Scott Sumner  

Why bubbles are so hard to spot

Ask What Changed... Markets Without Limits ...

There's been a lot of discussion about the recent Chinese stock market crash. Most observers view the events through the lens of the "bubble" model of financial markets. This view claims that there are sharp price run-ups caused by irrational exuberance, and that rational observers can spot when the prices are out of line with fundamentals. In contrast, I believe it is very difficult to spot market irrationality at the time, and sometimes even in retrospect.

In some ways the recent Chinese boom and bust looks a bit like the 1987 stock market in America, when prices soared over the first 9 months, and then fell very sharply. At the time almost everyone thought it was a stock bubble, and they seemed pretty confident that the "wrong" price was at the peak, not the subsequent trough.

You would think that with the benefit of hindsight we'd learn who was right. We'd learn which of the two prices were clearly out of line. But a comment left by Brian Donohue a couple months back shows that even today it's really hard to figure out who was right:

If you bought the S&P 500 at the peak (10/5/87) you've earned a 9.3% CAGR over the past 28 years.

If you bought at the subsequent trough (12/4/87) you've earned a 10.8% CAGR.

Note that the S&P500 fell 31.75% over that period. Both of those subsequent returns look fairly reasonable to me. If you forced me to guess, I'd say the 9.3% return seems a bit more consistent with market efficiency (recall that inflation and nominal interest rates have fallen since 1987). And if I'm right this would imply that prices were more "rational" at the peak of the "bubble." But either way, I think any fair observer would admit that even today it's really hard to know what stock market valuation (S&P500) was appropriate in 1987---328.08 or 223.92.

Now imagine you were thinking about where Chinese stocks would be 28 years from today. Also recall that rates on alternative investments are now much lower than in 1987. Does anyone seriously believe it's possible today to know which recent Chinese market valuation will be viewed as being more rational in 28 years, 5000 from a few weeks ago, or roughly 3000 today?

So I don't find the "bubble" hypothesis to be useful. But just to be fair and balanced, I do believe that 1987 provides a very powerful argument against the EMH, indeed one of the most powerful arguments that I have ever seen. Not because 1987 was a bubble (it probably wasn't), but rather because the more than 20% stock crash on October 19, 1987 was not accompanied by any new information that could justify such a sharp re-evaluation of equity values in 24 hours.

To summarize, there are pieces of evidence that seem inconsistent with the EMH. But the specific bubble argument is much weaker than most people assume.

PS. And don't forget that while seeing which price was clearly wrong in retrospect is a necessary condition for the bubble hypothesis, it is not sufficient. For instance, in retrospect NASDAQ was overpriced in 2000, but perhaps that reflects new information about growth in IT.

PPS. The closing price on October 19, 1987, was 224.84, so virtually all of the crash occurred in a shorter period than Brian considered.

PPPS. And notice that there was no recession in America after the 1987 crash. Will China have a recession? Maybe, maybe not.

Comments and Sharing

COMMENTS (8 to date)
Daniel Kuehn writes:

Would Friedman's plucking model help to arbitrate this? It's typically invoked to argue that business cycle peaks are right and troughs are the misalignment. I don't follow finance but would that test be appropriate here.

maynardGkeynes writes:

I think you may be focused on slightly the wrong issue, i.e., valuations, instead of excessive price movements. Bob Shiller showed pretty convincingly some years ago that whatever the range of "rational" valuations in equity markets might have been or should have been at any given time in the past, the actual price movements (volatility sort of, but on a larger scale) were significantly in excess of what one would expect in a rational market. So, even if valuations were in some sense "correct" before the 1987 crash in retrospect, the problem for the EMT is that actual price movement on that day probably were far in excess what they should have been. I think what most people call a "bubble" is more a manifestation of that, rather than "wrong" valuations per se.

Bob Murphy writes:

Scott, I'm not sure how to reconcile your views that the S&P500 1987 peak was not obviously overpriced, but the Nasdaq in 2000 was. Isn't a lot of this driven by your choice of time frame?

In other words, why are we picking 28 years in the future, to be our reference point? That's just because the 1987 crash occurred 28 years ago.

To see what I mean, take the Nasdaq and assume 4% average growth from today through 2100. If I did the math right, then that means from the trough in 2002 the Nasdaq would have earned 4.9% CAGR through 2100, whereas from the 2000 peak it would have earned 3.3%.

Is there some reason to prefer one number over the other?

More generally, yes, the farther into the future we go, the smaller the gap in growth rates before and after any given crash in the past. I'm not sure how we are supposed to use this.

Jose Romeu Robazzi writes:

Prof. Sumner

Leverage played a role in both events. In 1987, a lot of people mention portfolio insurance as one of the "causes" of the crash. Market makers who had sold stock market index options to investors rushed on the selling in order to balance their own position, who apparently were leveraged.

Right now, it looks like there was a lot of leverage on the chinese system, a lot of people buying stocks on margin...

Scott Sumner writes:

Daniel, I'm not certain about that. Although I see some similarities between market cycles and business cycles, there are also some really important differences. For instance, being able to predict recessions doesn't violate the EMH in the way that reliable predictions of market crashes would.

BTW, I don't buy the plucking model 100%, but I do think it might be roughly correct, picking up important asymmetries in the business cycle.

Maynard, If you reread my post I did look at both, and agree there is more evidence for the excess volatility claim. However I was focusing mostly on the bubble hypothesis.

There are also important weaknesses in Shiller's model, such as his apparent inability to use it effectively for market timing purposes.

Bob, Yes I suppose time frame makes a difference. But for bubble claims to hold up wouldn't they need to apply over any time frame? Thus even today gold at $850 an ounce in 1980 looks way overpriced, and ditto for tulips at the peak of that 17th century boom. My point is that if later data shows the price was reasonable (as at the peak in 1987) that's a sufficient condition for rejecting the bubble claim (or at least for establishing that it's completely unproven.) But it's not a necessary condition, as if the NASDAQ suddenly soars to 10,000 (which I don't expect), then the 5000 peak in the year 2000 might begin to seem plausible.

Jose, I've heard many explanations for the October 19, 1987 crash, and none are remotely plausible. For instance, they'd predict multiple such events, but it only occurred once in history. Explanations of unique events are almost always very weak, very ad hoc.

Glen W Smith writes:

When I was a kid we had a cat. She'd sit in the back yard while the squirrel just kept inching closer and closer until one was caught and killed by that cat.

Mike Rulle writes:

It is virtually impossible, even in retrospect, to guess which price is more rational, so I agree with your main point.

As an aside, you give the impression there is not much difference between 9.3 and 10.8% over 28 years. Reader Bob Murphy implies the same thing. The future value difference over 28 years between 10.8 and 9.3 is 51%. I find compound interest differentials fascinating. Reader Murphy's example results in a 5 fold difference in future value.

So, in future value terms the differences are enormous. I wonder if this helps us determine in retrospect, which price in fact was more rational? Don't know. To me the range of plausible rational prices is huge at any one point in time. Hence, I am an EMH guy.

Jose Romeu Robazzi writes:

@ Prof. Sumner
I believe portfolio insurance changed a lot after the crash, and besides, leverage manifest itself in may ways, in 2008, the investment banks were leveraged, that is why they went down.

Comments for this entry have been closed
Return to top