Saturday, June 18, 2005

still another example of bad causal inference for a good social end

From the NYT:
Gov. Tom Vilsack of Iowa announced yesterday that he would restore voting rights for all felons who have completed their sentences, ending what advocates for voting rights had called one of the most restrictive disenfranchisement laws in the country.

Mr. Vilsack, a Democrat who has been called a dark-horse presidential candidate for the 2008 election, pointed to research showing that ex-prisoners who vote are less likely to end up back in prison.
Something I wonder about from time to time is whether, should quantitative social science radically improve its understanding of and capacity for causal inference, there would even be a market for it. Or, at least, there is such a strong market now from all political sides for the ability to make causal-sounding insinuations of relationships that are almost certainly not causal. Not to mention commercial interests: Remember that Nike commercial where they had the little girls saying "If you let me play sports ..." followed by all kinds of dramatic assertions (based presumably on real studies with real correlations), including even something along the lines of "If you let me play sports, I will be three times more likely to leave a man who abuses me."

At least to me, it's interesting to imagine there being Fellow Liberals out there who think that it is ludicrous to suppose that the death penalty might deter some crime but that it is plausible that giving felons the right to vote will deter some recidivistic crime.

(For the record, JFW is anti-death-penalty and pro-voting-felons. The proprietor of this weblog would be pro-voting-felons even if it were not the case that he suspects felons will tend to vote Democratic. Indeed, he has an entire deeply skeptical spiel about the analyses that some sociologists have done to project how past elections would have gone differently if ex-felons were fully franchised, and he has even subjected a couple of his methods classes to parts of this spiel. As for the question of whether the death penalty deters crime, he has been interested in the vastly inconsistent results that have been cited for some time, and he even paid a graduate student a year ago to go look up this relevant research for him, but the graduate student has to date not reported back.)

18 comments:

Constance said...

I bet that's the graduate student you hired that wasn't me... that's you gettin' some get back!

Anonymous said...

For the record, that Nike line about women being 3x more likely to leave men who beat them if they play sports is actually based in real science. There is a slew of research indicating that girls who grow up playing sports have higher self-esteem when they are grown women than girls who did not play sports. There is also a slew of research indicating that women with high self-esteem are far more likely to leave partners who become abusive. Although playing sports may not directly cause a woman to be 3x more likely to leave an abusive partner, there is likely some degree of causality there.

jeremy said...

The whole point of my post is that I categorically reject the idea that just because researchers observe that women who play sports have higher self-esteem when they grow up than women who did not play sports, the difference in self-esteem is because they played sports. In my opinion, the existence of a slew of research in the social sciences on a question tends to indicate more about the appealingness about the question than whether the research has actually effectively resolved the problems with the causal inference. Perhaps there is more compelling research in this area than what I am aware. The problem, just to be clear, is that all kinds of individual characteristics are associated with playing sports in the first place--including, incidentally, self-esteem--and these characteristics are also associated with higher self-esteem in adulthood. My suspicion, about which I might be wrong, is that research on the issue recognizes this issue but does not adequately adjust for it.

Anonymous said...

If I were a violent thug just out of the joint, voting certainly would be a top priority for me.

Ben in Boston said...

Perhaps voting works to get people thinking about things that are bigger than themselves? Even if the relationship is not causal, we should give felons every opportunity to rejoin society once they have paid their debt. Why make them wear the scarlet letter any more than they actually do?

CM said...

I think you're making a logical leap here. You write that Tom Vilsack cites "research showing that ex-prisoners who vote are less likely to end up back in prison." Then you go on to say that you're skeptical that letting ex-prisoners vote will deter crime.

It sounds like you're talking about deterrence in the sense that ex-prisoners will stop and say, "Wait, I shouldn't commit this crime -- I don't want to lose the right to vote!"

But actually, there could be an alternate explanation -- for instance, ex-prisoners could become more involved in their communities by voting, which would make them less likely to go back to prison. Or it could be the other way around -- those ex-prisoners who do vote may have a newfound sense of civic inclusion, which also makes them less likely to commit crimes.

You can still be skeptical about the research, of course. But it sounds plausible to me.

Mike Sierra said...

I agree with Jeremy's point about weak causality, but I was wondering whether there's any good reason to deny felons their voting rights while they are still serving their sentences. If, as Ben suggests, voting may help felons to rejoin society as good citizens, shouldn't we get them started off early?

jeremy said...

chickenmagazine: I didn't mean for my skepticism to be restricted to just the narrow sense of voting providing a conscious barrier to recidivism, but I'm also skeptical of it having secondary effects. Again, that doesn't mean I don't support the extension of rights, however. Which is exactly part of the ethical problem for me: what if there are other people who are convinced by this claim, so they agree with me about the end policy but for reasons that I think about based on dubious social science?

mike sierra: I think the argument about extending voting rights to felons gains much more force when one is talking about felons that have been released. Personally, my main concern about felon voting from prison would be whether/how their vote might be manipulated in local elections. For prisons located in rural areas, prisoners could actually comprise a considerable voting blog relative to the total voting population.

DirtCrashr said...

If voting helps stop pathological felons from committing crimes, we should give them a few extra votes as an antidote!
"For prisons located in rural areas, prisoners could actually comprise a considerable voting blog relative to the total voting population."
Or with the example of Oakland, CA where many felons are returned to home-turf, a large, urban, inner-city voting bloc.

jeremy said...

I think there is a big difference between that example and the idea of giving a substantial number of votes in a local election to a group of people who only live in the locality because it is where they are incarcerated. Besides, what worries me more is that it would seem to potential for even more opportunities for the corruption of wardens than what already exists, especially in prisons in out-of-the-way places.

Mike Sierra said...

Jeremy says: "For prisons located in rural areas, prisoners could actually comprise a considerable voting blog [sic] relative to the total voting population." I have a solution for that: Provide all prisoners with their own representative regardless of geography, then in the interests of better serving constituents' interests, place him in jail.

christopher uggen said...

Hey Jeremy, As usual, you make a great point and your analysis is characteristically clear-headed. I'm not sure whether the gov is citing anything I've written, but I did a very simple law review paper on voting and subsequent crime with Jeff Manza in 2004. Given the data limitations, we knew we could not make strong causal claims. Here's how we talk about causality:

"Establishing a causal relationship between voting, or civic reintegration more generally, and recidivism would require a large-scale longitudinal survey that tracked released offenders in their communities and closely monitored changes in their political and criminal behavior. At present, no such data exist. Nevertheless, it is possible to bring some empirical data to bear on this question now..."

"We will first establish a correlation between voting and crime by examining whether those who voted in the 1996 presidential election had lower rates of arrest, incarceration, and self-reported criminal behavior in the years following this election than those who did not participate. We then conduct a logistic regression analysis to test whether this correlation is spurious due to factors that are associated with both voting and crime. By statistically controlling for prior self-reported and official criminal behavior and background factors such as race, gender, education, employment, and marital status, we can learn the extent to which the raw correlation is due to the self-selection of persons at low risk of crime into voting. While this approach cannot firmly establish political participation as a cause of desistance from crime, it allows us to adjust the observed effects of voting for some of the most important alternative hypotheses and sources of spuriousness..."

"We can take from this study both provisional support for the idea of civic reintegration through voting, and confirmation of the skepticism expressed by some of the prison inmates and probationers we interviewed. Voting appears to be part of a package of prosocial behavior that is linked to desistance from crime, but its unique independent contribution is likely to be small relative to pressing socioeconomic needs, family support, and other factors. With this caveat in mind, however, the right to vote remains the most powerful symbol of stake-holding in our democracy. To the extent that felons begin to vote and participate as citizens in their communities, it seems likely that many will bring their behavior into line with the expectations of the citizen role, avoiding further contact with the criminal justice system."

I think one could reasonably ask whether it makes sense to publish such a piece (even with the caveats). We decided to write it up for two reasons: (1) we wanted to address arguments that re-enfranchising felons would *reduce* public safety; and, (2) our qualitative interviews with felons provided a plausible rationale for both the null (no effect) and research hypotheses (some effect). We've now gathered another wave of crime and election data that should enable us to make somewhat stronger inferences.

My apologies for the lengthy comment. I am a bit defensive on this issue, but I'd argue that most felons are not violent thugs, most are not in prison, and almost all desist from crime. Also, many who favor expanded voting rights (and civil rights more generally) are libertarians rather than liberals. I've done some right-wing talk radio on this subject and I hear from conservatives who view disenfranchisement as another example of the tyranny of "big government." Most Americans in our 2002 Harris Poll agreed with Jeremy -- drawing a line between current prisoners versus people who have done their time. Still, reenfranchisement advocates argue that the "voting bloc" issue could be addressed by absentee voting from the inmate's home community. That is, Wisconsin inmates from Milwaukee could be counted as Milwaukee voters rather than Waupon voters if they were serving time in Waupon. Finally, not even "liberal" criminologists believe it is "ludicrous to suggest that the death penalty deters some crime." Instead, we argue about the marginal deterrent effect of capital punishment -- how much crime is deterred above and beyond that deterred by lifelong incarceration.

Akiva said...

It's amazing to follow a few blog links and find yourself in a group with such wildy divergent views.

I notice everyone seemed to be trying either to justify or dance around the no relationship possibility. I was very surprised no one said, "those felons who voted were felons who had put their lives back together and reintegrated as productive citizens." In other words, voting was a result of re-integration success, not a step of re-integration.

Here would be an easier statistic to track that could show an inference, do crime statistics decrease, stay level, or increase in the periods prior to elections? If voting is a deterence to crime, then one should see a drop in crime in the weeks or months prior to an election.

jeremy said...

Akiva: Your idea is based on a relatively narrow notion of what the causal effect might be, the idea that somebody is refraining from committing crime just so they can vote. One could argue that extending the right to vote (or subsequent civic participation) has salutary social psychological effects that extend beyond the election per se and wouldn't necessarily imply any especial pre-election drop. Besides, a contaminant of the test might be that police forces are often expanded in the period prior to an election, possibly making re-arrest more likely than it would otherwise be independent of behavior.

Chris: It's great that you commented (an actual expert weighs in on my blog!), but I'm going to have to respond to it. Just in reply to your last point, though, I was speaking in shorthand and did mean to refer to the different in the marginal effect over life-without-parole. I think that's pretty straightforwardly the important estimate for any moral implications of the question of whether the death penalty provides a deterrent. As for my "ludicrous" sentence, I didn't mean to be talking about liberal criminologists, but rather liberals-at-large, some of whom I think have used exactly a That's-Ludicrous line of dismissal to counter the deterrence argument for capital punishment.

christopher uggen said...

Jeremy, I thought that's what you meant. Most prisons are just such lousy places that I think it is important to add the "marginal" to deterrent. I'll offer another example that I think makes your point. Incarceration has risen as crime has dropped. Some "liberals at large" (and some liberal criminologists) dismiss as ludicrous the idea that mass incarceration has anything to do with the decline in crime in the last 25 years. I'd maintain that locking up two million people (overwhelmingly poor, mostly people of color, and many well past their prime offending years) is a really inefficient and inequitable way to reduce crime. Still, I admit that incarceration bears some relation to crime. It pains me to write this, but surely if we incarcerated half the males aged 16-25 we'd see at least a short-run reduction in crime outside the prison walls.

jeremy said...

Chris: Oops, I meant "I'm going to have to respond to it later" in the first sentence of my last comment. Ugh, today I'm sufficiently behind and torpid on everything that "later" might have been optimistic anyway. With regard to the comment you just made, I agree fully.

triticale said...

What is probably going on is that the sort of ex-offenders who get their voting rights restored are for one reason or another the sort less likely to re-offend.

Rhymes With Scrabble said...

Have you seen this open letter to the Kansas School Board? I got the link from my father; our favorite part is the graph at the bottom showing the inverse relationship between number of pirates and global temperature. They have t-shirts.