Self-esteem tracks the "belonginess", that is, social inclusion. It correlates with how valuable we are to other people in general, how strong our bonds with others are, and the level of support we can count on from them. Observations that support this theory: #1, #5, #6, #14, #18.
Self-esteem is a device that tracks our relative status within a group. It adapts to whatever type of group (friends, family, coworkers, sports buddies, strangers on the street) we find ourselves in at a particular moment, as well as to what subset of the group we are interacting with. Observations that support this theory: #2, #5, #6, #7, #11, #17, #18.
Feeling shame is a mechanism - a precommitment of sorts - for creating self-punishment. As not all social transgressions are punished (hard to detect, not socially acceptable to punish), propensity towards shame means others can trust us more, since they expect any transgressions to be self-punished. Self-esteem is then a measure of how strongly we are predisposed towards feeling shame (globally), or in general, how shameful we feel on aggregate. Observations that support this theory: #4, #5.
Doing things in public is an implicit leveraged bid for status. Doing things privately is an implicit call option on status. Self-esteem in a public setup translates into higher propensity towards more risky - leveraged - behavior, i.e., higher social risk tolerance. There is some optimal amount of risk tolerance. EV raises with risk, but so does the probability of ruin.. Observations that support this theory: #1, #9, #15, #17.
We have an internal image of what kind of person we should be. We also track what kind of person we currently are. The self-esteem is then the difference between the first and the second image. The hedonic treadmill heuristic suggests feeling good and bad should mostly depend on the gradient, not on the absolute level. An appropriate modification of this theory is that we increase self-esteem if we move towards (or over) the ideal self, and lose self-esteem if we move away from it (under). Observations that support this theory: #3, #4, #5, #9, #10, #15.
The level of self-esteem primarily regulates the investment-consumption tradeoff in all areas of life. This can also be rephrased as moving the slider between exploration and exploitation, between giving and taking, learning and teaching, low and high time preference, et cetera. Economics and control theory give simple models of optimal strategies in exploration-exploitation games, which could be useful to derive quantitative answers to some of the questions posed here. Observations that support this theory: #3, #9, #10, #15, #16.
None of the theories above seem to be related to or match observations #8, #12 (gender divide) and #7 (high-status advice), which is something I leave to think about later.
Different theories imply different trade-offs of developing high self-esteem.
Sociometer and hierometer theories suggest that artificially increasing one's self-esteem results in an inaccurate reading of a useful statistic. "Fake it 'til you make it" suggests a somewhat self-fulfilling prophecy, but on the other hand, Hanson's self-deception framework implies this being difficult in most situations. If the level of self-esteem is a -meter, hacking it means losing hygiene in personal epistemics.
Propensity towards shame theory makes low self-esteem an adaptive trait in situations or groups that lack effective policing methods, such as extended groups of colleagues, and maladaptive in those that do have them, such as close-knit communities or large-scale formal settings governed by established law.
Social risk theory makes hacking self-esteem equivalent to miscalibrating one's behaviour versus risk tolerance, which is clearly suboptimal. Alternatively, it might mean changing the risk tolerance itself. Since it is usually taken as a fundamental preference, I don't really understand what this would entail in detail.
Ideal-self theory says that low self-esteem might be beneficial as a motivating tool, while at the same time common-sense observations suggest that it still has to be kept above a certain critical threshold not to spiral into helplessness. Developing higher self-esteem is dangerous, since it might involve lowering the standards instead of rising up to the challenge.
Overall, while giving partial credit to each theory. I lean towards having high self-esteem being a bad trade-off, at least in my current situation. This is mostly for reasons of low time-preference, optimising for learning over teaching, being afraid of lowering my standards, and being in a novel and risky social environment with high penalties for slippage. Even if I wanted to increase my self-esteem, I would lean on doing it gradually, as the hedonic treadmill seems to kick in fairly easily, while loss aversion creates a ratchet-like dynamic which necessitates caution.