Is GDP caused by Verbal Ability or g? A Response to La Griffe du Lion
“Deeply gratified by Mentor's interest in smart fraction theory, and mindful of his appreciation of its good fit to observation, Prodigy offers a refinement to the theory that, notwithstanding the success of the original, should forever alter the way in which national wealth and IQ are perceived.” - La Griffe du Lion (2004)
La Griffe du Lion is one of the greatest and one of the first Human Biodiversity bloggers. In the early 2000s, he wrote about topics ranging from Why men are more likely to be serial killers to sex differences in mathematical ability. Best of all, he managed to write most of his ideas as Socratic dialogues between ‘mentor’ and ‘prodigy’. You can find his blog here.
His lasting influence was the creation of Smart Fraction theory. The earliest work on IQ and economic output tried testing for linear relations between the two variables. But a cursory plotting of the variables would show that the relationship was non-linear. La Griffe explained this by arguing that economic output was caused by the proportion of the population with an elite IQ (greater than 106). The theory makes sense, in many fields productivity follows a power law with the best individuals producing many times what others produce. We should expect the elite who control governments and corporations to be more important for GDP.
However, in practice, Smart Fraction theory is hard to distinguish from average IQ having a non-linear effect on GDP. In individuals each IQ point is associated with a constant percentage increase in income, so we should expect the effect to be non-linear anyway. It is standard practice for economists to use a logarithmic transformation of GDP or growth rates in GDP when they are modelling. But this was not entirely understood by Psychologists in the earliest work on national IQ. Moreover, average IQ correlates extremely well with the IQ of the elite within a country making the two effects difficult to pull apart.
Regardless, the theory is still popular today and academic papers are still being written trying to test whether elite IQ matters more than the average. La Griffe’s work has forever changed how we understand IQ and the Wealth of Nations, to paraphrase Lord Keynes “Practical scientists, who believe themselves to be quite exempt from any of anon influence, are usually the slaves of some based blogger.”
Despite the influence of Smart Fraction theory, there is one aspect of La Griffe’s hypothesis that has been completely forgotten and left untested. He did not believe the elite’s general intelligence caused economic output, but rather in a follow-up article, ‘prodigy’ refines the theory by arguing that the elite’s verbal intelligence is the true cause. Yes, La Griffe thinks wordcels are more important than shape rotators for economic growth.
In the early 2000s, North-East Asian countries appeared to be substantially underperforming given their intelligence advantage over whites. La Griffe argues that what makes these countries different is that they tend to perform worse on tests of verbal IQ, in fact, their verbal IQ is maybe 6 points or so lower than their general intelligence. After adjusting for this La Griffe finds the smart fraction intelligence correlates even better with economic output! So surely verbal intelligence is more important to economic output than general intelligence?
The theory suffers from a few problems. Firstly, it is an arbitrary, ad hoc attempt to improve the fit of intelligence with GDP. I’m not aware of any theory or previous evidence to suggest verbal intelligence is more important for productivity, so how do we know La Griffe’s theory is not just overfitting or a fluke? Secondly, there are alternative explanations for the underperformance of East Asia. Communism is a good explanation for China’s underperformance. Poor relations with China and distance from European markets has surely hurt the other North-East Asian countries. We could also speculate regarding culture and conformism, Asians make hard-working rich workers in the west but they are much less likely to take risks, starting and running businesses. Today East Asia’s underperformance does not seem dramatic, Singapore is incredibly rich, Japan and South Korea or roughly on par with the UK in GDP per capita.
But La Griffe’s theory is worth testing properly. Using the OECD’s Programme for International Student Assessments (PISA) 2012, I’ve downloaded the data for what proportion of students get each question correct. Because not all the same questions are given to each country I remove some questions and some countries that do very few of the question items. Then using principal component analysis I create a measure of general intelligence out of all the items, then mathematical, reading and scientific ability from the math, reading and science questions. Then I model these specialised abilities with our measure of general intelligence and take the residuals - ie. how impressive is a country at maths relative to their general performance across all questions?
I then run regressions of the log of GDP per capita (2017), explained by general intelligence and each of the residualised special abilities. As you can see in the regression table, whilst general intelligence is associated with GDP per capita the specialised abilities are not statistically significant.
It looks like having verbal or mathematical intelligence are not important for GDP, only general intelligence matters. We can go further and use Jensen’s method of correlated vectors. In the graph below, on the x-axis, we plot each question item’s correlation with the general factor of intelligence. On the y-axis, we plot each question item’s correlation with GDP per capita. These correlate at 0.74, the better a question item measures general intelligence the better it predicts GDP. GDP is related to general intelligence.
We can go even further by taking other principal components from the PISA items, capturing other sources of variance in the intelligence tests. If there is a component of intelligence, such as verbal ability, but not general intelligence, that is related to GDP, then one of these other principal components should predict economic output. Alas, when we run the regressions with other principal components only general intelligence matters.
The legend of La Griffe du Lion has changed how we understand IQ and economic growth, but it looks like he was wrong about the importance of verbal intelligence. But shouldn’t we be surprised that non-g sources of variance in national intelligence have no association with GDP? The psychological literature suggests that environmental effects (the Flynn Effect) are concentrated on question items that are less g-loaded. This means the non-g sources of variation in PISA scores should capture this environmental effect. And yet these non-g sources appear to have no relation to GDP. Wouldn’t that imply that economic growth does not increase intelligence? Wouldn't it also suggest environmental differences in national IQ don't cause GDP? Anyway, that topic is for another time.