Because my previous post on this subject caused quite a bit of controversy, I redid the regression analysis, this time using the 1991 income data in the General Social Survey (GSS). I’m looking at code RINCOM91, which is the respondent’s income from his job.
Unfortunately, the largest bucket here is only $75,000+, which I transformed to $85,000. As I wrote in my previous post on this topic, the GSS is unable to tell us what’s going on amongst the highest incomes. However, only 3.7% of respondents reported income of $75,000+.
As with my analysis of the 1998 data, after degree, age, sex and region (residents of New England, Northeast, and Pacific regions have slightly higher incomes) are accounted for, the results show no benefit to having an above average verbal IQ. At least, unlike in my 1998 analysis, the highest verbal IQ bucket (WORDSUM=10) doesn’t correlate with a drop in income. As in the 1998 analysis, there is a noticeable negative effect from having a below average IQ, but this effect is not quite as large as the effect from being a high school dropout.
One interesting finding in the 1991 analysis is that black people (RACE=2) are predicted to have higher incomes, but the correlation isn’t statistically significant. My interpretation in conjunction with the 1998 analysis is that being black doesn’t have a statistically significant negative impact on income.
Commenters who are familiar with the IQ literature, and that includes the visitors from the blog Gene Expression, are quite surprised at my findings. Everyone familiar with the IQ literature seems to somehow take it for granted that high IQ causes higher income independent of educational credentials.
In fact, the literature doesn’t say this. Someone gave me a link to an article by Charles Murray, “Income Inequality and IQ” (link to pdf file), but none of the findings in that article contradict my findings that high IQ doesn’t result in higher income after educational credentials are accounted for.
Charles Murray’s points, which are entirely supported by the GSS data, are that (1) IQ by itself is correlated with higher income; (2) IQ is correlated with higher educational attainment; (3) higher educational attainment is correlated with higher income. Clearly the means by which IQ correlates with higher income is through educational credentials; smarter people are more likely to get a college degree or a graduate degree. This is what the GSS data shows. But it’s the credential that employers seem to value and not the employee’s intelligence.
Murray’s article also has some confusing stuff about the correlation between IQ and occupational prestige. Occupational prestige is not the same thing as income. I can think of plenty of comparisons where the higher prestige job has lower pay. Most college professors are not especially well paid, but the occupation has a very high prestige. When comparing low IQ and high IQ people in the same income bracket, the high IQ person will more likely be in a more prestigious profession, so therefore IQ will have a higher correlation with prestige than it does with actual income.
The IQ literature is also full of studies which show that IQ is correlated with job performance. IQ has a higher correlation with job performance than any other indicator such as education, experience, or performance on job interviews. Some people naively assume that this means higher IQ people are paid more. But few employers actually use IQ tests to select job applicants so high IQ doesn’t help people land better jobs. Furthermore, it seems that after high IQ people get hired, their superior job performance doesn’t result in any raises or promotions. My previous blog post, Do companies really want to hire the best employees?, is relevant here.
A commenter pointed out that some doctors are better than other doctors, but this is truly an example of how job performance has no impact on income. Doctors get reimbursed the same amount by insurance companies whether they are great or just barely competent. If anything, the system favors the unethical doctor who lies on insurance forms and recommends unneeded but profitable procedures.
THE PROBLEM WITH USING A VOCABULARY TEST
WORDSUM is a vocabulary test. While mostly measuring g, it is also measuring the respondent's verbal ability independent of g.
Although a high score on a vocabulary test would also predict a high score on a math test because both tests measure g, the higher the measured verbal score the more likely it is that the respondent has greater verbal ability than math ability.
When a college student chooses a major, his ratio of verbal to math ability will probably have a big impact on his decision. The high vocabulary repsondents in the GSS will therefore be more likely to choose a verbal major like Roman history or journalism instead of mathematical major like engineering or accounting.
It is my observation that mathematical majors lead to higher paying career tracks than verbal majors, and I suspect this might be a reason why the regression analysis shows no benefit at all from having a higher vocabulary score. But this also affirms the conclusion that selecting the right track early in life is more important to income than being smart after one's course is already set.
Below is the output from the 1991 regression analysis:
|Regression Coefficients||Test That Each Coefficient = 0|
|Multiple R =||.563||R-Squared =||.317||Std Error of Estimate =||16,326.705|
|Allocation of cases (unweighted)|
|Cases with invalid codes on
variables in the model