A couple months ago, I did a back-of-the-envelope calculation which caused me to claim that Texas may have executed 54 innocent people. Soon thereafter, I vowed to review all 450 (now 451) lethal injections in Texas to see if I could identify those 54.
I used a coarse filter to rapidly exclude from consideration most of those who were executed. I do not dispute that most of the people executed were in fact guilty of a capital offense. I then designed and applied an Actual Innocence Scorecard to more carefully consider the remaining 69. I summarize below the ones I have scored so far.
Summing the scores of those executed, I claim so far to have identified 4.5 innocents executed. Note that my counting is decimal rather than binary. I have yet to claim that any one individual was absolutely innocent. I do argue, however, that if Texas executes 10 people, each having a 90% chance of being factually innocent, then Texas has likely executed 9 factually innocent people.
Notice that my decimal scoring is more conservative than careless binary counting, at least for those most likely to be innocent. Each of the first 7 people on the list have been declared innocent or probably innocent in various sites. A careless binary counter might count 7 people as wrongfully executed based simply on claims of a third party. I count 4.5 after a substantially more thorough review and analysis.
Now, as promised, an interesting observation. In each case where I have scored the person as having more than a 50% chance of actual innocence, there has been a viable (indeed likely) alternate culprit. In each case where I have scored the person below 50, there is no viable alternate culprit.
That surprised me when I realized the Alternate Culprit section of my Actual Innocence Scorecard has so far been a perfect predictor of likely actual innocence. Had the authorities pursued the alternative culprits, or not paid them with time off for their testimony, or not hidden their presence from the jury, I think it is likely Texas would have executed at least 5 fewer people.