the jonathan davis column
Past performance is not necessarily a guide to the future. The phrase is mandatory in all advertisements for unit trusts and other equity- based financial products. But what actually does it mean?

By trawling the historical record, and juggling with the start and end dates, there is always one period where the performance of a fund can be made to appear as good as anything else around. If you want to be precise about it, the predictive value of unfiltered past-performance statistics is nil.

The statistical evidence on this point is uncompromising - just as clear, in fact, as the finding that 75 per cent of all managed funds consistently fail to outperform the market averages over a five-year period.

So should one just revert to throwing a dart at a list of unit trusts? Clearly not. Common sense tells us that some fund managers are better than others. The good news is that the tools to make meaningful comparisons between different funds are becoming both more sophisticated and gradually more widely available. Raw data on the performance of funds is available from a number of sources.

But this still leaves the question of how to interpret the data. A pioneer of sophisticated fund performance analysis in this country is Fund Research, started around 10 years ago by two ex-Fidelity fund managers.

Peter Jeffreys, managing director of Fund Research, uses a combination of a sophisticated quantitative screening system and detailed qualitative assessments of individual managers to rank and rate the best funds in the UK. The main quantitative technique he uses involves looking at the consistency of a fund's performance and the frequency with which it appears in the top echelons of the performance tables. This in turn produces a weighted score which measures consistency. Fund Research also produces a series of statistical measures to measure each fund's risk profile, backed up by detailed interviews with fund managers.

The table shows how the top 10 unit trusts in one particular sector, UK equity growth, measure up. A score of 10 means a fund was in the top 10 per cent of all the funds in its sector in a particular period; a 9 means it ranked in the second 10 per cent, and so on down the list.

The method aims to weed out funds which, thanks to one particularly good year, suddenly appear in the top echelons of the league tables, only to disappear like a shooting star the next. They are the funds to be avoided. By contrast, lucky the investor who can find the few stalwarts, such as Fidelity Special Situations, which have been consistent outperformers for nearly 20 years.

How good is this method at picking winners? Well, that depends on your point of view. According to Fund Research, their methodology increases the predictive power of the performance figures from evens to about 2- 1 on. In others words, there is a two-thirds chance that a fund which scores well in the consistency tables will stay above average in the future - which is better, but not a clincher.

You would not know from the table alone, for example, that the number one trust, BGI UK Growth fund, had a very indifferent record in the five years immediately before the period covered by this analysis. At the time, it was known as the Barclays Unicorn Special Situations fund, and the new fund has been formed by amalgamating it with two other funds in the old Unicorn stable. It is now run in a quite different way from its predecessor (for the better in my view).

So, even with sophisticated tools, you or your financial adviser still have to do some detective work to pick out the best funds - and even the very best fund managers have occasional poor years. Mr Jeffreys makes the point that this tends to happen when markets are running away with themselves, (now being a case in point).

The wider point, however, is the more important one. Better screening facilities means buyers are becoming more sophisticated, which in turn is forcing providers to sharpen up their act.

The remarkable fact about the unit trust business, in retrospect, is not how few consistently good performers there have been - but how so many of the also rans have survived for so long, despite providing a rotten service to their users.