2,000 convicted criminals have been exonerated since 1989. Sounds impressive, right?

Well, this isn't as impressive as many believe. The Huffington Post has this story here:

More than 2,000 inmates and ex-cons have been exonerated since 1989, according to the database that aims to track all wrongful convictions in the United States. More than 100 had been sentenced to death.
"This is a beginning," said University of Michigan Law School professor Samuel Gross, one of the database's creators. "One of my great hopes is that this will lead us to learn more about exonerations."
The database, which was developed with members of Northwestern University's Center on Wrongful Conviction, focused on 873 individual cases. The researchers also identified 13 major police scandals that falsely netted 1,170 other people, although these are not included in the database because they are the results of a collective exoneration based on problems in individual agencies. . . .

But here is the problem. How many criminals are we talking about? We are talking about exonerations over 23 years, but the period over which the crimes many have occurred would have taken place over a much longer period of time. These data are for 2010, 2009, 2008, 2007, and 2006 for Table 29.

Murder 11,201 + 12,418 + 12,955 + 13,480 + 13,435 = 63,489
Rape 20,088 + 21,418 + 22,584 + 23,307 + 24,535 = 111,932
Robbery 112,300 + 126,725 + 129,403 + 126,715 + 125,605 = 620,748
Agg Assault 408,488 + 421,215 + 429,969 + 433,945 + 447,948 = 2,141,565

If this was the rate over 23 years, the number of arrests would equal 13,513,576, but given
how much higher crime rates were back then, the annual rate would be much higher.

At 13.5 million arrests and say 90 percent conviction rate, the total convictions would be
12.1 million. 2,000 out of 12.1 million (which again is an underestimate because of the

fact that the crimes that were eligible for exoneration were over a much longer period of

time) is only a mistake rate of 0.0165 percent. That seems like a remarkably low mistake

rate to me.

For murders alone, a similar analysis implies 292,049 arrests. At a conviction rate of 90
percent, that implies 264,844. 100 divided by 264,844 comes to a rate of only 0.038%
(but the upward bias problem mentioned above is even greater here because people are