Go to the Globe and Mail homepage

Jump to main navigationJump to main content

(John Foxx/Getty Images)
(John Foxx/Getty Images)

market psychology

Why computer models beat human forecasters Add to ...

R.B. (Biff) Matthews is chairman of Longview Asset Management Ltd.

Several years ago I was discussing with a colleague the well-known strategy of buying companies with low price-to-earnings ratios. He disparaged this approach as “rear-view mirror” investing, because it looked only at history and did not allow him to bring any judgment to bear upon the likely prospects for the company.

More Related to this Story

His observation was, of course, correct. What is less obvious is that investment decisions are generally hurt, not helped, by incorporating your predictions about a company’s future.

Of all the mental biases that humans possess, overconfidence in our ability to predict the future may be the most harmful.

Several studies have shown that predictions are most likely to be accurate when they are founded on a statistical base case rather than on the all too human tendency to extrapolate from particular situations. Models beat human forecasters because models consistently apply criteria that have proved over time to be relevant. They focus on the signal, not the noise.

Models don’t favour human-interest stories over boring statistical data. People, on the other hand, look to a limited set of remembered experiences, and then generalize to create a rule of thumb. While this works reasonably well in many aspects of our life, it cripples our ability to make rational investment decisions.

Behavioural psychologists have illustrated this unhelpful bias by posing the following question to groups of people: Fred lives in a small Midwestern city. He wears glasses and enjoys reading. Is it more likely that Fred is a salesman or a librarian?

Most people guess librarian. In fact there are, in any city, at least 40 times as many salesmen as librarians, so the correct answer is salesman.

Another study examined the results of medical diagnoses based on a series of medical test results. The diagnosis was performed, in one case, by a computer and, in a second, by experienced doctors who were presented with the same information. By now you will have guessed that the computer was able to diagnose the disease more accurately than the doctors.

What is most surprising, however, is the results of a second test. This time the doctors were given the results produced by the computer before being asked to make a diagnosis. The doctors’ performance improved but they still did worse than the computer.

So what should an investor do? To beat the stock market averages over the long run you must do two things. The first is to find a strategy that has been proven to outperform over time. The second is to follow it consistently.

Finding a proven approach to beating the market over the long run is not difficult. A good starting point can be found courtesy of Tweedy Browne.

In broad terms, most approaches that have been proven to be successful rely upon the fact that the relationship between the market price of a company and its current fundamentals (earnings, revenue, book value etc.) will revert over time to a long-term average.

One dead simple investment strategy that has been proven to work reasonably well is to own only the companies in an index, such as the S&P 500, that fall in the lowest decile by price-to-sales ratio (current market capitalization divided by revenue over the past 12 months), and adjust annually.

Continue this for 10 years and your chances of beating the market over that period will be greater than 80 per cent. (If you choose instead to buy an equity mutual fund, actively managed by a professional investor, your chances of beating the market over the same period, after fees, will be less than 20 per cent.)

Now, here is the hard part. Once you have identified a strategy, you must stick to it. You must not let your overconfidence in your ability to predict the future and your other biases interfere. If the formula requires you to purchase a buggy whip manufacturer, you buy it. And if the formula doesn’t work for two or three years you don’t abandon it. The odds are in your favour and the law of large numbers will eventually skate you onside.

Sticking purely to a formula through thick and thin requires more discipline than most investors, including this writer, possess. However, to the extent that you can base your decisions only on known facts, and follow a proven strategy in a disciplined fashion, your investment returns will improve dramatically.

 

 

In the know

Most popular video »

Highlights

More from The Globe and Mail

Most Popular Stories