No one can say with certainty how many innocent people have been killed by the death penalty in the United States, but the best peer-reviewed estimate points to dozens. A 2014 study published in the Proceedings of the National Academy of Sciences concluded that at least 4.1 percent of everyone sentenced to death in this country is likely innocent, and the researchers called that figure conservative.1Proceedings of the National Academy of Sciences. Rate of False Conviction of Criminal Defendants Who Are Sentenced to Death Since 1977, the United States has executed more than 1,580 people.2Bureau of Justice Statistics. Capital Punishment, 2023 Statistical Tables Apply the 4.1 percent rate to that total and the math suggests roughly 65 innocent people have been put to death. Individual cases with strong post-execution evidence of innocence, discussed below, support the conclusion that the real number is not zero.
Where the 4.1 Percent Estimate Comes From
A team of legal scholars and statisticians produced the estimate using survival analysis, a method borrowed from medical research. They modeled what would happen if every person sentenced to death stayed on death row indefinitely, continuing to receive the intense legal scrutiny that death sentences attract. Their conclusion was that at least 4.1 percent would eventually be exonerated.1Proceedings of the National Academy of Sciences. Rate of False Conviction of Criminal Defendants Who Are Sentenced to Death
That number is a floor. When a death sentence is commuted to life in prison, the legal attention and resources drop sharply. Fewer lawyers take on the case pro bono. Fewer investigative journalists dig into the facts. Some innocent people whose death sentences were converted to life terms quietly remain behind bars, never cleared. The 4.1 percent counts only the false convictions that would be caught with sustained effort. The real rate is almost certainly higher.
Applying the Rate to Executions
The 4.1 percent figure was calculated across all death sentences, not just those that ended in execution. That caveat matters, but there is no reason to think executed individuals were less likely to be innocent than the death row population as a whole. If anything, an execution eliminates the very legal processes that would have caught the error. Multiplying 4.1 percent by more than 1,580 executions since 1977 yields an estimate of about 65 innocent people put to death.2Bureau of Justice Statistics. Capital Punishment, 2023 Statistical Tables
That is a statistical projection, not a body count. It says the system produces false capital convictions at a measurable rate; it does not name the individuals. Naming them is harder, and in most cases impossible.
Why the True Number Can’t Be Confirmed
Once an execution is carried out, the legal machinery for uncovering innocence largely shuts down. Courts generally do not issue rulings on behalf of people who are already dead. Habeas corpus, the primary tool for challenging wrongful imprisonment, exists to free the living, not to vindicate the deceased.3Legal Information Institute. Habeas Corpus No standardized judicial process exists to review a closed capital case for factual innocence after the sentence has been carried out. A family member or advocacy organization can investigate, but there is no court to petition for a formal declaration.
Every execution that kills an innocent person also destroys the legal standing needed to prove the mistake. Defense attorneys move on. Evidence gets discarded or degrades. Witnesses die. The cases where post-execution investigations have surfaced strong evidence of innocence required extraordinary effort from outside organizations, academic researchers, or journalists willing to spend years on a case that could no longer save anyone’s life. They are the exceptions to a system built to close the book.
Cases Where the Evidence Points to Innocence
Carlos DeLuna (Executed 1989)
Texas executed Carlos DeLuna on December 7, 1989, for the 1983 murder of convenience store clerk Wanda Lopez. From the moment of his arrest, DeLuna insisted that another man, Carlos Hernandez, had committed the killing. Prosecutors dismissed Hernandez as a “phantom” suspect DeLuna had invented.4Columbia Law School. Columbia Law School Investigation Uncovers New Evidence Suggesting Texas Executed Innocent Man
Years after the execution, Columbia Law School investigators led by Professor James Liebman conducted a detailed reinvestigation. They found that Carlos Hernandez was very real, a man with a documented history of knife attacks who had bragged about the Lopez killing to multiple people. Police had failed to pursue him despite DeLuna’s repeated identifications. The investigation concluded that Texas likely executed an innocent man.4Columbia Law School. Columbia Law School Investigation Uncovers New Evidence Suggesting Texas Executed Innocent Man
Cameron Todd Willingham (Executed 2004)
Cameron Todd Willingham was convicted of capital murder in 1992 for setting a fire that killed his three young children in Corsicana, Texas, on December 23, 1991. The prosecution’s case rested almost entirely on a fire investigator’s conclusion that the blaze was deliberately set. Willingham was executed on February 17, 2004.5Office of Justice Programs. Rising From the Ashes – What We Have Learned From the Cameron Todd Willingham Case
After the execution, multiple fire science experts reviewed the evidence and concluded the original investigation relied on outdated indicators that have since been discredited. A leading fire dynamics expert determined the fire showed none of the hallmarks of intentional ignition under modern forensic standards. The Texas Forensic Science Commission investigated and confirmed that the arson determination did not meet scientific criteria.6Texas Forensic Science Commission. Texas Forensic Science Commission Complaint 09-01 – In the Matter of Cameron Todd Willingham The state executed a man for a crime the best available science says never happened.
Larry Griffin (Executed 1995)
Missouri executed Larry Griffin on June 21, 1995, for a 1980 drive-by shooting that killed Quintin Moss. The conviction relied heavily on a single eyewitness identification. After the execution, a year-long investigation by the NAACP Legal Defense and Educational Fund uncovered evidence that Griffin was not involved. A man injured in the same shooting said Griffin was not present. The first officer on the scene gave a new account that contradicted the identifying witness’s trial testimony. The investigation identified three alternative suspects, all of whom were in prison for other murders at the time the findings were published. The local prosecutor reopened the investigation based on those findings.
Marcellus Williams (Executed 2024)
Missouri executed Marcellus Williams on September 24, 2024, for the 1998 murder of journalist Lisha Gayle in suburban St. Louis. His conviction rested on the testimony of two witnesses who received incentives for their cooperation. Their statements were inconsistent with the physical evidence, with each other, and with their own prior accounts. No scientific or eyewitness evidence placed Williams at the crime scene.
Post-conviction DNA testing in 2016 excluded Williams as the source of male DNA found on the handle of the murder weapon. Three separate DNA experts reached the same conclusion. Williams was also excluded as the source of bloody footprints at the scene and unknown hairs found near the victim’s body. In January 2024, the St. Louis County prosecutor filed a 63-page motion to vacate the conviction, arguing that the evidence no longer supported it. The Missouri Supreme Court denied the motion, and the execution went forward despite the prosecutor’s own opposition. Even the office that originally secured the conviction came to believe the wrong man had been charged.
What the System Does Catch: 202 Death Row Exonerations
Since 1973, at least 202 people sentenced to death in the United States have been fully exonerated.7Death Penalty Information Center. Innocence Each was once found guilty beyond a reasonable doubt, housed on death row, and scheduled to die, only for the system to later determine the conviction was wrong. Some spent decades in a high-security cell before the truth emerged. These are not procedural dismissals; an exoneration means the charges were dropped, the person was acquitted at retrial, or a government authority declared them factually innocent.8National Registry of Exonerations. Understanding the Registry – Section: Definition of Exoneration
These 202 cases are sometimes offered as evidence that the system works. The exoneration process, though, depends heavily on factors that have nothing to do with guilt or innocence: whether biological evidence was preserved, whether a pro bono legal team took interest, whether a journalism project or law school clinic decided to investigate. For every person freed, the harder question is how many others lacked those lucky breaks and never got out.
Why Wrongful Capital Convictions Happen
The cases above are not random accidents. They share recurring patterns that show up across the broader body of wrongful capital convictions.
Eyewitness misidentification remains one of the most common drivers. High-stress incidents impair memory in ways witnesses themselves do not recognize. Cross-racial identifications are especially unreliable, and suggestive lineup procedures compound the problem. False confessions add another layer of risk, particularly when the suspect has an intellectual disability or is young enough to be overwhelmed by interrogation. Once a confession exists, juries tend to treat it as near-conclusive proof of guilt.
Incentivized witnesses drive many cases. Jailhouse informants receive reduced sentences, transferred housing, or other benefits in exchange for claiming the defendant confessed to them. Unless the defense can expose the underlying deal, juries have no way to weigh the testimony accurately. The Marcellus Williams case is a textbook example.
Forensic methods have also failed in capital cases. A 2009 report by the National Research Council found that, with the exception of nuclear DNA analysis, no forensic method has been rigorously shown to consistently connect evidence to a specific individual with a high degree of certainty.9National Academies. Badly Fragmented Forensic Science System Needs Overhaul Bite-mark comparison and microscopic hair analysis, once presented as definitive, have been largely discredited. The Willingham case turned on fire-origin analysis later experts called scientifically baseless. Ledell Lee, executed in Arkansas in 2017, was convicted partly on hair evidence using a method the broader forensic community has since abandoned.
Capital defendants who cannot afford private attorneys depend on court-appointed lawyers who may lack the experience, staffing, or funding to mount a serious defense. Under Strickland v. Washington, a defendant must show both that the attorney’s work fell below an objective standard of reasonableness and that the outcome would probably have been different with competent representation.10Justia. Strickland v. Washington, 466 U.S. 668 Courts give wide deference to strategic choices, so even glaring failures often survive appeal. And when prosecutors withhold favorable evidence in violation of Brady v. Maryland, the defendant goes to trial with an incomplete picture of the case against them.11Justia. Brady v. Maryland, 373 U.S. 83 Brady violations surface with regularity in post-conviction reviews, and many are never discovered at all.
What the Numbers Add Up To
The Innocence Protection Act, enacted as part of the Justice for All Act of 2004, created a federal right for convicted individuals to request post-conviction DNA testing and imposed obligations on the government to preserve biological evidence.12Congress.gov. S.486 – Innocence Protection Act of 2002 Those protections work only for the living. Once an execution is carried out, the window for testing closes and the legal incentive to preserve evidence disappears.
The 202 exonerations represent the cases where the system caught its own mistakes in time. The 4.1 percent estimate, applied to more than 1,580 executions, points to roughly 65 people put to death whose convictions would not have survived the same scrutiny. Four of those cases have names and documented evidence attached. The rest remain a statistical shadow the legal system is structurally unable to bring into focus.