Saturday, February 17, 2007

Smart people are only slightly less criminal: Plato thought that if people broke the law, they did it out of lack of understanding, since men naturally and rationally pursue happiness, and crime only brings misery. He would have predicted a strong inverse correlation between IQ and criminality, but the General Social Survey disagrees with him. I used logistic regression to estimate the relationship between a person's score on a vocabulary test and whether they have ever been arrested (1 for yes, 2 for no). The coefficient for whites is .052, and it's .054 for blacks. In English, what that means is that IQ explains not even close to one percent of criminality. Maybe Plato wanted to see a virtue in people like himself that just isn't there.


  1. I don't think you used a very good proxy for "criminality".

    I would expect that many of the higher IQ persons who answered yes were arrested for some political protest related activity, or for shoplifting as a kid and such. Minor things.

    I think if you did a regression on say served a total of 1 year or more jail time you'd get a FAR higher inverse correlation.

    As well much other data contradicts the general conclusion you're drawing. I really doubt to the point of near certainty that a literature survey would corroborate your finding of little correlation, other than to a spurious measure of "criminality".

  2. In The Bell Curve, the authors state that the average IQ of those who have been interviewed in jail was 90.

  3. Pp. 296-8 of Jensen's _The g Factor_ briefly review the IQ-crime lit. In his Table 9.1 is data from _The Bell Curve_ -- it shows that the percentage of those "ever interviewed in corretional facility (males)" is less than 1%, 1%, 3%, 6%, and 13% for those of IQ above 125, 110-125, 90-110, 75-90, and below 75, respectively.

    He also notes that, "Among 1780 enlisted men in the Army, delinquent behavior serious enough for court-martial conviction showed a (biserial) correlation of .31 with the AFQT, a highly g-loaded test. Other studies conducted in the armed forces show a similar relationship between mental test scores and delinquency."

    But I think you're right to point to flaws in Plato's reasoning, and the modern version that views the IQ-crime correlation as driven by defects in abstract reasoning per se. I haven't read a lot of this lit, but another (not mutually exclusive) hypothesis is that low IQ results in greater "misfit" status in modern societies, and through repeated failures as well as perhaps envy of those who are smart enough to succeed, lower-IQ people become more likely to say "nuts to the rules" and gain status by illicit means.

    For example, Eskimos / Arctic people have a mean IQ of 91, which is about what it is for non-white Hispanics -- except the Eskimos live outside of the modern US society, and thus don't need to complete high school / go to college in order to gain high status, nor are they constantly reminded of their below-Euro-average IQ, since they live in a homogenous area.

    And the Amerindians' mean is 86 -- just 1 point above the Af-Am mean, but again they largely live in a separate sphere from the modern US. Australian Aborigenes have the lowest mean IQ at 62, but as I understand, they also live separated from modern Australian society.

    I think it's only when low-IQ groups are forced to compete for status in a society that places high demands on intelligence, in particular a society that has a multi-modal IQ distribution so that the lower-mean groups are painfully aware of their disadvantages, that IQ and crime really start to correlate strongly.

  4. dougjnn: I imagine you are right that IQ is more strongly related to serious, chronic offending, but criminologists do not consider arrest a measure of minor offending since police usually don't haul people in for petty crimes.

    I am somewhat familiar with the research literature, and it shows a consistent but small effect. Researchers often make a big deal about their pet risk factor. The study that Agnostic cites reports a bit larger correlation than what I have usually seen, but even that finding explains less than 10% of the variation in court-martial convictions. The findings I report here are lower than usual, and you might have a point, but keep in mind that speculations can run in the opposite direction too: the correlation might be inflated because smart people are more likely to avoid detection.

    For example, when Malcolm X had just burgled a house and was seen by cops on the street, he didn't run. He approached them and asked them a question about directions. Clever.

  5. Don't buy this at all. Racial statistics, though general, point to blacks being the most criminal and asians the least with about a 20 point difference in IQ seperating them.

  6. vic: Of course you are right about racial differences in crime and in IQ, but that does not prove that low IQ is a major determinant of crime. Racial groups differ in many other ways as well (e.g., impulsivity, extraversion, aggressiveness, psychopathy) that might better explain the crime gap.

  7. The relationship between incarceration and IQ is curvilinear. There are very few retards and MENSA members in jail. A consistent finding is that the average inmate has an IQ of ~92.

    "A small amount of evidence, as well as common sense, suggests that despite the lower average IQ, the offender population contains relatively few very low IQs. In a search of police records for California county, Hirschi and Hindeland, for example, found a higher frequency of delinquency (defined as two or more officially recorded incidents) among white males in the next-to-the-bottom quintile of IQ scores (scores from the 20th to the 39th percentile) than in the bottom quintile. At each successively higher quintile above the next to the bottom, the delinquency level declined. Below some degree of intellectual deficiency, even the requirements of the most unplanned crime will be to great. The population of nonoffenders therefore extends farther down into the range of subnormal scores than that of offenders, yet still averages ten points higher, if not more. The relationship between intelligence and crime is curvilinear, with the most offenders falling in the low normal or borderline subnormal range, approximately from IQ 60 to 100; their relative frequencies decline on either side of this range."

    Wilson and Herrnstein, CRIME AND HUMAN NATURE (1985) pp.154-5

    Let us assume arguendo that crime rates are evenly distributed across each and every percentile on the scale on intelligence; that people with the lowest IQ are as likely to commit a crime as people with the highest; that inmates have a lower average IQ than non-inmates solely because people with higher IQs are more likely to elude incarceration; that "the system" is to blame for sorting by IQ. Were that the case, we would expect to see a lock-step, monotonicly inverse relationship between IQ and incarceration. We don't. Relationship is not monotonic, it is curvilinear.

    There is some evidence that people who are more intelligent are less likely to be incarcerated for their crimes.

    "In self-reports, there is less correlation between IQ and criminal behavior than official records [...]."

    Wilson and Herrnstein, CRIME AND HUMAN NATURE (1985) p.159

    Furthermore, different categories of crime are associated with different levels of IQ. On average, people incarcerated for embezzlement or counterfeiting have higher IQs than people who are incarcerated for rape or murder. And the clearance rates for embezzlement and counterfeiting are lower than the rates for rape and murder. The police are less likely to nab a guy for making funny money than they are for raping a college co-ed. But here again is an illustration of how personality shapes how people choose their perfered mode of crime as much as the criminal justice system sorts by IQ.

  8. Agnostic said--

    I think it's only when low-IQ groups are forced to compete for status in a society that places high demands on intelligence, in particular a society that has a multi-modal IQ distribution so that the lower-mean groups are painfully aware of their disadvantages, that IQ and crime really start to correlate strongly.

    That theory makes a great deal of sense.

    It’s not so different from a pet theory of my own. Which is that there’s a tendency for people who’s ambition and aggression and need to assert dominance exceeds their intelligence and ability to do WELL in society legally, to become criminals. That “exceeds” is of course a very fuzzy idea.

    As well there’s the factor that we’ve been incessantly telling everyone (and blacks are watching) in movies, tv entertainment and network news since the 1960s that most of black problems are traceable to pervasive and still continuing (if ok a little better but nowhere near good enough), white racism. Or it all goes back to slavery and Jim Crow. But anyway it’s up to whites to do more to make things better and easier, and it’s not the black “victim’s” fault.

    Black crime like out of wedlock childbearing first exploded in the 1960s and kept going until the 1990s, when much higher incarceration rates set in. Do blacks watch TV and say, hey, it’s OK if I want to be a playa? Well, yeah, sometimes, MTV and other shows give that message. But it also works in more indirect ways as well I think. “This game is rigged so I’m gonna make my own rules” or rather “I’m gonna follow the brothas rules – pulls in the honnies fast with the bling bling and all” kind of thing.

  9. Anonymous7:32 AM

    great article. I would love to follow you on twitter.


Conscsiousness explained naturally? Doubt it

From Amazon's description of a new book on consciousness: How can the seemingly immaterial experience of consciousness be explained b...