James Hamilton’s got my attention on housing bubbles. One of his many interesting observations is that population growth may explain a lot of the variation in real estate appreciation:

It’s noteworthy that over the last 5 years, the three states with the highest population growth rates as reported by the Census Bureau– Nevada, Arizona, and Florida– have also been among the locations that saw the biggest increase in home prices. Forces such as these, rather than a random distribution of irrational exuberance, seem a more natural explanation for why some communities got bubbled and others didn’t.

Getting data on 2000-5 turned out to be a bit of a chore, but data for 1990-2000 was easy to come by. Does Hamilton’s story work for the 90’s?

The answer is: Marginally. The simple correlation between a state’s population growth and its real estate price growth is .29, and a 1 percentage-point increase in population growth is associated with .57 percentage-point increase in real estate prices.

That’s just a bivariate regression, of course. To put matters in perspective, if you “race” population growth against variables like Kerry’s 2004 vote share or the level of (not change in) per capita income, population growth matters quite a bit less.

Admittedly, this is a lazy test of Hamilton’s claim, but at least his story does not leap out of data. I remain sympathetic to his skepticism, but I’m not convinced he’s right.

Since I bought a Northern Virginia house in 2000, though, I sure hope he is.

But that gets me thinking. What would I actually do if I knew for sure that my house was going to plummet in value one year from today? My ideal solution would be to sell my house to someone, rent it from them for a year, then buy it back. But that would be very hard to arrange. In practice, I’d have to sell, rent whatever’s available for a year, then use my nest egg to buy a comparable (or better home), pocketing the difference.

There’s a lot of transactions costs built in there. For starters, there’s moving costs x2, plus all the pain and suffering of changing my address and phone number twice, plus the loss of sentimental value in my current house. And don’t forget opportunity costs – I’d say that 100 hours each for me and my wife is a conservative estimate.

So how much money would I have to net to make me sell, rent, and buy again after the crash? Frankly, it would take $200,000 just to pique my interest. And that’s with certainty. Maybe I’m unusually averse to moving, but I can easily see people with kids in school being even less mobile than I am.

This hardly proves there’s a bubble, but it does make me doubt that speculation would stop a bubble before it got started.