There are some classes of cases with fee shifting and some without it.
The big disadvantages of the fee shifting where it is present are:
- In a significant subset of cases, particularly in tort law and in cases involving novel legal questions, it is very hard to predict in advance, even with the best available information, whether someone will win or not, and loser pays is a serious discouragement to people who in good faith think they have a claim.
This is a particular problem in U.S. litigation where decisions on the merits are deferred until late in the process and the judge does not announce resolution of particular factual or legal issues not sufficient by themselves to throw out a claim early on in the case. I have had cases in U.S. federal courts litigated for a year with full discovery and settlement before an answer to the initial complaint is filed because it can take a very long time for a court to rule on an initial motion to dismiss.
The American rule is one factor among many motivating people to settle cases before trial because litigation costs that can be avoided could not have been shifted to the other party even in the event of a win. It makes the size of the pie for a "winner" greater when both parties settle and the savings from avoiding litigation costs on both sides often psychologically "finances" the settlement.
Fee shifting often leads the parties to incur costs disproportionate to the amount at stake on the merits of the dispute. The underlying dispute may be just $2,000, but if each side runs up $40K litigating it and there is fee shifting, the resolution of the merits has $82K at stake.
Contrary to theory, in practice, almost nobody during litigation deliberately behaves in a manner designed to reduce litigation costs to the other side to reduce exposure to a fee award. Everybody in litigation is blinded by the belief that they will win.
Of course, the downside to the American rule is that it encourages presentation of costly defenses regardless of their pertinence to the merits of the underlying case or likelihood of success (so long as not frivolous or groundless) to defer judgment and to create an economic incentive to settle at a discount. This means that someone whose rights are violated when the American rule applies often is not made whole.
In some areas of law (like personal injury) there are work arounds to solve the problems of the American rule. The plaintiff has a lawyer who sues on a contingent basis and pain and suffering awards end up covering legal fees leaving considerably hard compensatory damages for the victim of a tort, and the prospect of non-payment to the lawyer discourages frivolous PI plaintiffs litigation. Defense costs in PI litigation are typically financed by insurance companies that bear the lion's share of the risk for paying the judgment.
Most contracts have an attorneys' fee clause (often one sided for the creditor but not the person sued in states where that is permitted) and this discourages frivolous defenses of debt collection suits and makes the creditor whole, but still creates a problem when there is a bona fide dispute over a narrow issue that can expose someone fighting slight overreach by a creditor to huge attorneys' fee exposure even if the defendant wins on the disputed issue.
In my professional experience, a one size fits all solution is not appropriate and Coase's law reflects that in the agreements reached by the parties to contact around the default rule.
Even where fee shifting is appropriate, the means by which the amount of fees due is determined and the way that the prevailing party is determined is not ideal. A German type rule would be better in many circumstances (and German legal fees are also much more regulated than in the U.S.).