Scott Sumner writes,

if the term “bubble” is to mean anything useful, it must contain an implied prediction of the future course of asset prices.

I read this as saying that a bubble is when someone can correctly predict that an asset price will fall. I think this is not a useful definition, for a number of reasons.

Instead, I would borrow from Robert Shiller the notion that a bubble is characterized by unreasonably high expectations for continued appreciation. When the housing market was bouyant, Shiller undertook surveys of people and found that their expectations for house price appreciation were 10 percent per year. One can raise questions about the survey methods, but the intent is to capture what to me is the essence of a bubble: prices rise because people expect them to rise.

In my high school econ class, I use an equation for determining the profitable of buying an asset:

profitability = rental rate plus appreciation minus interest cost

Whenever the expected appreciation exceeds the interest cost, the asset appears profitable to buy as long as its rental rate is not negative. If mortgage rates are less than 10 percent and people expect 10 percent house price appreciation, they are going to want to buy houses.

In general, I would say that when people expect an asset price to rise for a long period of time at a rate that is much higher than the long-term nominal interest rate, we are going to see a bubble in that asset. There will be strong demand for that asset until expectations change.

This issue of expectations is an interesting one right now when one considers gold and long-term bonds. As I have noted before, people buying gold probably have much higher expectations for inflation than people buying long-term nominal bonds. It would be interesting to see if surveys bear this out. If the discrepancy exists, it also would be interesting to try to figure out a way to arbitrage against it.