"So I took a subset of the people with the most unimpeachable IQ tests - ones taken after the age of 15 (when IQ is more stable), and from a seemingly reputable source."

I am a member of this population, and I lied. Although I have taken variants of the aforementioned tests, I have never done so in an academic or professional context (ie; Raven's via iqtest.dk). I suspect that I am not the only one.

"People were really really bad at giving their answers in millions. I got numbers anywhere from 3 (really? three million people in Europe?) to 3 billion (3 million billion people = 3 quadrillion)."

Two-thirds have a college degree and roughly one third are European citizens. Does this bode well for the affirmation about self-reported IQ?

"...so it was probably some kind of data entry error..." "Computers (practical): 505, 30.9%"

If people lie about IQ, why not just check Wikipedia and cheat on the Europe question? I lied about IQ, but I did not cheat for the Europe question. I suspect that I am not alone.

IQ is arguably as direct a challenge to self-appraisal as you can put to anyone who would self-select for an LW survey. Because mean for HBD was 2.7, many of the respondents may feel that IQ does not fall into predictable heritability patterns by adulthood (say, 27.4 years old). Could it be intertwined with self-attribution bias and social identity within a community devoted to rational thinking? Perhaps they don't realize that rational decision-making =/= improved performance on Raven's Progressive Matrices.

If I was a member of a health club for 2.62 years, ipso facto, would I be inclined to self-report as physically fit/strong/healthy (especially if I thought I had control over said variable, and that it wasn't largely the result of inheritance and environmental factors in a seemingly distant childhood)?

Self-reported IQ data via an online survey: robust? C'mon, you're smarter than that...