RULE 6

Analysts forecasts are usually optimistic. Make the appropriate downward adjustment to your earnings estimate.

Next well tum briefly to past studies of analyst and management forecasts to see if they were any more helpful to the investor.

Was Forecasting in the Past Any Better?

Let us start by looking at the management record. Because the thorough analyst or money manager carefully interviews senior corporate officials in making his or her projections, a number of past academic studies have measured management forecasts of their own companies eamings. As Table 5-3, taken from The New Contrarian Investment

Table 5-3

Forecasts, Analysts Versus Management Management Forecasts, One Year or Less*

Richards, Benjamin and Strawser, 1977

1969-72

18.1%

Mean error in analysts studies: 16.6%

Source: The New Contrarian Investment Strategy.

What Does It All Mean?

We have found that analyst forecast errors have been unacceptably high for a long time, and they have gone up over the past two decades. An

* The S&P Eamings Forecaster was one of the original services that measured eamings estimates in the 1960s and 1970s.

Strategy, indicates, the average error of four management studies between 1963 and 1971 was 14.5%. Therateofforecasting error by senior corporate officers (presumably in the Icnow, if not to some extent controlling reported earnings) is significantly higher than acceptable to practitioners.

Analysts also struck out from 1960 to 1976, as demonstrated by the seven studies charted in the second part of Table 5-3. As you can see, the mean error of the seven studies was 16.6%, and the studies were for a year or less. Could a couple of major errors or the odd poor forecaster have caused these results? Richards, and Barefield & Cominsky, agreed in previous studies that there was no significant differential among forecasters on any given company. The consensus was far off target, time and again.

Richards and Frazer concluded that a lack of expensive research doesnt put the small investor at a disadvantage. "It is unnecessary to pay large sums for certain services when others are available at low cost." The authors graciously did not add, "Particularly since they were so wide of the mark."

More confirming evidence by Richards, Benjamin, and Strawser indicates the average industry forecasting error made by analysts annually for the 1972-1976 period. Over the entire time, the error ran to 26.2% annually. More significantly, some of the industries supposedly having the "highest visibility," such as computers and retail stores, actually had the worst estimates. The average annual analyst error in the office equipment and computer group was an astonishing 88.8%. Since these two samples were taken from the S&P Eamings Forecaster,* which then carried the largest or most widely followed companies, and since they covered periods from five to nine years, the results can scarcely be considered an aberration.

We see, then, that analysts forecasts from the sixties through the mid-seventies are still wide of the mark, albeit the errors are smaller than more recently.

error rate of 44% is frightful-much too high to be used by money managers or individual investors for selecting stocks. Normally, stock-pickers believe they can fine-tune estimates within a 5% range. The studies show the average error is over eight times this size. Error rates of 10 to 15% make it impossible to pick out a growth stock (with eamings increasing at 20% clip) from an average company (with earnings growth of 7%), or even from an also-ran (with eamings expanding at 4%). What then do error rates of over 40% do?

Dropping companies with small eamings per share (to avoid large percentage errors) does not eliminate the problem. The error rate is still over 20%. Worse yet, analysts err often. Figure 5-2 showed that only one in four consensus analysts estimates fell within the cracial 5% range of reported eamings. Missing this range would spell big trouble for stock-pickers relying on laser-like estimates.

Unfortunately, the problems dont end here. Forecasting by industry was just as bad. Forecast errors averaged over 40%, with error rates almost indistinguishable between those industries with supposedly excellent visibility (for which investors pay top dollar) and those considered to have dull prospects. If eamings estimates are not precise enough to weed out the real growth stocks from the also-rans, the question naturally arises why anyone should pay enormous premiums for "high-visibility" companies.

Finally, we have seen two additional problems with analysts forecasts. First, the error rates are not due to the business cycle. Analyst forecast errors are high in all stages of the cycle. Second, and more important, analysts have a strong optimistic bias in their forecasts. Not only are the errors high, but there is a consistent tendency to overestimate eamings. This is deadly when you pay a premium price for a stock. The towering forecast errors combined with analysts optimism result in a high probability of disaster. As we saw, even a slight "miss" for stocks with supposedly excellent visibility has unleashed waves of selling, taking the prices down five or even ten times the percentage miss of the forecasting error itself.

The size and frequency of the forecasting errors call into question many important methods of choosing stocks that rely on finely tuned estimates ranning years into the future. Yet accurate eamings estimates are essential to most of the stock valuation methods we looked at in chapter 3. The intrinsic value theory, formulated by John Burr Williams, is based on forecasting eamings, cash flow, or dividends, often two decades or more ahead. The growth and momentum schools of investing also require finely calibrated, precise estimates many years into the future to justify the prices they pay for stocks. The higher the multiple, the greater the visibility of eamings demanded.