A judge who believes an earlier ruling was wrongly decided faces an uncomfortable choice: follow the reasoning she rejects, or correct what she sees clearly as an error. Legal systems built on precedent overwhelmingly choose the first course, and defend the choice as though it were self-evidently correct. Yet stated as a bare principle, this looks strange: why should the mistaken judgment of a predecessor, however careful and well-intentioned, bind the judgment of someone who has since seen the flaw more clearly? To answer that a rule is a rule settles nothing, since the whole question is why this particular rule -- follow the last decision, even the wrong one -- deserves the deference it receives.

Consider two disputes, arising a decade apart, that turn on an identical question of contract interpretation. The first court resolves the ambiguity in favour of the buyer; the second court, presented with the same clause and the same fact pattern, privately doubts that resolution but applies it anyway, because the earlier ruling is now the settled reading of that clause across the jurisdiction. Had the second court decided the matter afresh, it might well have ruled the other way. It does not, and the reason is not that the earlier court was more expert, more careful, or closer to the truth. It is that an entire commercial practice has since arranged itself around the first ruling -- contracts drafted on the assumption it would hold, prices set on the assumption it would hold, disputes settled out of court on the assumption it would hold. To reverse the rule now would not merely correct one case; it would unsettle every arrangement quietly built on top of it.

This is the strongest case for precedent, and it has nothing to do with the earlier decision being right. What precedent protects is not the correctness of any particular ruling but the reliability of the legal landscape as a whole -- the assurance that a rule announced today will still govern tomorrow, so that a person can plan, contract, and act without recalculating the law each time a new judge takes the bench. A legal system that reopened every settled question whenever a court found the old answer unpersuasive would offer great accuracy at the level of any single case and near-total unpredictability at the level of the system, since no one could ever be confident that yesterday's rule would survive tomorrow's more careful judge.

None of this makes precedent unconditional. Courts distinguish an earlier ruling by finding a genuine factual difference the first court never confronted, which lets a new question be resolved without touching the old rule at all. Higher courts, and constitutional benches especially, do overturn their own settled rulings, though rarely, and typically only where the earlier reasoning has proved unworkable in practice, has been eroded by later rulings on related questions, or rests on a factual premise since shown to be false. The doctrine, properly understood, is not a command to repeat every past error forever; it is a strong but rebuttable presumption, one that places the burden of justification on the court that would depart from precedent rather than on the court that would follow it.

Seen this way, the value of precedent lies not in guaranteeing that every individual ruling is correct -- no system can promise that -- but in supplying a structure stable enough that disagreement, correction, and change can happen within it rather than against it. A court that never followed precedent would answer each question freshly and correctly more often, perhaps, yet would leave behind it a landscape no one could safely build on. A court bound too rigidly would entrench error indefinitely. Stare decisis, at its best, is neither of these; it is the deliberate, engineered compromise between a system that must sometimes be wrong and a public that must always be able to rely on it.