Executing Innocents in the Age of Empiricism

In response to the SCT decision the other day in Kansas v. Marsh, where Justices Scalia and Souter were tussling over the extent and significance of innocents sentenced to death, there is a nice discussion in the 'sphere percolating that should be of interest to readers.  It began over at SLP, where Doug Berman averred skepticism toward Souter's claim that there is likelier to be a higher rate of erroneous conviction in the capital context than the non-capital context.  Doug writes:

Not only do I think that this assertion is wrong, but I think it is sad and dangerous that the four "liberal" Justices might actually believe it is true.

In response, over at Co-Op, Dan Filler has a thoughtful post detailing why pathology might arise in the death penalty context more frequently than the non-capital context.  Read the comments too.  In reply, Doug Berman is prepared to accept the possibility that the rate is higher in death penalty cases but the aggregate numbers in the non-capital context must be more worrisome; he reasons to this conclusion by examining the dynamic of the plea bargaining system, which governs most case dispositions in the country:

Imagine if only .01% of all defendants who plead guilty when facing felony charges are actually innocent and plead guilty due to a (reasonable) fear of getting a much harsher sentence after possibly losing at trial.  Even if 99.99% of all defendants who plead guilty are in fact guilty, we would still have roughly 1000 innocent people convicted of non-capital felonies each and every year.  (If only 99.95% of all defendants who plead guilty are truly guilty, then we have 5,000 wrongful convictions nationwide resulting from pleas even before we start looking at trial error in non-capital cases.)

Of course, this scary number crunching does not directly address the rate of wrongful convictions across crimes, and I suppose I was a bit rash when expressing exacerbation about Justice Souter's assertion on that point.  But I find the cumulative numbers far more disturbing than the rate, especially given the harshness of our non-capital sentences and the massive size of our criminal justice systems.  It is, of course, a great tragedy if 1 of the roughly 100 death sentences handed down last year involved an innocent person.  But I think we should be even more concerned that perhaps 1000, or 10,000, or even 50,000 other sentences were given to innocent persons even if the wrongful conviction rate in non-capital cases is lower than in capital cases.

I'll have something up later this week about the misuse of evidence in Kansas v. Marsh, but I thought this discussion should be highlighted.  As to this precise question about which context has a higher rate of error, I suspect the actual answer isn't quite available in our new golden age of empiricism.  Nonetheless, even if the ELS folks can provide one, I am skeptical that having the correct answer will bridge the gap among persons with different normative priors here regarding where to allocate resources. 

Btw, I was discussing the McGinnis article on the age of empiricism with a friend who does quantitative sociology yesterday.  She shared Dan Kahan's skepticism toward the McGinnis thesis.  She noted that in her field she knew of various academics who sit on studies with outcomes they're politically apprehensive about; moreover, she also thought that the design of empirical studies is such a normatively saturated project that it's hard to truly separate "the facts" from the "opinions" such that consensus about what the facts show will emerge. Reactions?