How Sure Does a Court Have to Be? Different Thresholds for Different Stakes

Legal systems use several standards and the differences between them are substantive rather than rhetorical. The criminal standard requires that the tribunal be sure, expressed in common law systems as beyond reasonable doubt, which is deliberately demanding and is not quantified. The civil standard requires only that a claim be more likely true than not, which is the balance of probabilities and which corresponds to anything above an even chance. Intermediate standards exist in some systems for particular matters, expressed as clear and convincing evidence. Lower thresholds govern preliminary decisions, including the reasonable suspicion required to stop somebody and the reasonable grounds required to arrest or to search, which are lower because the consequences are lesser and temporary.

The asymmetry follows from how the two kinds of error compare:

• •Convicting an innocent person imposes punishment and stigma that cannot be undone
• •Acquitting a guilty person leaves a wrong unaddressed, which is serious and different in kind
• •The state brings the case with resources the defendant usually lacks
• •The principle that it is better for several guilty people to go free than for one innocent to be convicted is very old
• •A civil case is between parties of nominally equal standing, so neither error is privileged
• •Setting the civil standard above an even chance would favour defendants systematically

Attempts to express the criminal standard as a probability are resisted by courts in most systems and the reasons are worth taking seriously. Juries are generally directed not to attach a percentage, and appellate courts have disapproved of directions that do. The objections include that jurors do not reason numerically about narrative evidence, that a stated threshold would invite arithmetic on quantities that cannot be estimated, and that combining probabilities across elements of an offence is not straightforward. Surveys asking judges and jurors what number they have in mind produce answers scattered widely, typically between eighty and ninety five per cent with substantial variation, which is itself an argument on both sides, since the scatter suggests either that the standard is inconsistently applied or that the verbal formulation carries meaning a number would lose.

The standard is only half the arrangement, since somebody has to bear the burden of meeting it. In criminal cases the prosecution bears it throughout for the elements of the offence, which is what the presumption of innocence means in practice, and a defendant who raises a defence usually needs only to put it in issue with the prosecution then disproving it. Some defences reverse that, requiring the defendant to establish something on the civil standard, and such reversals are scrutinised against fair trial guarantees. In civil cases the claimant bears the burden for what they assert and the defendant for any defence. Shifting burdens appear in discrimination and regulatory law, where a claimant establishing a pattern shifts to the defendant the task of explaining it, which is a deliberate response to evidence being in the defendant's hands.

The threshold matters most in cases where the evidence is genuinely balanced, and several categories recur. Cases resting on the word of one person against another, with no independent evidence, frequently fail the criminal standard while the same allegation could succeed civilly, which is why the same conduct can produce an acquittal and a civil judgement against the same person, a divergence that is confusing in public and legally coherent. Scientific evidence expressed probabilistically must be translated into the verbal standard, which is where the errors described as the prosecutor's fallacy arise. Regulatory and disciplinary proceedings use the civil standard while imposing consequences that can end a career. And the standard governs only the finding of fact, not the interpretation of law, which is decided differently.

The criminal standard demands that the tribunal be sure and the civil standard only that a claim be more likely true than not, with the asymmetry reflecting that wrongful punishment cannot be undone. Courts resist expressing the criminal standard as a percentage, and surveys find people mean widely different numbers by it. The same conduct can produce an acquittal and a civil judgement coherently.