Estimating the IQ Distributions of Languages

Will high IQ people switch to German as the new language of high culture?

It turns out it’s pretty easy to estimate the average IQ of a language, with one assumption: within nations, speaking a certain language does not correlate with IQ. If this assumption holds, all we need are the average IQ of each nation, as well as the number of speakers of whatever language we’re interested in.

The Model: Concatenated Gaussians

Let’s specify our model more explicitly. We want to get the mean and standard deviation of IQ for each language group from national IQ and speaking rate data.

To do this, we say our speaking group random variable is a mixture of Gaussians:

\( X = \{p_1 : N(\mu_1, \sigma^{2}_1), ... , p_n : N(\mu_n, \sigma^{2}_n) \}\)

Where to get, say, an English speaker, you draw from X, which then draws from one of the n mixture distributions with probability p_i.

It turns out that

\(\mathbb{E}[X] = \sum_i p_i \mu_i\)

and

\(\mathbb{V}(X) = \sum_i (p_i \sigma^{2}_i + p_i (\mu_i - \mathbb{E}[X])^2)\)

Here, every p_i is the number of language speakers in a country divided by the total number of speakers of that language, every mu_i is the average IQ of that nation, and we assume every nation’s standard deviation is 15.

Estimating the Average IQs of English, French, Spanish, and German

I pulled average national IQs from here and language speaking rates from Wikipedia articles.

Surprisingly, English and Spanish had roughly the same means. French had the highest standard deviation, which was also surprising, as I was expecting English to be more variant.

This chart also shows the per capita superiority of German. If you put 100 German speakers in one room and 100 English speakers in another, the English room would have a similar IQ distribution to a black ‘hood, and most of the smart people would be in the German room. This has interesting implications for the internet — what’s going to be smarter on average, a German social media site or an English one, given you’re interacting with a handful of random people on either?

This got me thinking about the high status languages of the past: ancient Greek, Latin, French, and maybe even English, until now. Back in early Modernity, the nobility of countries as far apart as England, Russia, Germany, Sweden, Italy, Poland, and Portugal all spoke French. Some think this is because French had the smartest population, and this attracted smart foreigners to the language, which added smartness, which attracted more smart people, and so on.

Before French, in the middle ages, Latin was the language of the educated. The clergy learned Latin and mass and the Bible were only in Latin. Again, this may be because most of human knowledge existed in Latin at that point, due to the Romans and the lack of intellectual output during the middle ages.

However, German is behind English when it comes to EHC, because of the small number of total speakers.

German may have the future on its side, however. Anglosphere whites are probably the worst whites in the world, coming in at the fattest and the most leftist, as well as the most isolated from selective pressure, while German speakers still have several meta-ethnic frontiers, including close proximity to Russia, which has enough vril to invade its neighbors.

English is also undergoing negrification, with ebonics becoming more and more popular. Eventually, English and its average IQ may decline to the point where it is no longer suitable for high IQ communication. At that point, German becomes the European language with some of the most high IQ speakers and the least low IQ speakers as well as one of the largest existing corpus of knowledge aside from English, so it becomes a natural choice for the new language of high culture. This is all the more true to whatever degree there is an incentive to run away from low IQ speakers. German pulls as much weight at the right tail right now as French and Spanish, which are much larger overall, without all the low IQ people to drag it down.

Getting away from low IQ speakers might become all the more important on the internet, where anyone can come onto any website and say anything they want. I think we all know the feeling of some <100 IQ person saying the dumbest stuff in an online comment thread. It happens now even under my tweets all the time — we’re talking several commenters that clearly can’t read, like the people who responded to my recent mutational load article with “did you control for race.” German is the biggest European language where this is the least likely to happen, given the data.

Who is making these language so dumb?

Why is the average IQ of English 85, anyway?

Here are the top 5 English speaking countries, and their English speaking population in relation to each other. As you can see, Indonesians, Pakis, Nigerians, and Indians added together outnumber Americans and any other national group of English speakers. English is becoming the language of saar and Nigerian princes.

French is the language of the Congo. There are over 100 million Congoids and they currently have a TFR of 6! Soon there will be more Congoids than Americans.

Did you know the US has the second most Spanish speakers of any country? The future of the US is one part ebonics, one part Spanish.

The vast majority of German speakers live in Germany, Austria, or Switzerland. Most of the rest are a part of the German minority in Brazil.

As a final note, these results will be off if speaking a language correlates with IQ within a country. This means that the mean national IQ is not the same as the mean IQ of the people who speak a certain language in that country. As of now, it’s hard to tell how much this biased the results. This could be an interesting topic of further research — especially for English worldwide.

Appendix: Model Proofs

For sufficiently large n:

\(\mathbb{E}[X] = \frac{1}{n} \sum_i x_i \)

For our random variable this can be split into:

\( \frac{1}{n} \sum_i x_i = \frac{1}{n}( np_1 \mu_1 + ... + np_n \mu_n) = \sum_i p_i \mu_i\)

Same trick for variance:

\(\mathbb{V}(X) = \frac{1}{n} \sum_i (x_i - \mathbb{E}[X])^2 := \frac{1}{n} \sum_i (x_i - \mu)^2 \)

\(\frac{1}{n} \sum_i (x_i - \mu)^2 = \frac{1}{n} \sum_i x_{i}^2 -\frac{1}{n} \sum_i 2\mu x_i + \frac{1}{n} \sum_i \mu^2\)

\( = \sum_i p_i (\sigma_{i}^2 + \mu_{i}^2) - \sum_i p_i 2 \mu \mu_i + \sum_i p_i \mu^2\)

And it’s easy to see this is the same as the variance equation given in the first section.