Scott Sumner  

Predicting bubbles

The Role of Unions... Against Winning...

MaynardGKeynes recently left this comment:

Simple fact is that Shiller correctly predicted 3 of the last 3 bubbles (2000, housing, 2007).
That's a widely held view, but is it correct? This is from Eugene Fama's Nobel Prize lecture (in the AER):
On the website for his book Irrational Exuberance, Shiller says that at a December 3, 1996, lunch, he warned Fed Chairman Allan Greenspan that the level of stock prices was irrationally high. Greenspan's famous "Irrational Exuberance" speech followed two days later. How good was Shiller's forecast? On December 3, 1996, the CRSP index of US stock market wealth stood at 1518. It more than doubled to 3191 on September 1, 2000, and then fell. This is the basis for the inference that the original bubble prediction was correct. At its low on March 11, 2003, however, the index, at 1739, was about 15% above 1518, its value on the initial "bubble" forecast date. These index numbers include reinvested dividends, which seem relevant for investor evaluations of "bubble" forecasts. If one ignores dividends and focuses on prices alone, the CRSP price index on March 11, 2003, was also above its December 3, 1996, value (648 versus 618). In short, there is not much evidence that prices were irrationally high at the time of the 1996 forecast, unless they have been irrationally high ever since.

The second "success" story is the forecast in the mid-2000s that real estate prices were irrationally high. Many academics and practitioners made the same forecast, but an easy one to date is Case and Shiller (2003), which was probably written in late 2002 were early 2003. To give their prediction a good shot, I choose July 2003 is the date of the first forecast of a real estate "bubble." The S&P/Case Shiller 20-City Home Price Index is 142.99 in July 2003, its peak is 206.52 in July 2006, and its subsequent low is 134.07 in March 2012. Thus, the price decline from what I take to be the first forecast date is only 6.7 percent. The value to homeowners from housing services during those nine years from July 2003 to March 2012 surely exceeds 6.7% of July 2003 home values. Moreover, on the last sample date, October 2013, the real estate index, at 165.91, is 16% above its value on the initial "bubble" forecast date. Again there is not much evidence that prices were irrationally high at the time of the initial forecast.

I see this sort of thing all the time. In January 1987 John Kenneth Galbraith claimed stock prices were a bubble. They then proceeded to rise by nearly 1000 points and then fall by nearly 1000 points, all within 1987. After stocks fell sharply, some people assumed Galbraith had predicted the crash. Back in 2003, The Economist magazine predicted that housing prices would fall in a number of specified markets. After the housing bubble burst, they ran an advertisement bragging about their prescience, even though their specific predictions were almost all incorrect. Housing prices rose strongly after 2003 in most markets, and remained well above 2003 levels even after declining.

If Shiller had lived in the UK, Canada, Australia or New Zealand, his housing price forecast would not have been just a little bit wrong, but rather wildly off base. He's lucky that he lives in a market that recently had a once in a century housing price drop, and yet as Fama shows his prediction was still somewhat off target. And of course his stock market model has done very poorly since 2010, when his model suggested the S&P500 was 20% overvalued. At the time it was at 1070!

We all make either implicit or explicit forecasts about the markets. If we later notice market movements that seem to align with our initial forecasts we tend the pat ourselves on the back and assume the forecasts were correct. This is just one of many cognitive biases that we human beings are prone to. My suggestion is to pay no attention to bubble forecasts. They are useless. Indeed the entire bubble concept is useless.

Comments and Sharing

COMMENTS (19 to date)
Robert writes:

Well said.

Another nail in the bubble coffin is to recognize that what appear to be price bubbles will always show up over time, albeit infrequently, even when prices move randomly. Prices should sometimes move up/down sharply in the face of positive/negative surprises - these are just the tails of the return distribution. Once in a while there is certain to be a series of ups followed by a down or three and *poof* you have a bubble. Flip coins long enough and you get a series of heads followed by a series of tails. It is just a question of when. And it does not mean anything.

Chris Hallquist writes:

"Pay no attention to bubble forecasts" seems reasonable, as plausibly bubbles are really, really hard to see in advance. If most people could see them in advance, they wouldn't happen. But how can the entire bubble concept be useless? That sounds like a denial that investors are affected by herd dynamics, when it seems pretty well-documented that they are. The examples you describe in this post seem like pretty clear examples of bubbles—just ones whose supposed predictors don't really deserve credit for predicting.

vikingvista writes:

"Indeed the entire bubble concept is useless."

Does this mean bubbles don't exist, that the term is not well defined, or just that there is nothing that can be done about them?

Does it mean the phrase "market correction" is a misnomer, or that the phrase only refers to appropriate responses to relatively recent events? Can there be a systematic bias in present relative prices that signals greater future profitability in some sector than can possibly be realized given actual conditions?

It is reasonable to believe that a simultaneous one-sided miscalculation involving billions of trades would be expected to happen randomly within anyone's (including Methuselah's) lifetime?

mickey writes:

I don't see why the decline must go to some specified past low. Decisions are reached, actions made, time passes, things change - you aren't comparing the same thing.

Lorenzo from Oz writes:

vikingvista: If turning points cannot be predicted, how can a "bubble" description be anything other than a claim that a given set of asset prices are unstable? But if turning points could be reliably predicted, how can you have a "bubble" (since no one would buy the asset at a price near the turning point)? So "bubbles" can only exist if the key fact about them cannot be known, except after the fact. So how is that a useful concept?

Even on the matter of asset price instability, if turning points cannot be reliably predicted--so we cannot reliably say whether assets prices will plateau, rise or fall in a given time frame--what are we asserting, in any useful sense? That, at some unspecified time in the future, asset prices may move in a different direction?

Michael Byrnes writes:

I think you nailed it, Lorenzo. All assets are a bubble, always!

Joe Teicher writes:

Bubble is a horribly overused concept but I'm not sure that it is completely useless. Do you think that "short squeeze" is a useless concept? You shouldn't because they do happen sometimes. And in those cases the value of the asset being squeezed can diverge dramatically from any reasonable notion of fundamental value. A short squeeze is essentially a mini bubble. If there can be identifiable mini bubbles, why can't there be identifiable larger bubbles?

Peter writes:

I think the bubble concept is useful, in the sense that it describes a situation where people are basing an inordinate proportion of their pricing expectations on a widespread belief that prices will not fall. As soon as they fall then the belief is shown to be a fiction and the price backing from that belief evaporates.

That's more likely in a market with a large number of new or inexperienced investors. Like housing in the 2000's. Or high-tech stocks in the late 90's. Anytime you hear a novice talking about how investing in a class of assets is a "sure thing" then there's a decent chance of a bubble forming.

One more thing about housing prices. The cost of carrying a mortgage on a home purchase has dropped tremendously since 2008. If you looked at what people are willing to actually pay to own a house, factoring in a 3.5% mortgage instead of a 7% mortgage, you'd see a 32% drop in the cost of paying off a mortgage over 30 years. Some of those savings might end up in the buyers pocket but my bet is a lot of it went to allowing sellers to avoid much larger drops in housing prices.

vikingvista writes:

“So "bubbles" can only exist if the key fact about them cannot be known, except after the fact. So how is that a useful concept?”

Even if there is nothing that anyone could ever do about a bubble, and even if bubble identities are nebulous both during and after the fact (which they obviously are), that doesn’t mean they don’t exist. And if they exist, they should be of interest to theoreticians even if they are of no practical use to businessmen or interventionists. The physicist’s dictum “Shut up and measure” leaves nothing for economics.

Further, if the turning point cannot be predicted, that does not mean that there is nothing that anyone can do about it--i.e. it is *not* necessarily useless to interventionists or businessmen. There is always enough confounding fog in empirical macroeconomics that certainty in inference is difficult or impossible to come by. That’s why economic theory is particularly important. And if there is a sound theory, e.g., that a particular type of intervention is likely to cause a discrepancy between price signals and future profitability, then such a theory would inform an interventionist to avoid such an intervention. One might expect then that avoiding such an intervention would *tend* to reduce the significance of such turning points over time, even if it is impossible to identify with reasonable certainty the cause of any particular turning point. And the businessman, as incapable as anyone else of identifying a bubble, could use such a theory to incorporate greater uncertainty in his predictions.

“Even on the matter of asset price instability, if turning points cannot be reliably predicted--so we cannot reliably say whether assets prices will plateau, rise or fall in a given time frame--what are we asserting, in any useful sense?”

Predicting any particular macroeconomic value may be a fool’s errand. But presumably within the immense fog of observational uncertainty there are at least some causal mechanisms taking place. Perhaps we could say with greater certainty than any attainable macro empirical inference, that certain actions can be expected to cause certain results.

Of course, I’m alluding in particular to Austrian business cycle theory, where central bank price manipulations in the market for loanable funds is said to alter signals in such a way as to bias widespread belief in future profitability in one direction, without altering the underlying conditions that would make such increased profitability true. In this theory, such interventions are expected to cause bubbles where absent such an intervention a turning point would be nonexistent or much less dramatic.

Scott Sumner writes:

Chris and Viking vista, How do you think bubble theories are useful? Just saying something looks like a bubble is not a use, it's a description.

Joe, How is your theory of short squeezes useful?

Peter, You say it's useful, but then don't provide an example of a use.

Vikingvista, That which has no practical implications has no theoretical implications.

Greg G writes:

The fact that bubble theories are often wrong and often misused does not mean they cannot be useful and cannot be improved. We get a lot of wrong weather forecasts but that doesn't mean there is no point in predicting the weather.

I would argue that Minsky offers a useful way of thinking about bubbles although it's been a long time since I read him and I don't remember if he has a stance on the word "bubbles."

If we think about bubbles as a sharp increase in riskier methods of finance for a certain asset class that could have some useful theoretical and practical implications.

Robert Simmons writes:

If you had asked Shiller at the time to predict the change in US housing prices from July 2003 to March 2012, how much lower than -7% do you think he would have been? Maybe he would have said -20%, maybe, at the worst. My quick math is that under normal conditions housing should rise ~50% over a 9-year stretch. I'm counting the stock market against him, but housing for him.

vikingvista writes:

Prof. Sumner,

"That which has no practical implications has no theoretical implications."

What of the the army of string theorists and the university physics departments that seek them out?

Also, since you think there are no practical implications to bubbles, does that mean you think Austrian business cycle theory is wrong? Does it generally mean that no economic theory can be more certain than a single (as in N=1) macroeconomic observation?

Joe Teicher writes:


As a market maker and arbitrageur, understanding short squeezes allows me to demand extra edge to trade products where they are more likely. Stocks that have small floats or that are hard to borrow have extra risk. Front month commodity futures have extra risk. In both cases I can be forced out of a position when the market is far from fair value, so I need to take that into account when deciding whether those products are worth the risk or if I need to adjust my strategy to trade those products. Squeezes are like earthquakes. Even though I can't predict them I want to know where they are likely to hit and make sure that my buildings there are reinforced against them.

Peter writes:


"You say it's useful, but then don't provide an example of a use."

I thought I did, but I'll rephrase.

If you hear novice investors talking about how a class of assets is a "sure thing" that can't go down, and the pricing history of that class shows enough variability that a moderate drop in prices can be expected in the next year or two then you have a situation where that moderate drop will turn into a more significant drop as the inflated "never drop" pricing support abruptly dissipates. In effect, information changes that would have caused modest drops can now be expected to cause more significant drops.

Does this prove bubbles exist? No. We don't have a formula where you punch in the market change and get an "expected drop in prices" to compare against. It's a judgement call.

Could you rephrase this to say there are times when markets are significantly more sensitive to downward pricing signals? Sure. But bubbles are a subset of that definition, with the additional requirement that the market be near a peak after a sustained run-up (and therefore perhaps easier to identify and anticipate). The goal is to try to find some predictive signals for that increased sensitivity, ahead of time.

For me, the bubble idea is about exactly that. About finding signals that can be identified that have some predictive value on one particular type of future pricing event. Maybe they can't be found. Maybe they're not predictive enough to be useful. I think they're still worth looking for.

I think people who are very experienced investors know this. But if they sit out the market while novices are jumping in, and this risk is developing, then they would miss out on the run up to the peak as the bubble builds. So they do the best they can. Nobody wants to be the first one to leave the party, or the last.

Another interesting clue is to identify what would allow a novice to believe things will never drop. They have to be lured into believing it. The market they are looking at has to experience an enticing mix of significant and/or sustained growth. It needs to seem like the "old rules" for pricing behaviour have been permanently changed, whether that's the technical and business revolution of the internet, or the changes to home loan availability leading to easy money. It starts with a lie, or at least an exaggeration. But when the market seems to support it, and people are getting rich following the "new logic" then more and more people jump in. It's a market. A lie isn't a lie if everyone believes it. Until people stop believing.

Tin Man writes:

Keynes warned us that "the market can remain irrational longer than you can remain solvent."

Jeff writes:

Bubbles have a very specific meaning in financial economics. I address the issues raised in this post here:

Mike writes:

I'm not an economist, and certainly not an expert on any of this, but it seems to me that the S&P 500 price history provides plenty of backing for Shiller's opinion in 1996. The S&P 500 was at about the same level at the end of 2008 as it had been at the end of 1996. 12 years with no increase. To say that stocks were overpriced in late 1996 was not to say that their prices couldn't go higher, but rather that chances were not good for long-term appreciation. It's only in the last couple of years that there seems to have finally been some real appreciation, but are prices going to stay up this time or go back down again? Once again stock prices are high, but the idea is, with P/E levels as high as they are currently, odds are not good for appreciation over the next 10 years.

Bob Murphy writes:

Scott, if you're still reading these comments, you may want to use my response to you/Fama as target practice.

Comments for this entry have been closed
Return to top