This is a key insight.

Most "rational actor" theories of human behavior actually only work in the large (where the average can dominate outlier behavior) and in systems where rational action is a positive feedback loop ("a fool and his money are soon parted").

If those assumptions break down (especially the second, i.e. if foolish use of money results in more money accruing, not less), what we perceive as rational behavior should not be expected.