The Volokh Conspiracy

Mostly law professors | Sometimes contrarian | Often libertarian | Always independent


How often are innocent persons convicted?

In a new law review article, I try to provide a realistic estimate of the rate. I come up with tentative range of somewhere between 0.016% and 0.062% -- well below the figure of 1% to 4% that is often cited as the conventional wisdom.


The Arizona Law Review has just published a mini-symposium on the issue of how often innocent persons are convicted in the criminal justice system. A number of scholars have suggested that this error rate might be at least 1%—and perhaps as high as 2%, 4%, or even more. Professor George Thomas and I challenge such estimates in our three articles—my opening article, his response, and my reply. We both suggest the error rate is lower than 1%, and by my calculations, the error rate is probably substantially lower.

In my opening article, I explain that a growing body of academic literature discusses the problem of wrongful convictions—i.e., convictions of factually innocent defendants for crimes they did not commit. But how often do such miscarriages of justice actually occur? Justice Scalia cited a figure of 0.027% as a possible error rate. But the conventional view in the literature is that, for violent crimes, the error rate is much higher—at least 1%, and perhaps as high as 4% or even more.

My article suggests a much lower estimate is appropriate. Based on a careful review of the available empirical literature, it is possible to assemble the component parts of a wrongful conviction rate calculation by looking at error rates at trial, the ratio of wrongful convictions obtained through trials versus plea bargains, and the percentage of cases resolved through pleas. Combining empirically based estimates for each of these three factors, a reasonable (and possibly overstated) calculation of the wrongful conviction rate appears, tentatively, to be somewhere in the range of 0.016%–0.062%—a range that comfortably embraces Justice Scalia's often-criticized figure.

If my article's tentative error-rate range is correct, it means that previous scholarship has significantly overstated the risk of wrongful conviction. Moreover, it is possible to compare the lifetime risk of being wrongfully convicted to the risk of being a victim of a violent crime. The relative risk ratio appears to be about 30,000 to 1. This decidedly skewed ratio suggests that reform measures for protecting the innocent may need to be cautiously assessed to ensure that they do not interfere with the important goal of prosecuting the guilty.

Professor Thomas' article then advances a slightly different view, partially in response to mine. Thomas argues that the DNA revolution has revealed that the conviction of an innocent defendant by the vaunted American criminal justice system is far from a freakish event. The National Registry of Exonerations now lists more than 2,200 cases of wrongful convictions. Thomas notes that 2,200 cases is a minuscule number compared to the roughly 1.5 million felons in state and federal prisons at any given moment. But the last quarter century has seen a vigorous debate about the error rate that leads to the conviction of innocent defendants. Estimates have ranged from 0.027% to 15%. Thomas finds most estimates are in the 0.5% to 2% range.

Professor Thomas then makes the first effort to draw on an existing data set of actual claims of innocence to estimate the overall error rate in the criminal justice system. The data come from the North Carolina Innocence Inquiry Commission (NCIC), a unique program that allows applications for exoneration from convicted felons. Conservative assumptions about the North Carolina data set produce a likely error rate of 0.125% to 0.5%. Though his estimates of the wrongful conviction rate are limited to North Carolina, Thomas finds no reason to think that the error rate is materially different in other states. His article is, in part, a response to my opening article (discussed above). Though Thomas' findings as to the estimated error rate are higher than mine, he agree that previous estimates tend to be generally too high.

In the last of the three articles, I briefly reply to Professor Thomas. I agree with him that determining an error rate for wrongful convictions remains among the most pressing problems in criminal justice research. And his use of the NCIC data provides an intriguing way to make that determination. My reply article reassesses Thomas's North Carolina estimate rate, concluding it to be somewhat too high. It then looks at another state—my home state of Utah—to find another possible jurisdiction-specific error rate. Properly calculated, the wrongful conviction rates for North Carolina and Utah support my earlier-offered suggestion of a wrongful conviction rate in this country much lower than the rates commonly suggested in other wrongful conviction literature. My reply concludes by underscoring the important point of convergence between Thomas's estimate and my estimates: both are much lower than the conventional wisdom on the subject suggests.