This is the fourth in a series of posts serial-blogging my new article, Prison Vouchers, forthcoming in the Penn Law Review. The first post in the series is here (I recommend reading it for an overview of the entire paper), and the second post is here and the third is here. Some of the concerns expressed in comments to past posts are addressed in this or later installments (though if you want to know everything immediately, with all the supporting footnotes, you should look at the full version on SSRN). Also, I see that many commenters are making the casual assumption that I’m in favor of the scheme I’m laying out; note, though, that I don’t make such a claim anywhere in the paper.
I’d appreciate comments (especially informed ones, which have a greater chance of making it into the final version, with thanks in the author footnote).
In this post, I discuss the disadvantages of the prison voucher idea. (Non-empirical arguments against prison vouchers were discussed in the previous post; this post only covers empirical disadvantages.)
* * *
“Market Failure” Arguments Against Vouchers
So let’s suppose that there’s nothing about a voucher proposal that’s philosophically inconsistent with the idea of incarceration. Would prison vouchers work?
The discussion above of how prisons would accommodate prisoners’ values if they had to compete for prisoners assumes that market forces would lead prisons to offer more valued amenities. Here, I discuss possible market failures that could prevent this from happening.
Barriers to Individually Maximizing Decisionmaking
Information. Convicted defendants may not know the actual quality of prisons, just as parents may not know the actual quality of schools. Does this argument apply with more or less force in prisons than in schools?
Someone sentenced to prison for the first time may not know much about different prisons. Even repeat offenders might have little direct experience with prisons if they’ve mostly spent time in jails instead of prisons; and even someone who’s been in one prison may not know about the others.
But one can find out about prisons from several sources.
First, one might have friends or neighbors who’ve been in prison, and information can spread by word of mouth.
Second, prisons can advertise, and reviews of prisons by current or former inmates may be available on the internet.
Third, there are already ways to evaluate prisons, such as reports of monitoring agencies or the Logan quality of confinement index. One can even require that prisons publish such information, as well as other information that would result from the voucher program, like the length of the wait list and the rate of transfer out of the prison, as part of their advertising. One possible model for prison information disclosure would be the federal government’s “Nursing Home Compare” site, which conveniently pulls together nursing home information already collected by the government.
Fourth, if the voucher program allows one to transfer out after a while, one will at least have some direct experience of one’s own prison, and if that experience is bad enough, it can be worthwhile to take the chance on another prison.
So far, the arguments look similar to those for schools. But the informational problem seems to be less severe in the prison context because the people who choose the prison are the same as the people who experience the service. With schools, there is an agency problem: The parents make the choice, but because they don’t directly experience the school, they have less of an incentive to become well informed (even if they’re altruistic toward their children) and they’re also less able to do so. Parents are imperfect agents of their children, whereas inmates are perfect agents of themselves.
A convicted defendant may also have some help from his lawyer, just as lawyers now try to get their clients into drug abuse programs or military enlistment. The government could also provide a default assignment or a list of recommendations, though the bureaucrats who currently assign an inmate’s prison may well be less favorably inclined to prisoners than school bureaucrats are to kids.
Ultimately, what convicted defendants lack in information, they may make up for in motivation, both to acquire information and to act on it.
Finally, even if information problems continue to be serious, this is an argument in favor of having prisons run by the nonprofit sector, just as information problems in education are often taken as an argument in favor of nonprofit schools.
Competitiveness. In addition to insufficient information, there might also be insufficient competitiveness; inmate choice might then still result in insufficient attention to inmates’ desires.
Even though I’ve stressed that prison choice needn’t involve private provision, making the market competitive enough to move in response to “consumer demand” might require the participation of private prisons. Public prisons, after all, may not even want more prisoners. Today, in fact, DOCs actually ask for fewer prisoners. Perhaps they do this because of inadequate funding—a problem that would be alleviated if an adequate voucher accompanied each prisoner (but not if total voucher funding merely matched the current funding of the correctional system). Or perhaps they’re just advocating a less incarcerative penal policy in general, not taking a position on whether they, as opposed to someone else, should get existing prisoners. But in general, it’s not obvious that public prisons benefit from having more prisoners, and in any event, public prisons may have less flexibility than private prisons to respond to market incentives. One should thus make sure to structure rewards and penalties for public prisons so that it’s in their interest to attract more people and also so that they have the freedom to experiment with different models.
One shouldn’t overstate the difference between the public and private sectors: private and public school curricula often look quite similar, perhaps because parents are uncertain about students’ ability and future job prospects, and the adoption of similar curricula allows parents to minimize this risk. This may well apply to prisons too: Prison is a traumatic place even under the best of circumstances, and inmates who will be spending a few years in one may not want to experiment.
But this doesn’t mean that competition is useless or that schools (or prisons) won’t differ. First, some schools (or prisons) will cater to those parents (or inmates) whose uncertainty or risk tolerance is different from the norm. This is why there are science magnet schools in large cities—some parents know that their children are gifted in science. Similarly, inmates may have heterogeneous rates of substitution between, say, the risk of violence and the quality of medical care. Second, schools still differ in “factors such as location, presence of religious instruction, and, perhaps most importantly, whether a school is a ‘good’ school or a ‘bad’ one in terms of technical efficiency.” All these factors are relevant for inmates as well, and it’s the third factor—the actual productivity of the prison—that may be the most significant in driving prison improvements.
External Effects of Individually Rational Decisionmaking
The previous market failure arguments suggested that prisoners (or prisons) might be unable to choose in a way that’s individually rational. Other market failure arguments could be based on externalities—even if everyone chooses rationally for themselves, the results may still be suboptimal, for instance if individual decisions have negative effects on other prisoners.
One obvious externality (this one on the taxpayer) could be the administrative burden on the system. I only flag this briefly, because the administrative costs of running the system seem minor relative to the overall advantages or disadvantages of the system.
Another effect of prison choice could be distributional. The better informed will tend to get better prisons—and these include repeat prisoners and those with the best connections in the criminal world. But this objection is strongest when there’s a single, unidimensional measure of prison quality. If “best” is different for different prisoners—if some prisoners prefer proximity to family while others prefer good medical care and still other prefer particular job training programs—choice serves a valuable matching purpose. In the extreme case, everyone could have their best choice. Moreover, even if there’s some redistribution from less to more informed, a rising tide could lift all boats: if prison vouchers lead to quality improvements, the uninformed may become better off as well, much like school choice can improve public schools for students who don’t use vouchers.
I’ll spend the rest of this section focusing on a particular dynamic that seems especially relevant in the prison context: self-segregation of prisoners based on level of violence, gang membership, or (relatedly) race.
In school, the quality of one’s fellow students is an important determinant of the quality of one’s education. Thus, students (or their parents) seek out schools with “better” students, and there may be strong pressures for stratification, though how important these pressures are empirically depends on how the particular voucher program is designed.
If such stratification occurs in schools, it can have various effects.
First, there’s the mere fact of segregation—as when students seek out other students of their own social class, ethnicity, religion, or race. This needn’t have any effect on educational outcomes, though some may consider segregation (even voluntary segregation) undesirable in itself.
Second, there are peer effects—when “good” students move from school A to school B, they may pull down educational quality in school A and raise it in school B. This process has systemic effects if selection is on characteristics that are correlated with high achievement, like socioeconomic status: socioeconomic stratification then leads to stratification of students by achievement. Since students benefit from being with “higher-quality” students, the departure of good students creates a negative externality for the remaining students.
Whether this applies to prisons depends on how important stratification pressures are, and how important peer effects are in the prison context. As an initial matter, note that the predicted stratification effect for schools comes about (if at all ) not only because of peer effects but also through tuition. Rich people are willing to pay more for a better education, and one’s own achievement increases when one is with smarter students; thus, a private school can benefit from attracting high-ability students with low tuitions and charging high tuitions to attract rich students to join them.
But prisons wouldn’t charge fees, so the tuition-based pressure for stratification would be absent. If prisons couldn’t be selective in their “admissions,” stratification would only be based on differences in tastes. If everyone had identical preferences for prisons, everyone would rank prisons equally, and a random cross-section of inmates would end up in the best prisons. If stratification occurs, it will be because prisoners themselves differ in what they value. Prisons with effective violence-prevention policies will acquire the prisoners who put the highest relative value on low violence (bearing in mind that prisoners can still only select within their security classification ); good-medical-care prisons will acquire the prisoners who put the highest relative value on good medical care.
Let’s consider self-segregation by violence level. Would a stratification of the system into “safer” and “more violent” prisons be beneficial?
First, let’s flesh out the concept of “safer prisons.” If a prison merely had low violence rates, relatively nonviolent prisoners would like such a prison, but so would violent prisoners who want to prey on nonviolent prisoners. So selecting for low violence rates probably isn’t a great strategy for a prisoner. Rather, a prisoner who values safety should select a prison that has policies in place that encourage lower violence, avoiding prisons that merely happen to be less violent because of, say, their inmate composition. A prison with effective security policies would tend to attract nonviolent prisoners and deter predators, who would prefer to stay in prisons with comparatively ineffective policies.
What would such stratification do to the overall “quality” of the inmates? Caroline Hoxby writes, in the context of “human capital segregation” in schools (the tendency of high-quality students to live around, and thus go to school with, other high-quality students):
[W]ithout knowing how spillovers work, we do not know whether the equilibrium has too much or too little segregation. Consider the “one bad apple” scenario. If a single household with low human capital in the district could make everyone else learn substantially less, yet would only experience small human capital gains itself, there would obviously be too little segregation. The converse scenario might be called “one shining light.” If a single household with high human capital could make everyone else learn substantially more, yet its own children would not learn any less (despite being surrounded by children from deprived backgrounds), there would obviously be too much segregation.
Prisons seem more like the “one bad apple” scenario. High-violence inmates are a bad influence on low-violence prisoners (because the low-violence prisoners are either brutalized by or learn bad skills from the high-violence prisoners), and “the bad eggs [or apples?] seem to have more of an influence on the good eggs than vice versa.” If this is so, then stratification by violence level may be desirable.
But what’s desirable under the assumption of perfect stratification may be less desirable if stratification is incomplete. Imperfect stratification might be extremely bad for some low-violence inmates who don’t get their preferred choice of prison and are stuck among a much increased population of high-violence inmates. This is less likely to be an issue for human-capital segregation in schools, because being stuck among a lot of bad students usually isn’t that bad.
So violence-based self-segregation may be beneficial in the extreme case, but could be either beneficial or harmful in the imperfect case.
What if inmates self-segregate according to race and ethnicity? This is similar to self-segregation according to gang membership, because race is an important driver of prison violence, with prison gangs organized along ethnic and racial lines.
Estimates of prison gang membership vary extremely widely. Some sources report that prison gangs account for no more than 5% or 6% of prison inmates. Others estimate that mean gang density at male institutions rose from 9.4% in 1991 to 24.7% in 1999 (3.5% to 7.5% at female institutions). The 24.7% figure masks variation by security level —16.1% in minimum-security prisons, 23.6% in medium-security prisons, and 32.7% in maximum-security prisons. And even these numbers mask substantial variation from area to area (numbers like 75% in California prisons and 90% in Illinois have been reported) and among individual prisons (some have reported 0% gang density and others 100%). Estimates of gangs’ contribution to prison violence also vary greatly, from less than 30% to over half.
The prevalence of prison gangs—even if the low estimates are right, there are some systems with high gang density—suggests that racial/ethnic and/or gang-based segregation is a realistic possibility. On the one hand, many prisons are already partially racially segregated, at least within cells, and also when inmates interact with other inmates in the general population. But on the other hand, explicit racial segregation—for instance, where inmates of different races were kept in different cell blocks—is on the wane, and in some systems there’s even integration at the cell level. So there’s still plenty of opportunity for both interracial contact. It’s thus plausible that inmates will seek out other inmates of their own race or ethnicity, if only to be less victimized by opposing gangs.
All this would tend to make voluntary self-segregation, either by race or by gang affiliation, a positive force for prison security. (Though, again, imperfect segregation could be bad for inmates of the minority racial or ethnic group or of the minority gang who don’t get their first-choice prison.)
This may be too quick a conclusion: whether racial integration actually increases violence is disputed. One study has found that desegregation doesn’t increase prison violence—in Texas, where the prison system has been progressively desegregated since the late 1970s, the inmate-on-inmate assault rate was generally less among inmates who were racially integrated in double cells; moreover, the rate of racially motivated inmate-on-inmate assaults decreased as the prison system became more desegregated.
This is an interesting result, though the study has some limitations. First, inmates were ineligible for placement in racially integrated double cells if they were members of racial or ethnic gangs or if they had had previous race-related problems in prison. (For instance, when Texas was starting to desegregate, gangs pressured inmates to resist desegregation efforts; “[i]n some cases, serious assaults were perpetrated on cell partners to earn a ‘racial restriction’ and be placed in a single-race cell.” ) Second, not all interracial assaults were coded as “racially motivated”; it’s possible that some interracial assaults weren’t considered to be directly motivated by race but were nonetheless motivated by, say, membership in opposite-race gangs (when such membership wasn’t known to the prison authorities so as to disqualify the member from integrated double-celling).
The bottom line of all this is that if prisons become racially more homogeneous, this may make little difference. To the extent it does make a difference, it might actually reduce violence, though life could become worse for some who are stuck in “the wrong prison.”
Finally, regardless of the effect on violence, some might still find the “social meaning” of such racial stratification objectionable. But this voluntary self-segregation probably wouldn’t be unconstitutional, and even liberal scholars have been much more accepting of some degree of racial segregation in prisons (for instance temporarily, after race riots) than in other contexts. Moreover, while there’s wide agreement that children benefit from being exposed to other types of children, this is much less obvious when it comes to prisoners.
“Market Success” Arguments Against Vouchers
But let’s suppose that the “market failures” discussed in the previous section aren’t too bad, and that this sort of competition really would improve prisons from the prisoners’ perspectives. Would this be good?
Some of prisoners’ preferences seem desirable. Prisoners probably prefer less violence and sexual abuse, better health care, and better vocational training, other things being equal. They also may prefer prisons close to their family, to facilitate family visits. These all seem unobjectionable, either on humanitarian grounds or on rehabilitative grounds, though many can be found who quietly—and sometimes not so quietly—favor more brutal and abusive prisons on retributive or deterrence grounds. To the extent we favor these preferences of prisoners—and to the extent the political process slights them—this is an advantage of prison vouchers, which would beneficially remove prison reform from the hands of unresponsive democratic majorities and put it in the more caring hands of impersonal market forces, that is, of prisons catering to the praiseworthy desires of prisoners.
But not all preferences of prisoners should be satisfied. (If prisoners had nothing but praiseworthy preferences, why would we have locked them up?)
First, let’s consider prisoner preferences that shouldn’t be satisfied, even though they’re morally neutral. Consider amenities like gym facilities and television choices. These are commonly derided as “country-club” amenities, though they are beneficial in maintaining prisoner discipline. Similarly, conjugal visits, far from being a frivolous luxury, may be important on rehabilitative grounds: conjugal visits may reduce prison rape, and helping the prisoner to maintain family connections and marital ties may also reduce recidivism.
But in general, the concern over prisoners’ enjoyment of amenities—for example, pleasant weather conditions—can be sound. Any amenity that improves the prison from the prisoner’s perspective also presumably dilutes its deterrent value, and therefore may be undesirable from an optimal deterrence perspective. Similarly, from a retributive perspective, even (and maybe especially) the small pleasant things in life are good candidates for elimination, precisely because they’re pleasant.
One can undo the effect of increased amenities by, say, increasing the prison term, but this is highly expensive relative to making prison stays shorter but less pleasant. In fact, strictly from a deterrence perspective—considering the high social costs of imprisonment (and ignoring any negative effects of bad prison conditions on rehabilitation)—it may be more efficient to concentrate on making prison conditions worse than to lengthen prison terms.
Of course the full policy analysis is more complicated: the partisans of prison brutality and rape may be right on deterrence grounds but wrong on rehabilitation grounds, as brutalization may make prisoners worse people when they exit prison. So having a high-amenity prison may, on balance, be socially desirable. But it’s still possible that some amenities should be kept low.
Second, let’s consider prisoner preferences that are affirmatively harmful to the outside world.
I’ve already mentioned that we may want to force certain prisoners into certain services, like psychological services for the mentally ill. They may not choose that for themselves; mentally ill prisoners’ distaste for psychological treatment would thus be socially negative. More generally, prisoners may not value rehabilitation, even if it works (aside from immediately useful rehabilitation like vocational training). Some sort of training may have to be mandated.
In a sense, this is similar to mandated curricula in the school context—maybe parents can’t always be trusted to choose what’s best for their children—but in a sense it’s worse: imagine how much more we might be mandating in the curriculum if the kids themselves were doing the choosing. Solving the agency problem—unlike with schools, those who choose prisons would be the same as those who have to be there —may exacerbate these negative externalities.
A related concern is that self-selected prisons could be breeding grounds for racially separatist or extremist religious movements. One does hear the complaint that prisons are already “fertile grounds for radical Muslim chaplains to recruit” adherents; should we adopt a system that could lead to a Muslim prison? (Compare this with fears about school vouchers based on the possibility that parents will have bad views of education and will send their kids to witches’ coven schools or madrassas or to schools that match their race or socioeconomic class.)
To the extent this concern is warranted, prison vouchers might actually alleviate the problem. Some of the concern does stem from the activities of Muslim clerics, but many blame primarily an inmate-driven “breed of ‘Prison Islam’ that distorts [traditional] Koranic teaching to promote violence and gang loyalty.” The heavy involvement of radical inmates in Muslim observance in prison, in turn, stems from an “acute [Muslim] clerical shortage”: as of 2006, there was one chaplain for every 900 inmates. Moreover, most Muslims in prison entered prison as non-Muslims, and about 80% of prisoners looking for a religion convert to Islam. Allowing Muslim prisoners to self-segregate may alleviate the clerical shortage (if there are economies of scale in chaplaincy), and may also reduce the amount of recruitment among non-Muslims.
So the concern might not be justified for the Muslim example; but perhaps the more general concern might be justified in other cases.
Would prisoners prefer prisons that help them escape? Why yes they would, but this preference is easy to control. Prison escapes are (eventually, even if not immediately) highly public affairs, and so private prisons could simply be heavily penalized (up to the firm’s losing the right to operate) for escapes. But while the escapes example is silly, it shows that, when the undesirable activity is hard to observe, voucher prisons may have broad scope to cater to prisoners’ antisocial preferences.
One example would be easy access to contraband. If the contraband is pornography, perhaps voucherized prisons wouldn’t have much of an incentive to control it; but perhaps this wouldn’t be terribly harmful. If the contraband is drugs or tobacco, perhaps again voucherized prisons wouldn’t have much of an incentive to control it, except to the extent drugs make prisoners violent, in which case prisons would have some incentive to control drugs in order to attract security-conscious inmates. The last point applies as well if the contraband is weapons—if prisons have good incentives to improve security, this translates into incentives to control the flow of weapons into the prison.
In any event, even if prisons would have few incentives to control contraband, they already do a bad job of controlling it now. It isn’t surprising that they don’t effectively control the flow of drugs, even into high-security prisons—inmates often have drug ties, and the stakes are extremely high, with illegal drugs and tobacco being sold in prison for up to ten times their street value.
Prisons may also choose to cater to criminals who want to run their criminal enterprise from within the prison; the prisons could become attractive to such criminals by loosely monitoring incoming and outgoing mail, phone calls, and visits. I take up related concerns in the discussion of gangs below.
Cell phone smuggling is a related problem—“[i]llegal cell phones are used to circumvent supervision of conversations, and can be used by inmates to orchestrate criminal activity, plan escapes, and be a menace outside of prison walls.” Inmates have always been able to plan criminal activity or intimidate witnesses through prison visits by confederates, calls on approved prison phones, and other means—but communication with the outside has its costs. La Nuestra Familia members resort to various methods like using the Aztec language, “micro writing,” codes, messages hidden in artwork (for instance, by writing invisible messages in urine), and long messages sent through paroled members. Cell phones can reduce these costs substantially. And detecting smaller and smaller modern cell phones has become harder in recent years.
To the extent cell phones encourage disorder and violence within the prison, perhaps we can already expect voucher prisons to police cell phone smuggling. Unfortunately, the same won’t be true if, as seems likely, the main disadvantages of cell phone smuggling flow to people on the outside, like intimidated witnesses or future crime victims. But cell phone jamming technology may be an adequate technological fix in this case.
The bottom line on smuggling is that prisons will have incentives to control some forms of smuggling and won’t have incentives to control other forms. It’s possible that the government can control contraband smuggling. Smuggling can still be measured, albeit imperfectly. The government could try to smuggle contraband into the prison—perhaps by using a fake inmate to try to acquire drugs or a cell phone, or by using a real inmate whose cooperation has been purchased—and see how often it succeeds. A prison that tolerates smuggling could thus expect to suffer penalties. But this strategy may not work very well. Some smuggling—the kind that is harmful but that isn’t contrary to the interests of prison administrators—will continue to be a problem.
Recall the discussion above of self-segregation by race or ethnicity. I argued above that such self-segregation may not be bad for inmates, since whatever the benefits of diversity for students, these benefits are probably small for adult prisoners, and given the correlation between race/ethnicity and prison gangs, such self-segregation may actually have the beneficial effect of reducing violence.
But this self-segregation has negative effects beyond prison. Being around members of one’s own community—and members of one’s own outside criminal community—makes prison a less undesirable place to be, and therefore qualifies as one of those amenities that reduce prison’s deterrent value. In addition, many gangs, like the Mexican Mafia, La Nuestra Familia, or the Nazi Low Riders, operate on the outside as well as in prison. (In fact, various outside gangs may have been fostered in prison. ) Thanks to telephone and mail monitoring and other measures, gangs that operate across prisons and in the outside world have problems communicating. Concentrating gang members in one place would reduce these communication costs. This is why various prison systems try to disperse gangs throughout many different institutions. Individual prisons also try to isolate gang leaders or members, for instance by transferring them among prisons or even out of state.
But concentrating gang members might not be all bad: dispersing a gang increases its recruitment opportunities, so self-segregation might reduce the gang’s power to some extent. Moreover, other aspects of the prison voucher system might also reduce the power of prison gangs: prison gangs control street gangs, for instance through the extortion of dues, in large part by threatening violence against gang members who don’t pay and by offering protection to gang members who do. So to the extent that prison vouchers lead to a more secure prison environment, this carrot and stick would become less valuable, which would tend to reduce gang power.
This whole discussion implies a possible continuing role for government regulation, to prevent prisons from offering amenities that are too attractive (without countervailing benefits for society as a whole) and to prevent prisons from catering to prisoners’ socially undesirable preferences. One way of implementing this would be to have an oversight agency with the ability to prevent prisons from offering, or competing on, particular amenities that it finds to be undesirable.
Note that this would be a total reversal of the current political dynamic, where political forces keep prison quality low, and it’s an uphill battle to improve prison quality.
But because not all undesirable amenities can be monitored, we will have to accept some amount of undesirable activity as a result of prison vouchers. The “market success” argument is a significant argument against prison vouchers, because social losses from the satisfaction of undesirable prisoner preferences may well be great. (The market success argument against prison vouchers is thus stronger than the corresponding argument against school vouchers, since inmates’ undesirable preferences are probably more harmful than schoolchildren’s parents’ undesirable preferences.) The question is whether this residual amount of undesirable activity outweighs the desirable consequences of prisoner choice—for instance, reducing prison assaults and rapes, improving medical care, alleviating overcrowding, providing better job training, and the like.