When it comes to predicting the market's overall future direction, conventional indicators are of next to no use at all

Who said that stockbrokers never produce any worthwhile investment research? Professional fund managers will quickly tell you that the tons of stockbroker research that pours into their offices every day is second- rate and worthless. The bulk of it ends up, unread and unwanted, in the wastepaper basket.

Is that fair? My impression is that brokers' research is actually now rather better than it was - certainly better edited and laid out, but also more rigorous and professional in its analysis.

There are occasional gems to be found amidst all the dross. This week, for example, I have been poring over a fascinating piece of work from James Capel, one of the stronger research-led brokers still left in the City.

What Capel's set out to test was what use conventional market valuation measures are in assessing the likely future direction of the stock market. Most investors are familiar with the traditional value indicators, such as a dividend yield, the P/E (price/earnings) ratio and the gilt/equity ratio.

More sophisticated investors may also now be looking at other indicators such as discounted cash-flow models, advance/decline ratios and so on. There is really no shortage of candidates and everyone has their own favourite.

But do any of them really have any value? For comparing individual shares, the standard valuation measures such as yield and P/E ratios are clearly helpful. But when it comes to predicting the market's overall future direction, the answer - Capel's research confirms - is that they are of next to no use at all. In fact, they may be positively misleading!

Take Wall Street, for example. The dividend yield of the market is now, as has been pointed out here many times, lower than it was before either the 1929 or 1987 stock market crashes. But that has not stopped the market powering ahead. Those who failed to be fully invested in Wall Street last year have paid a high price in missed opportunity since.

The same goes for the market's P/E ratio, says Capel's. In 1992, this reached a near all-time high, with prices on average nearly 23 times current earnings, roughly double the long-run historical average. Yet those who took that as an unmistakable sell signal would have missed out on an even bigger advance than those who waited for the dividend yield to drop below 3 per cent.

The reason is that whatever signal the P/E ratio may have been giving, in the event it was drowned out by other, more powerful influences - notably, in the case of Wall Street, the start of a strong surge in company profits and the relentless decline in both short and long-term interest rates.

A similar story can be told for nearly every other traditional market indicator in all the world's main markets. One that has done quite well, Capel's finds, is the ratio between the yield on short-dated gilts and the average dividend yield on the FT All-Share Index.

Every time, bar once, that this ratio has risen above 2.5 times, it has marked a peak in American share prices.

But, alas, even this seemingly robust indicator has its flaws. It's been good at calling the top of the market, but has given absolutely no warning of any of the market's troughs during the same 23-year period.

The general conclusions of Capel's statistical analysis are:

That no single indicator of market value has any real predictive power;

That although all measures eventually revert to their long-term average level, you cannot safely assume that they will do so on any one or two- year time horizon;

That the most powerful force at work in shaping equity values around the world is the level of interest rates, but even they only begin to have predictive powers if you already know where we are in the current investment cycle. In which case, of course, who actually needs them?

A blunter way of putting this conclusion is to say that "market timing" - trying to guess the future level of the stock market - simply does not work. Most investors, I suspect, are well aware of this, if only intuitively. The scientific evidence is certainly irrefutable. The Capel study is only the latest to underline this point.

But will it stop people trying to have a go at market timing, using whatever indicators they want? Of course not. Private investors do not have the inclination and professional fund managers have no freedom to stop trying to call the market's turns.

The latter are judged by and remunerated by their performance against the market as a whole and they have no choice but to try and beat it from year to year.

The reason I applaud James Capel for its outstanding latest piece of research - it is thoroughly analysed and full of fascinating historical detail - is the fact that the research has been produced by the broker's strategy team.

What is their job? To advise their professional institutional clients on where the market will be in six months to a year's time (for the record, the Capel's strategist Peter Oppenheimer and his team think that both Wall Street and the London market have further to go and are still in an earnings-driven phase).

It may be impossible, but at least the Capel boys are prepared to give this thankless, if not impossible, task their best shot.