Coerced confessions are dangerous evidence. They can appear powerful to jurors even when they are unreliable, and they can distort the jury’s understanding of the entire case. For that reason, courts must take coerced confession evidence seriously.
But not every improperly admitted confession changes the outcome of a trial.
The case of Texas v. Bobby Ray Hopkins is useful because it tests whether empirical trial-error analysis can identify a serious constitutional error that was nevertheless harmless. Hopkins was convicted of murdering two women in Texas. The prosecution introduced evidence of an involuntary confession, but it also presented strong physical evidence linking Hopkins to the murders. The legal question was whether the improper confession evidence made the trial unfair.
My research in Measuring Fairness uses the Hopkins case to test whether the method can correctly classify a coerced-confession error that did not cause substantial harm.
Measuring a Serious Error That Did Not Change the Result
The study compared two versions of the case: the actual trial condition, which included the challenged confession evidence, and a hypothetical fair trial condition, which excluded it. Respondents reviewed one version of the case and indicated whether they favored a guilty verdict. The analysis then translated those individual-level verdict preferences into estimated jury-level conviction probabilities under Texas jury rules.
The result was clear. The estimated probability of conviction was extremely high in both conditions. The study estimated the probability of conviction at .976 with the confession evidence and .985 without it. In other words, Hopkins was almost certain to be convicted whether or not the confession evidence was included. The estimated effect of the coerced confession was a decrease in the probability of conviction of about .010, with a small margin of error.
That result may seem counterintuitive. Coerced confession evidence usually helps the prosecution. But in Hopkins, the physical evidence was already strong. Adding a questionable confession may have done little to help the state’s case and may even have raised doubts about the investigation. As one respondent observed, if the police were capable of coercion during interrogation, they might also be capable of mishandling physical evidence.
The conclusion is not that coerced confessions are harmless in general. The conclusion is narrower and more important: the effect of a confession depends on the context of the case. When the remaining evidence is weak, a confession can be decisive. When the remaining evidence is overwhelming, the same type of error may not meaningfully change the probability of conviction.
Why Correctly Identifying Harmless Error Matters
The Hopkins study helps validate the method developed in Measuring Fairness. A sound method should not treat every serious trial error as outcome-determinative. It should distinguish between errors that likely changed the verdict and errors that did not.
That matters for both sides of the criminal justice system. When a trial error probably affected the outcome, the conviction should not stand. But when a defendant received a fair trial despite an error, justice is served by upholding the conviction. Fairness does not require reversal for every mistake. It requires careful analysis of whether the mistake likely mattered.
This is also why objectivity matters. A method designed only to help defendants would overstate harm. A method designed only to preserve convictions would understate it. A valid method must minimize both kinds of mistakes: wrongly treating harmless errors as harmful and wrongly treating harmful errors as harmless.
Hopkins is the coerced-confession counterpart to the Washington ineffective-assistance case. Both studies show that the method can identify errors and omissions that did not cause substantial harm. That is an essential part of the larger research project. Measuring fairness means correcting unfair trials while preserving fair convictions.
This case study reflects the broader values behind Fair Trial Analysis: rigorous methods, transparent assumptions, and practical analysis for real cases. The organization’s Solutions work is built around the same principle. Lawyers and courts should not have to rely on intuition alone when evaluating prejudice, harmless error, or the probability of a different outcome. They should have tools capable of measuring those effects in context.
The larger point is simple: fairness should be measured carefully enough to know when a conviction should be set aside—and when it should be upheld.