Weeding out the innocents


THE FIRST innocent American defendant to be exonerated by DNA evidence was Gary Dotson of Chicago. Before his conviction was overturned on Aug. 14, 1989, he’d spent 10 years in prison and on parole. This year, on April 23, Jerry Miller obtained the 200th DNA exoneration, also in Chicago. He had served 25 years for a rape he did not commit.

Two hundred innocent prisoners exonerated by DNA — plus more than 200 other exonerations that did not involve DNA. That sounds like a lot. But over 18 years in a criminal justice system that sends hundreds of thousands to prison each year? How frequent are wrongful convictions?

The truth is, we don’t know. But that hasn’t stopped prominent members of the legal profession from staking out a position.


More than a year ago, prominent Oregon prosecutor Joshua Marquis, commenting on my research on exonerations, wrote in the New York Times that the rate of erroneous convictions could be no higher than .027%. Last June, Justice Antonin Scalia endorsed that calculation in a concurring opinion in a Supreme Court case. This April, after the 200th exoneration, Colorado District Judge Morris B. Hoffman wrote in the Wall Street Journal that false convictions occur in fewer than .065% of criminal cases. Whatever the number, the message is the same: Not to worry, we get it right more than 99.9% of the time.

These reassuring words are nonsense.

Here’s how Hoffman and Marquis arrive at their numbers. Start with the number of known, proven exonerations (for Hoffman, the 200 DNA exonerations so far), then multiply that by 10 “to be safe” (Marquis’ formula). Take that product, divide it by an estimate of the millions of all criminal convictions over time, and you end up with something less than one-tenth of 1%.

This makes no sense. Imagine that a car company gets reports that 65 of its 2007 sedans have faulty steering columns, which sometimes lock up. What if the company said: “That’s no big deal. We have 10 million cars on the road, so that’s less than one-thousandth of 1%.”

But that’s ridiculous. The total number of defects could be 10 or 100 or 1,000 or 10,000 times greater than the first batch that came to light. Unless we investigate systematically, we just don’t know — not for steering columns and not for criminal convictions.

Further, the car company has divided by the number of all cars in service, when it should use the number of 2007 sedans only. Marquis and Hoffman make this mistake too. Hoffman, for example, divides the 200 DNA exonerations to date by his estimate of all criminal convictions — including check kiting, tax evasion and car theft. But DNA testing requires biological evidence; it has only been useful in a fraction of rape convictions and a scattering of murder cases (if the killer bled). Rape and murder account for fewer than 2% of felony convictions and a much smaller percentage of all convictions.

As it happens, we’re just beginning to learn something real about the rate of false convictions. The Virginia Department of Forensic Science recently found a large group of closed rape files with untested DNA, which will make possible the first systematic study of false convictions. So far, tests on a small preliminary sample are troubling: two previously unknown wrongful convictions out of 29, or an error rate of 7%.


We can also learn from death sentences, which are reviewed much more carefully than other criminal convictions, so more errors are caught. Of the 3,795 defendants sentenced to death from 1973 through 1989, 86 were freed because of DNA or other new evidence of innocence. That’s 2.3%. Of course, some of those freed may be guilty, while others still on death row are no doubt innocent. So last year, Michael Risinger, a professor at Seton Hall Law School, did a study of death row DNA exonerations only. His results? Among defendants sentenced to death between 1982 and 1989 for murders involving rape, at least 3.3% were innocent.

The good news is that the great majority of convicted defendants in the United States are guilty; the bad news is that a substantial number are not. Is an error rate of 2% or 3% or 5% high or low? That depends on your point of view and your purpose.

If 1% of commercial airliners crashed on takeoff, we’d shut down every airline in the country. That would be nearly 300 crashes a day. If as few as 1% of criminal convictions are erroneous, right now there are more than 20,000 innocent defendants behind bars.

Correcting false convictions is much harder than recalling automobiles, but we have to try. We’ll never save the innocent defendants who are already in prison — or keep others from suffering their fate — if we just wish the problem away.


SAMUEL R. GROSS is a law professor at the University of Michigan.