What is the probability that a criminal conviction is wrongful?
Evidence quality 4.63/5
Eight-dimension review score against the quality rubric . Each dimension scored 1–5.
- D1 Source grounding
- 5/5
- D2 Source authority
- 4/5
- D3 Arithmetic
- 5/5
- D4 Uncertainty
- 4/5
- D5 Scope
- 5/5
- D6 Prose
- 5/5
- D7 Perception honesty
- 4/5
- D8 Caveat completeness
- 5/5
Lifetime probability · lifetime, activity-specific
1 in 24
4.1% lifetime chance
range 1 in 36 to 1 in 17
● your factors — click this risk ▾ to reveal
≈ As likely as
Perceived
Wrongful conviction is one of those risks that generates intense concern when it surfaces in the news — an exoneration story, a DNA reversal, a podcast — and fades between headlines. There is no national poll tracking how often Americans worry about being wrongfully convicted, so the perceived risk is best characterized as an intuition shaped by media salience. Most people, when asked, guess the error rate is "very low" (well under 1%), which turns out to be optimistic relative to the best available estimate.
Rough estimate: Most guess well under 1%
Source: editorial intuition, not polled
Actual
~4.1% of death-sentenced defendants are innocent
US defendants sentenced to death, 1973–2004
Show derivation
The central estimate comes from Gross et al. (2014, PNAS), which used survival analysis on 7,482 death sentences imposed between 1973 and 2004. Of those, 117 resulted in exoneration by end-2004, and the model estimates that if all defendants remained under sentence of death indefinitely, at least 4.1% (95% CI: 2.8–5.2%) would be exonerated. This is a per-conviction probability, not a per-person lifetime probability, so scope is activity_specific_lifetime. The 4.1% figure is specific to capital cases, where post-conviction scrutiny is unusually intensive; for ordinary felonies, most researchers estimate 2–6%, but no comparably rigorous study exists. The normalized figure is used as-is because the unit of analysis is a conviction, not a person-year.
Caveats: The 4.1% figure applies specifically to capital cases, where the stakes drive ex…
The 4.1% figure applies specifically to capital cases, where the stakes drive extraordinary post-conviction review — multiple rounds of appeals, innocence-project involvement, and DNA testing that most felony defendants never receive. For ordinary felonies, the true wrongful conviction rate is almost certainly higher than the observed exoneration rate (since most wrongful convictions are never discovered) but may be lower than 4.1% if the capital-case selection process is unusually error-prone. Estimates for all felonies range from roughly 2% to 10%, depending on methodology and crime type. The rate also varies by offense category: wrongful convictions are disproportionately concentrated in homicide and sexual-assault cases, where eyewitness misidentification and forensic-science errors are most consequential. The figure should not be read as "4.1% of people in prison are innocent" — that is a related but distinct question with a different denominator.
Risks at similar odds
Other risks with roughly the same likelihood — useful for calibration.
What are the odds of a serious incident when hitchhiking?
What are the odds of a serious incident when hitchhiking?
Protest under autocracy
What are the odds of being physically harmed, arrested, or killed while participating in mass protests under an authoritarian regime?
Recently viewed on this device
Stored locally — clear anytime.
Pick challenger
The best available estimate of the wrongful conviction rate in the United States comes from a 2014 PNAS study by Gross, O’Brien, Hu, and Kennedy, which applied survival analysis to every death sentence imposed between 1973 and 2004 — 7,482 cases in all. Their conclusion: if all death-sentenced defendants remained under review indefinitely, at least 4.1% would eventually be exonerated (95% CI: 2.8–5.2%). That is roughly 1 in 24 capital convictions. For scale, if that rate held across the roughly 1.9 million people currently incarcerated in the US, it would imply on the order of 78,000 innocent people behind bars.
The number is both robust and narrow. It is robust because it uses formal survival modeling rather than raw exoneration counts, which understate the true rate by ignoring wrongful convictions that are never discovered. It is narrow because it applies only to capital cases, where defendants receive far more post-conviction scrutiny than ordinary felons. The National Registry of Exonerations, which has catalogued 3,698 cases since 1989, confirms that the machinery of exoneration is slow and selective: the average exoneree in 2024 spent 13.5 years in prison before being cleared. Most wrongful convictions never produce an exoneration at all.
Whether the 4.1% figure generalizes beyond death row is the central open question. Some researchers estimate 2–6% for all felonies; others argue the rate varies sharply by crime type, with sexual assault and homicide cases (where eyewitness misidentification and flawed forensic science are most prevalent) sitting well above property crimes. The figure here is best read as a well-supported lower bound on how often the system gets it wrong in its highest-stakes decisions — not as a universal error rate, and not as a number that any individual defendant can plug into a personal risk calculation.
Claim ledger
Every number below is what each source reported, with the verbatim quote we relied on and how we arrived at our figure. Click any link to verify directly.
-
[1] Proceedings of the National Academy of Sciences — Rate of false conviction of criminal defendants who are sentenced to death
Rate of false conviction of criminal defendants who are sentenced to death- Statistic
4.1% estimated false conviction rate among US death-sentenced defendants, 1973–2004 (95% CI: 2.8–5.2%)- Excerpt
“"We use survival analysis to model this effect, and estimate that if all death-sentenced defendants remained under sentence of death indefinitely, at least 4.1% would be exonerated. We conclude that this is a conservative estimate of the proportion of false conviction among death sentences in the United States." ”
- Source data from
- 2014-04-28
- Accessed
- 2026-04-12 · archived copy
- Calculation
- Gross et al. applied Kaplan-Meier survival analysis to 7,482 death sentences (1973–2004), identifying 117 exonerations. The 4.1% figure accounts for censoring (defendants removed from death row before exoneration could occur). The 95% CI of 2.8–5.2% is used directly as the uncertainty band, widened to 6% at the high end to account for the possibility that non-capital felonies have similar or higher base rates with less post-conviction scrutiny.
- Independence
- This is the only peer-reviewed study using formal survival analysis to estimate wrongful conviction rates. The National Registry of Exonerations provides independent case-level corroboration but uses a different methodology (cumulative count, not modeled rate).
-
[2] University of Michigan Law School / Michigan State University — National Registry of Exonerations — About
National Registry of Exonerations — About- Statistic
3,698 known exonerations in the US since 1989 (as of mid-2025); 147 exonerations recorded in 2024- Excerpt
“"The National Registry of Exonerations provides detailed information about every known exoneration in the United States since 1989. As of June 2025, the Registry lists 3,698 exonerations." ”
- Source data from
- 2025-06-28
- Accessed
- 2026-04-12 · archived copy
- Calculation
- The Registry is a case-level database, not a rate estimate. It records known exonerations but cannot estimate the denominator (total convictions) or the number of wrongful convictions that were never discovered. The 147 exonerations in 2024, with an average of 13.5 years of wrongful imprisonment each, illustrate the scale but do not yield a rate. Used here as corroborating evidence for the Gross et al. estimate, not as a competing figure.
- Independence
- The Registry is maintained by the University of Michigan and Michigan State, independently of the Gross et al. PNAS study, though Gross is a co-founder of both. The data pipelines are distinct: the PNAS study uses survival modeling on a closed 1973–2004 cohort, while the Registry is an ongoing, open-ended case collection.