Legal precedent

When is it reasonable for a court to depart from a decision it is otherwise bound to follow?

A judge who thinks yesterday's decision was wrong is nevertheless expected to follow it — which sounds like an instruction to do injustice on purpose. Yet no common-law system treats precedent as absolute either; courts do change their minds, and we do not call that lawlessness. So the rule cannot be "always follow", and cannot be "follow when convinced", because either would be trivial. What is the third thing?

Start with what following precedent buys. Suppose every judge decided each case afresh on its merits. Could you take advice before signing a contract, or know whether your conduct was lawful before doing it? The value is not that old decisions are wiser than new ones — often they are not. It is that people arrange their lives against a rule they can read in advance.

But notice how much less binds than the phrase suggests. Not everything in the earlier judgment — only the ratio decidendi, the reasoning necessary to the result on those facts. The rest is obiter dicta, carrying whatever persuasive weight its quality earns. And binding runs two ways: vertically, a lower court is bound by the court above it in its hierarchy; horizontally, a court may be bound by its own past decisions, which is the harder case.

So what may a judge who dislikes a precedent do? The commonest move is not defiance. She asks whether the material facts really are the same, and if a genuine difference exists she distinguishes the case: the old rule survives untouched and does not reach here. What is a court doing when it distinguishes on a difference that looks thin? Narrowing a rule without announcing it — honest work when the difference matters, quiet evasion when it does not.

Only when distinguishing is unavailable does the harder question arise, and it is one of power before merit. A court cannot overrule a court above it, whatever it thinks. Whether it may overrule itself varies: the House of Lords treated itself as strictly bound until its Practice Statement of 1966, while the United States Supreme Court has never claimed to be strictly bound by its own past decisions — which is how Brown v. Board of Education in 1954 could depart from Plessy v. Ferguson of 1896.

Where the power exists, what makes using it reasonable rather than merely willing? Courts name a recognisable cluster: whether the earlier reasoning was poor even by its own lights, whether the rule has proved unworkable, whether the facts or the understanding beneath it have changed, whether it now sits at odds with the rest of the law — and, cutting the other way, how heavily people have relied on it. That last carries the real weight. Overruling a rule about property titles unsettles deeds signed decades ago; overruling one about what a police officer must say unsettles little going forward. There is also a longstanding argument, put sharply by Brandeis in a 1932 dissent, that precedent should bind less tightly in constitutional cases, since the legislature cannot correct a constitutional ruling it dislikes. How much any of this weighs is genuinely contested rather than settled — reliance especially — and that disagreement is usually the substance of the case.

The tempting misconception is that overruling is what a court does when it thinks a decision was wrong. "Wrongly decided" is roughly the entry condition for the question, not the answer to it — courts routinely follow decisions they say plainly they would not have made. The operative question is whether the costs of the mistake now exceed the costs of unsettling what was built on it. And a court bound by a precedent it dislikes still has a move short of defiance — follow it, and say at length why it deserves reconsideration by a court that can act.

• Why courts sometimes overrule prospectively, sparing arrangements already made under the old rule.
• How civil-law systems, which formally deny precedent binding force, achieve consistency anyway.

Every term the collection defines is gathered in the glossary.