We have many clues that hint at the intelligence of the people around us. These clues include the size of their vocabulary, the wit of their jokes, and the insight of their observations. These clues, however, are less useful for distinguishing the intelligence of people who are smarter than us; we may not get their joke, understand their insight, or recognize their big word. We might rely on evaluations of people smarter that us, but this just moves the problem back a level.
This encourages assortative mating, because, for example, only the smartest women can can tell clearly who are the smartest men. And it often makes it hard to reward people who are smarter than elites. Yes, in some areas of life, like chess, intelligence may reveal itself quickly in better outcomes. But usually, one has relatively little to gain by acting smarter than elites; elites usually can’t tell the difference, and if they can they may resent you for making them look bad. You may well be better off hiding your extra intelligence, finding an area where the elites are smarter than you, or finding an area where better outcomes quickly show smarts.
Arbitrary fluctuations in who are the elites in an area can thus change whether smart people are attracted to that area. And the possibility of such fluctuations pushes smart people toward areas where better outcomes quickly reveal intelligence. This is a plausible explanation for why smart people tend to prefer mathematical areas. I quickly learned as a new teacher that my students were just as bad at math as at writing, but they preferred writing assignments because they could not as easily see that their writing was bad; math reveals intelligence more clearly. Of course this effect could induce people to rely too much on math; people may prefer to show their smarts at the expense of being useful.