At this point my indebtedness to GMU’s economics department is only slightly below the fast rising national debt. You have probably seen some very kind comments from Tyler and Bryan, and there have been some behind the scenes favors as well. I have a persistent feeling of guilt that I am not able to reciprocate. Thus I was horrified to recently discover that I had not even answered a question posed by Bryan a few months ago. He had asked whether investor panic might have been an independent shock hitting the markets last October. I responded, but never saw his follow-up:
In the comments, Scott graciously replied:
“As long as you define “panic” as “correctly ascertaining that the monetary authority was about to embark on a dramatically lower NGDP growth trajectory that would plunge the world into depression” then I am completely with you.”
I’m afraid I’ve got more in mind. Why can’t we think of the public’s mood as an independent – and highly volatile – causal variable?
I didn’t expect a panic in October, 2008. But when it happened, it sure didn’t seem to be due to news about nominal GDP. It seemed to be due to news about the public’s mood. A week before, the natives were calm. Then they suddenly got amazingly restless.
As long as this restlessness is unpredictable, it’s perfectly consistent with EMH. So this account seems to meet Scott’s objections to distant “root cause” theories. And it seems to fit our experience of actually living through those days, doesn’t it? So what’s wrong with blaming an exogenous panic attack?
I’d like to attack this question from several angles. I should first mention that for someone who has been linked to twice by Mankiw on EMH issues, I am actually pretty uniformed on the subject. I never studied it in school, but find it an interesting topic to think about. So I don’t know whether Bryan’s assumption about the EMH is right, indeed I think his hypothetical might violate the EMH. But that’s really a side issue, because it’s a hypothesis worth considering either way.
My first observation is that it is very hard to know what the market is thinking about, as the market is much smarter than any of us, and gets “news” long before we do. You can be sure that by the time we read that the Baltic shipping rates have plunged 90% (which I recall was in the papers last fall) the markets already knew about it. One of my visions of the stock market is Lucas’ archipelago model—people get local information before it is in the official statistics. Sort of like how the brain supposedly knows something before it knows that it knows it.
After a stock crash you will sometimes get news stories suggesting that information was creeping into the market before it was in the official statistics. And I am not just talking about “inside information” (although I have seen such cases), but also highly dispersed information. After the stock market crash of October 1929 there were suddenly a lot of very, very negative stories on various types of production. Causation could not have run solely from the crash to output, as the numbers referred to production declines that were already underway at the time. It seems undeniable to me that current and prospective output declines played a major role in the October crashes of 1929, 1937 and 2008. By “prospective” I mean information about the flow of new orders, etc. I would like to emphasize that I don’t think this fully explains any of these three crashes, but it was a factor.
A second factor is policy credibility. It may be coincidence, but right around the time of the three crashes there was a major loss of trust in the government. In 1929 and 1937, investors lost confidence in the President, and in early 2008 there was a loss of confidence in Paulson and Bernanke’s ability to manage the situation. (John Taylor makes a similar but slightly different argument.) Recall that Bear Stearns did not cause major damage to the markets in early 2008.
Also recall that although the market fell about 5% on the Lehman news, it then leveled off a bit for several weeks. The Fed statement on September 16th indicated that the threats of inflation and recession were roughly balanced. So things still looked “salvageable.”
By early October I think the markets came to the conclusion that the world economy was falling fast, and that the Fed would be unwilling or unable (take your pick) to prevent a sharp downshift in NGDP growth. The exact same thing occurred in October 1929 and October 1937. I can’t prove this, but based on a close study of those earlier crashes I am pretty sure that investor forecasts of NGDP going several years forward fell very sharply around the time of the crashes, which actually stretched from September to November in each case.
Of course it is very possible that the reason NGDP expectations fell sharply is because the stock market crashed, but I don’t like that sort of reasoning. Let’s take the worst case from my pro-EMH perspective, and assume the 1987 crash was completely irrational. Recall that the 1987 crash (comparable in size to the other three) did not cause even a tiny, tiny blip in GDP growth. Instead the economy stayed very strong for the next 32 months. Why? You can’t just pick an arbitrary reason; after all there are several hundred million consumers in America. Either their decisions are affected by stock prices, or they aren’t. If they are, there should have been at least some impact in 1987.
My theory is that stocks don’t have much independent effect on AD. When expectations for NGDP growth do not fall sharply, then we don’t see a recession after a stock crash. We often see stocks fall before a recession because the market often sees the recession looming ahead. (BTW, I know the recession had officially started by October 2008, I mean the severe intensification of the recession.) So I see the causation running from expectations of future recession to current stock price collapse.
Stocks are on a hair-trigger alert for trouble. You don’t want to be the one holding the bag when other investors have already seen trouble ahead. So it stands to reason that just as nervous police officers will occasional shoot an innocent person, the stock market will occasionally be spooked by an event that never actually occurs. So events like 1987 will occur on occasion.
Even so, I would be the first to admit that the crash of 1987 is hard to reconcile with the EMH. I believe part of the 1929 crash is explicable from actual events, although the most intense part may have involved a bit of panic. In contrast, I think the 1937 and 2008 October crashes are some of the most “rational” that I have ever seen. The market correctly saw that the world economy was falling off the cliff, and that the Fed had misdiagnosed the problem. If the Fed had cut rates on October 6th from 2.0% to 0.5%, investors would have been electrified. Instead they made one of the most contractionary decisions in Fed history.
My colleagues can tell you that at this time I was wandering the halls of the economics department muttering what the %#&@ is Bernanke doing? Why has the Fed suddenly stop targeting the forecast? The only reason you guys are reading these words is because early October 2008 turned me into a monetary crank. I don’t see it as an uneventful period at all. There were lots of bearish stories about the economy, and I am sure that people out in the “real world” saw lots of depressing indicators at the local level. Put it all together; the local news, the knowledge investors have that markets aggregate information, the bearish news out of Washington and the picture snaps into shape. And they were right. The early October crash took the S&P down to 900. That’s exactly where it is today. There was no “overreaction” as you’d expect if there had been panic. Panic doesn’t last forever, if there was panic a recovery would be expected by rational investors. There is no evidence a recovery was expected. The problem is very real (albeit caused by a nominal shock.)
I didn’t predict the stock and commodity crashes, but as they developed and the Fed didn’t respond it all made sense to me.
What about Bryan’s theory of panic attacks? I don’t like this theory for several reasons:
1. The law of large numbers. People don’t independently all become nervous at the same time, for no reason.
2. I presume Bryan had some contagion effect in mind. But what causes the contagion? Why does it occur in some stock declines and not others? There is a lack of serial correlation in stock prices. When prices have fallen for three straight days, there is still a roughly 50-50 chance they will rise on the 4th. So if the panic was internally generated, if the beginning of the crash created panic, which led to a bigger crash, then why wouldn’t stocks that had fallen three straight days be likely to fall on the 4th day?
3. If the panic was triggered by an external news event, why wouldn’t the crash have occurred on the day after some big event like the Lehman failure. Instead it occurred in early October when there was relatively little financial news. The only news I recall during that time period were persistent reports of very bearish forecasts about real growth and steeply falling commodity prices. Put the two together and you have very bearish forecasts about NGDP growth. The other news was a growing sense that Paulson and Bernanke were in over their heads, but that fits in with my “monetary policy failure” argument.
Having said all this, I should emphasize that I try not to be an ideologue. I don’t see any reason to have an emotional attachment to theories. I have found the EMH incredibly useful. But I don’t think it tells us much about 1987, and I think it is also plausible that there are some irrational elements in other unusual market events such as 1929 and 2000. But I also believe that many economists underestimate how quickly even highly rational markets could move on dispersed information. Again, the real economy saw a severe downturn appear suddenly in late 1929, late 1937, and late 2008. Would you expect markets to react slowly to the realization of how quickly things were getting worse? Of course not. So if the theory is very useful, and the theory fits the facts of 2008 reasonably well, why look elsewhere for theories that have never shown any utility at all, i.e. anti-EMH theories?