Correlations Are Real-Valued

Dave (non-Masten) on the importance of IQ:

Sure if they have brain damage that inhibits them from handling higher math they may have behavior problems also. If they have an IQ of 95 and do well as a clerk in a shoe store, is this a poorer outcome than being a slick high IQ type like Ted Bundy or Al Capone. Admittedly a shoe clerk might not have high social status but if you exclude persons from success this way there seems to be a certain elitism that is not scientific.

First off, no one's saying that there's anything wrong with being a shoe salesman. I think that someone with an IQ of 95 would probably be capable of--and better off by--learning a skilled trade, but there's a place for everyone, and if retail sales clerk is the best someone can do, then it's as honorable an occupation as any.

Second, low IQ isn't brain damage. It can be caused by brain damage, but undamaged brains have a wide range of ability levels, and AFAIK there's no evidence that the correlation between low IQ and antisocial behavior is due entirely to brain damage.

But the main point I wanted to address was this:

If they have an IQ of 95 and do well as a clerk in a shoe store, is this a poorer outcome than being a slick high IQ type like Ted Bundy or Al Capone.

This is a fallacy I see far too often: The idea that the only possible values of a correlation are -1, 0, or 1. Or, in layman's terms, the idea that a single counterexample invalidates a rule of thumb. It doesn't work that way. When we say that low IQ is correlated with undesirable life outcomes, we mean that people with low IQs are much more likely than people with high IQs to live in poverty, commit crimes, drop out of high school, receive welfare, and have children out of wedlock. That doesn't mean that people with high IQs never do these things, or that people with low IQs always do; it just means that people with high IQs are much less likely to.

The existence of high-IQ criminals like Bundy and Capone only proves that there's not a perfect correlation between IQ and criminality. But no one ever claimed there was.

Share this

Even high correlations allow for plenty of outliers

Thank you for making this point. I wrote a post about it here that includes graphs of what 0, 0.1, 0.2, and 0.9 correlations look like. Even really high correlations like 0.5 leave plenty of room for plenty of Ted Bundy's.

Correlation between brain damage and mood disorder

There is a correlation between brain damage and mood disorder, patients with moderate to severe TBI are 4 times more likely than the general population to develop a psychiatric illness in the six months following their injury.

Right. Like I said in the

Right. Like I said in the post, physical brain damage is one possible cause of antisocial behavior, but there are others.

Pobative Strength of Correlation in Question

"When we say that low IQ is correlated with undesirable life outcomes, we mean that people with low IQs are much more likely than people with high IQs to live in poverty, commit crimes, drop out of high school, receive welfare, and have children out of wedlock. That doesn't mean that people with high IQs never do these things, or that people with low IQs always do; it just means that people with high IQs are much less likely to."

The problems I see are two fold. First is that “poor outcome” is really a value judgments. In some societies everyone lives in poverty and having a lot of children with multiple women is considered laudable. But there is also a correlation between low socioeconomic status and these same indicators of poverty. The Irish and Italian immigrants were notorious for their violence, alcoholism, criminality and domestic unrest. Was this due to their low IQ?

Just how strong a correlation do we have? If it can be proved that blacks do have an average IQ that is under all conditions marginally lower than whites it would merely be an interesting if useless fact. If the correlation between and these “poor outcomes” is really robust to the point that being black has a 90% correlation of a “poor outcome” and being white has a 90% chance of “good outcome” the implications are much more serious. A weaker correlation is of less probative value. Well which is it?

I am fully aware of outliers and correlation coefficients. For example when we compare methods of clinical analysis we want a correlation coefficient of above .9 or you would have little confidence that method A was giving you the same information as method B. So if you have weak correlation such as an R value of .7 and even more .5, even if found consistently over many studies, you have little explanatory value.

With this data can you say that Willy, who is not doing well in school, since he is black, now we have the reason why he and most black people aren’t doing well and needn’t be expected to? If we are to use the IQ test to measure these things we need to know just how robust this correlation is, not just that it exists.

Dave

They aren't that high.

Dave,
Without looking up any particular studies, I will put money on the fact that the majority of respectable studies will show correlations well below 0.9 between IQ and pretty much anything else outcome-related like income.

When comparing "methods of clinical analysis", I'm guessing that you are talking about something medical. You guys feel you need 0.9 for a wide variety of reasons. These include:
1. It probably means you omitted a factor that you can pretty easily guess at and collect more data on.
2. You feel really bad and get sued when you kill people.

In other words, you have avenues available to you that will allow you to get to 0.9, and disaster strikes as you hit lower correlations.

In the social sciences, neither of these are true. Even the strongest of theories that have been tested out the wazoo will show correlations nowhere near 0.9. I would guess that even relationships like "Being mean to people causes them not to like you". result in correlations no greater than 0.8. In my field, getting 0.7 is really high.

And we don't kill people or get sued that often when we screw up.

This is not to say that *you* must lower your standards. However, realize that you pretty much can't participate in any discussion of social issues with such high standards. You will be so conservative as to disallow any sort of possible relationship between anything.

Correlations in social sciences.

The reason for this is that in social sciences virtually everything is multifactorial. If two independent factors each contribute equally to one particular outcome, the highest correlation you can get between the outcome and either factor by itself is 0.7. And that's only if those two factors together predict the outcome perfectly.

When you have many different factors interacting, a correlation of .5 is huge. A better analogy would be to epidemiology. I don't know what the correlation between exercise and longevity is, but I bet it's lower than the age-adjusted correlation between IQ and income.

The significance of a

The significance of a correlation is not directly related to its absolute level. It depends on the absolute level and the size of the population. For large sample, a correlation of 0.1 may be very meaningful.

In my field, finding autocorrelations as low as 5% may warrant opening a case of champagne.

Dave:In some societies

Dave:

In some societies everyone lives in poverty and having a lot of children with multiple women is considered laudable.

True, but even in societies where everyone lives in poverty, the vast majority would prefer to be richer than they are. And while men with many children by many women may be admired, women with many children by many men are, AFAIK, more or less universally stigmatized.

At some level, sure, everything's a value judgment. But these are value judgments I'm perfectly happy with, and with which I think most people would agree.

The Irish and Italian immigrants were notorious for their violence, alcoholism, criminality and domestic unrest. Was this due to their low IQ?

It may have been. In those days, many people were severely malnourished. The Irish, in particular, came to escape famine. I don't know about Italians. Or there may have been other factors.

Well which is it?

Somewhere in the middle. The mean black IQ is about one standard deviation less than the mean white IQ, meaning that the average black person is at about the 17th percentile of the white distribution, and the average white person is at about the 83rd percentile of the black distribution.

But the question under debate isn't the size of the black-white achievemnt gap. That's easy to measure, and pretty much everyone agrees on it. The question is how much of the gap is explained by IQ. And it looks very much as if the answer is "most of it."

Genetics and Social Factors

OK, to summarize, we agree that there is a black/white IQ difference which is multifactorial and is to an unknown extent modulated by genetics. As usual in social science, the exact etiology of observed social behavior is obscure. In discussing the subject there is lamentable tendency for one side to deny biology and blame things on “the legacy of slavery” and the other side to think as though IQ levels explain most racial stereotypes.

I am a Thomas Sowell fan. Sowell rejects your idea that genetics largely explain the pathologies of African American society. In his book Black Rednecks and White Liberals he stresses cultural factors. The history of various cultural groups such as American blacks, Southern whites, Jews, Lebanese, Koreans and Germans are explored. Of interest is what he calls middleman cultures, Jews, Chinese, Koreans for example. Clannish, hard working and economically successful, they are also despised and politically powerless. They stress family and education.

He explains the plight American blacks as due to the adoption of the ethos of lower class Southern whites followed later by misguided attempts at social engineering by white liberals. American black culture originally was influenced by a southern white culture of violence, laziness and disrespect for learning. After the release from slavery many blacks adopted middle class values. Many were fully capable of high achievement though there was a large overall lag. For example in Dunbar High School, a select academy in Washington DC the average IQ was 111 in 1939 and in 1950. Highly successful and rigorous schools were opened in Harlem. The subject matter was typical boring academic stuff that every student hates, but many black kids could handle it.

The stereotype of black illegitimacy and marital instability doesn’t hold up either. Quoting Sowell
“The raw facts are these: As of 1960, 51 percent of black females between the ages of 15 and 44 were married and living with their husbands, another 20 percent were divorced, widowed, or separated, and only 28 percent had never been married. Twenty years later, only 31 percent of black women in these age brackets were married and living with their husbands, while 48 percent had never married. By 1994, an absolute majority—56 percent—of black women in these age brackets were never married and only 25 percent were married and living with their husbands. Accordingly, while two-thirds of black children were living with both parents in 1960, only one-third were by 1994. While only 22 percent of black children were born to unmarried women in 1960, 70 percent were by 1994.”

I hear what you are saying but I still can’t explain how genetics which should be stable over a moderate time frame can explain rapid social change
Dave

Exactly

I agree with Dave here.

While it doesn't explain

While it doesn't explain change in such a small time frame, culture can influence genes through selection. If many blacks do not value learning as Sowell suggests, a low IQ and a high IQ will not influence many outcomes in their life since IQ often requires knowledge to be really useful. Thus there will be little selection for high IQ.

I don't know about him but

I don't know about him but I'm not saying genes can influence culture and vice versa.

Sowells not saying they don't value "learning" and that's the only factor. If a groups average age is around 20 vs. 40 for other ethnic groups then certain outcomes are going to be different.

"If many blacks do not value learning as Sowell suggests, a low IQ and a high IQ will not influence many outcomes in their life since IQ often requires knowledge to be really useful. "

Some just don't value certain kinds of learning. The kind of learning that gets you a job offer in our culture. Why wouldn't a person with higher IQ do better than someone with lower IQ given both have an aversion to learning. Might still outwit the other guy.

Who says brains and economic success is the only measure of getting genes into the gene pool. While your off at school he's screwing your wife.

Things are way more complicated than IQ = success. As I pointed out before one of the blackest and most successful groups is the West Indians. They either had "high IQ" all along or it doesn't matter.

Some just don't value

Some just don't value certain kinds of learning. The kind of learning that gets you a job offer in our culture. Why wouldn't a person with higher IQ do better than someone with lower IQ given both have an aversion to learning. Might still outwit the other guy.

I think the return of superior IQ dramatically increases with knowledge. A truck driver with an IQ of 120 will not do much better than a truck driver with an IQ of 80, an engineer will.

Tangent -- Why do people think Ted Bundy had a high IQ?

Is there any evidence to support this myth? At age 26, Bundy finally graduated from UW with a BA in psychology. He went to a night law school at 27 and flunked out after a year. The killings began around then. He was charming, but I've never seen an IQ score or any significant academic or professional achievement. He did odd jobs like grocery bagging, and a lot of volunteer work.

I don't know. I was just

I don't know. I was just taking Dave's word for it. Several web pages claim an IQ of 124, which would put him around the 90th to 95th percentile, but they don't cite the source. According to this site, this is about average for organized serial killers. His profile on the same site says that his college grades were "good" at first and then "great" after he returned from a hiatus. On the other hand, he took the LSAT twice and got a low score both times.