Monday, September 03, 2007

My computer guessed wrong


Sep 2nd 2007

The pitfalls of buying cheap stocks dearly

AUGUST'S market decline was almost the reverse image of the initial stages of the collapse of the dotcom bubble. Then, as now, one group of stocks had become overfavoured by investors. Sentiment changed, and turmoil ensued, as everyone dashed for the same exit.

Back in 2000, it was growth stocks (particularly those in technology, media and telecoms) that had been the flavour of the month. This time round, it was the value stocks. In itself, this sounds like a bit of any oxymoron. The definition of a value stock is that it is cheap and unloved by the market; how can it be popular?

The answer seems to be the computer-driven, or quantitative, funds that have come to dominate markets in recent years. They trawl through past data to find factors that appear to offer superior returns. The theory is that this approach removes the behavioural biases displayed by traditional investors.

The method has been enormously successful. But it tends to lead to a value bias; and while value stocks can hardly, by definition, be described as expensive, the range of valuations in the market is much narrower than it was in 2000. Value is certainly not as cheap as it used to be. As a result, in order to juice up the returns, the quant funds used leverage. A conference call with Goldman Sachs suggested its quant fund was geared six times.

That made them vulnerable when risk appetites changed. As they sought to cut back their gearing, the quant funds reduced their positions. But, as Rob Buckland of Citigroup has pointed out: “If everybody back-tests the same data, they will probably end up with the same strategy”. As everyone tried to sell the same stocks, prices were forced sharply lower. Hence market moves that quant models were simply not constructed to anticipate.

The problem is neatly illustrated by figures from MSCI Barra, a risk management firm. Traditionally, the correlation between value and momentum factors has been low, around 0.21. (Momentum stocks are those that have recently been rising fastest). Accordingly, many quant managers use momentum as a further criterion for picking stocks, since using two uncorrelated factors improves the risk-reward trade-off of a portfolio.

But, in the month from July 10 to August 9, the correlation between the two categories was 0.84, not too far from perfect (the highest possible correlation is 1). Both sets of stocks fell sharply, or to put it another way, value stocks had become (negative) momentum stocks. Quant managers suffered what MSCI Barra described as a “perfect storm”—what they thought were diversified portfolios turned out to be highly concentrated.

This had a powerful effect on markets. The quant funds had become central players in the financial system, effectively acting as market makers. As they were forced to sell, they withdrew liquidity from the system, at just the time when everyone was nervous about the subprime credit crunch.

What are the lessons of all this? It would be a mistake to believe that everyone should throw away his computer (indeed, some funds have already recovered a good part of their losses). But the quant guys will have to go away and work out how to allow for the effect of the size of their positions, both in terms of the amount of assets each firm can control and the effect of competitors' using similar strategies.

They will also have to rethink the issue of leverage. The value approach works in the long term, but can have horrible short-term periods, as those who were underweight technology during the dotcom bubble can testify. Using leverage is highly risky; better to accept periods of low but steady returns, than to risk blow-outs like the one we’ve just experienced.

No comments:

Lunch is for wimps

Lunch is for wimps
It's not a question of enough, pal. It's a zero sum game, somebody wins, somebody loses. Money itself isn't lost or made, it's simply transferred from one perception to another.