Nassim Taleb writes on this a lot, and I've seen some similar thoughts from Steve Keen . We expect to be able to take individual behaviors and basically linearly combine them into population behaviors, because it's easy to think that way, and it's easy to model in both computers and our heads. When that turns out not to work, we tend to...
... sorta ignore that, and keep pretending we can linearly combine these things.
"I strongly believe, and have horded anthropological evidence to support this belief, that humans at scale are completely irrational, and to expect otherwise is foolish and irrational."
Per your comment about anthropology, I think people in general are actually often surprisingly rational. In fact, humans will often act quite rationally if you seriously and carefully look at what they are doing when the people using the aforementioned linear models confidently proclaim the irrationality of the humans involved, because the local humans are more capable of understanding the nuances of the non-linear landscapes they are in that the models don't even acknowledge the existence of, and the local humans are constrained in their behaviors by the need to survive and reproduce whereas linear models will tend only to look at groups ("well, in this plan, this one person died literally 15 times but the group overall is doing great numerically, so that's rational behavior" - really easy to write that sort of crap in to a linear model and hide it in the aggregation).
But two things mitigate that in this case. First, in general, humans are just unspeakably awful, unbelievably atrocious, almost impossible to overestimate how bad they are at explaining their actions in rational terms. This is where you get the cynical, possibly to the point of misanthropic, view of the world that everyone is just some sort of raving psychopathic lunatic with as much connection to the real world as a bad surrealist movie. The problem is, that model fails to describe reality; if people were actually that disconnected, society would not be as bad as it is now, it would be worse, to the point of non-existent. There would be no technological society as we know it. So people are not as irrational as all that. One of the major keys to understanding human behavior, and I am 100% serious and using no sarcasm here, is to assign a very low priority to people's stated motivations or reasons. (Simultaneously, there is little to no evolutionary need for people's beliefs about their own reasons to be accurate and we've actively evolved deception, including deception about our stated motivations and goals. So we end up successfully deceiving everyone, even ourselves.)
And second, specific in this case, I think humans have a significant activation energy to do things like "move", for both rational and perhaps irrational reasons, so you see a lot of stickiness as to where people live as a result. I think people also have a hard time visualizing what it would be really like to pick up and completely change their life, which makes it difficult for the "change everything" option to be analyzed properly. Our rationality is much more optimized for the second, minute, day, week, and maybe month time scale. When it comes to making decisions that will be bad for us today but probably be very good for us in five years, we're not terribly well equipped for that. And one can observe that with a simple time-value-of-money calculation there is even a sense in which that is not entirely irrational. You can get stuck in traps where the locally rational thing to do at any given time leads to a long-term suboptimal outcome, and even from a purely rational point of view it can be difficult to escape such traps.
: https://smile.amazon.com/Debunking-Economics-Revised-Expande... - if you want this in intimate mathematical detail this is a good one
smile.amazon.com itself is a general charity program, but you have to set it up for your account. I have a plugin that autoredirects me there, since Amazon won't do it for you. I'm going to leave that; probably some people don't know about it.