Field Notes

concept

Wrongful conviction

StatusStableUpdatedStands on3 sourcesSupportsnotes

A wrongful conviction is a conviction of a factually innocent person. It is the central fact about the subject that there is no systematic way to count them: a false conviction that is never discovered leaves no trace in any statistic, and discovered ones — exonerations — are a selected sample of the errors the correction system happened to catch. Correction channel develops that selection problem generally; this note holds the empirical content of the largest discovered sample, the US National Registry of Exonerations.

Estimating the invisible rate

The only credible population estimate comes from death sentencing, the one corner of the system where scrutiny is intense enough that a large fraction of errors eventually surfaces. Gross and colleagues used survival analysis on death-row exonerations and estimated that if all death-sentenced defendants remained under sentence of death indefinitely, at least 4.1 percent would be exonerated — a conservative floor for the false-conviction rate among US death sentences.1 The estimate does not generalize cleanly to lesser sentences: capital cases get the most review, but they may also be more error-prone (high-pressure prosecutions of the worst crimes), so the true rate for ordinary felony convictions is unknown and plausibly lower. It nonetheless bounds the claim that wrongful conviction is vanishingly rare: in the most scrutinized cases in the system, at least one conviction in twenty-four was false.

The registry’s contents

The NRE has recorded more than 3,200 exonerations since 1989. Its 2022 race report documented the distribution: Black Americans, 13.6 percent of the population, are 53 percent of exonerees; among murder exonerees, the conviction was about 7.5 times as likely to be of a Black defendant as population share would predict.2 Official misconduct was found in 72 percent of murder exonerations, and police misconduct specifically in 58 percent of Black exonerees’ cases versus 38 percent of white exonerees’ cases. Murder is the largest single exoneration category, with more than 1,160 recorded. These are counts of discovered errors, and Correction channel’s constraint applies directly: the registry’s racial and crime composition reflects where scrutiny is concentrated (death-eligible and high-sentence cases) as well as where errors occur, so the figures understate total error while skewing its visible shape. The misconduct findings connect to Coercive interrogation and false confessions, one of the best-documented mechanisms by which an innocent person comes to be convicted.

Registries beyond the United States

The European Registry of Exonerations is the closest counterpart: 144 exonerations across 20 European countries as of 2025, a count that measures Europe’s thin detection infrastructure rather than its error. National registries also exist for Canada and for the United Kingdom (Exeter). The comparative lesson of all of them is the one this note started from: every registry observes the errors its country’s Correction channel caught, and the four channel architectures produce visibility that differs by an order of magnitude for reasons unrelated to accuracy (Hidden miscarriage risk and correction-channel opacity across democracies).

What this means for offender statistics

Wrongful convictions are absorbed into offender statistics invisibly: a falsely convicted murderer is counted as a murderer, and nothing in the count flags the error. The discovered-error evidence is therefore the only check on the reliability of conviction-based measures of who commits crime. The check is reassuring in one direction — even a 4 percent error rate leaves the large racial differences in conviction and victimization statistics essentially intact, far too big to be an artifact of wrongful conviction — and unreassuring in another: the same evidence shows the error rate itself is racially unequal, so conviction statistics slightly overstate the relative offending share of the group that is also more often wrongly convicted. The honest reading is that offender statistics are broadly reliable about scale while carrying an unequal error margin that the statistics themselves cannot reveal.

  1. Samuel R. Gross, Barbara O’Brien, Chen Hu, and Edward H. Kennedy, “Rate of false conviction of criminal defendants who are sentenced to death,” Proceedings of the National Academy of Sciences 111, no. 20 (2014): 7230-7235, doi:10.1073/pnas.1306417111; saved copy. ↩

  2. National Registry of Exonerations, Race and Wrongful Convictions in the United States 2022 (University of Michigan Law School, 2022). The registry’s site blocked automated download; figures are from the report’s published summary tables. ↩

Built on 3 sources (3 external).

Working out connections…