Friday, April 1, 2011

David Brooks and Malcolm Gladwell wrong about I.Q, Income and Wealth

In his book "The Social Animal", reviewed here, David brooks writes:

"Once you get past some pretty obvious correlations (smart people make better mathematicians), there is a very loose relationship between IQ and life outcomes."


Brooks further cites a study claiming that there is "no correlation between accumulating large wealth and high IQ."

Both claims are wrong. The result Brooks cites is after "controlling" for education and income. But education and income are themselves functions of I.Q, so you shouldn't control for them if the question you want to answer is how I.Q effects life outcomes.

I have not seen this graphed online, so let's visualize the relationship between an estimate of I.Q and income and wealth so you can see for yourself. The source is NLYS79, a dataset which tracks a representative sample of the U.S population. Intelligence is approximated by the military when the individuals in the sample were mostly teenagers, while income and wealth data is for the same guys in their 40s. The sample is restricted to non-Hispanic white men.

For this group the lowest decile is people with I.Q below 84, and the highest decile above 116, which is not a very high cutoff. So keep in mind that we are not talking about only super-geniuses, in which case the results would be even stronger. Also remember that the middle of the distribution have very similar I.Q scores, the 5th decile is around 101-104, and the 6th decile around 104-108.

As you can see Americans men lucky enough to be born either with genes or a home environment that facilitates high I.Q earn more and accumulate more wealth.




The strong link between I.Q and earnings is well known by labor economists, but perhaps not by the affluent and high-I.Q readers of the New York Times. Obviously most of it goes through education. As technological development makes I.Q more valuable and unskilled labor less valuable, this disparity is increasing.

Another common claim of Brooks and of Malcolm Gladwell is that I.Q may matter, but only until around 130, after which it becomes meaningless. This is also wrong. Many previous samples have had too few observations to make reliable inference about the effect of I.Q above 130. Of course not having sufficient data hardly justifies Gladwell confidently claiming that I.Q above 130 is irrelevant even for scientists in technical fields (which I and others who are not smart enough to handle advanced mathematics could have told you from personal experience was a bizarre theory). After all, 130 is not that high, around the mean for a Harvard or SSE student.

This recent paper by Heckman, Gensowski and Savelyev studies the life outcomes of the Terman sample, which entirely consists of American men and women with I.Q above 135 (in some cases far above 135). They find that I.Q has a significant effects on earnings and educational outcomes, also for those above the 135 I.Q threshold. Another Malcolm Gladwell myth busted.


There are some policy implications from this realization. One is that smart and successful people shouldn't congratulate themselves so much. They didn't so much "earn" their talent than were lucky in the gene/environment lottery. If you are born healthy, with high I.Q genes and with educated parents and a good home environment you are expected to earn more than a more disadvantaged child who exerts the exact amount of effort through life.

Unlike libertarians, Conservatives believe that those who were the recipients of good fortunate have a moral obligations towards the rest of society, in particular to the people who do their best but just have less marketable skills.

Another is that the left is wrong about the market allocating income mainly based on chance, connections or "power". In fact, earnings are strongly linked to intelligence, which indicates that they are linked to productivity, just as economic theory predicts. Poor people are on average less productive than rich people, a claim which may sound obvious (almost tautological) to an economist but which outrages a lot of people on the left.

Denying the link between productivity and earnings is very important for the modern left, as their entire source of outrage is based on the view that the capitalist system "exploits" the poor. More likely, because of the modern welfare state and because of the growing importance of human capital, more resources are transferred from the productive rich to the poor than vice-versa. There is so little demand in the labor market for unskilled people that the poor in industrialized countries increasingly don't even work full time.

The fact that the rich don't exploit the poor doesn't mean the rich shouldn't help the poor. But it's one thing to claim you are rich because you are stealing from poor people, and another to believe you have an obligation to help all members of society due to randomly having being granted more valued skills. Fairness perceptions are not only a function of the type of distribution we desire, but to an even greater extent a function of the process we believe creates inequality.

I suppose David Brooks and Gladwell give an inaccurate impression about I.Q and income/wealth in order to make their readers feel warm and fuzzy. But that is not an accurate depiction of the world we live in, we live in a much harsher and more unfair reality.

31 comments:

  1. I don't know that IQ research dispels the criticisms of the left. The correlations dissipate well before we get to the point of plutocracy.

    ReplyDelete
  2. I don't understand what you mean.

    ReplyDelete
  3. So far as I recall, IQ doesn't correlate well into the millions of dollars per year or even the multiple hundreds of thousands. Rather, it ends well within upper middle class lifestyles, especially once you control for cost of living.

    ReplyDelete
  4. Hyena:

    Recall from where? What data-set do you have with I.Q and millionaires?

    Incidentally, Billionaires, are 5 times more likely to have PhD:s than the average. By far the most common school for American billionaires is Harvard. Bill Gates scored a perfect 800 on SAT math.

    Are these coincidences?

    (obviously I.Q is just one of many variables that determined becoming a millionaire, but it is hardly irrelevant).

    ReplyDelete
  5. If I remembered, I'd not use the phrase "so far as I recall". That question can be set aside, though.

    Your argument doesn't succeed because the data you present doesn't reach into the criticisms of the left. You're not presenting poor people (the mean for the lowest decile is roughly 4x poverty line) and you're not presenting rich people (you cap around $160,000).

    So if someone is arguing that the poor don't deserve their poverty and the rich don't deserve their wealth, you can't counter that unless your range includes the poor and the rich.

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
  6. This comment has been removed by the author.

    ReplyDelete
  7. Hyena:

    If you don't recall and cannot even remember with effort, your argument doesn't have much weight. I doubt your claim since I have never heard of a American dataset with high-quality I.Q data and a large sample of millionaires. My guess would be you read this in Gladwell or some other unreliable source.

    You misunderstand. I haven't capped the data at 160.000 figure. That is the *average* for the highest I.Q decile, and obviously includes some people with much higher incomes income.

    There are rich and poor people both amongst the low I.Q and high I.Q sample. The results are driven from the fact that there are just more rich people among the high I.Q sample and more poor people among the low I.Q.

    As you point out, among 40+ year old white men in America there are few people below the official poverty line for the average of 2004-2008, even among the lowest decile (and obviously fewer still among the highest deciles).

    Let's add women, because at for that age group most poor people are women (single mothers).

    Of those with the highest I.Q decile, 0.8% earn less than $20.000 per year on average 2004-2008.

    Of those with the lowest I.Q decile, 31% earn less than $20.000 per year on average 2004-2008.

    ReplyDelete
  8. Is the correlation actually *statistically* significant? Also, confidence intervals would be nice.

    ReplyDelete
  9. To begin with, I've never read Gladwell and, in any case, "Outliers" was written two years after I stopped reading about IQ and outcomes in 2005-6. In any event, it's not an argument, it's a (mis)remembered fact. That there isn't such a dataset is good enough for me to accept that I'm misremembering. (Possibly misremembering a statement driven by lack of a good dataset.)

    I'm not misunderstanding: as I note that you're presenting means for each decile in my prior comment. The problem is that the focus of criticism is not what happens between deciles, but what is going on within the first and tenth deciles. Inter-decile comparison doesn't answer that question and deploying means destroys the information needed to drive your argument.

    I'd be wary of leaning on the representation of single mothers to drive income at the lowest decile. Since single mothers really do have a no-fault component--their partner has abandoned responsibility--we should be trying to eliminate them in the comparisons. White male incomes were a good choice to start with because it removes you from additional arguments about single motherhood or the disparate impact of drug laws or whatever else.

    My point though, is that to make your argument as a criticism of the left, you really need to be presenting correlations within the first and tenth deciles rather than between them.

    ReplyDelete
  10. Yes kiernan of course the correlation between I.Q and income is highly statistically significant, p<0.001.

    I don't see the point of adding confidence intervals for such as tight distribution other than to make it ugly. For instance twice the standard error of the mean of the lowest docile is $5000 (mean $45.000) and the highest $12000 (mean $160.000)

    ReplyDelete
  11. Hyena:

    Inequality in society is hardly restricted to differences within I.Q deciles. Or is your claim that the left only complains if people with identical I.Q have different earning?

    And a correlation between I.Q and income within a I.Q deciles is absurd, since that already restricts the range. I think you mean to say we should look at the variance of income whiting deciles?

    I added women to the sample to make if more fair to *your* point, not to cheat. With only men, 0% of the highest I.Q docile earn less than $20.000, and more than 30% of the lowest decile.

    ReplyDelete
  12. I don't think the idea of hunting correlations within deciles is "absurd". NLSY79 should contain 1,200 members in the 10th decile. You also cite Genkowski, et al, which operates entirely in the 99th percentile. To answer one prong of left criticism, you'd need to show that the Terman data allows you to completely explain very high incomes ($500k to many multiples of that), especially since the Terman data has the other variables of interest.

    If it doesn't, then you're going to end up looking largely at luck.

    ReplyDelete
  13. Hyena:

    No-one claims that I.Q "entirely" determines income. That is confusing a strong correlation with a perfect correlation.

    There are only 180 observations per deciles for white men.

    ReplyDelete
  14. Absoluely: no one claims that, present company included. The Terman data, if you recall, covers other personality factors and education, allowing controls for "innateness" and approximates for "hard work" or whatever.

    ReplyDelete
  15. I think blogspot.com ate my comment. Let's see, if it shows up or if I have to do it all over again.

    Anyway, IQ heritability and meritocracy wouldn't undermine the redistributionist project. It ruin the contrete social democratic utopia that equality of opportunity - to be brought about by further social democratic reforms - would lead to equality of outcomes.

    Redistributionists could simply skip straight to redustribution of income, based on that you misattribute as a conservative obligation.

    ReplyDelete
    Replies
    1. Equality of opportunity is meritocracy. Equality of outcomes is not meritocracy. The current rich are benefiting from a half-assed equality of outcome that is only applied to their social class. They do not operate by meritocratic standards a since far more meritorious people born to poverty would have to invest far more energy to get to simple parity in competition with the established rich.

      Delete
  16. I didn't get a captcha for my first comment, so I presume it is gone.

    Here is a scatterplott
    http://thesocietypages.org/socimages/2008/02/06/correlations-of-iq-with-income-and-wealth/

    Wikipedia has r between 0.4 and 0.5:
    http://en.wikipedia.org/wiki/Intelligence_quotient#Income

    So the correlation is weak, especially within the human experience. So David Brooks and Malcom Gladwell are correct, if they are talking about individuals, which I presume they do.

    ReplyDelete
  17. dietervie,

    0.4 is weak?!? It is the strongest single explanatory variable for income!

    Your *own* income from one year to the next only correlates 0.55 with each other. Your income and your parents one year income
    correlated 0.1-0.2.

    Income varies a lot from year to year, and is measured poorly. This is why it correlates so low with most variables. It is remarkable how strong it correlated with I.Q.

    If you aggregate the correlation goes up
    dramatically, state I.Q and state income correlate 0.7-0.8. The discussion is not only about individuals, it's about groups such as social classes. Even for individuals I.Q is only "weak" if you consider every other variable on the planet, including age, education, race, gender, parental income, your own past income, a "weak" predictor.

    I already wrote that the fairness and inequality aversion arguments don't go away. But those are far weaker than "you are poor because the rich exploit you" argument.

    ReplyDelete
    Replies
    1. @Tino: "0.4 is weak?!? It is the strongest single explanatory variable for income! Your *own* income from one year to the next only correlates 0.55 with each other."

      You realize that your example only serves to emphasize how weak the correlation is. It's more important what year it happens to be than what your IQ is, lol. That should tell you something.

      Delete
  18. The correlation between being white and wealth is this sample is 0.18, and the correlation between being male and wealth 0.03.

    Age and other continuous variables also correlate far lower than I.Q.

    Correlation can only be compared in relative terms. Only if you think race, age, parental income and gender are unimportant for wealth distribution can you claim that I.Q is unimportant.

    ReplyDelete
  19. Hyena—

    Since single mothers really do have a no-fault component--their partner has abandoned responsibility--we should be trying to eliminate them in the comparisons.

    This is a completely whacked though common way of looking at it.

    He probably would have wanted her to have an abortion or give the child up for adoption. As for the divorce route to single motherhood, women file for divorce at 2.5x the rate men do these feminist days, and most often for “grew apart” type reasons, which really amount to her not feeling fully “in love” anymore, often after she has an affair. Contrary to feminist myth making serious abuse is rarely the real reason. As well American wives emotionally abuse their husbands far more often than the other way ‘round, due to feminist excuse making for “strong” (nagging, strident, yelling) wives. “Your wife is always right or you should act as if for marital harmony’ and ‘men are all dogs’ are strong current feminist memes.

    What business did she have getting pregnant without getting married first to a well chosen man who wanted to be a dad to her kids? It’s not like birth control is hard to find. Even the stupidest (almost) girls know about birth control. Many are just too irresponsible to use it reliably. All but one reversible forms of birth control are within the exclusive knowledge and control of women, and the other one greatly diminishes sexual pleasure.

    As well I’m convinced that a very large percentage of unwed mothers are that intentionally, both consciously and subconsciously through an oops pregnancy. She claims she’s on birth control but the “forgets” to take her pills. Depending on social level they may hope to lure the reluctant guy into marrying her, they may want to get his state extracted child support=also stealth alimony since it was jacked up under a federal mandate in the 1990s, or they may want welfare.

    ReplyDelete
  20. (Sorry to reawaken an old post -- it's new to me). You say . . .

    "In fact, earnings are strongly linked to intelligence, which indicates that they are linked to productivity, just as economic theory predicts. Poor people are on average less productive than rich people, a claim which may sound obvious (almost tautological) to an economist but which outrages a lot of people on the left."

    Where's the link between intelligence and productivity? Your data may show the economic system rewards high IQs financially, but that doesn't link it to productivity (at least as that term is commonly understood).

    ReplyDelete
    Replies
    1. What does productive mean?

      I keep getting the impression that people seem to think making millions on the stock market is productive or that rent seeking is productive (this includes people who charge rent for other people to be productive on their property).

      Delete
  21. I do think you have to be more than a tad smarter than average to be "in the game" at all. That doesn't mean that people don't squander their gift (or at least don't make large sums of cash with it). And study after study shows that intelligence is largely a randomly bestowed gift, rather than something earned. We -hate- that idea, for obvious reasons, but it does seem to be true.

    ReplyDelete
  22. That article/argument was based a resentment against the liberal opinion that everyone should be treated with equal opportunity because often the most intelligent are a diamond in the rough that never gets appreciated. IQ is based on tests written by people with average IQs but were priveledged with a college education which may have been provided through grants that some people would like to deny the public now that they already got their education. An IQ test would have disqualified Albert Einstein and based on some more affluent people's idea of what a test should ask, this is always biased. Asking for opportunity isn't asking for privilege, as this suggests. And the only people I hear talking in that manner are the privileged by birth rich kids that inherit their prosperity and grow up in a success oriented environment, who also go to work for the best firms in the best jobs based on a higher priced education paid for by their affluent parents.

    ReplyDelete
  23. This comment has been removed by the author.

    ReplyDelete
  24. "An IQ test would have disqualified Albert Einstein"

    That is a myth. Einstein never took an IQ-test.

    IQ-tests are not perfect, but have high predictive power for life outcomes such as efficiency at job task, education, income, decision making etc. The predictive power of IQ-tests is similar for the rich and the poor. IQ-tests have far higher predictive power than parental socio-economic background. The fact that IQ-tests "work" disproves the theory that IQ-tests are mostly noise or "privilege".

    ReplyDelete
  25. Thanks for the information. I'd love to travel there some day. idateasia review | idateasia scam

    ReplyDelete
  26. Thanks for sharing this informative article about iq. Also you can learn more about a young lady who has iq level 225 which is very much high. Vishalini iq level is recorded as a world record. You can also get more information about her while visiting this website. Also share some updated iq news in your website
    vishalini iq

    ReplyDelete
  27. Just FYI, the curve is exponential, that can be seen by how great the difference is between 90 and 80 percent, compared to the other intervals. So if this was expanded to percentiles, it goes farther up, I believe if I remember correctly though I have trouble finding documentation of this right now, that by the 98th percentile the average income is already at nearly 1 million.

    ReplyDelete

Google Analytics Alternative