In the New York Times this week, Adam Liptak takes a long overdue and somewhat tepid look at the fuzzy math Justice Scalia used in his concurrence in Kansas v. Marsh when he concluded that "The rate at which innocent people are convicted of felonies is less than three-hundredths of 1 percent - .027 percent, to be exact". Scalia sleeps well knowing our system works so brilliantly.
The problem, of course is that .027 percent is a hoax, and reading the piece, I was stuck once again that a justice generally considered to be so bright could get something this important so fundamentallywrong. But one need only look at the study Scalia cites (by Joshua Marquis , a stalwart of the prosecutorial lobby) to understand the error of his ways.
Marquis came up with the number that Scalia adopted much like a toddler solving a problem in a kindergarten math workbook: He took the total number of exonerations, (north of 200 now) picked a gratuitous multiplier (10 purely for rhetorical purposes), and then divided by 15 million — the total number of convictions during the period of years he considered.
As I've previously pointed out, here's why that's a ludicrous methodology:
Comparing exonerations to felony convictions is like arguing that the Ford Pinto was safe because compared to the total number of automobiles sold in the United States, not many of them blew up . The proper way to determine the failure rate of the Pinto is not to use the total number of cars sold as the denominator, but rather the number of Pintos sold. Likewise, the denominator in Marquis' fraction shouldn't be the 15 million felony convictions over the past 15 years, but rather the number of similar cases in which innocence is actually disputed.
Marquis' most glaring error is his failure to acknowledge the fact that most felony arrests aren't contested. In fact, 95 percent of them are resolved by plea rather than trial. Thus in 19 out of every 20 felony cases, there is no contested issue of guilt and no real claim of error.
Only trials in which someone is convicted while maintaining his innocence should be consideredin computing an error rate. Of Marquis' 15 million felony cases, 14.25 million were pleas. When the denominator in his fraction is changed from 15 million to 750,000, the error rate jumps from the arguably ignorable 3 in 10,000 to more like 50 in 10,000.
And Marquis's numbers become even more disturbing with further analysis. Because of the overwhelming demands involved in reinvestigating a crime with an eye toward exoneration, it is almost exclusively defendants sentenced for rapes and murders whose cases get scrutiny from groups like the Innocence Project. The chances that a drug defendant is going to interest them are virtually nil. Thus the only people who have any meaningful access to the possibility of exoneration are a tiny subset of criminal defendants. Murders constitute only 0.8 percent of all felony cases,and rapes less than 2 percent. In other words, less than 450,000 of Marquis's 15 million felony convictions came in cases where the defendant has had a real shot at exoneration.
It is true that murder cases go to trial far more often than run-of-the-mill drug sales or check forgeries. In fact, some 44 percent of murder cases actually go to trial, with an average conviction rate of about 85 percent. But even taking this into account, of the 150,000 murder cases in Marquis's 15 million, only 66,000 homicide defendants maintained their innocence through a trial, of which just over 56,000 were convicted. Using similar trial and conviction rates for rapes yields somewhere south of 200,000 contested convictions in serious cases.
So what are the real numbers? Liptak cites a small Virgina sample of closed rape files from 1973 through 1988, which suggested a false conviction rate of between 6 and 9 percent. Another study , (by my colleague at Seton Hall, Michael Risinger) which looked at death row DNA exonerations among defendants sentenced to death between 1982 and 1989 for murders involving rape suggested that at least 3.3 percent were innocent.
A few percent may not sound like a lot, but imagine a failure rate of just 1 percent elsewhere in our society: If 1 percent of heart stents failed upon implantation, an additional 100,000 people a year would die, If commercial jets had a 1 percent chance of crashing after take-off almost 400 planes a day would fallout of the sky. Quite simply 1 percent is not an error rate we are be willing elsewhere and it's not ok to accept it in our criminal justice system. Of course as I've just finished explaining, the actual number is almost certainly much higher than that anyway.
In the end, whether the actual number of innocent people wrongly convicted is 3 percent or 9 percent, what is certain is that many thousands of innocent people are languishing in our nation's prisons tonight. And that is not a number Justice Scalia or anyone else should be willing to accept.