A lot of people are scared of the markets these days.  Most people seem scared because they don’t know what the markets are going to do next.  But the thing that scares me the most is the fact that anyone would think they should be able to figure out what the markets will do next.  The simple fact of the matter is that markets are always wildly unpredictable.

The Efficient Market Hypothesis

Most financial markets follow the model of the stock market. The selling of stock was initially devised as a method for firms to raise money.  They sell stock certificates, which represent a small percentage of ownership in the company, in exchange for returning some of the firm’s profits to the shareholders in the future.  The transaction is structured almost exactly like a bank loan, with the shareholders playing the role of the bank.  The profits are distributed to the shareholders in the form of annual dividends, which are paid on a per share basis.

This structure means that every stock has a fundamental value, which is based on the current value of the expected dividend payments, with a bonus paid to shareholders willing to give money to firms where the potential return is riskier.  Firms usually don’t sell stocks until they are reasonably well-established, as it is nearly impossible to place a value on their stock until people have a good idea of what their prospects are like.

Most analysis of the stock market then is based on the idea that every stock being sold has an underlying fundamental value.  The value of the stock will go up if the firm does things that suggest that profits will increase in the future – for example, when a pharmaceutical company announces the discovery of a promising new drug.  The value goes down when information arises that causes concern about the future prospects of the firm.  It has been well documented that stock prices usually reflect the value of new pertinent information within seconds of the discovery of the info.

An important implication of this view is that stock prices are pure information, which is referred to as the Efficient Market Hypothesis. When we view the markets from this perspective, bubbles and crashes are pretty hard to explain.  Bubbles and crashes require some kind of systematic mispricing of stocks.  The orthodox view of economics (and consequently, of stock pricing) is that this is impossible.  However, if the price of a stock precisely reflects the current value of future dividend payments, then why would anyone ever buy or sell shares?  There are at least three important reasons that stock prices may well deviate from their fundamental value.

Problem:  Errors in Risk Estimation

The first is that we consistently underestimate risk.  The best explanation of this comes from Nassim Nicholas Taleb – his two books, Fooled by Randomness and The Black Swan are both outstanding, and this talk summarises his views quite nicely.  His basic premise is that nearly every model of stock valuation consistently underestimates risk.  Which in turn means that nearly every single stock price is wrong.

Problem: Who holds the risk?

The second issue is that even if we do somehow correctly estimate the risk involved in a particular venture, innovation in the creation of new financial tools means that we have no idea who is actually holding the risk.  This video by Paddy Hirsch explains how this happens:

We’ve seen both of these problems in the current sub-prime crisis.  First off, the banks severely underestimated the default risk on these marginal loans, but more importantly, they even more seriously underestimated the risk of the housing market going flat or declining (this happened both in the US and the UK).  Secondly, the use of Collatoralised Debt Obligations (CDO) means that the risk from these outcomes was not carried by the banks that made the initial loans, but by anyone that bought (often unknowingly!) a CDO that was based at least in part on these loans.  That is how the problems caused by mortgage defaults jumped from banks and other institutions involved in the housing market to the stockmarket.

Problem: What do stock prices really mean?

The first two problems are serious. The transfer of risk through CDOs and other financial instruments can be controlled pretty well through legislation and increased oversight.  I’m not sure of any good way to get around the problems with risk estimation, although there are certainly tools available that can help people do this more accurately.  So these problems are scary, but they’re not terrifying.  The thing that genuinely terrifies me about the stock market is the third problem: we don’t know what the prices mean. This misunderstanding is so deeply embedded in the way we think about it that I’m not sure this can be corrected.  Didier Sornette addresses this issue in Why Stock Markets Crash: Critical Events in Complex Financial Systems.

This is a wonderful book that you should never read yourself.  It is part of the growing literature in ‘econophysics’ – the branch of economics that uses statistical techniques borrowed from physics to analyse economic questions.  It is a very technical book, which makes it unsuitable for a general audience.

The key point is that stock prices actually have two components:  one is the fundamental value discussed above, the second is a distortion induced by people making bets.  Sornette’s models are a real leap forward.  The idea that prices reflect not just a fundamental value, but also a measure of popular sentiment is a breakthrough. Adding a price component that is not closely correlated to value helps explain several market anomalies.

How do we explain market movements?

If stock prices are pure information, then there must be a reason behind all of the normal up and down movements that we observe in prices.  That’s why we end up with meaningless headlines like ‘Dow up 50 points after upbeat housing report.’   If you look through the pages of the financial press, you will find that all of the ’causes’ of price movements are regularly invoked to explain changes in both directions.  So upbeat housing reports will explain a market rise one month, and a decline the next.

However, by adding a component to prices unrelated to fundamental value, Sornette creates a model of stock prices that are not pure information.  In this view of the markets, price movements can reflect not just information, but also changes in belief, betting trends, herd behaviour, fashion, or even just random chance.

Why do market crashes occur?

Normally, the presence of betting doesn’t cause a problem because people are likely to place equal amounts of money on both sides of the bet.  So when bookies choose point spreads for football games, they pick a number designed to split the bets evenly between both teams.  However, there’s no point spread in the stock market, and we get bubbles and crashes when all the bets start going the same way.

The beauty of Sornette’s book is that he provides good mathematical models of bubbles and crashes.  Not only is the math in the models sound, but the stories that he tells to explain the models are plausible.  These stories are quite simple – bubbles occur when people start coordinating their bets – either consciously or unconsciously.  The scary thing is that it doesn’t take a lot of coordination to create a bubble.  In Sornette’s models these bubbles tend to be self-reinforcing as well.

As the bubbles expand, the likelihood of a crash increases nonlinearly.  This means that when bubbles are small, the chance of a crash is also quite small.  It remains small even as the deviations from fundamental value increase.  Eventually, the deviations will become large enough that the chance of a crash crosses a tipping point, and suddenly becomes quite likely.  This model seems to reflect reality reasonably well: most bubbles deflate slowly, with no adverse effects, while a small number crash dramatically.

A Short Digression

The idea that no one has modeled this before may seem incredible.  And Sornette is obviously building on previous work.  Keynes was probably the first to discuss the role of sentiment in pricing when he talked about the influence of ‘animal spirits’ on stock valuation.  In the 1960s, Benoit Mandelbrot found many of the statistical regularities on which Taleb’s work is built.  And George Soros has argued quite forcefully for the influence of sentiment for over 10 years now.  But none of this work was turned into models, so it has not had much influence on the theory underlying trading.  The arguments of Keynes and Soros were purely descriptive.  Mandelbrot’s research had the opposite problem – it relied on non-linear math that was not part of the standard financial analytic tool kit at the time.

Sornette has formalised this previous work very nicely, in models that can be worked with.  They also provide a level of prediction as well.  The predictions are not so specific that they say ‘a crash will occur next month’, they are more like warning signals.  Really loud warning signals.

The consequences of poor financial modeling

Crashes lead to problems that are most apparent in the financial derivatives markets.  Most derivatives are built on models designed to find small deviations between prices and fundamental values.  The profit comes from exploiting these deviations, and they are based on the premise that the deviations will never be large and that they won’t last very long.  In other words, bubbles don’t exist in these models.  A consequence of this is that the derivatives markets are caught by surprise when a crash occurs – and we are seeing the impact of this right now.  The value of financial derivatives is  five times the value of the real global economic output.  Since derivatives are built on shaky models, this is a major potential problem.

The currency market is a good illustration of this, where less than 10% of the trades are made by people that actually require funds in a different currency.  The rest of the trades are just bets on which direction a particular currency is likely to go.  If we think that we are just betting on whether or not there is a small deviation between the price and the fundamental value of an asset then all this gambling really isn’t a problem.  However in the real world, where many of these bets are highly correlated, the volume of gambling actually increases the threat of the crash.  These gamblers have created a very poorly engineered casino – where the players on average make a small profit on every bet they make.  The tradeoff is that there is always a chance of hitting the jackpot, but in this case the jackpot is a crash.  And unlike a slot machine, eventually everyone that stays in the market long enough will hit this jackpot.  Taleb likens this approach to an investment strategy based on snatching pennies from in front of a moving bulldozer.

Here’s Paddy Hirsch again explaining the nature of some of these bets:

This should show you why the idea that we should be able to predict what the stock market will do is flat out crazy.  Such predictions are based on the idea that prices are pure information.  As soon as we allow for the fact that prices contain more than simply data, three key problems come into play.  The first problem is that we consistently underestimate risk.  The second is that we don’t know where the risk of the system lies.  And the third is that prices reflect not simply information but also sentiment.  These three factors make financial markets inherently unpredictable.  It’s not the unpredictability that scares me, it’s the people that think that they can eliminate the unpredictability simply through juggling assets.  The fact that these people have created a financial derivatives pyramid worth many times the value of the real economy is the thing that should really terrify you about the markets right now.