> I find this article to be embarrassingly bad. Just find metro areas and compare neighborhoods and observe density levels, and see the effect on price.

This sketch of an analysis is significantly worse than what's in the article. You can't estimate effects by correlating levels. You need to correlate deltas, at minimum. You also need a reason to think those deltas are causally linked, in that you're not suffering from omitted variable bias, which is why you need a natural experiment. The article follows this correct methodology by discussing the Austin natural experiment.