Three months ago shares of online bookseller were trading above $70, and a survey of Wall Street analysts by the First Call research organization found 13 recommending the stock as a "strong buy" and 11 as a "buy." Only seven held less enthusiastic "hold" ratings, and not a single analyst recommended selling the stock. Since then, Amazon shares have fallen to around $35, meaning investors who followed those buy and hold recommendations lost half their money.

So the analysts must be jumping out the windows, or at least urging investors to sell and cut their losses, right? No. In the last week of June, Amazon still had 13 "strong buys," 11 "buys," seven "holds" — and still not one "sell" recommendation.

Perhaps the analysts all think Amazon is a bargain at today’s lower price. Or perhaps analysts aren’t the savvy stock pickers investors expect them to be. After all, Amazon isn’t a rarity. Missed price targets or off-the-mark earnings forecasts occur all the time. At the end of June, Unisys shares fell 38% in one day when the computer company forecast second-quarter earnings that were half what analysts had predicted. Late in 1999 analysts were tripping over each other to tout Internet stocks, leading investors to enormous losses as that entire sector collapsed this year.

It’s not that analysts are stupid. In part, it’s that they have a tough job. It’s hard to beat the competition in spotting promising stocks when so much corporate data is so widely available. New information is quickly "built into" a stock’s price. But there are some less benign factors in poor analyst performance as well. "A number of [academic research] papers have found strong conflicts of interest," says Wharton finance professor Franklin Allen.

What do analysts do to earn annual salaries that typically run in the six and seven figures at top brokerages and investment banks? Generally, an analyst is assigned to "cover" a handful of companies in a specific industry. He or she is expected to understand how the industry operates and how it adapts to changing economic and financial conditions. Zeroing in on individual companies, analysts look at sales and profit figures, outstanding debt and other fundamental data, and they evaluate company management. Good analysts have ready access to executives who can shed light on strategic thinking, top-level personnel changes and other internal factors that affect a stock’s price.

Big and small investors rely on analysts’ written reports to make decisions about buying and selling individual stocks. Analysts polish their reputations and earn raises, bonuses and partnerships by making accurate predictions and winning top rankings in an annual poll by Institutional Investor magazine.

Since only half of all stocks can perform better than the median stock, at least half of analysts recommendations should be to sell — which, for those who don’t already own the stock, means "don’t buy." Indeed, if analysts were effectively steering investors only to top-performing stocks, most recommendations would be to sell, and only a handful of stocks would be worthy of buy recommendations.

Yet research has long shown the opposite. Kent L. Womack, associate finance professor at Dartmouth College’s Amos Tuck School of Business Administration, found that in the early 1990s analysts made about six buy recommendations for every sell recommendation. Today, Womack says, the ratio has grown to around 50 to one. Perhaps it’s human nature for analysts to hope for good things rather than bad, or to become enamored of companies they follow. There also is an element of self-interest, as there have been cases of company executives freezing out analysts who issue critical reports, Womack says.

Many Wall Street firms have three major sources of income: Commissions and fees from executing customers’ trades, profits from trading securities in the firm’s own account, underwriting fees from helping companies sell new blocks of stock in initial public offerings. Analysts were once house intellectuals separated by a "Chinese Wall" from the rough-and-tumble of making money. Now an analyst may have conflicting roles in all three money making operations. The analyst who feels customers should avoid shares in Company X might be reluctant to issue a negative recommendation if his employer owned a large block of Company X shares.

"In the last 20 years, the role of the analyst has changed–at a pace that has been accelerating," says Paul Irvine, assistant finance professor at the Goizueta Business School at Emory University. "Now, instead of being kind of a cost center to the entire brokerage, the analyst is expected to contribute with revenue-generating products."

This includes, of course, the regular analysts’ reports offered to customers for a fee. More importantly, during the stock boom of the 1990s it has meant producing analyses that help the investment bank’s underwriting operation market new stocks through initial public offerings. Indeed, says Irvine, a company going public will pick an investment bank that can guarantee a positive report from a highly regarded analyst capable of influencing a stock’s price. A report that makes investors salivate over an IPO enables the underwriter to boost the offering price, raising more money for the client company. It would be difficult for an analyst to release a negative opinion on Company X if his employer was helping the company market a new block of shares. This concern can even infect recommendations on stocks in companies that are merely prospects for future business.

Securities and Exchange Commission Chairman Arthur Levitt has expressed concern about the imbalance between buy and sell orders, but the agency does not have the legal power to regulate analysts. The SEC does, however, set rules governing corporate practices. It is considering requiring that information disclosed to analysts be released to the public as well, eliminating the inside advantage companies give favored analysts. No longer needing to go easy on companies to preserve access, analysts would then in theory be more willing to issue sell recommendations.

Womack and Roni Michaely, associate finance professor at Cornell University, looked at these issues in a 1999 study called "Conflict of Interest and the Credibility of Underwriter Analyst Recommendations." They noted that analysts who help to attract underwriting to their firms may receive "bonuses that are two to four times those of analysts without underwriting contributions." In looking at 391 initial public offerings from 1990 and 1991, they found that analysts consequently tended to be more optimistic about the prospects for stocks underwritten by their employers than they were about stocks underwritten by competitors.

In the first month in which underwriters are legally allowed to comment on their new IPO issues, analysts for underwriters made 50% more buy recommendations on their IPO stocks than did other analysts. On average, stocks recommended by underwriter analysts had fallen in price before the recommendations were made, suggesting some recommendations were "booster shots" to push stock prices up. In contrast, the average stock recommended by non-underwriting analysts was already rising in the month before the buy recommendations were made, suggesting these analysts were making more objective appraisals.

Were the underwriting analysts’ positive views due to better insights gleaned from special access? Apparently not. In the two years after the IPOs, stocks pushed by underwriting analysts returned 15.5 percentage points less than those recommended by non-underwriters.

Were the underwriting analysts just bad analysts? No. The analysts who made poor recommendations on stocks their firms underwrote had better records when analyzing other stocks. Poor recommendations by underwriting analysts were therefore caused by bias, not incompetence, the researchers concluded.

An analyst, of course, cannot continually give bad advice without destroying the reputation that makes him or her valuable to the firm. So analysts should be expected to do the best work they can when there are no major conflicts. But how good is the best work?

A recent study by Brad Barber of the Graduate School of Management at the University of California, Davis, and two colleagues, looked at 360,000 analyst recommendations from 269 brokerage houses between 1986 and 1996. They found that stocks receiving strong positive recommendations beat the average stock return by a healthy 4.13 percentage points a year, indicating that analysts are indeed capable of identifying good stocks.

Unfortunately, the study found an ordinary investor could never match these results, which required owning a large block of stocks and immediately buying or selling when new recommendations were issued. The result would be annual portfolio turnover of 400%, causing such enormous transaction costs, such as taxes and commissions, that all of the extra gains would be destroyed.

Although stock-price predictions can be faulty, many academics believe investors should nonetheless read reports by analysts who have good track records, mainly for the detailed discussion of the companies.

Buy and sell recommendations from underwriting analysts appear to be of little value, and even recommendations from more objective analysts should be scrutinized. Analysts have adopted lots of odd-sounding verbiage to survive in an industry that frowns on anything other than a buy recommendation. Sophisticated investors know that if an analyst uses terms like "strong buy", "buy", "accumulate", "hold", and "reduce", everything other than "strong buy" means "sell."