Wouldn't the perception track the statistics in the scenario you've lined out? More people being stabbed, for example, results in more friends and friends of friends being unwilling to take a walk in that area.
I mean, you can have an outlier, a very popular person who triggers more friends/friends of friends, but on average, it seems like perception wouldn't be the worst way to measure this under the scenario you've laid out.
But, like, wouldn't a neighborhood with a lot of muggings be more likely to have more people doing that?
> So for instance when a group of well off people move into a depressed neighborhood, there might be the same number of muggings, but 10 times as much talk about it because these people go to the police and write to the paper about it.
That seems like a valid explanation for a single neighborhood that just started gentrifying, but seems unlikely to be common enough to disrupt the correlation that I'm talking about globally.
> The decline goes beyond homicides: Violent crime overall ticked down in 2022 across the country, the FBI numbers showed, returning the U.S. pretty much to the level of 2019, before the COVID-19-era increase.
It looks like the original survey was asking about whether there is anywhere near the respondent's home where they'd be afraid walk at night. That sounds to me like it's very much a question about perception rather than statistics. Of course, people rarely base that kind of belief on hard evidence: a single incident that happened to a friend of a friend might be enough to make you say "I'm never walking in that neighborhood at night".