It happens like clockwork. Every few years, researchers contrive yet another study to prove that conservatives are mentally deficient. This time, the attack comes from evolutionary psychologist Satoshi Kanazawa in his current paper, Why Liberals and Atheists Are More Intelligent. To be fair, gunning for conservatives does not appear to be his primary motive. Instead, he seems to have tied himself in knots trying to affirm a pet theory. Either way, he has recklessly disparaged millions. The methodology is atrocious.
Truth be told, I prefer to write on other topics. But every so often, I simply reach my limit with the gratuitous abuse heaped onto conservatives by members of my profession. (Our motto: celebrate diversity!) If you are in the mood for some rich, curmudgeonly goodness, then read on. Otherwise, check out my more upbeat entries. Either way, thanks for the patronage!
My Psychology Today blog response to Dr. Kanazawa is here.
Just Another Rigged “Study”
In psychology’s latest assault on American conservatives, evolutionary psychologist Satoshi Kanazawa has determined that “very liberal” people possess nearly a 12 point IQ advantage over those who identify as “very conservative.” Not only is the hard left more intelligent than the hard right, but they are smarter than everyone else, as well – mainstream conservative, middle-of-the-road, and even mainstream liberal.
In a blog posting at the Psychology Today website, Kanazawa baldly asserted that,
“…apart from a few areas in life (such as business) where countervailing circumstances may prevail, liberals control all institutions. They control the institutions because liberals are on average more intelligent than conservatives and thus they are more likely to attain the highest status in any area of (evolutionarily novel) modern life” (emphasis in original).
This 11.6 point IQ difference (see figure 1) is an astounding development. It is unimaginably improbable. In fact, it does not pass the smell test. There are a massive number of variables describing the American population. Dividing this population along a single variable will certainly yield some differences corresponding with that variable. Overall, however, the division will result in two or more heterogeneous groups that very much resemble the original population. Slight differences aside, the new groups will average out to be pretty average, if you’ll pardon my circularity.
For example, suppose we divide the American population based on their preference for either football or baseball. We will probably notice some personality trends that correspond with sport preference. However, it would be astounding to discover something as profound as a 12 point IQ difference. That would be like discovering that baseball fans are three inches taller than football fans. It could be true, but results like that would require that we dismantle the research and examine its methodology from the ground up.
Ideally, that happens during the peer-review process before any study is published. Unfortunately, research that maligns conservatives seems to earn unquestioning approval in peer-reviewed psychology journals. You can be sure that any flawed study suggesting the intellectual inferiority of liberals would not see the light of day.
So how, exactly, did Kanazawa discover that “very liberal” people are smarter than the rest of us? And what prompted the study in the first place?
Searching For the Savanna Principle
Satoshi Kanazawa’s discovery is a secondary point in his most recent paper, Why Liberals and Atheists Are More Intelligent, published in the current issue of the peer-reviewed journal Social Psychology Quarterly.
Kanazawa’s main point was that intelligent people are more likely to solve “evolutionarily novel” problems. He defined these as problems that fall outside the typical daily experience of our ancestors, such as how to respond to fire caused by lightning, or what to do when a flood separates a person from his clan. According to what he calls the Savanna Principle, “The human brain has difficulty comprehending and dealing with entities and situations that did not exist in the ancestral environment.”
It’s an intriguing theory requiring evidence in contemporary humans. That’s where Kanazawa turned to politics, religion, and sex. He suggested that people who endorse liberal ideology, are atheists, or are sexually monogamous are more intelligent than their counterparts because these behaviors contradict the instincts handed down by our ancestors.
I will forego his stance on atheism and monogamy and focus on the methodology behind his assertion that liberals are more intelligent than conservatives because A) there are only so many hours in the day, and B) I seem to have taken up the cause of debunking my profession’s large collection of bogus anti-conservative literature.
If you are curious about my political outlook, I am libertarian. I tend to agree with conservative fiscal principles and liberal social principles. Given the subject matter, you have a right to know where I stand.
Kanazawa used existing data from the National Longitudinal Study of Adolescent Health (hereafter referred to as Add Health; see here and here). In the Add Health study, 20,745 adolescents were interviewed in 1994-95 (Wave I), again in 1996 (Wave II), then a third time when the participants were between the ages of 18 and 28 (Wave III). Naturally, there was some attrition along the way, but the data are solid and respectable.
The Add Health study collected a variety of data ranging from each subject’s height to their education history (see here). In demonstrating that very liberal people are the most intelligent among us, Kanazawa used two pieces of Add Health data:
- From Wave I, when participants were adolescents, he gathered results from the Peabody Picture Vocabulary Test (PPVT). He used this data to estimate each subject’s level of intelligence. This was his independent variable.
- From Wave III, when participants were young adults, he used responses to the question “In terms of politics, do you consider yourself conservative, liberal, or middle of the road?” Subjects were able to choose very conservative, conservative, middle of the road, liberal, or very liberal. This self-report measure was his dependent variable.
To sum up: Kanazawa compared intelligence, measured during adolescence, with political orientation, measured during young adulthood, to see if the first variable affected the second. Nothing wrong with that.
However, there are serious problems with the way he interpreted the data at both ends. First, the intelligence data, which was extrapolated from a simple vocabulary test, factored out participants unlikely to offer the desired response. Second, the political orientation data was gathered at a time when participants were most likely to offer a response that supports Kanazawa’s desired outcome.* Let’s look at each.
Intelligence: The Independent Variable and Its Broken Data
Kanazawa repeatedly refers to IQ in his paper. There’s a problem: he never measured IQ. The PPVT, at best, provides a quick and dirty estimate of verbal intelligence. It is not a test, or even an indicator, of general intelligence, despite his assertions to the contrary.
Here’s how the PPVT works. The examiner says a word, then shows the examinee four drawings. The task is to identify the drawing that goes with the word. Examinees do not need to speak during the test, they can simply point to the drawing they believe to be correct. The test takes about 20 minutes to administer and score. It is rather perfunctory, as verbal tests go.
According to the publisher of the PPVT, the current version boasts a .91 correlation with the Verbal Intelligence Quotient (VIQ) of the WISC-III. Translation: a kid’s score on the quick and dirty PPVT will probably be close to the score that child would obtain on one portion of the Wechsler Intelligence Scale for Children (WISC).
What is the WISC, you ask? It is a gold-standard intelligence test for children up to the age of 16. It consists of 15 subtests that combine to provide scores for verbal comprehension, perceptual reasoning, processing speed, and working memory. These scores can be combined to generate a single IQ score, provided there isn’t too much spread between them.
The WISC and the PPVT serve different purposes. They are not comparable (see figure 2). The PPVT, in its entirety, is roughly proportionate to one of the 15 WISC subtests.
When psychologists speak about IQ, they typically break it into two major components: verbal intelligence (VIQ) and performance intelligence (PIQ). PIQ refers to abilities like spatial reasoning that cannot be measured by verbal tests, and certainly are not captured by the PPVT.
And this is where we encounter a major problem with Kanazawa’s data. By using a study that measures only verbal ability (and very little of it), he is ignoring PIQ altogether. A child whose PIQ is higher than his or her VIQ – a common occurrence – is effectively tossed out of the pool of “more intelligent” people. That skews the data in favor of Kanazawa’s hypothesis.
It is commonly accepted that children who excel on verbal scales are likelier to succeed in school than those who do not. Children who have high VIQ scores, and therefore perform well in school, are likelier than their verbally challenged classmates to go to college. It is common among the highly educated for VIQ to be significantly higher than PIQ (see for example Kaufman & Lichtenberger, 1999).
Moreover, children who struggle with VIQ but who have perfectly solid PIQ scores are more likely than their counterparts to drop out of school (Romi & Marom, 2007). That does not make them unintelligent, but simply intellectually mismatched for a word-based academic environment. These children may even be more intelligent, on balance, than their classmates.
Kanazawa’s measure of intelligence is skewing the data in favor of children who are the likeliest to go to college, even though they may not be more intelligent when all abilities are taken into account. In the next section, I’ll describe the ramifications to the dependent variable.
Lest you think this is a minor point, VIQ-PIQ discrepancies are more common than not, and they run in both directions. In fact, this discrepancy is one of the first and most important statistics that psychologists calculate when scoring an IQ test.
How common is it? Hsu et al. reported that 45.5 percent of the population within the Average IQ range display a VIQ-PIQ discrepancy of 9 points or more. Scoring manuals, such as that for the Wechsler Adult Intelligent Scale-III (p. 207) corroborate. It reports that 42.7 percent of the population display a nine-point difference, and the mean difference across IQ scores is 8.6.
By relying on a test that merely approximates VIQ, Kanazawa ignored a major component of intelligence and hopelessly tainted his data. Had he instead tested subjects’ ability to rebuild a carburetor, he would have arrived at a different (and equally flawed) pool of “more intelligent” people. Instead, his definition of “more intelligent” is biased toward those with relatively higher VIQs and away from people with relatively higher PIQs.
Poisoning the data even further, Kanazawa interpreted the PPVT as a measure of general intelligence while openly acknowledging that “the PPVT is properly a measure of verbal intelligence, not general intelligence.”
Even that much is debatable. Sadock & Sadock’s (2003, p. 191) Synopsis of Psychiatry classifies the PPVT as a test of one-word receptive and expressive language – not verbal intelligence, and certainly not general intelligence. In my experience, most clinical psychologists regard the PPVT as little more than a screening tool.
At best, one can say that the PPVT provides a decent estimation of VIQ in a pinch. But Kanazawa engaged in dodgy statistical contortions to tie the PPVT to general intelligence. For example, he asserted that the PPVT has a good correlation with Raven’s Progressive Matrices, which he said is “widely regarded as the best measure of general intelligence.”
Wrong. Raven’s Matrices is a test of nonverbal reasoning. Sadock & Sadock (2003, p. 191) classify it as a test of executive functioning, not general intelligence. I know of no clinical psychologists who turn to Raven’s Matrices for a measure of general intelligence.
But let’s say, for the sake of argument, that the PPVT and Raven’s Matrices could somehow be tied together to provide an estimate of general intelligence. The best evidence Kanazawa could muster was from a study that found correlations between the tests of .22 (weak) among first graders and .52 (moderate) among both third and fifth graders. He provided no correlation for the two tests among adolescents, dismissing the lack of evidence by claiming that “It appears that the PPVT becomes a better measure of general intelligence as children get older.”
It is customary to provide evidence for such assertions, especially when stretching data well beyond its intended use. But that really doesn’t matter. Irrespective of any presumed correlation, propping up a vocabulary test with a test of nonverbal reasoning is sketchy business from the start. They simply measure different things, and neither measures general intelligence.
Apologies for the jargon. I’ll spare you the rest of the gory statistical details and refer you to page 42 of his study. While engaging in nimble leaps of logic, Kanazawa even went so far as to provide IQ ranges (see figure 1) based on this single, paltry vocabulary test.
All of his explanations and contortions aside, Dr. Kanazawa based his intelligence data on a limited measure of verbal ability, and nothing more. Returning to my point about VIQ and academic success, his methodology thus far has factored out the very children who would probably contradict his theory. How might the little rascals do that? Read on…
Liberalism: The Dependent Variable and Even More Broken Data
Kanazawa’s liberalism data is at least as muddled and biased as his intelligence data for two reasons.
First, the “more intelligent” kids – because of the way they were selectively funneled into this category via the aforementioned methodological mess – are more likely than their classmates to end up in college. Anyone who has paid attention during the last 50 years can attest to the fact that most colleges and universities overwhelmingly encourage liberal thought while discouraging conservatism. Kanazawa himself has acknowledged that academia is dominated by liberals. The data back him up.
In an in-depth study of the political orientation of college professors, Gross & Simmons (2007) reported that 62.2 percent of professors identified as liberal, while only 19.7 identified as conservative. That means that Kanazawa’s “more intelligent” people were more likely than their counterparts to be immersed in liberal ideology at the time of Wave III data collection.
Second, the timing of Wave III data collection corresponds nicely with Kanazawa’s desired outcome. Recall that Wave III data was collected when subjects were between the ages of 18 and 28. According to the Pew Research Center (2003), young adults in the present population who choose a party affiliation (most do not choose an affiliation) are about evenly divided between Republican and Democrat. With increasing age, affiliation shifts toward the conservative side, most dramatically among men.
There is only one reasonable explanation for this: people lose IQ points as they move toward conservatism and gain points if they become more liberal. That has been well documented in The International Journal of Shrinks Who Say Conservatives Are Doody-Heads.
Another possibility is that “more intelligent” people, regardless of their ultimate ideological destination, were more likely to identify themselves as liberal at the time that they were measured. It seems that Wave III data was collected at a time in the subjects’ lives when their answers would be most beneficial to Kanazawa’s theory.
Kanazawa’s dependent variable is a perfectly stacked deck: select the people who are likely to confirm the hypothesis, then time the data collection to the height of confirmatory likelihood.
A Word about Definitions
Though I am no expert on politics, I would be remiss if I ignored the glaring issue of Kanazawa’s ideological definition:
“I provisionally define liberalism (as opposed to conservatism) as the genuine concern for the welfare of genetically unrelated others and the willingness to contribute larger proportions of private resources for the welfare of such others.”
Got it? Liberals care about unrelated others, and conservatives do not. This is perhaps the most simplistic, unsubstantiated, self-serving definition I have ever encountered.
Liberalism and conservatism are not measures of generosity. They are ideologies concerned with the organization of societies. Denver-based political commentator Mike Rosen (1986), citing James Burnham, suggested that liberal-conservative differences lie in the prioritizing of societal values. For example, liberals might put the following values in this order of importance:
Peace, Justice, Freedom, Liberty
…whereas conservatives might prioritize those same values differently:
Liberty, Freedom, Peace, Justice.
That is just one way to think about the differences. There are others. My point is that Kanazawa is dealing with complex ideologies. To delineate them along a single heartstring sentiment is plain silly and clearly tendentious.
His contrived definition is also demonstrably wrong. In an extensive research review, Brooks (2006) found that American conservative households donate 30% more of their personal income to charity than do liberal households. I am not suggesting that American liberals are stingy. Perhaps they express their compassion in other ways. Rather, I am pointing out that the existing data flatly contradict the very foundation upon which Kanazawa has built this house of cards.
You may be wondering how Kanazawa deduced that liberals are more generous than conservatives. In his own words,
“In the modern political and economic context, this willingness [to care for unrelated others] usually translates into paying higher proportions of individual incomes in taxes toward the government and social welfare programs.”
This is a shining bit of irrationality. Anyone who voluntarily donates her hard-earned resources to others can truly be called altruistic. On the other hand, someone who supports a tax increase may or may not be the target of that increase. They may, in fact, be the beneficiaries. It hardly qualifies as altruism to support a tax increase on someone else, especially if that increase transfers resources into one’s own pocket. I am not pointing fingers; I am saying it has been known to happen.
This absurd definition has no direct effect on Kanazawa’s data collection, but it certainly undermines his hypothesis, and that is unfortunate.
Why the Methodology Matters
Satoshi Kanazawa’s latest paper is a parade of tendentious, self-serving methodological problems. In spite of that, I do not believe that he harbors any special animosity toward conservatives. Frankly, he seems like a nice guy. I enjoy much of his work – I’ve even cited it in the past – and I take no pleasure in beating up on his methodology. I do this because ideas have consequences. Whether it was intentional or not, Dr. Kanazawa has exacerbated a serious problem in the field of psychology.
My profession has a shameful history of beating up on conservatives. For example, a contrived study from 1970 used methodology strikingly similar to Kanazawa’s to reach the same conclusion. Surprise! Our side is smarter than theirs (see Harvey & Harvey, 1970).
More recently, John Jost et al. (2003) concocted a remarkably biased literature review which judged conservatives to be dogmatic, closed-minded, and nearly incapable of grasping shades of gray. Their methodology, which I’ve detailed here, is a shameless collection of rigged samples, shifting definitions, confirmation bias, and other methodological sins.
These are just two of many examples. Based on the sheer number of these studies, you might reasonably suggest that they are correct. Maybe conservatives are, in fact, intellectually inferior to liberals.
That is possible, but every conservative-bashing study I have encountered has been absolutely biased and broken from the ground up. Perhaps that is because they are so incestuous, with each one compounding the errors of the studies that preceded it. Anti-conservative researchers cite each other’s work with unquestioning faith, just as Kanazawa, in this very paper, used the intellectual garbage produced by Jost et al. to prop up his own findings.
I am certainly not the first to notice psychology’s animosity toward conservatives. As Richard Redding (2005) pointed out,
“To date… psychological research has been strongly biased toward validating the ‘flattering’ psychological portrait of liberalism and the ‘unflattering’ portrait of conservatism.” (p. 309).
This mounting body of literature results in a helping profession that is not very helpful to those who sit outside its narrow ideological comfort zone. As Redding put it,
“In effect, psychology’s pervasive liberal Zeitgeist may adversely affect treatment or program effectiveness with politically conservative clients and communities” (p. 311).
…To say the least. We psychologists bill ourselves as healers, but why would any conservative trust us in the face of such clear hostility? And don’t think that those dumb Republicans don’t notice. Surprisingly, they know how to read, and “findings” like Kanazawa’s are leapt upon by an eager press.
Let’s look at some headlines to see what the mainstream media has gleaned from Kanazawa’s latest research on the Savanna Principle:
|National Geographic: Liberals, Atheists Are More Highly Evolved?Fox News: Are Liberals Smarter Than Conservatives?CNN: Liberalism, Atheism, Male Sexual Exclusivity Linked to IQ
Toronto Star: Are Liberals and Atheists Smarter?
Hey media guys, you’re missing the point. He was trying to illustrate his theory called the Savanna Pri… oh, never mind.
Trying to mitigate the needless harm caused by this type of research is like trying to remove urine from a swimming pool. One does what one can, but the damage is done. Ultimately, this study will become one more cheap ideological bludgeon. I doubt that is what Dr. Kanazawa meant to achieve, which makes it all the more disappointing.
* In Study 2 of Why Liberals and Atheists Are More Intelligent, Kanazawa generalized his findings to the population at large through a different vocabulary test and similar self-report measures of ideology. There is no point in discussing Study 2, as he simply replicated the errors from the first part of his paper.
Brooks, A.C. (2006). Who Really Cares: The Surprising Truth About Compasionate Conservatism Who Gives, Who Doesn’t, and Why It Matters. New York: Basic Books.
Gross, N. & Simmons, S. (2007). The Social and Political Views of American Professors. Working paper. Downloaded March 30, 2010 from: http://www.wjh.harvard.edu/~ngross/lounsbery_9-25.pdf.
Harvey, S.K. & Harvey T.G. (1970). The effects of intelligence as an independent variable. Midwest Journal of Political Science, 14(4), 565-595.
Hsu, L.M., Hayman, J., Koch, J., & Mandell, D. (2000). Relation of statistically significant, abnormal, and typical WAIS-R VIQ-PIQ discrepancies to full scale IQ. European Journal of Psychological Assessment, 16(2), 107-114.
Jost, J.T., Glaser, J., Kruglanski, A.W., & Sulloway, F.J. (2003). Political conservatism as motivated social cognition. Psychological Bulletin, 129(3), 339-375.
Kanazawa, S. (2010). Why liberals and atheists are more intelligent. Social Psychology Quarterly, 73(1), 33-57.
Kaufman, A.S. & Lichtenberger, E.O. (1999). Essentials of WAIS-III Assessment. New York: John Wiley & Sons, Inc.
Pew Research Center for the People & the Press (November 5, 2003). The 2004 Political Landscape. Downloaded March 28, 2010 from: http://people-press.org/report/?pageid=750.
Redding, R.E. (2005). Sociopolitical diversity in psychology: the case for pluralism. In Wright, R.H. & Cummings, N.A. (eds.) Destructive Trends in Mental Health: The Well-Intentioned Path to Harm (303-324). New York: Routledge.
Rosen, M. (1986). The fundamental differences between liberals and conservatives. The Denver Post, August 6.
Sadock, B.J. & Sadock, V.A. (2003). Synopsis of Psychiatry: Behavioral Sciences/Clinical Psychiatry. Philadelphia: Lippencott Williams & Wilkins.
Wechsler, D. (1997). WAIS-III Administration and Scoring Manual. The Psychological Corporation, Harcourt Assessment Company.