Sure, II's Sentiment Indicator Works. So Does Flipping a Coin.

For people who follow investor sentiment, there are few things as pretty as this:

It is a chart of the S&P 500 against the ratio of bears to bulls-plus-bears in the weekly Investors Intelligence poll of investment newsletter writers -- the most widely followed sentiment index. It shows that when the market hit its July top, there were not too many bears out there. This, for sentiment watchers, flashes danger. It means that the market is ignoring potential pitfalls and, conversely, that it has priced in just about any bad news coming down the pike.

As the market declined, more and more of the writers got bearish. On Sept. 16 bearish sentiment peaked. The S&P 500 was at 1037. It was not the best time to buy -- better to have waited until Oct. 8. But in retrospect it was by no means a bad time to buy.

The current thinking on using sentiment is not to wait for it to reach some set level, but to watch when it hits extremes and then wait for it to turn (as the market keeps climbing, people can keep getting more bullish, after all). This makes right now a particularly worrisome time, first because bearish sentiment is very low, and second because it looks like negative feeling troughed on Feb. 10. Take a look at the II poll, and it seems like now is a good time to get out of the market.

The only problem is that the II poll isn't a very good indicator at all.

This, according to Santa Clara University finance professor Meir Statman. Statman first took a look at the II poll in 1988. In a paper -- "How Useful is the Sentiment Index?" -- published that year, he and colleague Michael Solt looked at the poll from 1963, when it started, to 1985. They examined the idea that the Dow Jones Industrial Average would rise when bearish sentiment (the ratio of bears to bulls-plus-bears) rose above 0.52 and that the Dow would fall when it dropped below 0.29. But they found that the "difference between actual realizations and expected realizations does not differ by a statistically significant amount from what would be expected if chance governed the process."

(Note: TheStreet.com does a weekly, purely anecdotal sentiment poll of its eclectic readership, but it doesn't pretend to be scientific.)

Last spring Statman re-examined the II poll in an article written with Roger Clarke, chairman of Analytic TSA. "We extended the time period through 1995 and found that the passage of time did nothing to improve the forecasting ability of newsletter writers," they wrote. In other words, the II remains about as valuable as, say, flipping a coin.

Rodman Has Many Things, but Never a Hot Hand

The coin flip is a useful analogy because it brings up the idea of the gambler's fallacy -- the belief that if I've just thrown heads, my next throw will be tails. Or that if I carry a gun onto an airplane, that will lower the chances that someone else on the plane will be packing.

Related to the gambler's fallacy is the belief that something besides simple probability is happening when we see a run. The best example of this is the hot hand in basketball, something that just about everyone -- NBA players, Spike Lee, the guys at West Fourth Street, specialists on the New York Stock Exchange floor and this writer -- believes in, but statisticians do not.

How else do you explain the game in which Dennis Rodman (whom coach Chuck Daly called an inspiration to kids everywhere who can't shoot), a 52% career shooter, had against the Denver Nuggets in 1991, when he hit 15 of 21, scoring 34 points? Yet statisticians have shown again and again that a player is no more likely to hit a shot following a made shot or a miss.

What we have in our chart might be the stock market equivalent of Rodman's big game against the Nuggets. Statman thinks that the continued belief in the II poll's usefulness (more than 10 years after he demonstrated that it wasn't all that good) is because we're selective about what we remember. We remember the 1991 Nuggets game. We do not remember all the times Rodman hit 2 shots and then choked for the rest of the night. (Though statistically speaking he may not have been "choking.") We remember when the sentiment index worked. We do not remember when it didn't.

"There is a great tendency for people to be anecdotal in their claims," said Statman. "To me it sounds like, 'Remember it was Friday when the market was down, and therefore the market is bad on Fridays.' [People] just pick a few examples that work their way. Whatever works their way, they call evidence. Whatever does not, they call it a fluke."

This is, of course, nothing like good science. "When you do medical research, people do double-blind tests," said Statman. "And here, people do seat-of-the-pants hypothesis!"

An Indicator That Works! Usually.

It seems a shame to give up entirely on sentiment: The idea that everyone piles into the bullish camp as the market tops and that everyone goes bearish as it bottoms is a hard one to give up.

Not to worry. Statman has recently found that while newsletter writers are "merely mediocre" when it comes to calling the market, Wall Street strategists and individual investors are genuinely horrible. In a forthcoming paper, written with Kenneth Fisher of Fisher Investments, Statman found there is a genuine negative relationship between sentiment among those groups and where the market's heading.

Perhaps what's most surprising about that is the poor performance of the strategists. Yet Statman found that the "Sell-Side Indicator" developed by Merrill Lynch quantitative strategist Rich Bernstein, which uses strategists' asset allocations to measure sentiment, genuinely works.

Though that isn't necessarily good news. Starting in late September, Bernstein's indicator showed that strategists had hit the bullish levels that for him are a sell signal. "That's clearly not the best call the model has ever made," said Bernstein -- indeed, the thing pretty much called the bottom. But there's a problem: Going back to 1985, the only time the indicator gave a sell signal and the market didn't follow was in early 1987.

But Statman points out that just because Bernstein's indicator works well, it doesn't mean it's always right. "When people are looking for market timing, they're looking for something perfect," he said. But there is no such indicator. Instead, you are merely playing with the odds. With a good indicator, you may find you shoot more like Larry Bird than like Dennis Rodman. You will not shoot 100%. Counting cards, you can win more games of blackjack. You will not win all of them.

"If you have 55-to-45 odds, these are wonderful odds," said Statman. "But you will be wrong very many times."

For investors who think that they have something better, perhaps it is best to listen to Statman's advice to Bernstein on the recent failure of the Sell Side Indicator: "I would say to him, 'Richard, take heart. Keep your expectations low, and you won't be disappointed.'"