Arnold Kling  

The Income Heritability Puzzle

Public Goods and Pedagogy... What to Learn from the XM-Siri...

At GNXP, Herrick writes

let's just go all the way to perfect heritability of IQ and perfect assortive mating on IQ. In other words, let's see if "IQ clones" will be have enough similarity in wages to match the 0.4 intergenerational correlation of income.

Will the IQ clones have similar incomes? Not so much. (0.3^2)*1 still equals something small: 0.09. Less than 1/4 of the intergeneration correlation in income. Medium-sized potatoes, but we had to make a ton of ridiculous assumptions to get there.

This refers to work done several years ago by Bowled and Gintis. The puzzle is that income is fairly highly heritable, .4, while the link between IQ and income is less so, .27. That makes it very unlikely that IQ explains much of the heritability of income.

In my view, if there is a swindle in the original Bowles-Gintis paper, it is that they measure income in different ways. When they look at the effect of IQ on income, they look at studies that appear to measure one year of labor income. This is a highly volatile number, which means that much of the variation is unexplained.

On the other hand, when they look at the heritability of income, they seem to prefer multi-year measures of total income. This has lower variability.

My guess is that if you do it apples-to-apples, the amount of income variation you can explain by parental income is about the same as the amount than you can explain by IQ. That is, if you used one year's labor earnings as the dependent variable, then IQ might explain 20 percent of the variation; and if you instead used parents' income, you would explain about 20 percent of the variation. If you used both variables together, you might explain 25-30 percent of the variation.

Next, consider using a 5-year average of labor income as the dependent variable. My guess is that the explanatory percentages would be more like 40, 40, and 50-60 percent, respectively. Five-year averages smoothe a lot of the noise, so you can get better results.

Even if I am correct, the Bowles-Gintis point that there is a lot of income heritability that is not mediated by IQ still stands.

Note that the comments on Herrick's blog post are very high caliber. Although some of them appear to be riddled with errors, it's a lot better than your typical drive-by snarkfest.

Comments and Sharing

CATEGORIES: Income Distribution

COMMENTS (7 to date)
floccina writes:

I wonder if this is related to the question of why people in rich countries work so much when we have our needs met. The relationship being that we are more interested in not loosing standard of living than we are in gaining standard of living.

liberty writes:

Even if the heritability of income and the heritability of IQ are perfectly correlated, that doesn't mean that intergenerational income is explained by IQ.

For one thing, we inherit or learn other things, like laziness. We also grow up with our parents, and often learn their trades.

I think a not insignificant proportion of intergenerational low mobility can be explained by the desire to, and ease of, learning your parent's trade.

This is obvious in the case of coal miners and plumbers (and it really often is that they want to follow in their father's footsteps). It is maybe less obvious, but just as true that the children of lawyers want to be lawyers, that the children of businessmen inherit the business and the talents and also want to make their dad proud, and that the children of doctors go into medicine.

It is even true when following in their footsteps will earn you less than you could easily get elsewhere. This is often true for blue collar type kids who choose to do that instead of going to college - by choice - but it is also true of the well educated and wealthy kids. Artists are often kids of artists, writers of writers, non-profit workers are often from families of the same, etc.

my 2c

Kurbla writes:

I think liberty is right. And it is not only about learning - it is about nepotism as well.

MouseJunior writes:

That could easily be accounted by other traits: sociability, ability to focus, boredom threshold, risk tolerance, risk seeking, etc.

None of those will affect IQ, but most of them affect income to some degree or other.

BGC writes:

I get the feeling that economists are looking at this the wrong way.

The first point is that IQ is important in predicting socio-economic status - whether SES is measured by social class, income, years of education, level of educational attainment or however you choose to measure it. IQ is important - this is established by c.100 years of work.

Second, literally nobody is claiming that IQ is the only important heritable psychological trait in influencing SES. The claim is that IQ is important.

For example, the Big 5 personality trait Conscientiousness (C) is also a powerful and heritable psychological predictor of examination success and occupational success - however at present C is mostly measured by a self-rated scale, which is not ideal. (The same could be said for Eysenck's factor called Psychoticism - which is oversely related to SES)

In the GNXP comments, gc comments on the flaws in using straight line correlations measures the size of effects - and this is correct.The matter has been well covered by Dierdre McClosky in her statistical writings.

It strikes me as bizarre that people use left-over 'unexplained' variance as if it was a measure of something (usually 'the environment' - in fact much of it is probably genetic). People take a pathetic delight in the fact that IQ does not give a perfectly accurate estimate of SES - they then insert into this uncertainty their favourite pet theory.

But the problem runs deeper than this. When an author takes a data set and (using whatever stats) suggests that such a number 'explains' so much of the variation - this is just a post hoc hypothesis; and both the method and the specific number must then be tested on further independent (and preferably better, more specifically-controlled) data sets to see if it stands up to this testing.

Even more importantly, statstical models are not really theories, because they are not open to multi-method testing. Instead they just lead to more and more measurements of the same type (on different populations of different sizes). The measurements always differ, then people get exercized over whether these differences are important and in what way. Usually there is no resolution. This is pretty much just doing measurements ad infinitum, not testing a theory - it keep people employed but not much else.

I mean, does anyone _really_ believe that IQ could have the same predictive value for income in all nations at all times in history? Of course not, the predictive power of IQ for future income would be different according to a host of other local, national and who knows what factors and would vary over time. So what exactly is at issue?

The reason that IQ research has moved ahead is that it has been testing a complex hypothesis using multiple methods.

The numbers generated by a single study are not much more than post hoc hypothetical suggestings for further study - it doesn't make sense to keep analyzing and re-analyzing the same data set. What is needed are more data. But what is needed much more than this is complex hypotheses with ramifications observable in many types of scientific discipline; not atheoretical effect size estimates posing as hypotheses.

Mr. Econotarian writes:

I believe the income heritability has much to do with culture. Those with similar culture and higher IQ would do better, but those with a lower IQ who come from a culture that respects learning, despises violence, and expects to go to college will do better than those with a high IQ who come from a culture that respects joining gangs to obtain local prowess and drug income, accepts violence in times of "disrespect", or becoming a single mother.

The problem is that culture is highly heritable - how often do we feel ourselves "talking like our parents"? It is probably evolutionarily important that culture can be learned and passed on, so that behavioral traits that lead to survival success in the past get passed on. And culture is linked in our brains with closely with ethnicity and language.

Now I believe that lots of bad government policies encourage bad culture (such as the drug war), and government policies also isolate culture (zoning, unequal treatment under the law by ethnicity, etc.)

But that doesn't mean that bad culture isn't bad culture, from an economic results perspective.

Mr.goodpoint writes:

I think Liberty is right too. Excellent points. And Kurbla as well.

Comments for this entry have been closed
Return to top