RULE 7

Most current security analysis requires a precision in analysts estimates that is impossible to provide. Avoid methods that demand this level of accuracy.

Hey, Im Special

What do we malce of these results? If the evidence is so strong, why arent more investors, particularly the pros, aware of it, and why do they not 1 1 it into their methods, rather than quiclc-marching into an ambush? Why do Wall Streeters bhthely overloolc these findings as mere curiosities-simple statistics that affect others but not them? Many pros believe their own analysis is different. They, themselves, will hit the marlc time and again with pinpoint accuracy. If they happen to miss, why it was a simple slip, or else the company misled them. More thorough research would have prevented the error. Next time, it wont happen.

Lets examine why this mentality is prevalent in the face of overwhelming evidence to the contrary.

Some Causes of Forecasting Errors

As we just saw, investors either ignore or arent impressed by the statistical destruction of forecasting, even though the destruction has been thorough and spans decades. There are a number of reasons, some economic, some psychological, why investors depending on finely cah-brated forecasts are filcely to end up with egg on their face. Cragg and Malkiel did an early analysis of long-term estimates, published in the Joumal of Financed They examined eamings projections of groups of security analysts at five highly respected investment organizations, including two New York City bank tmst departments, a mutual fund, and an investment advisory firm. These organizations made one- to five-year

If the average forecast error is 44% annually, the chance of getting a bulls-eye at a distance of ten years seems extremely slim. Which brings us to another rule.

estimates for 185 companies. The researchers found that most analysts estimates were simply linear extrapolations of current trends with low correlations between actual and predicted eamings.

Despite the vast amount of additional information now available to analysts, say Cragg and Malkiel, and despite the frequent company visits, estimates are still projected as a continuation of past trends. "The remarkable conclusion of the present study is that the careful estimates of security analysts ... performed litde better than those of (past) company growth rates."

The researchers found that analysts could have done better with their five-year estimates by simply assuming that eamings would continue to expand near the long-term rate of 4% annually.*

Yet another important research finding indicates the fallibility of relying on earnings forecasts. Oxford Professor I. M. D. Little, in a paper appropriately titled "Higgledy Piggledy Growth," revealed that the futures of a large number of British companies could not be predicted from recent earnings trends. Littles work proved uncomfortable to theoreticians and practitioners alike, who prompdy criticized its methodology. Litde accepted the criticism, carefully redid the work, but the outcome was the same. Eamings appeared to follow a random walk of their own, with past and future rates showing virtually no correlation. Recent trends (so important to security analysis in projecting earnings) provided no indication of the future course."

A number of studies reach the same conclusion: changes in the eamings of American companies fluctuate randomly over time.

Richard Brealey, for example, examined the percentage changes of eamings of 711 American industrial companies between 1945 and 1964. He too found that trends were not sustained, but actually demon-sti-ated a slight tendency toward reversal. The only exception was companies with the steadiest rates of earnings growth, and even their correlations were only mildly positive.

Juxtaposing the second set of studies to the first provides part of the explanation of why analyst forecasting errors are so high. If analysts extrapolate past eamings trends into the future, as Cragg and Malkiel have shown, and eamings do follow a random walk, as Little and Brealey have demonstrated, then one would expect sizable errors. And large forecast errors are what we have found consistently through the chapter.

Thus, once again and from quite another tack, we see the precarious-ness of attempting to place major emphasis on eaming forecasts.

104 The Expert Way to Lose Your Savings RULES

It is impossible, in a dynamic economy with constantly changing political, economic, industrial, and competitive conditions, to use the past to estimate the future, r -

There are several other economic reasons that can cause eamings forecasts to be off base. One is what Harvard economist Richard Zeckhauser calls "the big bath theory." In a paper coauthored with Jay Patel of Boston University and Frangois Degeorge of the School of Management in Paris, the researchers provide evidence that many companies try to manage eamings by trying to show consistent, gradual improvements. Analysts have an appetite for steady growth, and that is what management tries to serve up. When they cant do it, they take the "big bath," writing off everything they can, perhaps even more than is necessary (accounting again), in order to show a steady progression of eamings after the bath. The big bath could be another unpredictable effect that throws analysts forecasts off.

Reviewing the evidence makes it appear that forecasting is far more art than science and, like the creative fields, has few masters. Excluding the highly talented exceptions, people simply cannot predict the future with any reliability, as the figures starkly tell us.

Career Pressures and Forecasts

There are some substantive factors that affect the analyst directly, the most important being career pressure. These can result in forecasts missing miserably. After surveying the major brokerage houses, John Dorfman, then editor of the market section of The Wall Street Joumal, whom we met earlier,* provided a list of what determines an analysts bonus, normally a substantial part of his or her salary. In Dorfmans words, "Investors might be su rised by what doesnt go into calculating analysts bonuses. Accuracy of profit estimates? Thats almost never a direct factor. .. . Performance of stocks that the analyst likes or hates? ... It is rarely given major weight." The ranking of seven factors determining an analysts compensation places "accuracy of forecasts" dead last.

What is most important is how the analyst is rated by the brokerage firms sales force. Many firms conduct a formal poll of the sales force, which ranks the analysts primarily on how much commission business they can dram up. At Raymond James, the sales forces rating accounts