Field of Science

Atheism and IQ: explained by "The Savannah Principle"?

The correlation between atheism and IQ is being talked about again because the study I blogged about last year (by Richard Lynn) has just been published. PZ Myers simply dismisses it - there's no correlation at all, he thinks (although he hasn't read the paper).

But there's another, rather more interesting paper out on the topic. It goes further than Lynn by actually trying to take into account some of the other factors that could explain the correlation.

The author, Satoshi Kanazawa, is an evolutionary psychologist from University College London with some fairly controversial ideas. One of these is what he calls the 'Savannah Principle'.

The idea behind the Savannah Principle is that we have a range of cognitive biases that adapted us for survival in small groups of hunter gatherers, but which are maladaptive for modern life. The controversial bit is that he goes further than that, claiming that people with high IQ can over-ride their evolutionary heritage.

He claims that intelligence is linked to accepting 'correct' but evolutionarily counterintuitive ideas. In his new recent paper, he shows that nations with higher average IQ have higher tax rates and lower income inequality, which suggests that people in these countries are more willing to redistribute wealth to people unrelated genetically. What's more, he shows that higher IQ is linked to less polygamy.

And he also shows that high national IQ is linked to less religion.

Kanazawa concludes that people with a high IQ are more able to overcome our innate, in-built tendency to invent god-like beings.

That's all well and good, but the devil is in the detail. IQ is strongly influenced by environment, and it's quite possible that those environmental influences also affect religion. The web of cause and effect can be pretty complex.

Kanazawa controls for a number of factors, shown in the figure above. Even after taking into per capita GDP and years of education, IQ remains a potent correlate of religiosity.

But I think that there is a factor he's left out. It's well known that, although high national wealth predicts less religion, the connection is not as tight as you might expect. Back in 2004, Norris and Ingelhart showed that wealth distribution is key. Nations with highly unequal wealth distribution are more religious.

In fact, I have a paper due for publication which shows that income inequality is a powerful predictor of religiosity even after controlling for a number of other factors, like GDP, urbanisation, and religious pluralism.

Since Kanazawa has found a link between IQ and income inequality, this immediately begs the question of cause and effect. Is it that high IQ leads to low income inequality and low religion?

Or is it that democratic, homogeneous nations are more likely to invest in other people. As a result, they are more able to develop intellectually and also are more secure (leading to less religion)?

Kanazawa, S. (2009). IQ and the values of nations Journal of Biosocial Science DOI: 10.1017/S0021932009003368

Creative Commons LicenseThis work by Tom Rees is licensed under a Creative Commons Attribution-Share Alike 2.0 UK: England & Wales License.



  2. Ecological fallacy writ large.

    This sort of rot gives athiests a bad name.

  3. Yes the fundamental problem with the Kanazawa's hypothesis is omitted variable bias. I would like to see how IQ relates to susceptibility to the kinds of faulty thinking that underpins supernatural beliefs. That to me would be evidence for the hypothesis.

  4. I can't take you seriously when you misuse 'to beg the question'.

  5. I must admit that I am very disturbed at so many scientists categorically dismissing research that they haven't read or considered in detail just because, as it appears, the topic at hand is politically incorrect. And IQ is always that. Even the way they phrase their statements change, from the usual, and proper, uncertain one, to one where they can admit of absolutely no doubt. It's pathetic.

    Having said that, PZ talks only about causality. Strictly speaking, I would also guess that there is no direct causal relationship between atheism and IQ. Causality can be very hard to prove. But correlation is a different matter, as in the present paper where less religion correlates with national IQ (I suppose that means mean IQ of a sample of that nation's members). This is something that cannot be dismissed easily, and shifting focus from correlation to causality is borderline devious.

    Lastly, what is up with all you anonymous guys? You flame authors for things you aren't willing to explain properly for the rest of us, and you flame Tom for using the wrong phrase, when you perfectly well know what he means anyway. That's pathetic, too.

  6. What about religion being an enabler for wealth inequality. If people have something non logical to base their success on they are going to look the other way at things that are actually ruining their success. Let them eat communion wafers.

  7. Bjorn

    Anonymity is not a method for refuting our arguments.

    Ecological fallacy you could have googled. Anytime you aggregate data that is fundamentally at a smaller level you create a possiblity of ecological fallacy.

    IQ is an individual level characteristic. Not a national one.
    Athiesm is an individual level characteristic. Not a national one.

    The correlation between athiesm and IQ is therefore meaningless at a national level. Even if there is a real correlation (in millions of individuals) aggregating data at a national level causes correlations to appear stronger than they really are.

    And that's completely ignoring the possibility of confounding which Tom Rees mentions.

    Telling a series of people that they havn't considered in detail something you don't understand just because they have provided a pithy takedown is oh, I don't know? Some sort of P word that one should use to describe professional opinions?

  8. I am not refuting your argument of the ecological fallacy. I didn't know it, and find both anonymity and discussion by googling annoying. That's all.

    Telling a series of people that they havn't considered in detail something you don't understand just because they have provided a pithy takedown is oh, I don't know? Some sort of P word that one should use to describe professional opinions?

    I'm confused. What does this refer to? I told who that they hadn't considered what? That I don't understand? Who provided a pithy takedown? You?

    As for anonymity, I don't expect everyone to use their own name, but fly-by commenting as anon is often a conversation killer. I'm thankful you're returned. How about 'AnonymousX,' or something?

  9. Anon: using aggregate-level data to investigate individual-level characteristics does introduce the possibility of ecological error. But it is necessary if the causal link between the independent and dependent variables is mediated by a societal-level factor.

    For example: in the current analysis, have IQ may lead to individual atheism. But a nation with more high IQ individuals may generate a more atheistic culture. So low IQ individuals in that culture might be more likely to be atheists.

    If that were the case, individual religiosity would be driven (in part) by aggregate IQ.

    There are other considerations too: some of the independent factors exist only at the aggregate level (GDP). That doesn't mean it isn't a useful marker of effects expressed at the individual level.

    Yes, there is likelihood of ecological error but any study you do is incomplete and has to be looked at with caveats. The hypothesis needs to be looked at one many levels before it can be accepted/rejected.

    Pointing out potential methodological weaknesses is strengthened if you can put your finger on how they might effect the analysis. Don't just say 'ecological error' - say how ecological error might manifest in this particular analysis.

  10. Bob Conan: I think think that's very likely. More religious nations have less social welfare investment, and religious people are less in favour of state support.

  11. Countering religion requires greater knowledge, not intelligence. It's just happens that people with greater intelligence have more knowledge.

  12. Tom

    Your example to refute my argument that any correlation strength will be partly due to ecological fallacy is itself the textbook example of ecological fallacy.

    Low IQ individuals in a high IQ and high atheistic could well be more likley to be athiests than similar IQ people in a country with lower average IQ and lower average athiestic tendancy.

    This will, as I pointed out, cause the correlations at a national level to be overstated in their strength. This is ecological fallacy. I stated what effect that would have on the correlation in the response. I did this clearly.

    Confounding. I agree that GDP or any other societal factors need to be taken into account. However, you are incorrect to state that this cannot be done with individual level data. Simply analyse data in individuals taking into account societal level factors. Don't go putting it in the too hard basket as it's done every day.

    Just because we as athiests like the idea that High IQ causes athiesm as an hypothesis doesn't mean we shouldn't scream bullshit when tripe like this is served of as some sort of self-congratulatory feel-good exercise.

  13. Anon: then it's only a fallacy depending on your hypothesis. Kanazawa's paper is looking at societal-level effects, which probably do magnify individual effects via cultural feedback. If you try to strip out societal-level synergies, then you will have another type of analysis entirely that will tell you something different.

    Incidentally, Kanazawa has told me he has another paper in press (Q1 2010, apparently) looking longitudinally at an individual level, also showing robust effects. No doubt this paper will also have methodological issues - no paper is perfect.

    Regarding doing a individual-level analysis of country-level effects. It is simply not possible in this case because there are no data that link individual IQ to individual religious belief on a cross-national basis.

    And of course you can factor in GDP into such an individual-level analysis but this is itself generating potential ecological fallacies. You either do everything at a personal level (i.e. personal wealth, but that will not tell you the same thing) or run this risk.

    So the question is: is such a study worth doing? I think that it is (and analyses using aggregate-level data are also done every day). You have to approach the results cautiously, and understand what they mean. But that goes for every study. It doesn't invalidate the results.

  14. Anon: Incidentally, the question is not whether atheists have, on average, higher IQ - this was fairly well established prior to Kanazawa's paper. The question is why. That's why looking at the data on several levels - including aggregate, cross-national data - is interesting.

  15. Consider the utility of reason for the person of 70 IQ, or 80 IQ, or 90 IQ,. . .
    Some trend watchers note that this segment of the population is growing faster than the >100 IQ group.
    At what level of intelligence is reason of greater utility than faith?

    Here is a quote:
    "A religion is sometime a source of happiness, and I would not deprive anyone of happiness. But it is a comfort appropriate for the weak, not for the strong. The great trouble with religion - any religion - is that a religionist, having accepted certain propositions by faith, cannot thereafter judge those propositions by evidence. One may bask at the warm fire of faith or choose to live in the bleak certainty of reason- but one cannot have both."
    [Robert A. Heinlein, from "Friday"]

    Those with a 135 IQ who "choose the bleak certainty of reason" may not find it as bleak as the person with the 85 IQ.
    It has been said that religion is the opiate of the masses.
    It may turn out to be a cheaper opiate than the gov't check.

  16. I frequently hear the argument that you can't argue the existence of God without a definition of God. How can you correlate atheism and IQ without a definition of IQ? And how can you measure the average IQ of a country? Or compare the IQs of people of widely differing cultures? as a special ed. teacher, I have yet to see an intelligence test that reliably compares intelligence among people of one culture. We're talking gobbledygook here.

  17. I highly disagree with these statements, I live in South Africa, which has a relatively high income-inequality ratio, and yet, crime, (which is considered to be highly prevalent in poorer economies) defies a religious system, in fact, I have yet to personally meet a criminal with a judicial and honest belief in Religiosity. In other words, these not-so-smart, poor people, are atheists.



  18. Monica, if it wasn't reliably measuring something, then it would be random scatter and so wouldn't correlate with anything. However, it does correlate. Now, it might not be measuring 'general intelligence', but it's certainly mesuring something, and that something must surely be related to intelligence (what else could it be?).

    Dyllann, I don't know the stats for South Africa, but whenever I see stats on prison inmates the rule is that they are more religious than the general population. The reason for this is that religious people tend to be poorer than average, and poorer people are more likely to be convicted and imprisoned.

    I suppose you could argue that they are not truly religious, because truly religious people don't commit crimes. But hen you are just defining away the data.

  19. I think we can postulate, that prisoners resort to Religion as a superstitious belief system: that they are "being delivered from evil" or any of the many axioms of religion.

    In reality, we need to define how; all of this evidence is correct, that this information will lead to something less nonsensical.

    in other words, what goal are we trying to achieve, by these statements?

  20. There are so many different religions and belief systems, that it would be hard to prove this.

    The ten commandments, for example have been a strong set of rules to keep society from chaos (seems like a higher than average IQ was needed here). The bible lays out history, like any other book to be learned from. It is people that interpret "gods" will & set up internal principles for their own behavior---often based on community mores' and culture.

    This is a discriminatory piece of work--- (Religious people are dumb.) Your article doesn't show a higher thought process at work

  21. I haven't read Kanazawaa's paper, but I am familiar with his name. In 2006 the distinguished statistician Andrew Gelman of Columbia University took the unusual step of writing to the Journal of Theoretical Biology to point out major statistical errors in the paper "Engineers have more sons, nurses have more daughters"
    by Kanazawaa. I'll be interested to see if there are flaws in this one.


Markup Key:
- <b>bold</b> = bold
- <i>italic</i> = italic
- <a href="">FoS</a> = FoS