October 28, 2003

Vernon Smith on Vulnerability to Bubbles

Tyler Cowen writes about Vernon Smith's echoing of Charles Kindleberger's rule-of-thumb that you can't get a bubble within a decade of the past bubble:

Marginal Revolution: Once burned, twice shy?: Our colleague Vernon Smith argues that new traders are most susceptible to asset market bubbles, largely because of their inexperience. If the market rises again after a bubble bursts, we should take the run up in prices seriously:

Smith points out that market double dips don't occur in rapid sequence. Indeed, since 1926 the space between down years for the broad market has always been at least two years, and usually much longer, according to Ibbotson Associates data. The closest sequence in the recent past was the two positive years between the 1973-74 bear market and the 7% downturn in 1977.

In other words, Smith argues that the second round of high prices is usually for real. Experience has beaten the traders down into a state of fearfulness, so presumably there is good grounds for their optimism.

I would like to see the more systematic time series evidence, in the meantime here is the article from Forbes.

Here is my favorite part from the article:

To Professor Smith, bubbles perform a great service for capitalism. "Every bubble is driven by great innovations, and they all leave behind a lot of long-term value," he says.

Posted by DeLong at October 28, 2003 09:37 AM | TrackBack

Comments

Carolta Perez in her recent seminal work, Financial Capital and Technological Revolutions, details this phenomena in wonderful detail. Highly recommended.

http://www.corante.com/brainwaves/archives/000429.html

Posted by: Zack Lynch on October 28, 2003 10:00 AM

____

I really do not understand the notion that if something didn't happen in US markets from 1926 to today it can never happen.

This isn't nearly enough data to get that level of comfort. Conditions today are different than in the past. Among other changes, traders today are aware of past data and take it into account.

Posted by: richard on October 28, 2003 10:18 AM

____

Maybe it's the reporter, but I don't even understand what point V. Smith is supposed to be making. Since 1926 to 2002 there were altogether 23 negative years for the S&P. Twelve of them--i.e. more than half--occurred in clusters of consecutive down years: 29-32 (4 years), 39-41 (3 years), 73-74 (2 years), 00-02 (3 years). Is the point that you have to wait long after each such cluster for another down year? That not too convincing either: 34 was a down year after 29-32, 46 was down after 39-41--admittedly a while, and 77 after 73-74; we still don't know what's going to happen after 2000-2002. Anybody care to enlighten me?

Posted by: maciej on October 28, 2003 10:55 AM

____

And the more people that believe the market doesn't have a large drop soon after one has occurred, the faster and sharper the rebounds will be and the more likely it will be that a second drop occurs.

Isn't that the problem with any technical analysis? The observation gets fed back into the market and eventually defeats itself. Anyway, since there is nobody here to tell me what the fair value of the market is as was done in his experiments, I think I'd rather just figure out for myself the fair value rather than trying to guess what everyone else thinks the fair value is.

Posted by: snsterling on October 28, 2003 10:56 AM

____

Maybe it's the reporter, but I don't even understand what point V. Smith is supposed to be making. Since 1926 to 2002 there were altogether 23 negative years for the S&P. Twelve of them--i.e. more than half--occurred in clusters of consecutive down years: 29-32 (4 years), 39-41 (3 years), 73-74 (2 years), 00-02 (3 years). Is the point that you have to wait long after each such cluster for another down year? That not too convincing either: 34 was a down year after 29-32, 46 was down after 39-41--admittedly a while, and 77 after 73-74; we still don't know what's going to happen after 2000-2002. Anybody care to enlighten me?

Posted by: maciej on October 28, 2003 10:57 AM

____

Markets can do whatever they like.

The question related to bubbles is always one of valuation. If the deflation of a bubble restores the market to fair value or below, the market will eventually start to rise. If popping the bubble only partially relieves the pressure, then the tendency will be for a second event to trigger further bubble deflation.

None of this is particularly deep. It's a restatement of the concept of equilibrium and the restorative forces that exist when it is perturbed.

How is the present market valuation? About right, perhaps 10-20% overvalued, in my estimate. 10 years ago, the Dow closed around 3700. At 7% growth, it would be at 7200. That's within the vagaries of value estimation and represents a conservative valuation benchmark (as I remember, forward P/Es were about 10-12 when it last was in that territory).

On the other hand, bursting the bubble has never driven the value below an equilibrium price. Therefore, restorative forces have to favor a further fall in the market. For the Dow to fall to 5000 is not impossible.

Posted by: Charles on October 28, 2003 11:52 AM

____

Much of the justification for market capitalism is that it is efficient in allocating capital (honest people tend to say "the most efficient mechanism").

It seems to me that bubbles give the lie to this statement. How much value was lost in the tele-dot.com buble. Lest you say that it left a lot of real capacity at cheap prices, well, first the price was cheap at the bankrupcy sale, but not to the original buyers and second a lot of that investment never got transformed into anything beyond vapor.

Posted by: Joshua Halpern on October 28, 2003 12:09 PM

____

Remember, the changes in dividend and capital gains taxes has made stock holding for the wealthiest investors considerably more valuable.

The market may be able to sustain a higher p/e ratio than history would suggest.

Posted by: anne on October 28, 2003 12:41 PM

____

maciej writes:
>
> Maybe it's the reporter, but I don't even understand what
> point V. Smith is supposed to be making. Since 1926 to
> 2002 there were altogether 23 negative years for the S&P.
> Twelve of them--i.e. more than half--occurred in
> clusters of consecutive down years: 29-32 (4 years), 39
> -41 (3 years), 73-74 (2 years), 00-02 (3 years). Is the
> point that you have to wait long after each such cluster
> for another down year?

I took that at his point, but it is spectacularly unconvincing. Thanks to your data, I do some back-of-the-envelope stuff to show just how unconvincing this is.

Here is a depiction of the losing streaks involved:

1 @ uddddu
2 @ udddu
1 @ uddu
11 @ udu

You tell me there are a total of 23 down years, and 54 up years. P(up) = .701 and P(down) = .299, so I'll round to 0l7 and 0.3. Now you can compute the empirical counts of each of the 4 possible 2-year sequences:

dd = 8 # compute dd and du from above streaks
du = 15 # same as above
ud = 15 # there are 15 losing streaks total
uu = 39 # 54 up years - 15 ud years

Now we can compute the expected number given p(up) = .7, and compare it to the observed values:

# Observed Expected
dd 8 6.93
du 15 16.17
ud 15 16.17
uu 39 37.73

So in other words, there is about *zero* support for the idea that losing years are especially spread out from the two-year data. (Mendel couldn't have done much better for a fit and he was cheating. :-)) But maybe there's some weird thing happening with 3-year sequences. Fine; I can get at that from your data, too, since I guess I know that "dud" does not occur:

# Obs Expected
ddd 4 2.025
ddu 4 4.725
dud 0 4.725
duu 15 11.025
udd 4 4.725
udu 11 11.025
uud 15 11.025
uuu 22 25.725

So this time if I were to do a chi-square test on the data (although of course I really shouldn't for obvious reasons) I actually get a number close to what you should expect from chance. Yeah, that empty cell for "dud" does stick out, but I would not concoct an "invulnerability to bubbles" theory based on that, which is what Vernon Smith is at least quoted as doing.

So I remain baffled. A strong random-walk person would have placed a bet on the market going up in 2003, since that's what it does 70% of the time. But not because down periods need some "breathing space" between them.

Posted by: Jonathan King on October 28, 2003 12:44 PM

____

>>Remember, the changes in dividend and capital gains taxes has made stock holding for the wealthiest investors considerably more valuable.>>

This makes sense, but does not seem supported by the data. A while ago I read a study which attempted to find correlations between tax rates and stock metrics, such as p/e ratios. The study was not able to find any relation.

One reason might be that a whole lot of stock is held in IRA's, pension plans and other non-taxpaying entities.

Posted by: richard on October 28, 2003 12:55 PM

____

"Smith repeated his experiments numerous times, sometimes using graduate students and sometimes using finance professionals. The financial types, he noted, made bigger bubbles."

Strange, aren't the finacial people supposed to be immune? The should have lived trough a bubble? Somethings wrong or more complicated...

Posted by: GB on October 28, 2003 03:19 PM

____

If Kindenberger's Law is an empirical one about whether people learn the first time around or not, it might not be true. Maybe people are getting dumber. OR -- maybe they believe in Kindenberger's law, and thus produce the second bubble.

Posted by: Zizka on October 28, 2003 03:26 PM

____

Secondary bubbles aren't that uncommon. But hey, whatever turns his crank. Go back and look at the early to mid seventies.

Posted by: Ian Welsh on October 28, 2003 09:25 PM

____

Ian made my point for me. We had a classic speculative blowout in 1966-68 -- the Go Go Years -- then a rather nasty crash and bear market in 1969-70.

But that led to the great Nixon-Burns reflation (closing the gold window, slapping on wage and price controls and cranking up the printing press) which, among other things, produced the "Nifty 50" bubble of 1971-72.

I think the moral of the story is that it is possible -- with heroic effort -- to reflate burst bubbles, if the initial bursting doesn't entirely destroy investor confidence in New Era valuation theories. As Anne demonstrates above, this hasn't happened -- yet.

Of course, Nixon's economic lunacy eventually caught up with him (or rather with Jerry Ford), just as Bush's economic lunacy no doubt will catch up with either him or whoever has the misfortunate to follow him into the Oval Office.

Admittedly, the data is limited. The only other example of a secondary bubble that comes to mind is the 1993-95 rally in the Japanese market, which was supported by every reflationary trick the Ministry of Finance could throw at it.

But, based on the outcomes (the 1973-74 near financial meltdown in the US, and the post-1995 near implosion of the Japanese banking system) it seems like reflating old bubbles is at best a temporary remedy, and may result in a crash that's even worse than the original one.

Posted by: Billmon on October 29, 2003 08:00 PM

____

Post a comment
















__