Using the General Social Survey, I transformed the respondent’s 1998 income variable (RINCOM98) and then did a multiple regression analysis.
I expected to discover that verbal IQ (based on the 10 item Wordsum test) would be a bigger determinant of income than the respondent’s degree. I was completely wrong.
It turns out that verbal IQ has hardly any impact at all on income after the highest degree completed is added to the regression analysis. The biggest impact is at the bottom of the bell curve, where having a very low IQ causes a drop in income of $5,300 (although not as big as the $7,700 drop associated with being a high school dropout).
Having an IQ slightly above average (WORSDUM 7-8) yields only a $3,200 boost to income, in comparison to $13,900 for a bachelor’s degree and $26,900 for a graduate degree.
Now here’s the kicker. Having an IQ in the highest bucket (WORDSUM=10) results in a $2,500 lower income than just having an average IQ, and a $5,700 lower income compared to having a slightly above average IQ.
Does this mean IQ is useless? Not exactly. IQ has a strong impact on whether or not a person obtains a bachelor’s or graduate degree in the first place. However, the regression analysis indicates that the credential is more highly valued in the workplace than actually being intelligent.
Of course this analysis is riddled with problems. First of all, although the ten item Wordsum test appears to be a pretty good test considering how short it is, it’s still only a ten item test and therefore has a high margin of error. If respondents’ IQs based on a longer and more comprehensive test were included in the GSS, then the more reliable IQ score would surely show a stronger impact in the regression analysis.
However, the degree data is also lacking because it does not distinguish between degrees. A degree from the College of Staten Island is given the same weight as a degree from Harvard. And a degree in something with little employment value such as sociology is given the same weight as a degree in engineering.
Furthemore, Wordsum has shown to be highly predictive elsewhere in the GSS. For example, the higher the Wordsum score the more likely it is that the respondent believes the Bible to be just a book of fables. And Wordsum of 9 or 10 is more highly predictive than having a graduate degree.
The unreliability of Wordsum and the unreliability of the degree designations probably approximately cancel each other out, so it’s more likely than not that above average IQ has little impact on income besides helping the high IQ person get credentialed.
And does having a very high IQ actually lower one’s income? Possibly. Perhaps, once a person is able to demonstrate that he can do a job with a minimum level of competence, social factors such as being liked by your boss are more important to getting a raise or promotion than performing at a level beyond minimum competence. And perhaps the high IQ person doesn’t get along as well with his boss.
The GSS is not able to shed any light on what’s going on at the very top. How are the highest salaries (such as $150,000 and beyond) related to having high IQs beyond the measurement of Wordsum (such as top 2% or higher)?
Below is the regression analysis, which may not be entirely comprehensible:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
FOLLOWUP POSTS
More IQ followup
High IQ does not result in higher income, part II
I think what you're detecting is more or less that someone who is intelligent but fails to complete school often has other attributes that hinder success in the working world. While completing a college degree is not the same thing as working at a job, I can imagine many things (alcoholism, laziness, lack of 'future time orientation' or whatever) that would make either endeavor difficult. I'm not trying to rag on the dropouts, either - I'm one of them :-).
I agree that wordsum doesn't shed much light on the very high end. WORDSUM = 10 would correspond roughly to an IQ cutoff somewhere in the lower 120s, which is not that high.
Posted by: bbartlog | June 20, 2006 at 09:31 AM
It could be people with high IQs are just smart enough to realize that after a certain point, quality of life tradeoffs aren't worth it. I know that I personally chose a job which doesn't offer the highest possible salary but makes up for it in many other ways.
Posted by: JewishAtheist | June 20, 2006 at 09:41 AM
I tend to agree that bbartlog's suggestion has merit: that personality factor X causes both completion of more degrees as well as higher salary independent of degrees, but that factor is only being measured in the regression through the existence of completion of a degree, causing the value of the degree to seem higher than it really is.
What is factor X? I also agree that it probably has something to do with conscientiousness, a known Big Five personality factor.
I don't agree with Jewish Atheist's proposition that smarter people value QoL more than money--this just hasn't been my observation.
Posted by: Half Sigma | June 20, 2006 at 10:05 AM
Unfortunately for theory, 5 factor items show only a low correlation with life outcomes compared to IQ. OTOH, 5 factor tests are poor measures. Better tests would surely give larger correlations. The same is of course true for wordsum.
One question I really want answered with respect to 5 factor tests is "controlling for "g", does Openness correlate and negatively with life outcomes?" It is often said that O correlates with g, but it is also often said that it doesn't correlate with life outcomes.
I wish GSS had an item for "facial symmetry" and some measure of muscle mass and vocal quality. Does it have "height" and "weight"?
Did you look at degree controlling for age? Wordsum controlling for age? Does the data discard outliers for income or class income in large categories including an X+ category where X isn't that large?
Posted by: michael vassar | June 20, 2006 at 11:06 AM
Michael,
I am still searching for questions in the GSS that might in some way correlate with any of the Big Five personality factors.
Unfortunately the GSS has no entries for physical attributes. While facial symmetry, muscle mass, and vocal quality would be hard to measure, it would be pretty easy to ask repsondents their height and weight. It's a shame that it's not asked. I am sure that height would have a positive correlation with income. And being fat is known to be correlated with low social class (a fact which Paul Fussell wrote about at length in his book Class).
The income data has no outliers for income because the highest category is $110,000+. I transformed this to $125,000 for the regression analysis.
1998 seems to be the last year with income data and has the best income data. In earlier years, the largest bucket was set at too low of an income. It seems that the social "scientists" who decide what questions to put into the GSS are more interested in studying poverty than studying wealth. This is unfortunate because wealth is a lot more interesting than poverty.
The reasons for poverty really aren't that difficult to figure out. Just turn on the Jerry Springer show and think about how his guests are different than the white collar college educated people you know.
Posted by: Half Sigma | June 20, 2006 at 11:18 AM
Half, there is probably less difference than you think, if you count people from community colleges and the state school I went to.
I apologize in advance for my statistical clumsiness. My theory is that high verbal IQ people are more likely than average people to gain higher education levels than are typical for their family's income and educational background. However, perhaps background is more important to one's future income than is education. Then, the educated people with average IQs would be more likely than the educated high-IQs to have the important background characteristics for financial success.
Therefore, the high-IQ educated people would earn less money than the average-IQ educated people -- but because of their lower-status backgrounds, not because of their higher IQ. An average or slightly above average person from a well-connected family will earn more as a Harvard grad than some nerdy schoolteacher's kid who bookwormed his way in.
Posted by: Nancy Spungen, Esq. | June 20, 2006 at 02:19 PM
Halfsigma: do you have the list of Wordsum questions? That would be really useful to those who aren't experienced with this sort of test.
Nancy: what you said is surely true, but is irrelevant because the Wordsum test we are discussing is only concerned with differentiating truly average from pretty low and pretty high, not with distinguishing both from the sort of nerdy kid who gets into Harvard. Even the connected people who get into Harvard will have wordsum scores of 9 or 10, and probably 10. A wordsum score of 10 is probably about equivalent to a SAT verbal of 600 or 650, comparible to Kerry and GW.
Posted by: michael vassar | June 20, 2006 at 02:37 PM
Careful now. None except WORDSUM 0-3 are significant.
This is something I should have done during my own analysis, so I apologize. But remember that if you are making 20 comparisons, you will get on average one where p=.05 just by chance (a p=.05 means a 5% chance a correlation is due to chance).
Posted by: SciFiGeek | June 20, 2006 at 04:36 PM
Let me correct. None of the WORDSUM results except for 0-3 are significant.
I do not disagree with the basic thesis that you can be too smart for your own good. However, from what I have read the optimal IQ is about 130, which probably corresponds to WORDSUM 10.
Posted by: SciFiGeek | June 20, 2006 at 04:40 PM
The exact content of the Wordsum questions are not in the codebook.
I would make an educated guess that a Wordsum of 10 is equivalent to an SAT score of no higher than 1300 on the new post-1995 scale.
The median SAT score at elite schools (such as Harvard) is higher than 1300, so the majority of students at elite schools would score 10 on the Wordsum.
I agree with Nancy that parents' social class would have a positive impact on income independent of both IQ and education. And also that the variables are correlated with each other because a higher social class predicts college attendence independently of IQ, and also predicts IQ because IQ is hereditary.
I don't think that adding parents' social class would have much of an impact on the results of this particular regression because it's already being indirectly measured twice--as a component of WORDSUM and a component of DEGREE. I also think that the U.S. is much more of a meritocracy in the range of five figure incomes than it is for higher level incomes.
In the 1998 sample, 75% of respondents with WORDSUM(10) aged 25 and older had at least a bachelor's degree. Perhaps in the future I will attempt some analysis to find out who are the 25% who are smart enough to excel at college but somehow wind up not getting a degree.
Here's a another guess about very high IQs: the problem is that the data doesn't capture high salary jobs (above $110,000) where people with WORDSUM(10) would make up a disproportionately high percentage of the workforce. In those jobs, top 6% IQ is an advantage or even a necessity to perform the job at a minimum level of competence, but at lower level jobs high IQ people may actually be poorer quality workers because they are bored with their work and have a poorer attitude towards work because they perceive themselves (correctly) as being underemployed.
scifigeek: "None except WORDSUM 0-3 are significant."
The magic of the computer output says that there's a 73% chance that WORDSUM(10) has a negative correlation with income. Yes, that doesn't prove very well that the coefficient is negative. I think there's a greater chance than that that the coefficient for WORDSUM(10) is lower than for WORDSUM(9) but I'm not sure how to calculate that probability.
Posted by: Half Sigma | June 20, 2006 at 04:54 PM
About 1/19 of the people who answered had a WORDSUM of 10, corresponding to the top 5% of the population or so, which is 1.6 SDs above normal; an IQ of 124 or higher.
1.5/19 have a WORDSUM of 9, which is the next 7%, or an IQ of 117 and up.
Statistically speaking you have two treatments WORDSUM 9 and WORDSUM 10, so you'd use a two-sample t test (practically speaking a z-test with this big sample). Not sure how to do that in GSS either.
Posted by: SciFiGeek | June 20, 2006 at 06:46 PM
95% confidence intervals for the RINCOM8 means for WORDSUM 9 and 10 overlap. But there are better ways to do this...
Posted by: SciFiGeek | June 20, 2006 at 10:30 PM
My suspicion is that people with lower IQs, when obtaining a degree, compensate for their lower intelligence by working harder than their high IQ counterparts. As a result, when they are out in the workforce, they not only have the same qualifications as their more intelligence peers but also superior work habits.
Posted by: tommy | June 21, 2006 at 04:30 AM
"However, the regression analysis indicates that the credential is more highly valued in the workplace than actually being intelligent."I'd really try to be more careful when you make statements like this. Especially in a blog which is purportedly trying to cut through the clutter with facts and logic.
Posted by: jult52 | June 21, 2006 at 08:32 AM
Remove 'the regression analysis' from that and I think nobody would disagree.
Posted by: SciFiGeek | June 21, 2006 at 09:10 AM
I suspect that smart people who get college degrees are more motivated and organized and wise than smart people who do not get college degrees. So the performance of college grads is not so much due to the credential as it is due to the qualities that caused them to get a credential.
Posted by: Randall Parker | June 24, 2006 at 12:39 AM
I've taken numerous IQ tests over the years and I have always scored exceptionally well on tests in high school. My lowest IQ score ever was 128 and my highest 150. My SAT was 580 math 640 verbal which I forgot to bring calculator too, otherwise it may have been closer to a 1300. My high school GPA was roughly an 86 in 2002 when I graduated. In college after my first two years I've done pretty horribly my currently GPA graduating this fall is roughly a 2.3. But let me offer some insight into other possible cuases for IQ/Income descrepency. yes the underworked theory is indeed plausible, I used to work for a large grocery store chain my final years of high school and well into college suffice to say it was one of the worst decisions of my life to stay there so long. I try not to think about how working 30 hrs a week and going to school hindered my education. Also working doing the most menial jobs was so "boring" that getting to work on time was a minor factor, and getting along with the "lifers" was a depressing and annoying task to say the least. Perhaps this lead me away from my studies even more so... which led to going out with friends and drinking and partying too much. Oh yeah did I mention I go to SUNY Albany former #1 party school two years in a row. Which makes me regret even more not just shelling out the money for a private, more prestigious institution. If I didn't have to work to pay for my transportation, insurance, food, and other bills would I have done better in school? Probably, at least alittle better. I still would have most likely partied alot, and gone out even on weekdays which I did quite often. But financial factors did play a large part in hindering my time constraints. So now I'm stuck at the end of my undergrad... with my miserable 2.3 index praying to my agnostic god(s) that I will annihilate the LSAT so that some feeble law school might admit me.
Posted by: Steve Rossello | September 27, 2006 at 12:37 AM