As the U.S. District Court in Manhattan noted in a 1986 opinion in a securities case, in the law business it is a familiar axiom that "a bad settlement is almost always better than a good trial." There's less grief, less expense and less chance of being saddled with a crippling judgment.

But another key reason that at least 19 out of 20 civil cases end in settlement is the belief among lawyers that they can guess which party a jury will side with and how much will be awarded in damages.

New research from an associate professor at Stanford Law School has shattered those assumptions. If Janet M. Cooper is right, the merits of a case have almost nothing to do with the amount of the settlement.

Her study looks at securities fraud litigation, but she suspects that the same results could be found in private antitrust suits, in attempts to stop a tender offer, in stockholder actions to prevent a merger and in other business matters.

What Cooper did is take the opportunity to examine a bunch of unrelated but similar cases: allegations that computer industry concerns that went public in the new-issue market during the first half of 1983 did not adequately warn buyers of risks. Nine such companies whose stock price plunged were sued.

"To a remarkable degree, the sample cases were a small world for legal representation," Cooper wrote in a paper published by Stanford's Olin Program in Law and Economics. The law firm of Milberg, Weiss, Bershad, Specthrie & Lerach was involved in all nine cases; another firm was active in seven of the nine and a third firm in six.

In Cooper's analysis, the claim that stock buyers were duped was significantly weaker in some cases than others. Nonetheless, there was an undeniable pattern of the cases being settled according to the same formula, regardless of how good a case the plaintiffs had. In five of the eight cases that had been settled when Cooper did her study, the company agreed to pay out about one-quarter of the total loss suffered by stock buyers between the initial public offering and the filing of the suit.

If it isn't the relative strength of the plaintiffs' and defendants' cases, what is the key element in deciding a settlement amount?

Here Cooper moves from statistics to speculation, but her answer makes a lot of sense. "First, the incentives, for all participants, to avoid a trial are unusually strong -- so strong that it is almost unthinkable that there will be a trial," she said.

Typically, there will be a large number of defendants. In addition to the corporation issuing the securities, the individual officers and directors, the underwriter of the issue, the lawyers and, often, the accountants who prepared the financial statements all will be named as defendants. It's tough to hold those diverse groups together in a trial, where there may be a goad for each class of defendant to point the finger at the others. They are afraid of the possibility of astronomically high damage awards that could put the company out of business. And liability insurance policies often will pay for a settlement but not for damages imposed by a court.

From the plaintiffs' table, a trial looks no more appealing. The lawyer is working on a contingency arrangement, and if the trial exonerates the company, finding that the offering statement contained no substantial misstatements, the lawyer walks away with nothing -- not even compensation for out-of-pocket expenses. Even though different judges use different formulas to figure out how much the lawyers should get when they win on behalf of stockholders, the take tends to be between 20 percent and 30 percent of the total award. So, Cooper argued, putting in all the extra time required by a trial may actually diminish the per-hour earnings of the law firms involved.

In this sort of case, Cooper concluded, separating the settlement from the underlying merit of the claims means that stock buyers with no legitimate gripe sometimes get a windfall, and that those who were really misled may be undercompensated. But fixing the system, she warned, has the potential of doing harm in other kinds of cases where settlements really are fair.