Tuesday, June 1, 2010

Are bubbles observable?

The study of bubbles has come into fashion lately, and of course identifying them is high on the priority list. The thing is that they are very difficult to find, by definition. A bubble appears when the value of an asset deviates significantly from its fundamentals. But the latter are not directly observable, as they are mostly formulated in terms of expectations, and most of the time they are actually measured using the asset price. You see how this is circular and makes the exercise impossible.

Friedrich Geiecke and Mark Trede claim to have solved this. The use the prices of futures to determine dividend expectations, specifically future contracts on the Dow Jones Euro Stoxx 50 DVP Index, which is a traded index that measures all cumulated dividends paid on stock listed in the Dow Jones Euro Stoxx 50 Price Index. From this, Geiecke and Trede back out the fundamentals, then compare them to the last Price Index and find indeed that there are rational bubbles.

This is an interesting approach. I am worried though that the market for those futures may not be that liquid so as to tease out market fundamentals. We will see with time whether this approach holds water

4 comments:

AlabamaFor You said...

Very interesting paper from a very unlikely place. How did you find this paper, and others on the blog?

rosserjb@jmu.edu said...

The paper is interesting, but has some limits. A big one, subtly admitted in it, is an assumption of Gaussianity of the underlying distribution. If the underlying distribution is skewed, then one cannot infer anything from means of futures markets. This has been known for a long time in forex futures markets and goes under the name of the "peso problem," first identified by Ken Rogoff in his Ph.D. thesis in 1979 at MIT, although Krugman has claimed that this was common knowledge among the grad students in the cafeteria at that time. The forward market for the Mexican peso regularly underforecasts the actual future spot price of the peso because of the asymmetric downward tail on the long-run distribution.

BTW, a better example where the misspecified fundamentals problem is overcome is closed end funds. There the net asset value is the fundamental, correcting for some minor matters such as tax effects and management fees, assuming the underlying assets are liquid (which is not the case for country funds for nations that do not allow foreigners to trade in their stock markets). Because of the caveats, one should expect closed-end funds to run at small discounts, and that is what one generally observes.

So, when one sees a premium appear on a closed-end fund, one that rises sharply and then collapses suddenly, one can be pretty sure one has really and truly observed a speculative bubble. Examples include closed-end funds in the US in the late 1920s and closed-end country funds in 1989-90. Both of these ran up to premia of about 100% before spectacularly crashing. There is an established lit on this, and size of discounts or premia on these funds are a commonly used measure for "sentiment" in financial markets.

Cyril Borusyak said...

The paper is interesting indeed but results don't seem robust - with respect to discount rates, for example, and other details of the calculation technique. Increase r slighly or let them change over time, and you'll get the opposite results easily, I guess.
I hope though that longer data series will help to overcome these problems.

Btw, the dividend futures they analyze seems to be useful not only in determining bubbles. I expect to find it in various papers soon.

Stephen Williamson said...

Keep up the good work. Glad to know I'm not alone.

Steve Williamson