Rational manias

by John Q on July 19, 2004

There’s a cottage industry within economics involving the production of historical arguments giving rational[1] explanations of seemingly irrational historical episodes, of which the most famous is probably the Dutch tulip boom/mania. This Slate article refers to the most recent example, a complex argument regarding changes in contract rules which seems plausible, but directly contradicts other explanations I’ve seen.

Once opened, questions like this are rarely closed. Still, articles of this kind seem a lot less interesting in 2004 than they did in, say, 1994. In 1994, the efficient markets hypothesis (the belief that asset markets invariably produce the best possible estimate of asset value based on all available information) was an open question, and the standard account of the Dutch tulip mania was evidence against it. In 2004, the falsity of the efficient markets hypothesis is clear to anyone open to being convinced by empirical evidence.

We have seen billion-dollar valuations placed on companies that proposed to home-deliver dogfood at prices lower than those charged in discount stores. We’ve seen unimportant subdivisions of profitable companies valued at more than the companies themselves. We’ve seen a dozen different companies simultaneously priced at levels that made sense only if they were each going to monopolise the industry in which they were competing. And don’t even get me started on the US dollar bubble (now burst) or the bond bubble (still inflating).

In summary, is contradicted by our own recent experience far more thoroughly than by anything that might or might not have happened in Amsterdam in the 1630s. Everybody who cared to look at the numbers coming out of markets in the late 1990s knew they were crazy, but it didn’t matter. Those who bet on an early return to sanity (George Soros for example) lost their money. The only sensible course was to withdraw to the sidelines and wait the madness out.

It’s true that dramatic episodes like the dotcom mania don’t happen all the time. But even one such episode, occurring in a well-developed and sophisticated financial market like that of the US in the late 1990s is sufficient to undermine the assumption that asset markets ever yield the best possible estimate of asset values, except by chance.

fn1. That is, explanations consistent with individual rationality as defined by economists



James Surowiecki 07.19.04 at 3:06 pm

John, I don’t see how your last sentence follows from your evidence. It’s certainly true that the EMT is not right. But if you had a system that offered the best possible forecast 90% of the time, would the fact that it was wrong 10% of the time — even if when it was wrong it was really wrong — mean that its success the other 90% of the time was due to chance?

Take horse-racing. We know that the odds on horses predict almost perfectly how likely it is that a horse will win (over, say, 100 races subjective probabilities are essentially identical to objective probabilities). In the Belmont Stakes, it seems pretty clear that the market overvalued Smarty Jones and undervalued Birdstone. (I don’t mean just that Birdstone won, but that the odds on these horses did not reflect their objective chances of winning.) Should we really conclude from this that the accuracy of odds is due to chance?

There’s no doubt that markets suffer from enormous periods of irrationality. But it seems possible that, although we don’t have the analytical tools to do this yet, these periods are objectively distinct from the way the market works most of the time, and that while they shed light on the potential perils of markets, they don’t invalidate the use of markets to set asset prices. The fact that someone occasionally gets sick is not proof that they’re not healthy.

I’m not sure, in any case, what follows from your conclusion. Every day, the US stock market values more than 7000 stocks, which entails, roughly speaking, predicting what the future will be like for those 7000 companies (and their competitors, etc.) for the next 15-20 years, depending on valuation. That’s a near-impossible task. The difficulty of the task isn’t going to change, but what’s a better way of doing it?


dsquared 07.19.04 at 3:13 pm

In general, (and while once more procrastinating my review of “The Wisdom of Crowds”), I would plead with both John and James to join my crusade to break with normal statistical usage and not to use “best estimate” as a synonym for “most accurate estimate”. If you are speculating with borrowed money, for example, it is much worse to overestimate the value of an investment than to underestimate. In general, I suspect that most real-economy forecasting problems face non-symmetrical and often quite complicated loss functions.


Jonathan Goldberg 07.19.04 at 3:22 pm

I assume that “This Slate Article” was meant to include a link. Would you please give it?



q 07.19.04 at 3:22 pm

In 1994 many of the _limitations_ of the efficient markets hypothesis were evident by reading Keynes’s General Theory. Of course, not so many people actually READ the general theory, just a lot of people spout off about it.


James Surowiecki 07.19.04 at 3:53 pm

You could argue — and actually many people did argue — that the crash of 1987 was an even more definitive refutation of the EMT.

Daniel, judging from your post my best course of action would probably be to advise you to keep procrastinating on that review. After having written 320 pages using “best estimate” in the sense you don’t like, I can’t quite see my way to changing to your usage. But I do think there’s an interesting question implicit in your theory about loss functions: are markets (as opposed to other decision-making systems) more likely to overestimate the value of an investment than to underestimate them? In this regard, I do think the dearth of shortselling in the US stock market has seriously bad effects.


Bill Tozier 07.19.04 at 4:24 pm

A follow-up on dsquared’s comment:

Having (a) built functional automated portfolio management systems for equities traders, and (b) later discussed those systems with academic economists, I’m always struck by one thing: none of the economists even know what “maximum drawdown” is, or why the traders care about it. Even some financial engineers of my acquaintance have said understood the concept, but regard it as a quirky sort of thing far too intractable to be interesting.

Yet the traders all insist on it. Indeed, the traders insist on multi-objective decision-making tools. Which — as all of us will agree — ultimately demand subjective judgments.

Whereas all my economist friends prefer to apply a handy linear weighting to those nasty multiple objectives and collapse them back down into something nice and tractable….

It’s an interesting dichotomy, and not unrelated I suspect. But then, I’m just a by-stander I suppose. :)


Jason 07.19.04 at 4:57 pm

Without knowing much about tulips in general, or even the formal statement of EMT (what does it mean to be the “best” estimate when we are talking about utility functions of individual estimates of discounted dividends), it seems to me your evidence doesn’t contradict the thing you want it to.

The strange valuations you describe could all be reasonable in particular contexts. During the dotcom bubble, there was the possibility of huge efficiency gains, and it was impossible to tell just how much various new markets would be worth. Sure, in 20/20 hindsight the valuations seem high, but if you’d had 20/20 foresight you’d now be a very rich man. If George Soros was so sure the bubble would burst, but wasn’t sure when (suppose he knew to within +/-2 years). He could have formed a portfolio making considerable amounts of money by buying forward puts and calls. Investment banks have whole departments devoted to allowing these sorts of bets.

I’d say better evidence would be along the lines of psychological/experimental evidence that people are “sheep” in that they will buy into pyramid schemes and that they will buy into fads rather than actually looking at what their returns are. I am sure this evidence exists, and so your conclusion justified, but I just don’t think your “obvious” tag really is.

d2, I don’t see why borrowed money is needed in your claim. Are you saying it is morally wrong to not be risk averse with someone elses money? Even if we took this as a given, I don’t see why “best” is worse than “most accurate.” Best depends on context, if your goal is to maximise a risk averse utility, then the best amount to invest would be to shift towards less risky outcomes. If your goal is to provide the best risk neutral estimate, then you won’t.


Matt McGrattan 07.19.04 at 5:50 pm

“During the dotcom bubble, there was the possibility of huge efficiency gains…”

I see this sort of thing said a lot but has there ever been any empirical justification for it?

Having worked in companies both prior to- and post-adoption of a range of various IT technologies including widespread adoption of PCs, local area networking, dial-up Internet, wide area networking, high-speed Internet access, and so on: I’ve never really seen any spectacular efficiency gains in any working environment in which I’ve worked. Small incremental gains and sometimes radical changes in very narrow areas of working practice, yes, but widespread gains, no.

I used to be an IT consultant and was also involved with support – including a period as head of tech support – for a couple of different ISPs both in the late 80s and mid to late 90s so I’m not totally ignorant here. I do remember, from the inside, the companies I worked for charging unsustainable loss-making prices because everyone else in the market was doing the same and no-one wanted to be the first to flinch and everyone assumed they’d be the last one left standing and reap the rewards of the ‘efficiency gains’.

No snark intended. It’s a serious question — is there any solid evidence of even the remotest likelihood of the kind of efficiency and productivity gains that would justify the hyperbolic claims being made during the dotcom boom?


Jim Harrison 07.19.04 at 6:23 pm

About IT efficiency gains: the assumption seems to be that internet-related technologies produced the bulk of productivity gains. In my corner of the economy, though, it wasn’t the internet but older computer technology that drastically increased productivity. The outfits I work for finally figured out how to integrate homely applications like word processing and electronic databases into their workflow. The benefits promised in the 80s materialized in the 90s. For example, the turn-around time for the production of research reports shrunk drastically in that period, but most of the speedup resulted from organizational changes that used technology that had been around for quite a while. The effortless transmission of text materials over the internet certainly helped, but it wasn’t the main thing–technically with it people were already getting most of the benefits of that back in the late 80s through the use of email and modems.

I think we generally underestimate how long it takes for business to figure out how to use technology efficiently. If I’m right, much of the benefit from the internet remains to be realized. One of the recurring problems of evaluating technology is to guess which lump in the snake corresponds to which pig.


Matt Weiner 07.19.04 at 8:44 pm

Having no expertise whatsoever, I’ll try to be the link fairy:
This looks to be the relevant Slate article.
According to this, “Drawdown” represents the maximum percentage loss for a given period. This represents the greatest percentage difference between index highs and lows for a selected investment period.
I might guess why equities traders care about it, but if Bill T or someone else wants to explain that would surely be more accurate.


John Quiggin 07.19.04 at 9:32 pm

To everyone, link is there now. Sorry about that.

d2: I agree with your definition of “best”, in principle at least. It’s the standard Bayesian definition and the correct one to use in reading my post. Of course we don’t generally know which loss function to use – minimum-variance is a neutral choice.

James: I agree that you can’t systematically otupick the stock market most of the time. But the strong EMH hypothesis is important in lots of contexts, such as arguments about unrestricted international movements of capital, use of markets rather than regulation in electricity and other infrastructure industries, privatisation and so on. It’s mostly implicit, but if you work through the argument you’ll find that it makes a big difference.


still working it out 07.19.04 at 9:41 pm

If you’re going to talk about EMT then you should really mention George Soros’ principle of Reflexivity (which he explained, rather poorly, in Financial Alchemy, I think. Can’t remember the title exactly) as it provides a pretty good reason why EMT is bunk from someone who can claim to understand financial markets better than most.

People keep seeming to assume is that if EMT fails, it is because humans are irrational and emotional creatures. This is not true. As George Soros explains, markets are inherently unstable, regardless how rational the market participants are.

EMT assumes that the market has taken into account all information and processed it into a price. One of the fundamental problem’s with EMT is that one of the pieces of information that is used to calculate an asset’s price, is the previous price of the asset. If that price has been rising, then that will add to the value of the asset. Vice versa if it has been falling. This creates a self re-inforcing feedback loop that is not the result of irrational behaviour. If the price has been strongly trending, then it is entirely rational, whether you are a highly emotional manic-depressive or an ice cold supercomputer to assume that it will continue to trend for one simple reason. You know the market is smarter than you. Once you accept that a price is trending then the logical thing to do is buy or sell the asset based on that fact. Something any successful trader understands very well. Psycholoy probably plays a role, but trending markets are easily explained, even when it is assumed that all participants are completely rational as long as price history is part of the information used to determine the current price, which in practice it alway is.


James Surowiecki 07.19.04 at 9:42 pm

John, I agree that the use of strong EMH — often, as you said, implicitly — as an argument-stopper when it comes to the kinds of problems you mention is both common and dangerous. One place, at least in the US, where this is very much the case is corporate law, with pernicious effects.

I think many of these questions — privatization, international flows of capital, etc. — are still open ones, and that market solutions may in some of these cases be the best ones possible (which is not to say they’re optimal). But the superiority of those solutions needs to be demonstrated empirically, rather than simply making recourse to a rational-expectations-based idealization of market efficiency.


John Quiggin 07.19.04 at 10:11 pm

James, I agree with you that these are empirical questions. Any model that shows either that the market uniformly outperforms the state or vice versa is inconsistent with our experience.

On the 1987 crash, I didn’t regard it as definitive for a couple of reasons. First, although the change in market valuation was sudden, it wasn’t huge compared to changes that often occur over a year or two. So if you accept that these changes are consistent with EMH, a sudden adjustment is not much of a problem.

Second, even if the crash was a violation of EMH it could be (and was) explained by fixable institutional problems, such as rules about program trading and so on. No such explanation is possible for the dotcom boom.


James Surowiecki 07.19.04 at 11:15 pm

That’s an interesting argument about the price change in the 1987 crash. I’ll have to think more about it.

It’s true that the dot-com boom tout court can’t be explained by institutional problems, but I do think that if shortselling Internet stocks had been easier (or, in some cases, even possible), and if Net-stock floats had been larger, the pricing of Internet stocks would have been significantly better. Even so, you would undoubtedly have been left with a significant deviation from intrinsic value, so that I’d say the structural problems magnified, but didn’t cause, the market’s collective irrationality in 1999 and early 2000.


dsquared 07.19.04 at 11:24 pm

James: but the “dot com” boom wasn’t only in dot com stocks. It was perfectly possible to short America Online, the telco sector and large-cap software companies.


dsquared 07.19.04 at 11:41 pm

I’d also note that regulatory explanations for the dot com boom have to cope with the fact that it also occurred in the UK, France, Germany and Italy under very much more lax regimes.


James Surowiecki 07.20.04 at 12:31 am

Daniel, absolutely right — I was going to add a caveat saying that shortselling limitations can’t explain why the price of, say, Cisco got completely out of whack. Again, as you know, I believe that the market was deeply irrational in the late 1990s.


Jason 07.20.04 at 12:48 am

James, again I think it’s easy in hindsight to say we all should have shorted Cisco, but there were plenty of people who thought Cisco was going to own all the machinery that ran the internet, and they’d be able to price as much as they like. They bid the price up, and at the time Cisco was growing like crazy. I know plenty of people who bought them, and they weren’t all stupid (not me, thankfully, although looking at their share price now, if you’d bought in 99, you’d still be even now, which isn’t really that bad).


LeftCoast 07.20.04 at 8:27 am

The best argument I know of against the EMH applying to the stock exchange is closed-end mutual funds. Many of them have traded for years at either a premium or discount to the value of their underlying securities.


Shai 07.20.04 at 8:54 am

jason writes:

“Sure, in 20/20 hindsight the valuations seem high, but if you’d had 20/20 foresight you’d now be a very rich man.”

um, no. the tech bubble resembles powerball if it were a pyramid scheme. the information was out there that valuations were crazy.

and i’m not just talking about the tail end of the bubble when everyone I talked to knew a 4 billion Akamai IPO was ridiculous, and the economics of VA linux was a sham (if you don’t believe me, go look at old threads on slashdot).

sure, we do remember the fed, aka alan greenspan testifying to congress that IT productivity suspended the laws of economics, but there was also widespread talk about hundreds of ipo’s that were nothing more than some idiotic idea with e- tagged onto it.

Here’s an example from pets.com, from a time period placing it inside the bubble:

David Schehr, an analyst at the Gartner Group, based in Raleigh Durham, N.C., says he has some problems with the whole concept.

“It’s possible there can be a price advantage, and it might be attractive if they build a good consumer experience online,” he says. “But shipping 50 pounds of dog food over UPS may become problematic over time. And if Fido runs out of food, you’re not going to be wanting to wait for it.”

For those reasons and others, Scherhr says he thinks the online pet-supplies market will eventually be characterized by vicious price wars that over time will leave most of the competitors broke and howling at the moon.

(it really wasn’t economic at all, they were just eating the shipping with hitherto unlimited cashflow, but still)

Let me use myself as an example. In high school economics class my final presentation compared IPO’s and market valuations of several companies against future earnings potential derived from basic economic analysis of markets we learned earlier in the course (minimally mathematical super simple textbook). my conclusion was that this was possibly a sign of a bubble. (we had been discussing the recent asian bubble the previous week). this information didn’t come from nowhere. there was a lot of talk about ridiculous valuations in the media.

but the point i wanted to get to is this: this information did nothing to prevent me from investing in qwest (internet backbone) or @home (cable internet, now bankrupt). I did happen to make money, but that’s luck: I was itching to spend money, thinking I could invest more and get rich later. the general idea was that everyone was irrational, so you may as well take advantage of them. but that’s obviously a self defeating idea. as james says here ( http://www.randomhouse.com/features/wisdomofcrowds/Q&A.html ):

“… instead of worrying about how much a company is really worth, investors start worrying about how much other people will think the company is worth.”


Jason 07.20.04 at 5:50 pm

Shai, as always the best comeback is to ask if you made your millions in predicting the crash?

Yes, I had some VC friends at the time who thought things were massively overpriced (compared to historically), but they still had to invest at these prices. They did fine by managing their portfolio in a way that was bubble-burst resistant, while still managing to grab a lot of upside. However, even they still weren’t sure where all the tech was going to lead us, and weren’t so confident in their prediction that they bet the fund on the phenomena being a bubble.

I didn’t lose a dime in the crash (nor make much either), so I’m not trying to justify my own stupidity here. I’m just saying that, even though there were pyramid scheme buyers like you, there were also plenty of smart investors who were thinking of the internet as a landgrab, and that it was important to grab as much territory as possible, and then figure out the money later.

For a few companies this has worked. Look at Yahoo! or Google. In a different environment, neither of these companies would have been able to exist.

If you really were so sure it was a bubble and going to burst, you *really* should have bet on it. You’d be a very rich person now. Of course, if you’d managed to get enough people along with you, there wouldn’t have been a bubble in the first place.


Nelson 07.21.04 at 3:56 pm

Keynes worried about the implications for our society when “the conventional valuation of stocks is established (by) the mass psychology of a large number of ignorant individuals.” The result, he suggested would lead to violent changes in prices, a trend that would be intensified as even expert professionals-who, one might have suppossed, would offset these vagaries by their perspective and judgement-follow the mass psychology, and try to foresee changes in the public valuation. As a result, he described the stock market as “a battle of wits to anticipate the basis of conventional values a few months hence rather than the prospective yield of an investment over a long term of years.”

Comments on this entry are closed.