|Almost 75 years ago, Alfred Cowles III was the first to study the accuracy of financial analyst forecasts. Cowles was interested in applying scientific research methods to economics, and in particular he devoted a lot of attention to the stock market. He made major advances in the field and developed a stock market index that tracked prices back to 1871, serving as the predecessor to the Standard & Poor's 500. Working with other professors, Cowles was instrumental in publishing a textbook that defined how to apply statistics to economics. |
THE TRUTH ABOUT FORECASTING
In detail, Cowles examined the recommendations of 36 investment firms that provided research on individual stocks between 1928 and 1932. On average, their recommendations underperformed the market by about 1.3% a year. Market forecasts were available from two dozen publications at that time, and Cowles found their accuracy to be about 50%; half the time they were right and the other half they were wrong, no better than a coin flip.
Updating his work in 1944, the conclusions were no different. Eleven years later, Cowles found that analyst forecasts were still no better than random market predictions. He wrote: "The records of 11 leading financial periodicals and services since 1927, over periods varying from 10 to 15.5 years, fail to disclose evidence of ability to predict successfully the future course of the stock market. [Figure 1] indicates that six of the 11 forecasters met with some degree of success and that five were unsuccessful in their forecasts. The 11 forecasters were on the average only 0.2% a year better than the random forecasting record."
|The chart shown here, which is adapted from the original paper, shows an interesting symmetry. Half of the analysts did better than the market and half fared worse. As a group, the results are no better than random. The best forecaster was almost exactly offset by the worst.|
Some readers will doubtless take the wrong message from the chart; they will want to know who the best performer was so that they can follow that research and earn those excess profits. John Bogle, the mutual fund pioneer who made low-cost index mutual funds available to the public, has looked at what happens when investors chase past performance.
In his first book, Bogle On Mutual Funds, he looked at the records of each year's 20 best-performing mutual funds in the subsequent year. From 1982 through 2005, he found this group of funds delivered only average performance in the following year. In statistical terms, he demonstrated that mutual funds exhibit mean reverting behavior, with good years just as likely to be followed by bad years as bad years are likely to be followed by good years.
IN MODERN TIMES
Looking only at more recent data, Dreman studied 108,000 analysts' quarterly consensus estimates of widely followed companies from the beginning of 1973 to the end of 1998. For most stocks, the consensus forecast consisted of the average from 10 to 40 individual analysts' estimates. The study used the forecasts published two weeks before the release of actual earnings.
The results showed that analysts were still just as prone to errors as when they made forecasts using only pencil and paper. The consensus forecasts over the entire 25-year period were off from reported earnings by 42% on average. Dreman did note an improving trend, however. "To be fair, analysts have gotten better lately at picking the right numbers: In 1998 they were off by only 36%," he wrote.
He noted that most of the analysts erred on the more optimistic side. This also agreed with the results that Cowles had obtained decades before. During the Great Depression, there were more than four times as many bullish estimates as bearish earnings projections. During the time span Cowles studied, more than half of the years included the greatest bear market in history, and stocks, in general, suffered a decline of almost 70% during that time span.
Given the bullish bias of analysts and the horrific track record of short-term forecasting, we can only question why anyone puts their faith in annual forecasts or even the five-year projections offered by some firms.
Earnings surprises are actually so common that many analysts track them and have developed a technique to quantify the magnitude of the surprise. This indicator, called "standardized unexpected earnings" (SUE), offers a way to compare the degree of a company's most recent earnings surprise to the company's longer-term track record of earnings surprise.
For example, some companies seem to always beat earnings estimates by a penny. When the quarterly announcement reveals that the company did in fact beat the consensus estimate by just that amount, it shouldn't really be unexpected. SUE offers a method to quantify the degree of the earnings surprise and compare one company's surprise to another.
As a formula, SUE is equal to the earnings surprise for a given quarter divided by the standard deviation of earnings surprises measured over some historical period such as the previous 20 quarters.
For example, consider a stock that has just announced a 0.05 earnings surprise and has a standard deviation of past earnings surprises of 0.08. The most recent surprise is actually smaller than normal, and the standardized earnings surprise would be 0.05/0.08 = 0.63. SUE values greater than 1 would indicate that the surprise is greater than normal.
Many academic studies show that earnings surprises can be useful in identifying stocks that should deliver excess returns. According to these studies, buying stocks after they have delivered a positive earnings surprise has consistently delivered positive returns. Similarly, shorting stocks after a negative surprise has been shown to be a profitable strategy over the long term.
When a company initially announces better than expected earnings, it is often signaling a change in the fundamentals of the business. The price of the stock usually moves higher, and analysts raise future earnings estimates a little. Investors and analysts tend to underestimate the impact of the news, and the price moves and revised estimates don't fully capture the positive changes within the company. The next quarter may display the same pattern -- a positive earnings surprise followed by an upward move in the stock price, and then the analysts increase their future earnings estimates.
In other words, analysis tends to be focused on the most recent past and everyone expects the future to be much like the past. In reality, analyst errors are an indication of fundamental changes within the company and the market initially underreacts to these changes, expecting mean reverting behavior to prove the original analysis correct in the long run.
STRENGTH OF RELATIVE STRENGTH
Numerous other studies have validated the utility of relative strength, usually relying solely on a mathematical examination of price action. It is interesting to note that fundamental factors such as analyst estimates can also contribute to an understanding of why relative strength is an effective investing tool.
Bogle, John . Bogle On Mutual Funds, Dell.
Cowles, Alfred, III . "Can Stock Market Forecasters Forecast?" Econometrica, July.
Dreman, David . Contrarian Investment Strategies, Free Press.