 1) Asians are very clustered in certain geographic areas where not all job types are found. 2) Asians have substantially different job interests (no idea if this is true), which therefore throws the correlations off.

Note that the reliability of Asian% was just as high as that for White%/Black%, so it cannot be a reliability issue.

I'd like to authors to use multiple regression to predict the racial%'s. Perhaps the Asian% finding is due to Asians having some large preference for not working with people or preference for working with things. Multiple regression may be able to shed some light on this. Does the DOT supply full RIASEC scores by job or only the three datapoints listed?

Quote:Consistent with past research (Gottfredson, 1986; 2003; Schmidt & Hunter, 2004), we found that job complexity and IQ are strongly related. The DOT element, Data, correlated very strongly with job IQs as reported by the Wonderlic (2002). The People (but not Things) element was also a non-trivial predictor of job IQ. Again, job complexity and IQ are intrinsically linked.

I would like the authors to provide the effect sizes of the prior research so readers can see whether the effect sizes are similar. This is needed for cumulative science.
http://www.phil.vt.edu/dmayo/personal_we...esting.pdf
 Reply
#3
Thanks for the prompt and fair review. I will be addressing concerns piecemeal over the next few days. Here, I attach the data file. Based on your comment, I decided to indeed include the IQ scores. It's SPSS and I hope it's easy to convert.

Bryan


Attached Files
.sav   job_iq2.sav (Size: 19.25 KB / Downloads: 457)
 Reply
#4
Sounds good.

I have converted the data to CSV format (program neutral) and uploaded the files to an OSF repository. If the authors make users on OSF, I will transfer control of the repository to them.

https://osf.io/pxmjc/files/
 Reply
#5
This was a straightforward, clever analysis detailed by a remarkably lucid write up. I just have a few nitpicks:

Quote:Race groups differ... Galton (1892) first documented Black / White (e.g.) differences on cognitive tests...The Asian IQ mean is estimated at 106 (Rushton & Jensen, 2005)... individuals of different races.

There is some semantic sloppiness here regarding the race/ethnic classifications. For example, Galton's Blacks (Sub-Saharan Africans) are not the same as U.S. Blacks either culturally or genealogically -- the latter being admixed. And U.S. Asians, a gallimaufry of peoples from a vast continent, are not the same as Rushton & Jensen's (2005) North East Asians. Reword this section.
Maybe additionally clarify the groupings e.g., "race and ethnicity as defined by the Office of Management and Budget, groups defined in term of regional ancestry" or however this is typically done in IO psych, if it is.

Quote:A parsimonious explanation for race / IQ differences is Spearman's hypothesis (Spearman, 1927; see also, Jensen 1985). It proposes that IQ differences are g differences.

In practice, SH refers to "weak" SH according to which differences "mainly" reflect g differences.

Quote:The literature strongly supports Spearman’s hypothesis (for reviews, see, e.g., Jensen, 1998; Rushton & Jensen, 2005).

To cover bases, maybe add a cf. and cite e.g., Dolan, C. V. (2000). Investigating Spearman's hypothesis by means of multi-group confirmatory factor analysis. Multivariate Behavioral Research, 35(1), 21-50.

Quote:If Spearman’s hypothesis is correct, as job IQ increases, the percentage of Black workers holding the job should decrease (but the percentage of White and Asian workers should increase).

"Should" assuming, as Gottfredson noted, minimal social interventions.

Quote:We coded all but percent-Hispanic, as the BLS considers this group to be an ethnicity versus a race. That is, people in the BLS’ Hispanic category could be either White, Black or neither race.

So you couldn't code for Hispanics or you didn't? I don't see why the ethnicity/race distinction would matters, as SH is thought to hold for Hispanics as an ethnic group. Or was the relevant data missing?

Maybe note whether the associations were stronger or weaker than what Gottfredson found. I recall she predicted that they would be attenuated owing to e.g., affirmative action for medical school. (In Gottfredson (1987) she notes that she found a r(cognitive complexity x B/W ratio) of -.5 or so.)

Gottfredson, L. S. (1986). Societal consequences of the g factor in employment. Journal of Vocational Behavior, 29, 379-410.
Gottfredson, L. S. (1987). The practical significance of black-white differences in intelligence. Behavioral and Brain Sciences, 10 (3), 510-512.

It would be interesting to see the fields for which there were strong deviations from the expected employment ratios. Maybe a figure with B/W employment ratios on one axis and cognitive complexity on the other. (I'm not asking you to include such a figure; but just noting that I would be interested in seeing one.)
 Reply
#6
Emil:

I typed my reply in Word, all nicely formatted. Pasting it in here, though, was a mess, and so I attach the Word file. Hope this is ok.

Bryan


Attached Files
.docx   odp.docx (Size: 17.17 KB / Downloads: 487)
 Reply
#7
Thanks for the review!
-Attachment

(2016-May-25, 00:41:14)Chuck Wrote: This was a straightforward, clever analysis detailed by a remarkably lucid write up. I just have a few nitpicks:

Quote:Race groups differ... Galton (1892) first documented Black / White (e.g.) differences on cognitive tests...The Asian IQ mean is estimated at 106 (Rushton & Jensen, 2005)... individuals of different races.

There is some semantic sloppiness here regarding the race/ethnic classifications. For example, Galton's Blacks (Sub-Saharan Africans) are not the same as U.S. Blacks either culturally or genealogically -- the latter being admixed. And U.S. Asians, a gallimaufry of peoples from a vast continent, are not the same as Rushton & Jensen's (2005) North East Asians. Reword this section.
Maybe additionally clarify the groupings e.g., "race and ethnicity as defined by the Office of Management and Budget, groups defined in term of regional ancestry" or however this is typically done in IO psych, if it is.

Quote:A parsimonious explanation for race / IQ differences is Spearman's hypothesis (Spearman, 1927; see also, Jensen 1985). It proposes that IQ differences are g differences.

In practice, SH refers to "weak" SH according to which differences "mainly" reflect g differences.

Quote:The literature strongly supports Spearman’s hypothesis (for reviews, see, e.g., Jensen, 1998; Rushton & Jensen, 2005).

To cover bases, maybe add a cf. and cite e.g., Dolan, C. V. (2000). Investigating Spearman's hypothesis by means of multi-group confirmatory factor analysis. Multivariate Behavioral Research, 35(1), 21-50.

Quote:If Spearman’s hypothesis is correct, as job IQ increases, the percentage of Black workers holding the job should decrease (but the percentage of White and Asian workers should increase).

"Should" assuming, as Gottfredson noted, minimal social interventions.

Quote:We coded all but percent-Hispanic, as the BLS considers this group to be an ethnicity versus a race. That is, people in the BLS’ Hispanic category could be either White, Black or neither race.

So you couldn't code for Hispanics or you didn't? I don't see why the ethnicity/race distinction would matters, as SH is thought to hold for Hispanics as an ethnic group. Or was the relevant data missing?

Maybe note whether the associations were stronger or weaker than what Gottfredson found. I recall she predicted that they would be attenuated owing to e.g., affirmative action for medical school. (In Gottfredson (1987) she notes that she found a r(cognitive complexity x B/W ratio) of -.5 or so.)

Gottfredson, L. S. (1986). Societal consequences of the g factor in employment. Journal of Vocational Behavior, 29, 379-410.
Gottfredson, L. S. (1987). The practical significance of black-white differences in intelligence. Behavioral and Brain Sciences, 10 (3), 510-512.

It would be interesting to see the fields for which there were strong deviations from the expected employment ratios. Maybe a figure with B/W employment ratios on one axis and cognitive complexity on the other. (I'm not asking you to include such a figure; but just noting that I would be interested in seeing one.)



Attached Files
.docx   odp2.docx (Size: 34.85 KB / Downloads: 384)
 Reply
#8
"Reply: Would this be something like affirmative action? The revision didn’t address this issue."

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.

"Reply: Unfortunately, we have no data on Hispanics. The USBLS allows people of any race to self-identify as Hispanic. So, that category included both black- and white-Hispanic people."

I don't see a sound justification for the exclusion.

(1) "Hispanics" are defined as a multi-racial ethnic group by the census (Directive 15). Both national samples, such as the NLSY79/97, and IQ standardization samples adopt this understanding.
(2) Spearman's hypothesis has been proposed to hold for and has been tested on "Hispanics" (a cultural groups), not just Amerindians (a racial group).
Quote:Warne, R. T. (2016). Testing Spearman's hypothesis with advanced placement examination data. Intelligence.
The nature, source, and meaning of average group score differences between demographic groups on cognitive tests has been a source of controversy for decades. One possible explanation is “Spearman's hypothesis,” which states that the magnitude of score differences across demographic groups is a direct function of how strongly the test measures g.... In this study I used the method of correlated vectors to examine the relationship between racial/ethnic group differences of Advanced Placement (AP) exam scores and the correlation between those AP exam scores and a test of general cognitive ability, the PSAT. Results are consistent with Spearman's hypothesis for White-Black and White-Hispanic comparisons, but not for White-Asian comparisons.

(3) The "Hispanic" group is not atypically defined, relative to common social science practice; rather "Whites" "Blacks" and "Asians" are since these typically refer to e.g., "non-Hispanic Whites" in the literature. (Or is the situation different in IO?)
(4) The relatively problematic delineation of e.g., "Whites" is not a problem for SH, since SH, as now formulated, simply proposes that "group differences are a function of g loadings", thus it is applicable to all sorts of culturally and politically delineated groups, which is why you can see if it hold for "Asians" (a clearly politically defined group, which has no cultural, linguistic, or genetic basis).

Unless you can come up with a better reason, add Hispanics.
 Reply
#9
Asians, as a group, does have a genetic basis, but it would be complicated to use because one would then have to include Native Americans and Eskimos (who are sometimes, sometimes not considered part of the US "Asian" category). Furthermore, many migrants from Latin/South America would also be classified as majority Asian by genomic analysis. So, yeah, it would be complicated, but doable to some degree.
 Reply
#10
(2016-May-30, 23:33:03)Emil Wrote: Asians, as a group, does have a genetic basis


No.

"Asian", in this context, is a class (one which includes: South Asians and East Asians) in a socio"racial" classification, one which has the following other divisions: Blacks and "Whites" (meaning: West Eurasians). By what set of genes could one arrange (classify) almost all of our "Asians" into their census class as opposed to that of Blacks and West Eurasians?

(One would have to reduce the class k to 2: (a) Blacks and (b) (East + South) + West Eurasians (i.e., out of Africans)).
 Reply
1 2 3 4 5 6 Next 
 
 
Forum Jump:

Users browsing this thread: 2 Guest(s)