NEW YORK (MONEY Magazine) - If you think all analysts are charlatans and their research isn't worth the paper it's printed on, tell that to the AOL Time Warner shareholders who heeded the warnings of Merrill Lynch media analyst Jessica Reif Cohen last fall.

On the morning of Oct. 17, Cohen and Henry Blodget, her much maligned Merrill Lynch colleague, downgraded AOL to neutral from buy, citing AOL's mounting debt load and softening core Internet business. Cohen's biggest concern: that slower online growth at AOL would erase the premium that investors had been willing to pay.

Her worries were soon borne out. Over the next two quarters, not only did the company's long-term debt balloon 37 percent to $28 billion, but its Internet subscription revenue grew at a mere 9 percent clip. Investors reacted just as Cohen had anticipated, driving down AOL's stock from $33 to below $20. (see more on Cohen's latest opinion on AOL)

There's been much hue and cry in recent months about the miserable state of Wall Street research. Analysts' failure to foresee the tech wreck or warn investors about the implosions of Enron and Global Crossing has given credence to longstanding criticism that sell-side analysts are captives of their investment banking departments.

Also in this series       How to use Wall Street research Analysts: Just get rid of them

What the Merrill downgrade of AOL shows, however, is that Wall Street research can add value even when analysts serve two masters. Indeed, if any Wall Street firm were going to be beholden to AOL, it would've been Merrill Lynch. Its investment bankers had advised AOL on the acquisition of Time Warner two years earlier, pocketing tens of millions in merger fees in the process. Yet that didn't stop Cohen from raising red flags about AOL (which is the parent of this Web site, and publisher of MONEY magazine).

Moreover, the evidence that research still matters isn't just anecdotal. A study prepared for MONEY by Zacks Investment Research shows that investors who read between the lines and don't take buy recommendations at face value can profit greatly from analysts' work.

Zacks found that from August 1991 through March 2002, an investment strategy based on buying stocks with the most analyst upgrades (as opposed to those with the most buy ratings) would have topped the total return of Standard & Poor's 500-stock index by eight percentage points a year. On a $10,000 investment, that eight-point gap over the 11 years would have meant the difference between $42,000 and $89,000.

There's little doubt that the mad rush for investment banking fees crushed the objectivity of Wall Street research. To win lucrative underwriting and merger-and-acquisition assignments, many firms appeased corporate clients by toning down negative research and playing up positive news. "There was way too much stenography going on," says Chuck Hill, research director at First Call.

But even before analysts became de facto investment bankers, research wasn't immune from outside influence. Analysts have long faced pressure from institutional investors who complain when their holdings get downgraded.

How do you use Wall Street analysis?   For earnings estimates   For recommendations (buy, sell, hold)   For background information   All of the above   None of the above View results

Sallie Krawcheck, chief executive of Sanford C. Bernstein (a research house not affiliated with an investment bank), remembers a flood of angry phone calls after Bernstein analyst Paul Sagawa downgraded Cisco and Nortel in September 2000. (Sagawa was the first prominent analyst to predict the crash in telecom-equipment spending.)

Says Krawcheck: "Downgrades like that one take particular courage, not just because it upsets company management -- which we are always okay with doing -- but because it's telling clients things they don't necessarily want to hear."

For analysts at investment banks, the cost of courage grew exponentially in the 1990s. Research used to be jointly paid for by investment banking and brokerage. Over the past decade, however, brokerage commissions have fallen so dramatically that the onus of paying for research has fallen disproportionately upon investment banking.

This alone would not have compromised research had it not occurred at a time when investment banking was becoming much more competitive. Foreign banks like UBS and regional firms like Robertson Stephens began using research to woo investment banking clients who believed high-profile analysts could make or break their stocks.

As a result, the pressure to pander to corporate clients became immense. "I spent 18 years as a sell-side analyst in the 1970s and '80s," says Hill, "and back in those days I put sells -- not holds that mean sell -- on investment banking clients and I didn't hear anything from investment bankers or the companies. You do that today, you get threatened by investment banking, and you risk being cut off from the company communication-wise."

A low point seemed to come in June 2001 when Qwest Communications CEO Joseph Nacchio publicly accused Morgan Stanley analysts of raising questions about Qwest's accounting only because Morgan had investment banking relationships with Qwest's competitors.

Wall Street is now paying the price for all this in the form of plummeting credibility, evaporating retail trading volumes and the threat of civil -- or even criminal -- liability.

More on Merrill Lynch       Merrill settles charges Merril alters ratings

Earlier this year, New York State Attorney General Eliot Spitzer accused Merrill of defrauding investors and released subpoenaed e-mail messages in which Blodget, Merrill's former Internet analyst, seemed to be privately denigrating stocks he had publicly rated a buy. Merrill reached a settlement with Spitzer's office in late May, in which the company acknowledged no wrongdoing but agreed to pay a $100 million fine and change the way analysts are compensated.

Further change is in the air. Discount broker Charles Schwab recently jumped into the research game, offering A through F letter grades for the 3,000 largest U.S. stocks. Full-service firms like Morgan Stanley and Prudential have adopted more transparent rating systems. Morgan Stanley now rates 22 percent of its stocks underweight (vs. overweight or market-weight).

Having this many negative ratings would've been unthinkable two years ago, when only 1 percent of the ratings put out by Morgan and its peers carried sell, underperform or any other negative tag. "It's a crisis having the public believing they cannot trust Wall Street research," says Dennis Shea, Morgan's director of global equity research, of the impetus for change.

Analysts themselves are getting back to basics, digging further into the numbers behind the numbers found in annual reports and other corporate filings. "It's amazing how fast the invisible hand of competitive marketplaces is correcting these problems," says Oakmark manager Bill Nygren. "Two years ago you saw almost no new research come out after the annual 10-K filing. Now when a company comes out with its 10-K, you see half a dozen analysts immediately coming out with a report analyzing all the details."

This year, for example, eight different analysts who follow General Electric published research delving into the minutiae contained in GE's annual report. That's up from zero in 2001, when analysts were content to rely on the less substantive earnings press release. "Everybody now is digging a little deeper," notes Joseph Kalinowski, senior equity strategist at Thomson Financial.

Another sign of change comes from the initial public offering market, where JetBlue received low ratings from its own underwriters after surging 67 percent its first day of trading. The analyst for lead underwriter Morgan Stanley initiated coverage with a tepid "market-weight." Even more surprising was the "reduce" JetBlue received from UBS Warburg, a co-manager on the IPO.

Analyst Samuel Buttrick issued a 12-month target of $38 for the start-up airline -- a price 30 percent below where JetBlue was trading at the time. Why the sudden burst of analyst straight talk? "No analyst wants to become the next Henry Blodget," says Zacks director of research Mitch Zacks.

Wall Street analysts aren't the only ones who need to change their ways. Investors do too. Individuals, in particular, need to get a lot smarter about how they use equity research. They should take a cue from the pros and pay less attention to recommendations and more to what analysts are saying in the body of their reports.

"When you're reading a research report or talking to your broker, the question you should be asking is not, Did the analyst say buy it?" says Firsthand Funds' Kenneth Pearlman, himself a former sell-side analyst with CIBC Oppenheimer. "What you should be asking is: How much is the company growing? Who are their customers? What's the valuation? If you stop reading once you have gotten to the recommendation, you almost deserve what you get."

Much of the criticism being leveled at analysts lately deals with these buy and sell recommendations. According to Thomson Financial, two-thirds of all stocks still have buy ratings and only 2.5 percent are rated sell. (The increase from last year's 1 percent is attributable almost entirely to Morgan Stanley.)

Politicians such as Sen. Joseph Lieberman have cast these numbers as proof positive that Wall Street is misleading investors. "One of the most stunning facts I have learned is that...no matter what the market does, analysts seem to just keep saying buy" said Lieberman at a committee hearing in February.

Pardon our cynicism, but we have a hard time believing that most investors -- or most senators, for that matter -- were unaware that ratings inflation is as persistent a problem on Wall Street as grade inflation is at Harvard.

The American public is not dumb. Consumers understand that tall coffees aren't always tall and that large eggs aren't always large. And if they'd been paying attention to the financial media, they'd have known that buy-rated stocks aren't always buys. Any investor who didn't know this had no business dabbling in individual stocks in the first place.

Even if some investors failed to see how pervasive ratings inflation had become, there's still no excuse for buying a stock based solely on a one-word recommendation. Institutional investors have never used research this way, because they know that the ratings and earnings estimates are frequently wrong.

"Analysts have always been overly optimistic," says fund manager David Dreman. His own research shows that from 1998 through 2001, analysts' earnings estimates were wrong by an average margin of error of 37 percent. This sounds awful, but it's actually better than the 41 percent average since 1973. And back in the early 1990s -- when research was still supposed to be objective -- the average earnings estimate was off by a whopping 57 percent.

Dreman, Nygren and the other pros we spoke with all report that they pay little or no attention to buy or sell talk. What they look for is detailed information about companies and industries. "Analysts are great at providing data and information that the average buy-side shop doesn't have the capacity to get on their own," says Angela Kohler, manager of Federated Large Cap Growth. "But no one on the buy side was ever under the illusion that these people were objective."

You hardly need an M.B.A. to read between the lines of most Wall Street research. Jessica Reif Cohen, for example, may only have downgraded AOL stock to neutral, rather than to sell, last October, but even the most cursory reading of her report would have shown her opinion to be anything but neutral. The opening page alone is peppered with language like "underlying concerns," "weaker than expected" and "not optimistic."

Indeed, what this indicates -- and what the Zacks study confirms -- is that when it comes to picking stocks, changes in ratings are far more telling than the ratings themselves. Zacks, which publishes rating-change information for most stocks at its Zacks.com website, analyzed the 3,000 largest U.S. stocks for MONEY and found that the 10 percent with the most upgrades returned 22 percent a year from August 1991 through March 2002.

That compares with 14 percent a year for the S&P and 5 percent for the 300 or so stocks with the most downgrades. Stocks with the highest absolute ratings also beat the market, but their margin of victory was much smaller -- two percentage points, an amount that could easily be consumed by trading costs.

Ratings changes are even better predictors of market-beating returns during bear markets -- something that's generally not true of buy ratings. Incredibly, stocks with the most buy ratings greatly underperformed both the market and their sell-rated peers in 2000 and 2001. In contrast, those stocks with the most upgrades bested the S&P by a wider margin during the recent bear market than they did during the broader 11-year period.

From March 2000 through March 2002, the most frequently upgraded stocks returned 6 percent vs. a 5 percent loss for the S&P and a 7 percent loss for the most downgraded stocks. What's clear from these numbers is that analysts are observing and reporting fundamental changes in companies' prospects before that information has been priced into the market.

Obviously it would be better if investors didn't need a decoder ring to get useful information. Yet one of the ironies of the political and regulatory efforts to reform the current system is that the proposed solutions could make the system even worse. Take the Securities and Exchange Commission's new rule requiring underwriters to wait 40 days before publishing research on companies that they take public. Given that underwriters are often the only firms that cover newly public companies, many IPO investors will be flying blind for the first 40 days.

Think about the implications for an IPO like JetBlue. The stock declined 17 percent from the morning of May 7 -- when the UBS reduce rating was issued -- to the end of trading on May 17. Had the SEC rule been in force, UBS' analyst would have been unable to write about the stock until the end of May.

Both the SEC and New York State Attorney General Spitzer have said they'd like to see research and investment banking kept separate. Sounds reasonable, but it flies in the face of Wall Street's economics. Topnotch research is expensive. Given how much broker commissions have declined since the 1970s, trading alone cannot support existing staffing levels in equity research without help from investment banking. Brad Hintz, a securities industry analyst with Bernstein, predicts that if Spitzer gets his way, Wall Street firms will respond by slashing their research budgets and limiting coverage to the large-cap stocks that generate the most trading volume. And Hintz used to be chief financial officer at Lehman Brothers, so he knows better than most where big brokerages will cut if research can't be mixed with investment banking.

In the end, the fate of analyst research lies with the investing public. Wall Street has a long history of giving clients what they want and telling them what they want to hear. That's what happened during the bull market, and that is what's happening right now. So long as investors continue to demand a full accounting of their favorite stocks -- warts and all -- the trend toward tougher, more objective research will continue. The only question is who's going to pay for it.