Fortunately, experimental economists have looked into the question, reproducing bubbles in the laboratory and playing with the factors that drive bubbles.
Both Vernon Smith and Charlie Plott have contributed to this research.
In this 2003 paper [pdf], Porter and Smith review experiments that look at bubbles and conclude that:
bubbles seem to be due to uncertainty about the behavior of others, not to uncertainty about dividends... futures markets help to dampen (but do not eliminate) bubbles... limit price change rules make bubbles worse.In this 2002 paper, Plott et al. [pdf] find... "that errors in decision-making [not speculation] are a primary source of bubbles."
In sum, then, bubbles form when someone guesses high (makes a mistake) and someone follows. They form for assets (not consumption goods), because asset values depend on people's opinions on future value. Note that short sellers can dampen bubbles, but only if they turn the market before they are buried by a wall of "dumb" money.
Bottom Line: We cannot ban bubbles in the same way as we cannot ban celebrity gossip. All we can do is protect ourselves (if you don't get in, you don't have to worry about getting out) and protect others (money managers should be fired -- or worse -- for trying to time bubbles with other people's money).
* (via BB), John Mauldin writes a scathing critique of EMH.