Plea bargaining

Why does a system that guarantees everyone a trial settle almost every case without holding one?

A criminal justice system spends enormous care on the trial: the jury, the rules of evidence, the presumption of innocence, the right to confront witnesses. Then it uses that apparatus almost never. In United States federal courts in recent years, roughly ninety percent of defendants pleaded guilty, about eight percent had charges dismissed, and only around two in a hundred were tried.

The reflex explanation is congestion — too many cases, not enough courtrooms, so corners are cut. Hold that thought: it explains why trials are rationed, but not why the rationing takes this shape, and not why the parties agree. Nobody is forced to bargain. Both sides prefer it. Why?

Begin with what a trial is from inside the case. It is expensive for both sides in time, money and attention, and its outcome is uncertain — but two competent lawyers looking at the same file will form similar estimates of it: this evidence, this defendant, this jurisdiction, conviction perhaps four times in five, sentence in a certain range.

Now notice what follows. If both sides can estimate the trial's result, the outcome is available to them without holding it, while the cost is pure waste. So there is a surplus on the table equal to the cost both would avoid, and a range of settlements leaving both better off than fighting. That range is why cases settle, and it is the same reason most civil suits settle — bargaining in the shadow of the trial, as the phrase has it.

This reverses the usual complaint. The guaranteed trial is not going unused; it is the price list. Change the strength of the evidence and the offer moves; suppress a search and the offer moves. The right is exercised in every case, as a threat rather than an event.

Then ask what sets the discount. In an ordinary market price emerges between two parties of symmetric power. Here one party writes the menu: the prosecutor selects the charges, may stack several arising from one act, and in many systems those charges carry mandatory minimums that bind the judge. So the gap between the plea sentence and the trial sentence is not merely discovered — it can be manufactured to whatever size is needed. In Bordenkircher v. Hayes in 1978 the Supreme Court considered a prosecutor who warned a defendant that refusing a five-year offer would bring a recidivism charge carrying life, then filed it when he refused, and held this permissible bargaining rather than vindictiveness.

Now put the caseload back in, because it decides where the gap settles. Suppose the trial share rose from one in fifty to one in ten: the system would need roughly five times its present trial capacity, which it does not have and will not be given. The differential therefore cannot be small. It must be large enough to keep the plea rate at whatever the courtrooms can absorb — which makes that rate an equilibrium quantity, held in place by adjusting the penalty for insisting on the right. The system is not straining against its caseload; it has settled where the two balance.

And that is where the bargaining model starts to leak. If pleas simply tracked the expected verdict, the innocent would rarely plead, since their expected verdict is acquittal. But a detained defendant offered time served walks out today by pleading and waits months in a cell by refusing, whatever the evidence says; a defendant with an overloaded appointed lawyer cannot price the case at all; and people are risk-averse where a mistake means years. Stephanos Bibas's well-known argument is that these structural distortions — detention, resources, information, the gap between a lawyer's interests and a client's — move outcomes at least as much as case strength does. How much weight the shadow-of-trial model deserves against them is a live empirical dispute.

One historical note stops this looking inevitable. Plea bargaining is not ancient: it grew up in nineteenth-century American practice and was not squarely endorsed by the Supreme Court until 1970 and 1971, when it was called an essential component of the administration of justice. Several civil-law systems ran for a long time with nothing equivalent, and Germany's courts practised negotiated judgments informally for years before a 2009 statute regularised them. The practice follows caseload and sentencing structure, not the nature of criminal law.

Three refinements hold this together.

First, congestion and bargaining are not rival explanations. Bargaining explains why any given pair of parties settles; capacity explains where the discount must sit for the aggregate to clear. Remove the caseload pressure and pleas would not vanish — they would get cheaper for defendants.

Second, the trial penalty and the plea discount are one number seen from two directions, and which name you use is close to the whole controversy. Defendants convicted at trial do receive substantially longer sentences on average, though that raw comparison is confounded, since cases that reach trial are not a random sample.

Third, this is not a story about corrupt actors. Every participant may behave impeccably and the pattern still appears, because it is generated by the arithmetic of costs and capacity rather than anyone's intent — which is why urging prosecutors to bargain less does nothing, while changing mandatory minimums, pretrial detention or defence funding moves the equilibrium at once.

• What a system looks like when it does not permit bargaining — and what it does with the caseload instead.
• Why an innocent defendant might rationally plead guilty, and what safeguards attempt to catch that.

Every term the collection defines is gathered in the glossary.