At a time when finance executives have become adept at managing Wall Street expectations, a new hurdle has entered the realm of investor relations: whisper numbers.

And unlike the estimates of a cadre of analysts, whisper earnings expectations have a secret life of their own. “With the whisper-number culture, you are expected to exceed expectations,” says one CFO, who prefers to remain anonymous. “But I can’t manage what I can’t see. You end up chasing your tail. The only way you can possibly obtain market satisfaction is to publish infinite earnings. We can’t meet that consensus. We’re going to come in below infinity.”

As unpredictable and unreasonable as they are, whisper numbers seem increasingly ingrained in the corporate-earnings scene. As the unofficial, off-the-record forecasts of what a company will earn in a quarter, traditionally whisper numbers have referred to the revised expectations certain Wall Street analysts share with favored clients but don’t put in their widely disseminated research reports. “The whisper number is what you think will really happen,” says one analyst, who requested anonymity. “I’ll tell investors, but I won’t change my official estimate.”

More recently, individual investors have been getting into the game with earnings estimates posted on the Internet that are now being aggregated by Web sites, such as WhisperNumber.com and EarningsWhispers.com. Although this companion set of predictions is typically more optimistic than the analyst consensus published by First Call, IBES International Inc., and Zacks Investment Research, it is often more accurate and can serve as a reasonable predictor of stock movement.

During the April earnings season, for instance, Wall Street had Yahoo Inc. making 9 cents a share, which it beat by a penny. But the 19 “whispers” compiled by WhisperNumber.com had the Internet portal company at 10 cents, and Yahoo shares fell 18 percent over the next five days. In contrast, Sun Microsystems Inc. saw its stock price soar 13 percent after it posted earnings of 26 cents a share; it beat both the whisper number and the analyst consensus, though the individual investors’ forecast of 24 cents was closer to the actual than the analysts’ 23 cents.

“We want to tap into what online investors are thinking and feeling, because we believe they are starting to define market sentiment,” says Paul Hauck, co-founder of WhisperNumber.com, who charges that the First Call consensus has lost credibility as an unbiased forecast because of the earnings-management game that analysts and executives play. “We want the market to view the whisper numbers we publish as a valid number, not just some figure that comes out of the blue.”

Critics, however, contend that such sites offer nothing more than an unscientific poll of self-selected respondents who don’t put their names on their estimates or reveal how they came up with their numbers. “If you think of these whisper numbers as a poll of market watchers, and use it as a sentiment indicator, fair enough; it’s an interesting number,” says Scott Rosen, director of research at IBES, a market-data collection firm in New York. “To consider them a true predictor of anything is unwise.”

Speak in Low Voices

On a recent visit to WhisperNumber.com, the April earnings season is going full bore, and the din of CNBC on a nearby television fills the room, part of a one-story strip of offices on a rural road in Central New Jersey. By 11 a.m., a computer has scoured some 1,600 messages on four Web sites — Silicon Investor, Yahoo: Finance, Motley Fool, and Raging Bull — and saved 155 after searching for such words as “earnings,” “EPS” and “my estimate is.” (On some days, it may scan as many as 300,000 postings.) Several data “miners” sit at terminals reading through all the saved messages to determine which have whisper numbers that can then be added to the Web site’s database.

Increasingly, the estimates WhisperNumber.com compiles are submitted directly on the Web site, which has about 20,000 unique visitors daily. As of January, making your entry requires registering your name, address, and E-mail address (12,000 have signed up since then), and the system permits only one whisper entry per company each quarter. The system checks Internet protocols, so those who have tried to manipulate the published whisper number by registering multiple times under different names have been caught, Hauck says.

What the system doesn’t do is validate the submissions or postings. Were these whispers the products of rigorous analysis, numbers that inflated the analyst consensus by a few cents, or just some loony guesses? Do the submissions come from short-sellers?

To WhisperNumber.com’s Hauck, that shouldn’t matter. A quartile analysis of all the numbers compiled on a company is used to set a possible range of legitimate whispers. Those that are way off base are thrown out of the sample, and the average of the rest is the “official” whisper that’s posted on the Web site.

In the April quarter, Advanced Micro Devices Inc. had the most whisper estimates, 72, and the site’s consensus had the chipmaker earning 64 cents a share. That was 18 cents higher than the analyst consensus, though both were way below the blow-out actual of $1.15. Hauck boasts that the investor estimate could have been even closer, because there were many whispers on the high side that were deemed outliers by the site’s analytic approach. “If we had broken our rules, we would have had a whisper in the high 80s or low 90s,” he says.

On the Money?

Hauck may have a point about the relative accuracy of whisper numbers. In 1999, Purdue University professor Susan Watts, along with two co-authors, found that the First Call consensus tended to be more pessimistic and less accurate — about 6.1 percent below reported earnings per share, whereas the postings were 4.9 percent above the actual results. “The whispers tended to be too optimistic, but closer in absolute terms,” Watts explains.

Hauck confirms that the whispers on his site show a similar bias on the high side, but contends that the merit of whisper numbers is not in how well they predict actual earnings. “Stock movement is what will deem us a valuable indicator to investors,” he says.

Overall, the stock rises for a company that beats the whisper number and falls for a company that misses more than 60 percent of the time, according to statistics compiled by WhisperNumber.com from the whispers published on the site. In sectors such as biotechnology, the correlation is closer to 100 percent. Also, in the finance, Internet, and technology sectors, a stock will fall more than 80 percent of the time over a five-day period when earnings beat the analyst consensus but fall short of the whisper number.

Scott Rosen of IBES finds such data less than convincing in a historic bull market, where companies have tended to surprise on the upside. “The question is, Is there any value in the whisper number?” he asks. “Or is it simply made up of people who are more optimistic than analysts, and because the market is generally doing well, their predictions become a self-fulfilling prophecy?”

Given the recent market volatility, it will be interesting to see how well whisper numbers can predict stock movement during the upcoming earnings season this month. As for CFOs who may feel they have no way of managing the mass of online investors who might submit a whisper number, perhaps what is needed is a further democratization of disclosures.

“To manage the process, companies need to be more open,” says WhisperNumber.com co-founder John Scherr. “They fear they have no way of managing the whisper number, but all they have to do is step up their disclosures to the public via the Internet and market them better. I can envision a company with an accounting system that can report real-time, live on the Internet, their key performance indicators. What would that do for our business?”