Tuesday, December 13, 2011

Market Forecasts, And Regression to the Mean


I've been reading Daniel Kahneman's excellent new book entitled Thinking, Fast and Slow.

The book was recently listed as one of the best non-fiction books of 2011 by the New York Times. Based on my reading, I would agree with the Times.

Dr. Kahneman won a Nobel Prize in 2002 for his work on behavioral psychology. However, his book is very readable, and targeted for a larger market. For anyone interested in the quirkiness of how our minds work, and how we often make decisions that sometimes seem totally irrational, it will make an excellent addition to your holiday reading list.

One of the points that Kahneman makes in his book is that we often place too much time looking for causality in trying to explain events.

Sometimes events are just random - coming up tails nine times in a row when you're flipping a coin doesn't necessarily mean that the tenth flip will be heads; the odds are always 50/50, regardless of prior results.

At other times, though, changes occur that are the result of simply regression to the mean.

Kahneman discusses the fact in a large sample size you will often get data points that seem far out of the norm. However, overall results will very often return to the longer term averages, and that any interpretation of random results that doesn't include regression to the mean are usually incorrect.

I have been thinking of Kahneman's work when I am trying to come up with the best advice to help clients structure their investment portfolios.

The last 10 years have been relatively poor ones for stock investors, while bond investors have enjoyed a very healthy run.

However, over longer periods of time, stocks have produced much higher rates of return than bonds.

If regression to the mean holds, then, the next decade should be much better for stocks than bonds, simply because the magnitude of stock underperformance relative to bonds has been so unusual relative to historic norms.

Note that this forecast isn't based on guessing on economic outlook, Fed policy, euro, etc. No, all it's based on is simple statistics.

Think back to the end of 1999: Stocks had just completed one of the most remarkable runs in capital markets history, while bonds had been consigned to those poor souls who didn't understand the bull case for stocks.

As we now know, this was exactly the time that stock investors should have been heading for the exits and piling into bonds based solely on regression to the mean. And yet if you look back you will see few, if any, advisors were recommending an overweight in bonds.

Note that even if I had told you in late 1999 that the US economy would continue to thrive for most of the coming decade, you still would have been better off in bonds, not stocks.

Or, put another way, regression to the mean, and not economic or market forecasts, may be a more useful investing tool for investors truly focused on longer term results.