Forecasting stars wide of the mark (CORRECTED)

CORRECTION (PUBLISHED 23 AUGUST 1992) APPENDED TO THIS ARTICLE

ONE firm of City brokers estimates it sends out nearly 1,500 kilograms of research material to its clients every week.

Geoffrey Osmint, founder of Extel's Annual Survey of Investment Analysts, calculates that for the City as a whole, the weekly output of analysts' papers and circulars could reach 30 to 40 tons.

That, to put it mildly, is an awful lot of trees. If Nigel Taylor, head of investor relations at ICI, is anything to go by, it is also a monumental waste. Analysts, he said, 'produce too much paper, a lot of which is reproduced, and most gets dumped'.

Even fund managers, the people analysts are directly meant to serve, claim not to read the stuff. In the 1991 Extel survey, 60 per cent of fund managers responding said they did not read three-quarters of the research they received.

If that was not bad enough, according to the forthcoming issue of Management Today magazine, the analysts the fund managers rate most highly are by no means the most accurate.

The Extel survey gives the top-polling analysts a 'star' rating, based on fund managers' opinions. By comparing their profit forecasts for the two biggest companies in their sector, as published in The Estimate Directory (a monthly compendium of brokers' predictions), with actual results, it is possible to judge just how good they really are (see table, right).

In the engineering-aerospace sector, for example, Rolls-Royce recorded profits last year of pounds 51m. The top team of analysts in the 1992 Extel survey, and by some considerable margin, belongs to Kleinwort Benson. But its forecast for Rolls-Royce of pounds 70m makes it the second most inaccurate of the 17 brokers which cover the sector. Top place would be shared by Nikko and Strauss Turnbull, which both forecast a close pounds 55m.

The pattern is repeated elsewhere. In food manufacturing, SG Warburg holds the star spot. However, the firm's estimate that Unilever, the largest company in the sector, would make profits last year of pounds 2.05bn was the 10th least accurate out of the 14 firms that research the industry.

The dismal performance of Hoare Govett, though, takes some beating. Its team, ranked fourth by Extel, said Unilever would make pounds 2.205bn - pounds 413m wide of the mark.

As for those analysts who do seem to get things right, Management Today points out they may not necessarily be at the best-known firms.

Strauss Turnbull, for instance, appears twice in the best forecast list, and the most accurate estimate for Racal profits came not from the giants of BZW or SG Warburg, but Gilbert Elliott.

Extel invited fund managers to base their assessments on more than profit forecasts. But among analysts themselves, it is the main criterion they use for judging relative success or failure.

Perhaps Alan Sugar, the Amstrad boss, got it right when he declared in 1990: 'Most of the time they talk through their backsides.'

CORRECTION

Management Today is correcting a table of analysts forecasts attributed to the magazine in last week's Independent on Sunday, following a number of complaints. The table attempted to rank the performance of top analysts by comparing their forecasts with actual profits reported by companies.

Warburg Securities and Hoare Govett have asked us to point out that the forecasts of Unilever's profits attributed to the firms referred to 1992, and not 1991 as stated in the table and an accompanying article, headlined Forecasting stars wide of the mark. It was therefore incorrect to state that Warburg's forecast for last year was the 10th least accurate of the 14 firms that research the sector.

It was also incorrect to attribute the worst forecast of any firm to Hoare Govett. In fact Hoare Govett's forecast of Uniliver's profits for 1991 was relatively close to the actual outcome. Henry Cooke Lumsden was incorrectly placed as the worst forecaster of Racal's profits. Its forecast of pounds 148m for 1991 included the profits of Racal Telecom, which was since demerged.

(Graphic omitted)