Data Mining Can Produce Fool's Gold

By David McEwen

Monday 19th November 2001

Past returns are no guarantee of future performance. We've all heard that phrase before yet many people still ignore it.

Last year's top fund performers always seem to attract the most new customers and shares that have risen dizzily in price still attract the most attention. Unfortunately, those who invest solely on the basis of recent buoyant performances can end up buying expensive assets.

Others who tend to learn about past returns the hard way are investors looking for patterns they hope will be repeated in the future.

Chartists are in this category but so are the so-called "data miners".

These are people who crunch numbers, looking for correlations or themes that other investors haven't discovered. This has become much easier with the speed and power of modern desktop computers.

One famous example of a data mining outcome is the "Dogs of the Dow" strategy. Around ten years ago, US funds manager Michael O'Higgins discovered that the worst performing shares on the Dow Jones Index tended to outperform the index in the year following. He revealed all in a book and now the strategy is widely implemented.

A more recent theory getting a lot of attention is where an investor buys shares with the highest dividend yields in the market. History has found a portfolio of these shares also tends to beat the market in the year following.

While these strategies have shown results, sometimes over extended periods, they still attract sceptics. That's because it is extremely rare to find a pattern than consistently repeats.

An example is a dice that rolls a six ten times in a row. The consistency of past results does not increase the likelihood of another six on the next throw.

In many cases, events can seem correlated but are not. For example, imagine if research found that over the past ten years the best performing class of shares on the New Zealand share market were those that started with the letter 'B'.

Obviously, there is no way there can be a link between the first letter of the name of an investment and its performance.

In a celebrated study a couple of year ago, a US researcher spent many hours analysing United Nations statistics looking for money-making patterns. At last he found an uncanny correlation between butter production in Bangladesh and the movement of the US share market. Again, it would be a brave or foolhardy investor who bet that a steep rise in butter production would automatically foreshadow the next bull market.

One of the ironies of data mining is that, should an overlooked indicator of price rises be discovered, it doesn't take long before enough people get to know about it to cancel the effect out.

Take the Dogs of the Dow theory. This has become so popular that investors now pour tens of billions of dollars each year into target shares. As Mr O'Higgins pointed out in 1997, after he had stopped using it as an investment tool, too much demand for unloved shares quickly pushes up their price to the extent that they are no longer cheap.

Forget data mining, profitable investment comes from realising that markets are inherently unpredictable and taking the appropriate steps to spread your risk.

David McEwen is an investment adviser and author of weekly share market newsletter McEwen's Investment Report. He is commissioned by the New Zealand Stock Exchange to write an independent personal investment column. He can be reached by email at davidm@mcewen.co.nz or by mail care of this newspaper.

Comments from our readers

No comments yet