CAMBRIDGE, Mass. - A large body of evidence suggests analysts' earnings estimates are unreliable, said David N. Dreman, chairman, Dreman Value Advisors Inc., Jersey City, N.J.
Speaking at the Behavioral Economics Forum at Harvard University last month, Mr. Dreman said investors depend on fine-tuned earnings estimates today more than ever before, yet too many variables and changing conditions lead to inaccuracy.
He cited a study he conducted of analysts' consensus earnings estimates from 1973 to 1990 in which the average annual error was 46%. When small companies were removed from the sample, the average annual error was 23%.
Further, he said, the forecast error as a percent of reported earnings appears to be trending upward.
"We can't fine-tune the estimates," he said. "Only 26% of analysts get within plus or minus 5% of the actual earnings. And 57% of the estimates are outside plus or minus 10% (of actual earnings) and 45% are not within plus or minus 15% of actual earnings.
"We demand precision that isn't there. The error rate is far too high (for earnings estimates) to be the sole determinant of stock prices."
Growth/value distinction blurs
In addition, Mr. Dreman said the high error rate makes it hard to distinguish between growth and value companies.
Studies suggest two explanations for the errors: Earnings over short periods follow a random walk, or analysts project past growth rates into the future.
Further, he said, current theory assumes that rational decision-makers should learn from past mistakes, but studies indicate they do not. Even slight earnings surprises, despite the lack of precision, can result in major stock price movements.
Mr. Dreman said studies demonstrate an asymmetric reaction to earnings surprises by "in favor" stocks and "out of favor" stocks. Negative earnings surprises hurt "in favor" (or high p/e stocks) more than they hurt "out of favor" (or low p/e stocks). Positive earnings surprises helped low p/e stocks more than they helped high p/e stocks.
The net effect of surprises is positive for low p/e stocks and negative for high p/e stocks, he said. And the surprise effect lasts a long time.
Manipulating earnings
Richard Zeckhauser, professor of political economy at Harvard University, told the forum that most major American corporate managements manipulate their earnings in recognizable ways, because earnings matter for investors.
Some of the manipulation is done by postponing expenses and costs, and some by accounting, he said.
"If earnings are lower than expected, your stock is going to fall," Mr. Zeckhauser said. But earnings expectations are set by analyst estimates, and there is a network of interdependent relationships between the analysts and corporate management, he said.
Stock prices are affected by positive or negative earnings and whether they beat analysts' consensus estimates and earnings for the same quarter last year.
"If you are below the analysts' forecast you lose 3% of your (market) value," Mr. Zeckhauser said. "The magnitude of how much you are below doesn't affect the stock price. However, if you are above the forecast, the size of the excess does matter.
"If you fall below where you were four quarters ago, that costs you another 5% (of market value)."
Bettering analysts is 'nonsense'
But surpassing analysts' consensus estimates is nonsense, he said, because analysts get the information from company management, and company management wants to outperform the analysts' consensus forecast.
Mr. Zeckhauser noted that behavioral finance says individuals have a hard time making rational decisions especially in the financial arena, where they should make the best decisions.
In the financial arena, he said, the objectives are fairly well defined, a lot of information is available, lots of professionals are in the market, and the investor is engaging in the decision-making process over and over.
In addition, he said, the decisions are mostly ones in which the individual can stop and reflect. The individual does not have to make them in a moment of time.
Psychological tendencies listed
Mr. Zeckhauser said individuals have a hard time making rational decisions because of a number of psychological tendencies. One is status quo bias; people tend to stick with what they have.
He said the overwhelming majority of people who inherit a portfolio tend to stick with it, apparently because they would feel terrible if they sold what they had and the stocks then rose.
Fewer than 3% of investors rebalance their portfolios every year, even though finance textbooks say an investor should rebalance when one asset rises a lot.
Another psychological tendency is "barn door closing," Mr. Zeckhauser said. People tend to make the decision today they would like to have made yesterday, allowing them to avoid confronting past mistakes, he said.
Investment behavior is also affected by the "expert" or "reliance" effect. People rely on the advice of people they perceive to be experts. Mr. Zeckhauser noted that Harvard University faculty members generally allocate their retirement account exactly as they are told most of their peers have allocated.
Another example of the expert effect, he said, is the use of mutual funds by investors, though mutual funds perform modestly and they are relatively costly to use. However, mutual funds are supposedly managed by experts.
Stock's price and value differ
Richard H. Dreihaus, chairman and chief executive officer of Dreihaus Capital Management, Chicago, told the forum that a company's stock price is rarely the same as its value.
"The reason for that is the valuation process is flawed, " he said. "Stock prices are heavily affected by market dynamics and by investors' emotions. These emotions swing widely from pessimism to optimism.
"Also, many investors buy stocks with the intention of holding them for one to five years, based upon information that really only applies to a short-term time horizon. While the information they are using to invest may be valuable, it is often the wrong information for their investment time frame.
"If people invest in a company based on current information, they have to be prepared to act on any changes in that information in a much shorter time frame than most investors are prepared to do."
3 behavioral phenomena
Langdon Wheeler, president, Numeric Investors L.P., Cambridge, said there were three behavioral phenomena in the market:
Short-term overreaction, which lasts three to six days;
Medium-term underreaction, which can last from three to six months; and
Long-term overreaction, which can last three to six quarters.
Short-term overreaction is caused by too-urgent traders creating short-term imbalances in supply and demand, Mr. Wheeler said. It often provides a 15% return reversal over a few days, but it is too fast to capture.
The medium-term underreaction arises when analysts revise estimates in small steps, thus forming trends in the revisions. Past revisions forecast future revisions which forecast relative returns, Mr. Wheeler said. The effect is smaller than the short-term overreaction, but it can be captured.
The long-term overreaction is caused by investors overreacting to news, moving prices away from stocks' intrinsic values, he said. Prices gradually return to equilibrium, generating excess returns.
Mr. Wheeler said the markets respond to overreaction or underreaction at different times, but one or the other most of the time.
