Wednesday, 8 July 2009

Bubble-detection technology

Pointing out a speech by William C. Dudley, president of the New York Fed, Simon Johnson says:
Dudley says that the Fed can pop or prevent asset bubbles from developing. This would represent a major change in the nature of American (and G7) central banking. It’s a huge statement - throwing the Greenspan years out of the door, without ceremony.

It’s also an attractive idea. But how will the Fed actually implement? Senior Fed officials in 2007 and 2008 were quite clear that there is no technology that would allow them to "sniff" bubbles accurately - and this was in the face of a housing bubble that, in retrospect, Dudley says was obvious.
But is that true?

If we define a bubble as "overvaluation of assets relative to their future returns" then to spot one, we would need to compare asset prices with future returns. But although asset prices are measurable, future returns are not - and this is why people generally think that bubbles are unspottable. We wouldn't want to put the brakes on, say, a fast run-up in the shares of biotech companies just because we don't know what their profits are going to be in ten years.

But there's another word in the definition, apart from "assets" and "future returns": and that word is "overvaluation". How about if we could directly measure overvaluation? If we could determine directly whether investors are putting an irrationally high value on the things they buy?

As it happens, we can.

Experiments have been designed to measure investors' attitudes to asset valuation - for example lots of Vernon Smith's work, including Caginalp, Porter and Smith (1998) (details of other experiments also available here). This experiment was particularly interesting in understanding recent events, as it showed that the valuations of identical assets with identical returns were strongly correlated to the amount of cash available to investors.

That may be exactly what took place in the 2000s, as a loose monetary policy found no outlet in consumer prices but instead flowed into asset valuations (Scott Sumner has argued that money did become tighter later in the decade, more so than signified by interest rate movements; if true this could be one reason that asset prices are not recovering despite very low interest rates).

Outside of the laboratory, precise knowledge of the returns of some assets does become available at times, and it would be possible to measure investors' behaviour with regard to those assets. If investors, in aggregate, become overconfident about returns it will be possible to spot this from certain types of price change.

If we can't find suitably measurable instruments in the market, it may still be possible to set up the appropriate conditions in a laboratory; but that would be less reliable than using market price signals. So a preferable next step is to determine which kind of assets can be tested and under what circumstances, to reveal the risk attitudes of different classes of investor. These measurements would then be combined with data on the cash balances of those investor classes to detect bubbles in their inflation rather than collapse phase.

What regulators can do about these bubbles is the next question; consumer-level financial regulation is one area to consider, but there are other tools too. More on this later.

(see also my VoxEU papers [1], [2] on this subject)

1 comment:

OSR said...

If we can't find suitably measurable instruments in the market, it may still be possible to set up the appropriate conditions in a laboratory; but that would be less reliable than using market price signals.

And if all that fails, there's always common sense.