[maxlybbert]

If cities have become more dangerous in recent years, people will notice and change their behavior. It doesn’t matter that today’s cities haven’t matched their historic peak, only that they’re worse than they were ten years ago.

[ghaff]

Furthermore, a non-trivial number of people are just looking for reasons to move out of cities right now. Consider that many cities that we consider "elite" were losing population well into the 1990s. There's nothing written in stone about college-educated young people continuing to migrate to cities (which is pretty much the demographic that has moved the numbers) as they have over the past 20 years or so.

[dragonwriter]

> If cities have become more dangerous in recent years, people will notice and change their behavior.

Dubious.

Certainly, behavior is driven by perception of danger, but whether perception of danger generally tracks with actual danger or not is…less certain.

I know for child abduction and child assaults by those outside of the family that has historically not been the case; and perceived danger has increased as media focus increased despite decreasing actual danger, for several decades in a row.

I don’t see why similar (or opposite, if media focussed shifted away while actual danger increased) trends would occur with other forms of danger.

[maxlybbert]

I stated that if cities become more dangerous, people will change their behavior. A case of people changing their behavior even though cities aren’t more dangerous isn’t really a counterexample. A valid counterexample would be a case of cities becoming more dangerous and people not changing their behavior.

[dragonwriter]

> A case of people changing their behavior even though cities aren’t more dangerous isn’t really a counterexample

It wasn't presented as a counterexample, it was presented as grounds for doubting the mechanism for the effect you describe.

Increased danger -> increased perceived danger -> behavior which attempts to mitigate danger is a nice theory, but it relies on deltas in perceived danger corresponding to deltas in actual danger.

If deltas in perceived danger that manifestly do drive behavior are driven by processes that are independent of actual danger, your argument no longer makes sense.