“The thing I have noticed is when the anecdotes and the data disagree, the anecdotes are usually right.” -Jeff Bezos

Being data driven is not necessarily a bad thing, but the design of the study, the kinds of data collected, and all the other factors that go into how the data is measured need to be considered too.

This article about there being no “Californian Exodus” just doesn’t seem true, much in the same way I discount all those reports that say something along the line of “No, crime is actually down in the Bay Area”. Despite people living there saying it’s gotten way worse. The answer probably lies in changing what gets counted as “a crime”, coupled with police not responding to things they used to report, and citizens feeling like it is futile to call the police. But the data would paint the picture that crime is objectively down. But when you apply some other analysis you can see the conclusion is flawed.

Many people moved their nexus to their parents house while physically going elsewhere with the uncertainty. USPS data has no way of accounting for that if the physical human being did not tell USPS to forward mail to where they physically were.

Anecdotes are a signal to say “look over there and see where the hypothesis goes”, if people havent done that and simply invalidate anecdotes because meme, then the answer is not known

And for many questions the data doesn’t exist, just because someone elevated USPS change of address filings as a signal doesn’t mean it is the strongest