How reliable are earnings estimates?

[John Mauldin pointed out that] David Dreman pointed out in a study that he conducted from 1982 through 1997 analysts were ON AVERAGE wrong by 200%. (Forbes, July 8, 2002)

Dreman tells us, "Earnings performance for 2002's first half was a sorry one. Company after company was forced to lower expectations or restate past results downward. How can the consensus justify such a healthy-looking multiple for the year as a whole? By forecasting a second-half profit boom that gushes up from nowhere: a 48% gain (from a year earlier) in the third quarter and a 45.7% one in the fourth, according to S&P analysts' forecasts. Included in the forthcoming profit explosion, as reported in First Call, are a 127% income increase in technology stocks in the third quarter and a 73% jump in the fourth and a hardly modest 19-fold rise in transportation earnings in the third quarter (mainly airlines), with an even larger gain forecast for the fourth."

Another longer-term study published by the National Bureau of Economic Research shows that analysts typically overstate earnings by at least a factor of 2. From the report: "Analysts predicted a five-year growth for the top 20% of companies to be 22.4% which turned out to be only 9.5%. [The researchers also pointed out the actual return rate should be lower because many companies actually failed over that period.]

They created sample portfolios based upon analysts' forecasts. Predictably, the top portion of the portfolios actually returned only about half of what the analysts predicted: 11% actual versus 22% predicted. "These results suggest that in general caution should be exercised before relying too heavily on long-term forecasts as estimates of expected growth in valuation studies."

Anecdotally though, it seems to me that the majority of estimates are pretty close. It's fairly rare that an earnings surprise is way out in left field. [Just look at the InvestorGuide reported vs. expected earnings for example.] In some cases it happens, so that would skew the averages. Five year estimates may be something different though. It's pretty hard to forecast accurately five years out. For example, a few years ago they were probably forecasting CSCO to earn 40% a year. Now it's more like 15%.



[9/5/08] Dreman studied and wrote a good deal about analysts and their predictive powers (or lack thereof). In his book "Contrarian Investment Strategies," he wrote: "There is only a 1 in 130 chance that the analysts' consensus forecast will be within 5 percent for any four consecutive quarters. . . . To put this in perspective, your odds are ten times greater of being the big winner of the New York State Lottery than of pinpointing earnings five years ahead."

... [Dreman] used it as a reason to invest in beaten-down stocks. For highflying stocks, a good earnings surprise doesn't help that much, because the stock is already riding on great expectations. A negative surprise, however, can send its price plummeting.

Beaten-down stocks, on the other hand, have such low expectations that a negative earnings surprise won't hurt them too much, while a positive earnings surprise can send them soaring. The message for Dreman: Since analysts are often wrong and earnings surprises are frequent, it makes sense to focus on beaten-down contrarian plays.