My posts below on IQ, politics & religion resulted in a fair amount of blogospheric response, and weird comments. A few quick points
1) I think results on standardized tests are informative and correlate reasonably with a host of life outcomes. If you don’t think they do, that’s fine, I don’t particularly care. But just do note that your dismissal of IQ carries no weight with me. Additionally, though I believe IQ to be substantially heritable, even that’s not necessarily important in this case (since I am not focusing on evolution). Rather, realized test scores correlate with other outcomes.
2) As I said, I’m not too hung up on Kanazawa’s paper itself. I’m not too fixated on the causal mechanism either, and barely paid attention to the evolutionary model. Rather, I focused on the description of the distribution. The particular results are not novel, they are in line with findings which have come before.
3) If you poke around the World Values Survey (use education as a proxy for intelligence, go back to #1 and note that I’m more concerned about realized intelligence in this case, not a reified construct of g) and the GSS you can see similar trends.
4) Some readers point out that the relationships may not be monotonic. This is evident with education and political party voting in the United States, as Democrats are the most and least educated segments. This is fair, and I would like to see surveys of the political views of the super high IQ vs. the above average. I have found data which shows that political moderates are actually the least intelligent, probably because smart people have clear & distinct ideological viewpoints, while stupid people express wishy-washy moderation since they’re just stupid.
5) Correlation is not transitive and Kanazawa measured verbal intelligence. Correlation is not necessarily transitive. But it might be, and in any case, as I suggest above I wasn’t too concerned with being psychometrically precise here in terms of g. Additionally, caution in this case is modulated by other results which point in the same direction.
6) IQ has a fat tail, so the normal approximation is “nonsense.” No, the fat tail is really powerful extremely deviated from the median. At IQ 130 there are only 10% more than you’d expect, and at IQ 145 only two times as many. Of the people with IQs above 130 only 5% or so have IQs above 145. So though the fat tailing is a problem, it becomes a problem precisely in the proportion which is the smallest of the high IQ segment.
Some conservatives and religious people are pointing out that just because smart people believe X it does not follow that X is true. No shit. I don’t know who is stupid here, the people who believe this, or the people who believe that people believe this. Others point out that we haven’t established the causal chain here, so it doesn’t matter. Actually, it does. As a non-liberal person who has conversations almost exclusively with people whose IQs are north of 130, and sometimes north of 145, it is very rare that I come across individuals with non-Left liberal or libertarian viewpoints. Part of this is the selection-bias of social milieu, but I think that it goes beyond that, and at the commanding heights of the elite, which tends to be high IQ, there are certain views which are extremely overrepresented and that distorts public policy (whether the distortion is good or bad depends on your viewpoint). In any case, as someone who is a “person of the Right,” and has viewpoints which are outside of the liberal/libertarian arc I am sensitized to the biases which the cognitively endowed seem to take for granted (I’m not politically interested enough to really care too much, this is more an observation and fact of my life, not an injustice which needs to be corrected).
It is important that when it comes to politics people need to make a distinction between liberalism on social issues and general political affiliations or party voting. The view that IQ tends to correlate positively with social liberalism seems much more robust to me than that IQ correlates with ideology in general, or that IQ correlates with party voting (there is some data which suggests that Republicans may be smarter than Democrats, though it depends on how you slice the data).
But enough chatter. Here are some data from the GSS. The variables:
ROW: ABANY GOD BIBLE FREETRD, MORETRDE
COLUMN: WORDSUM
I filtered so that only whites in the sample were included (blacks invariably have much lower WORDSUM scores no matter other variables).
I combined categories for WORDSUM, so 0-4 = dumb, 5-8 = average, and 9-10 = smart. 18.5% are dumb, are 66.4% are average, and 15.2% are smart in this sample. The N for the total sample is over 20,000, but some variables have smaller N’s, so I invite you replicate and poke around the data yourself (which is why I provide link and variables). You can look at the confidence intervals and how robust these findings are yourself.
The rows below add up to 100%. So for example in the first row of data, 23.8% of atheists are smart. What you’re seeing are the proportion of smart, average and dumb people in each class.
| Dumb | Average | Smart | |
| Confidence in the existence of God | |||
| Atheist | 20.3 | 55.9 | 23.8 |
| Agnostic | 6.2 | 66.6 | 27.2 |
| Higher Power | 12.6 | 64.8 | 22.6 |
| Believe Sometimes | 18.3 | 66.9 | 14.8 |
| Believe but Doubts | 14.1 | 71.7 | 14.2 |
| Know God Exists | 18.2 | 70.2 | 11.2 |
| Feelings about the Bible | |||
| Word of God | 29.1 | 65.4 | 5.5 |
| Inspired Word | 12.4 | 71.6 | 16 |
| Book of Fables | 13 | 59.2 | 27.8 |
| Think of self as liberal or conservative | |||
| Extreme Liberal | 21.3 | 51 | 27.7 |
| Liberal | 17 | 59.3 | 23.7 |
| Slight Liberal | 15.7 | 63.6 | 20.7 |
| Moderate | 20.4 | 69.5 | 10.1 |
| Slightly Conservative | 13.9 | 68.9 | 17.2 |
| Conservative | 16.8 | 67.3 | 15.9 |
| Extreme Conservative | 24.9 | 62.8 | 12.4 |
| Abortion if woman wants for any reason | |||
| Yes | 13.6 | 65.7 | 20.7 |
| No | 22.4 | 66.6 | 11 |
| Homosexual relations are…. | |||
| Always Wrong | 23.7 | 66.3 | 10.1 |
| Almost Always Wrong | 13.8 | 68.5 | 17.7 |
| Sometimes Wrong | 8.5 | 64.6 | 26.9 |
| Not Wrong at All | 9.6 | 63.9 | 26.6 |
| More trade, less jobs in USA | |||
| Created More Jobs | 10.3 | 64.7 | 25 |
| About the Same | 10 | 70.3 | 19.8 |
| Taken them Away | 19.1 | 77.8 | 3.2 |
| Free trade leads to better products | |||
| Strongly Agree | 10.9 | 65.2 | 23.9 |
| Agree | 12.8 | 73.1 | 14.2 |
| Neither Agree nor Disagree | 9.4 | 72 | 18.6 |
| Disagree | 9.9 | 74.7 | 15.4 |
| Strongly Disagree | 11.4 | 81.9 | 6.6 |
One thing to take away: the intelligent are more liberal, broadly. That is, they support liberal trade policies, even though the modern Left-liberal party, the Democrats, is more hostile to free trade than the Republicans, who have traditionally operated as a Right-liberal party. It is not coincidental I think that the Democratic party in the United States, led by cognitive elites, have done little to rollback the push of globalization which has resulted in an erosion of the employment possibilities of working class Americans. Similarly, it is not coincidental I think that the modern Republican party, led by cognitive elites, has done little to rollback the push toward liberal social permissiveness in the United States. I recall back when George W. Bush was elected that Bob Guccione’s lawyers were drafting a new porn code which would protect them from obscenity prosecutions (something with John Ashcroft was actually working on before 9/11). That all seems quaint now.
Razib Khan’s degrees are in biochemistry and biology. He has blogged about genetics since 2002, previously worked in software development, is an Unz Foundation Junior Fellow and lives in the western US. He loves habaneros.
March 4th, 2010 at 7:27 pm
As if you haven’t worked this enough, is there a difference between men and women?
I think the wordsum might favor women.
March 4th, 2010 at 7:32 pm
yes. in the sample above there are notably more men in the dumb category than in the average, outside of 95% confidence intervals. but about the same in the smart category. men exhibit more variance.
March 4th, 2010 at 8:06 pm
My point is that maybe the cutoff for smart women should be 10 not 9.
Kind of like the SAT. Women do about as well on verbal but not math.
Math requires more reasoning than remembering words. Since the wordsum is just a proxy, it might need to be adjusted for gender in order to correlate to thinking ability. With only ten questions, moving the threshold by a point up for women or a point down for men might overcorrect. I think what you are trying to tease out of the data are correlations between thinking ability and thoughts on topics that require thinking rather than just correlations to vocabulary knowledge. It just seems that you end up with similar numbers in the smart category which is eerily similar to SAT data for verbal but not math.
Just a thought.
March 4th, 2010 at 8:16 pm
i controlled for sex after you asked. i didn’t see much of a difference except that men exhibited more variance. i’ll look at it in more detail tonight when i have time.
March 4th, 2010 at 9:05 pm
So, Razib, are you on the right, using either of the two examples by which the intelligent are liberal? Do you oppose liberal trade policies? Favor stricter pornography laws?
March 4th, 2010 at 9:19 pm
Free Trade is a good example of how liberal attitudes require a more nimble mind. “Comparative Advantage” is a difficult concept, and so we wouldn’t expect the duller citizens to grasp it. That they intuitively sense that it’s a crock and is never actually realized in practice, that’s of course another matter entirely:)
March 4th, 2010 at 9:25 pm
Do you oppose liberal trade policies? Favor stricter pornography laws?
i am liberal, though i am more respectful of non-liberal attitudes toward trade and open to them than a typical liberal. over the long term, and all things equal, trade is beneficial and a boon. but proximately and contingently it may not be so beneficial, there are winners and losers, and if you conceive of society as an organic whole you can make anti-trade arguments.
as for porn, i’m pretty liberal. but when it comes to speech the correlation between intelligence and hostility to censorship is really strong. i think it’s because the smart tend to think they have a lot interesting to say.
March 4th, 2010 at 9:27 pm
i think within the past generation the perception is that trade has hurt the lower classes, while the professional licensed sector and business elites have accrued gains. so the intuition could derive from simple economic interest.
March 4th, 2010 at 10:59 pm
Someone who believes that trade is beneficial in sum, but has hurt the lower classes, might favor strengthening the social safety net over further restricting trade. Or to put it another way, people can be driven to liberal social policies from the combination of respecting the economic power of trade and understanding its benefits vary quite a bit across different groups.
I wonder if you would be perceived as a “person on the Right” from the perspective of the Tea Parties rather than from more intellectual circles. Socially liberal atheists are more their rhetorical demons than political allies, no matter one’s stance on trade and particular foreign policies.
March 5th, 2010 at 1:43 am
Intelligence should also correlate with the depth of policy vision. I would guess that the more intelligent would tend to have considered issues less commonplace than whether homosexuality is moral, such as, say, the overall effect of fiscal expansion on GDP. It would be interesting to see whether, at that level, the more liberal are still overrepresented, or whether it evens out.
March 5th, 2010 at 5:21 am
The Wordsum test covers vocabulary and thus how widely read you are. Smarter people will tend to read more, but religious people will tend to spend a lot of time on their religion’s scriptures. A smart religious person will have a smaller vocabulary than a smart secular person, or may have as large a vocabulary but with specialized words that might not show up on Wordsum.
One example that comes to mind is Dubya versus Kerry; Bush had better grades at Yale, but naming Jesus as his favorite philosopher makes him look “dumb.” It’s actually less widely read rather than dumb.
March 5th, 2010 at 7:18 am
education could well be the reason people are more liberal. College campuses are not friendly to conservative ideas.
March 5th, 2010 at 8:53 am
Byron, isn’t that just saying that a certain set of religious are stupid _because_ of their values? If you spent your day consuming potato chips and television you wouldn’t say you only _appear_ less healthy than you would if you jogged and had a balanced diet. You actually would be less healthy.
Of course there are cultural attitudes that could lead you to express things more plainly that wouldn’t reduce your intelligence, but I doubt that “not reading” is one of them.
March 5th, 2010 at 1:21 pm
I’d like to bring up something that hasn’t been mentioned in this debate but I immediately start thinking about when I read people talking about this.
This is going to be extremely politically incorrect, but have you considered that being very religious or very conservative actually lowers people’s intelligence?
I grew up in the Deep South where people are very conservative and very religious. They also scored very low on standardized test scores. It didn’t matter what their ethnic background was (usually either white or black); so I really doubt that genetics played a role. The only thing that crossed ethnic/cultural boundaries was being very religious and very socially conservative. Being very religious and very socially conservative required them to reject things that were observably true. Two examples I would give are evolution and that gay people weren’t part of some grand demonic conspiracy. In order to remain in good standing in your community you had to either accept this nonsense or accommodate it. People who “accommodate it” generally do things such as say “while I personally believe that X is an accepted provable fact, I would never tell some one they were wrong because of their religious beliefs.”
Let me illustrate this in a metaphor that takes out a lot of our present culture war language. Imagine you have a population. 75% of this population believes that the earth is flat and 2 + 2 = 5. The remaining 25% of this population believes that it is wrong to attempt to disprove these conclusions false to the other 75%. Would you really be surprised that this population on the whole scores worse on standardized tests than a population that’s not hampered by these beliefs?
March 5th, 2010 at 4:44 pm
I have found data which shows that political moderates are actually the least intelligent, probably because smart people have clear & distinct ideological viewpoints, while stupid people express wishy-washy moderation since they’re just stupid.
Looking at WORDSUM versus POLVIEWS, it looks like political moderates are from the middling segment (3-6, the lower end). The least intelligent (0-3ish) tend to fall to the extremes. So, it’s more that the smart have clear and distinct viewpoints, the very dumb are easily led to the extreme, and the not-so-bright (3-6) aren’t bright enough to figure out clearly where they stand, and simply follow the (local) herd.
March 6th, 2010 at 12:13 pm
What does it even mean to say in this context that IQ is not normally distributed, and has a fat tail?
It may be that certain tests of IQ are such that there are, say, twice as many people at 130 IQ than a normal distribution would entail, under the assumption that each IQ point across the range of IQs is equal in what it is measuring.
But the problem has always been — hasn’t it? — that there’s just no way to know whether each IQ point is equal in what it’s measuring. This is the well known basic issue in statistical measurement of whether a given metric is an ordinal scale, an interval scale, or a ratio scale. Basically, IQ as any kind of a ratio scale has been pretty much rejected. Moreover, there’s no known way to make sense of it as an interval scale either. So we are left with an ordinal scale. Problem is, an ordinal scale leaves us pretty much nowhere with the question as to how to fashion a distribution from raw scores on IQ subtests.
My understanding is that, for the most part, psychometricians adopt the simplest possible such mapping: one that maps those raw scores into a normal distribution.
If so, I don’t see how one can come up with more than expected IQs at any given range if one has a large representative sample from the underlying population over which IQs are originally defined.
March 6th, 2010 at 1:25 pm
“The nature of Truth-the-Fact cannot be described by means of verbal symbols that do not adequately correspond to it”
Aldous Huxley
Of the ontological, experiential, and symbolic truths, which do you utilize the most to surmise your assessment of intelligence? Balance is key.
March 6th, 2010 at 3:19 pm
16.
But the problem has always been — hasn’t it? — that there’s just no way to know whether each IQ point is equal in what it’s measuring. This is the well known basic issue in statistical measurement of whether a given metric is an ordinal scale, an interval scale, or a ratio scale. Basically, IQ as any kind of a ratio scale has been pretty much rejected. Moreover, there’s no known way to make sense of it as an interval scale either. So we are left with an ordinal scale. Problem is, an ordinal scale leaves us pretty much nowhere with the question as to how to fashion a distribution from raw scores on IQ subtests.
If so, I don’t see how one can come up with more than expected IQs at any given range if one has a large representative sample from the underlying population over which IQs are originally defined.
IQ scores are definitely only ordinal, so the comments above are good. Even worse, the items discriminate better in the fat part of the curve (97 versus 99 could be meaningful)and less well outside (the difference between 140 and 142 is probably not meaningful). This I think is a practical matter– to discriminate at the tails one needs different tests (tests that most people would get all correct or all wrong).
Elementary cognitive tasks, however, are indeed ratio scales so one could get more precision studying the whole spectrum of performance.
March 6th, 2010 at 5:52 pm
“If so, I don’t see how one can come up with more than expected IQs at any given range if one has a large representative sample from the underlying population over which IQs are originally defined.”
I think it’s important to remember that IQ is not a “measurement” in the empirical sense. It’s an operational outcome. I’m not sure why we should expect a perfectly normal distribution. Particularly, I think, because IQ test performance likely involves several different cognitive traits, and probably everyone is not using–or has–the same subset of cognitive “modules” (or whatever) when they take a test.