Sunday, September 20, 2009

Public views on actual innocence driven by perceptions, not data

Here's a particularly interesting result from last week's reader poll regarding how many actually innocent people are currently residing in Texas prisons: 70% of readers chose answers outside the range of the various statistical estimates of innocence. The poll asked readers to choose between an array of innocence estimates from various sources (as applied to the Texas prison popuulation), plus an option to choose a greater (5,000 or more) or lesser (less than 1,000) number than any data-driven estimate.

Nearly half of readers (47%) thought there are 5,000 or more innocent inmates (out of nearly 160,000) currently locked up in Texas, while 23% thought the total was fewer than 1,000. None of the statistically derived innocence rates received more than 8%.

While I could speculatively make arguments that, if correct, would drive the estimate of incarcerated actual innocents in Texas higher than 5,000, from a mathematical perspective I could not justify it based on available data. (FWIW, I selected 3,500 for my vote.) Similarly, the lowest rate estimate came from a source at the national district attorney's association and is almost certainly low-balled. Yet nearly a quarter of respondents thought the real number of innocents was really even lower than that.

We all know statistics can be manipulated, so a lack of faith in one or another statistician's model, especially when the results vary considerably, is no big surprise. More interesting to me, these results appear to reflect a growing divide between people whose belief in our current system has been truly shaken by recent discoveries -DNA exonerations, junk arson investigations, or perhaps the recent NAS study of junk forensic science generally - and a hard core of folks who believe that relatively few innocents ever go to prison regardless of the evidence, statistical or otherwise.

Those in the 5,000+ group are probably waiting for the next shoe to drop. Will it be hundreds of dog-sniffing cases found to rely on junk science? Will it be false allegations based on bad urinalysis from a corrupt lab in Bexar County? Shaken baby cases? More drug-war scandals like in Tulia or the Dallas fake-drug fiasco? Such things cannot really be incorporated into a statistical model.

By contrast, those in the "under 1,000" group may believe so strongly that most defendants are guilty that they don't want to upend the apple cart over just a few people. Perhaps coming to grips with the idea that thousands of inmates in prison might be actually innocent would require contemplating bigger changes to how the system operates than these folks are comfortable with making. I'm interested for these readers in particular to say in the comments why they think the number of actual innocents is lower than existing estimates.

Granted, this was not a scientific poll, and you, gentle readers, are not a representative sample on this particular topic. So I'll readily admit I'm speculating here on what you think and what it means. If I'm misinterpreting, by all means please correct me in the comment section.


Anonymous said...

Although I don't have the answer either, your last paragraph about this venue not being a representative sampling is on the mark.


Anonymous said...

As an observation, the volume of comments on various news reports of wrongful convictions (and now executions)seem to differ in the DMN and Hou Chron. Those that comment in Dallas seem to be more rapid against upturning the Texas criminal justice apple cart. Merely a lay observation since I have never viewed Houston or Dallas as liberal bastions.

Anonymous said...

My estimate of over 5,000 innocent in prison is based upon a belief that far too many innocent people agree to a plea bargain rather than wait two years and then roll the dice in hopes a trial will correctly find them innocent.

Most of the recent cases of actual innocence are for higher level felonies. My thought is that the system gets it wrong on lots and lots of lower level offenses. Because the length of sentences in these cases is probably shorter than the time in jail waiting for a trial and there is no appeal, actual innocence is never discovered.

Francis Holland said...

People, especially white people, trust police and the criminal injustice system, and its frequently all-white juries - far more than is warranted.

It would surprise me if only three or four percent of prison inmates were innocent, considering that more than half of the inmates are probably Black and Latino.

I would focus not merely on "innocence but also on whether they would be in prison if they were white and engaged in the same behavior that brought them to the attention of police. I don't think you have to be "innocent" to be in prison unjustly.

I think that no Black or Latino person should go to prison where a white person who committed the same acts would not go to prison. That brings up all sorts of issues like police's disproportionate attention to the behavior of Blacks and Latinos, the prosecutors' tendency to overcharge in the correct belief that all-white juries will believe whatever they are told about Blacks and Latinos, and then the issue of disparate sentencing, and even disparate parole violation treatment.

Anonymous said...

You would also see a lot more middle and upper class white people in prison if white collar crime were pursued and prosecuted as aggressively as drug crimes.

I've said it before...I don't believe its a product of any grand conspiracy to make it that way but we've allowed our criminal justice system to become a tool that the wealthy and powerful use to oppress the poor and the weak.

Travis said...

I picked "below 1000" because I don't know how many innocent people are in prison. I do know there have been 39 exonerations in Texas. Dallas has the highest incidence, and faulty witness identification is the leading cause. I haven't seen the "statistics" on the speculation about how many innocent people are in our prisons. Seems like a ridiculous guess to me actually. It doesn't matter one way or another; 39, a thousand, 5000. One is too many.

People chose "above 5000" and "below 1000" because those are broader categories than the other set number choices. It doesn't have a thing to do with not wanting to "upend the apple cart over just a few people." It has to do with the nature of guessing.

My personal opinion is that we need witness identification reform in Texas police departments but didn't get it because of the bullsh*t of the 81st legislature. Throw the damn cart over, let the apples spill all over the floor. How many innocent people have to go to prison to justify convicting the guilty? How many DA's will testify that we don't want to shake the "fragile coalition." F*&k the coalition. When we convict innocent persons on eyewitness testimony despite overwhelming evidence to the contrary the system is delegitimized. And all it takes is some manila folders. But we couldn't get it done because whoever thought it a good idea to limit the session to less than half a year every two years.

Here is a questionnaire.

How long should the Legislative session be?

Less than Half a year


Long enough to get sh*t done.

sunray's wench said...

Scott, what definition of "innocence" are you using for this? The TX Law of parties can convict people who are innocent of the crime they are being tried for, yet they are guilty of other things.

Gritsforbreakfast said...

"what definition of "innocence" are you using"

I'm speaking of people who are actually innocent of the crime for which they were incarcerated: people who flat-out did not commit the offense.

It's certainly possible that some people are "innocent of the crime they are being tried for, yet they are guilty of other things." As Paul said to the Romans, all have sinned and fallen short of the glory of God. But that's beyond the ability of the justice system to decipher. Guilt and innocence in this context are a defined based on whether they committed the crime they are charged with.

Mr. Anxiety said...

How did the number of responses to this poll compare to the number of responses you've gotten to other polls?

I didn't participate in this one because I don't have a clue how many people are wrongly incarcerated and I didn't want to blindly speculate.

Mike Howard said...

From talking to juries and venire panels (pre-jury selection), it seems to me your under 1000 crowd is most likely based on the idea that "they're guilty of something." The attitude I run into most often centers around the failure to adhere to the presumption of innocence. I've even heard people blaming people found to be innocent - "they would've have found themselves in prison if they wouldn't have been running with the wrong crowd..." etc.

Karo said...

This just in:
Most views driven by perceptions, not data.

This just in:
Texas courts don't find people "innocent."

The only thing we can ever really say about an aquittal is that the evidence was not sufficient prove guilt beyond a reasonable doubt. This is possibly, but not necessarily, because the accused is actually innocent.

Anonymous said...

Excellent comment by Francis L. Holland -- thanks.

Gritsforbreakfast said...

Mr. Anxiety, FWIW most Grits polls get in the range of 400+ respondents and this one had 371 - so just a little low but not unusually so. I've had reader polls with as few responses as 250 and as many as 2,000 once when linked to a poll.

And Karo, your point about acquittals is well taken. That's what's so extraordinary about all the recent DNA exonerations - for once we have a category of cases available for analysis that avoids that conundrum and focuses exclusively on actual innocence.

Unknown said...

I plead guilty to a minor assault crime simply because I didn't have the money to pay my attorney, and the fine was much less than the attorney costs. By the way, I didn't commit the original offense as charged. I just didn't appolgize to a woman I bumped, after she intentionaly stepped infront of me to block my path. No, I didn't knock her down or injur her. Bottom line.. by definition I AM now GULITY, of "unwanted" physical contact.

Soronel Haetir said...


I have some problems with your definition of innocent, since it would also cover people who were convicted of the greater charge when in fact they only committed the lesser included offense. As I have said numerous times this is one area of plea bargain research that the raw conviction stats just do not explore very well. Yet those conviction rates after charges are brought are generally the only data anyone tries to correlate. I do not know that any better method exists however.

Fransis L Holland,

I would ask how much of that is based on race and how much on family showing up to court and other factors? I have seen numerous studies showing that white defendants have a much greater level of support that is visible to the court during criminal proceedings. It would not surprise me if there is some level of engrained racism in the system, yet it would also not surprise me if it is far weaker than many beleive.

Gritsforbreakfast said...

Soronel, so you think the definition "people who flat-out did not commit the offense" is too vague for "innocent"? Really?

To be clear, this discussion is occurring in the context of DNA exonerations, which are actual innocence cases that don't fit the scenario you describe at all. The estimates referenced in the poll were based on data derived from the number of DNA exonerations, not some different category of cases you're speculating about.

Anonymous said...

Help me out here...I must be missing something, but this is how I am looking at the issue. The problem of the viable denominator in trying to estimate number of innocent people who are actually convicted can be approached as follows, using the numbers that Grits gave from the article 'Texas criminal justice system has major flaws': Estimating

innocent people in prison at

major.html .

My understanding is that we can divide the 38 exonerations out of 400 available DNA sample that are available to be tested ( see ). That's 9.5% . Now, if we use the Grits methodology from his article entitled "How many innocents in prison: Exonerations make up 3% of Texas DNA case resolutions" (see ), then we would use percentage and apply it to the 160,000 Texas convicts, which gives 15,200. It gets even scarier if you look at the Pew Center Report "1 in 31 U.S. Adults are Behind Bars, on Parole or Probation" at . It says that 1 out of every 22 people in Texas is under the correctional control (prison, probation, parole, jail). For the year end of 2007, that was 797,254 people. If you multiple by our innocent rate of 9.5% , that's more than 75,000 people.

Soronel Haetir said...


The problem I am trying to work through is that I don't see the forces that lead to false convictions in the studied cases being applicable to the vast majority of cases. In murder, rape and other serious offenses there is enormous pressure to resolve the case. Sometimes I will admit to the detriment of finding the actual offender.

I don't see that being true for more typical burglary or assault cases. If one more burglary slips through the cracks no police chief has to worry about the next election cycle.

I am sure there are officers who are padding their stats by planting evidence, but again I just have a hard time applying evidence for one type of case against the system as a whole.

Anonymous said...


The profit motive explains alot, as well as other forms of benefits, and you can look at the different players involved to see how they benefit specifically. The cop who wants the stats and promotions ( Dallas cops like Mark Delapez who throw chalk into people's cars , and then bust them for having "drugs"), the informants and snitches who get the bounty, jailhouse snitches , scared crime suspects who "cooperate" with the police in order to avoid punishment, ect... There is plenty to look at, it's just that people don't take a second look unless it happens to them or someone they care about.

Gritsforbreakfast said...

Soronel, while the pressure to get a conviction may be less in lesser offenses, bad eyewitness ID contributed to 3/4 of DNA exonerations and eyeID is used in all sorts of cases - robberies, assaults, etc.. And as 10:28 notes, there are big incentives to inflate arrest numbers in drug cases, and the widespread use of informants has led to some serious abuses - the Dallas fake drug scandal being the most iconic.

Admittedly, some types of cases are more prone to convicting innocents than others. The percentage of false arson convictions, e.g., is probably a lot higher than other crimes. But when sloppy or faulty techniques are used by law enforcement, they can result in errors whether it's a "big" case or a small one.

Gritsforbreakfast said...

To 10:13, for a variety of reasons I'm not sure it's valid to use 400 as the denominator, though I can see why you would propose it. In Dallas, e.g., the Innocence Project of Texas and the Dallas DA's Conviction Integrity Unit vetted cases before doing DNA testing, so the cases where they decided not to test would also have to be included in the denominator.

Of the first 1,000 cases solved by DNA in Texas, 33 (or 3.3%) were exonerations for actual innocence. So if you're going to try to compare the DNA exonerations to a DNA-specific denominator, arguably it should be that one (but even that needs caveats).

Anonymous said...

Why would you want to use a data pool of 1000 cases? Why not 100, or 500 , ... it seems arbitrary, especially when there are only about 400 DNA samples that can be reexamined...and the effect of choosing the larger denominator is to arbitrarily lower the percentage. It seems more logical to use the data pool of the available samples, and let the hard science of DNA testing be the judge and arbitrator. The best types of studies are those that involve scientific data, wouldn't you agree? If your 3.3% estimate was correct, it should be reflected in the results of the sample size of 400 that I am proposing as well. If not, then your denominator of 1000 would be more suspect since there were no means to test all 1000 cases to see if they are indeed valid. The 400 number which I propose does not face that problem, since it is simply the number of cases for which DNA samples are available for testing. Like your 1000 number , I would use it as a random sample of the larger population of violent crimes, etc... In other words, I like the design of your method, but I think that you have , as you admitted in the article, a problem with your choice of denominators. So far you have not given any concrete reasons to support your choice of a denominator, and you vaguely dismissed my choice with the phrase of something about a variety of reasons, none of which you deliniated. That's not an attack, just an observation.

Gritsforbreakfast said...

"Why would you want to use a data pool of 1000 cases? Why not 100, or 500?"

Because that's all the cases that have been solved in Texas by DNA, or it was this spring when I posted that stat (both the total cases solved and number of DNA exonerations has increased since then). You can't use a larger data pool than exists.

Also, you write that I "vaguely dismissed [your] choice with the phrase of something about a variety of reasons, none of which you deliniated" That's incorrect. Here's what I wrote critiquing the denominator you suggested:

"In Dallas, e.g., the Innocence Project of Texas and the Dallas DA's Conviction Integrity Unit vetted cases before doing DNA testing, so the cases where they decided not to test would also have to be included in the denominator." I don't know the total, but including those cases would boost the denominator pretty far beyond 400.

You may disagree, but I gave you my reason.

Anonymous said...

I see your point on the reason you chose the 1000 number,
but I think my previous comment is still on target. Your point about vetted cases is well-taken, however, and I thought that the 400 number represented ALL the available data samples. If a case was vetted, and there was in fact an available data sample for that case, I agree that it would have to be added into the count, assuming that it was in the violent crimes category, since that is what was being examined. You wouldn't happen to have the exact numbers on the vetted cases handy by any chance?