I hadn't thought of it like this before, but I think he's arguing that there is a network effect. High density is more convenient, which induces demand.
I realize he is arguing for that but I think its pretty silly to have the chart there showing the correlation if so. He does address the point I made in the article but I think he dismisses it too fast without much evidence.
I suppose to expound on my point more - the graph is by people not housing units. Presumably if a cheap city, like Detroit, had a bunch of unused units the graph would not credit it for the built-in density because if its not being used.
It is harder for us tech workers to fathom, as we work for the global market directly or indirectly, the but a lot of people working in a city are directly serving other people in that city. A person coming in might create more than a jobs worth of “CityGDP” by their simple economic activity.