More ostrich

I had a transfusion yesterday. It took six hours. By the time I was done, I had no energy to reply to the extremely thoughtful comments I received in relationship to More about AG Holder as an ostrich. I feel better this morning, so I will use this post to reply to the comments and, more importantly, concentrate on the 16 other social science variables specified by Dr. Oleson to get at the issue of a defendant’s likelihood to reoffend.

Let’s face it friends, “race” is low-hanging fruit. It is too easy to attack, although the social science data on race when used as a predictive metric for sentencing is not really about genetics (“race”) as a causative factor in crime. It is about being correlated with crime, and there is a huge difference between the two (causation and correlation). But the word “race” is too freighted with the notion of “discrimination,” so let’s just agree for the sake of argument that empirical data on “race” will never be used at sentencing.

OK. But what about the other social science predictors? Things like gender. Or socio-economic background. The things that our delicate AG Holder seemed so frightened about.

Dr. Oleson’s complete list, with the exception of race, is set out below together with the page of his second article (attached to yesterday’s post) where the discussion about the variable is found initially:

1. Criminal Companions ……………………….. 1353
2. Criminogenic Needs …………………………. 1354
3. Antisocial Personality ……………………….. 1354
4. Adult Criminal History ………………………. 1355
5. Race .……………………………………… 1356
6. Pre-Adult Antisocial Behavior ………………… 1359
7. Family Rearing Practices …………………….. 1359
8. Social Achievement ………………………….. 1360
9. Interpersonal Conflict ……………………….. 1361
10. Current Age ………………………………… 1361
11. Substance Abuse ……………………………. 1362
12. Family Structure …………………………….. 1363
13. Intellectual Functioning ………………………. 1364
14. Family Criminality ………………………….. 1365
15. Gender …………………………………….. 1365
16. Socio-Economic Status of Origin ……………… 1366
17. Personal Distresss …………………………… 1366

Why shouldn’t a federal judge take these 16 other variables into account when deciding what sentence to impose, particularly when the judge is trying to reduce the federal prison population be selecting out the “non-violent?” Go ahead, make my day!


32 responses

  1. I don’t see how you can just slice “race” out of the list and pretend we’re not still talking about systemic bias. The point is that black Americans are arrested disproportionately, and the reason they’re arrested disproportionately has to do with the history of race relations in this country and the continued uneven enforcement of the laws to this day — including, I would argue, the largely unconscious use of race by trial judges (especially in state courts) at sentencing. The simple fact is that if a black kid growing up in Englewood and a white kid growing up in Highland Park behave in exactly the same way between the ages of 12 and 17, one of those kids is much, much more likely to have a four-page juvie rap sheet by the time he reaches adulthood. So which one of these hypothetically identical 18-year-olds is going to have identifiable “criminal companions” and “pre-adult antisocial behavior”? And when that black kid gets picked up at age 22 with an ounce of weed (which his white counterpart smokes freely when he gets home from his new banking job), and gets locked up for seven years, will that have an effect on his newborn son’s behavior? And when his son grows up, we’re now dinging that kid for his “family structure”, “family rearing practices”, and “family criminality”.

    The fact is, people from low-income communities commit more crimes than other people. They are much more likely to be arrested when they commit a crime. They are much more likely to be prosecuted once arrested. I’m going out on a limb and guessing they’re also at least somewhat more likely to be convicted once they’re prosecuted. The correct policy response to this is not to lock those people up longer (on the basis of the very things that led to the systemic problems in the first place). It is to ameliorate their crushing poverty to the extent possible, to get the police to start treating them as a community to be protected rather than an enemy force to be occupied, to get serious about the idea of rehabilitation and reform our penal system. I say again — I am making a moral argument, not an empirical one. If we don’t share the same basic framework it’s not surprising that we wouldn’t wind up seeing eye to eye on this, but I can’t tell for sure whether you disagree with me that, at least to some extent, it’s the criminal justice system that causes these disparities in the first place.

    Best wishes, Judge. Hope you’re feeling a bit better today.

  2. Judge,

    Why can’t you consider (most of) these factors now? In other words, why can’t you consider recidivism data as evidence for the level of punishment needed to provide “adequate deterrence” and to “protect the public from further crimes” under 18 USC 3553(a)(2)(A) and (a)(2)(B)? Judges like Posner have already stated that Section 3553 covers the “whole range of penological theories.” US v Dean, 414 F3d 725 (7th Cir 2005). And statistical studies of this sort would be admissible in a trial, so they could surely be used in a sentencing proceeding where the Federal Rules of Evidence don’t apply. The trick is how far you can go to encourage (or even require) the parties to submit this sort of evidence.

  3. Criminal companions? Nothing racial in there, because so many poor people of a certain darker skinned race hang out with corporate CEOs, Nobel Prize laureates and the occasional Bishop, winter in Cannes and summer in Southampton, none of whom have any significant criminal history points.

    Oh wait, that’s us. Never mind.

  4. I guess that I’ve missed something in this discussion, but, it seems to me, that all of these factors (excepting race) are parts of a defendant’s “history and characteristics” and, therefore, must be considered per 3553(a). Most but not all are mitigating, as I have argued many times and as has been recognized by more than one court. On the other hand, they are also correlative to increased possibility of recidivism. But I don’t and can’t agree that the presence of these factors (except for adult criminal history) supports a longer term of imprisonment for whatever crime or crimes has brought the defendant up for sentencing. I don’t see how a sentence of 7 years as opposed to say 4 can be justified based on these factors, excepting, again, adult criminal history. They do suggest that post-release supervision might be more effective in addressing these factors; so how that might be better done seems the more compelling issue.

  5. I agree Mr Abell. At some point i believe i have argued each of these in allocution. Strong family support, lack of violence on his record, good reports from treatment, etc. makes him/her a good candidate for probation.

  6. Criminal companions: depends what is meant by this term, could have first amendment free association problems if the companionship was not itself criminal.

    2-4: sure, you’re allowed to consider these, in fact for some, you’re required to

    6-13: you can consider these.

    14: No dice. You’re not going to convince me that it’s just to punish people based on the behavior of their family. Even though the corruption of blood rule in Article III is written solely regarding treason, its reasoning provides strong guidance in determining equal protection under the 5th and 14th amendments. And in the unlikely event that the family member has been convicted of treason, then it’s explicitly banned.

    15: Again, no dice. It is fundamentally unjust to punish someone for their genes. Even if you could sneak this by the Supreme Court (I doubt it), the Congress would ban it the next day. And for good reason, it’s immoral.

    16: First, this shares the problem of 14-15 in being an unjust ground for punishment. Second, in as much as it highly correlates to race, it can be accurately seen as race by another name. And third, the use of this to calculate reoffense chance directly works against the “just deserts” aspect of sentencing, since to the extent that socioeconomic status is a factor of criminality, it is a factor cutting against harsh punishment for moral responsibility.

    17 sure, you can consider mental illness.

    Ultimately, the problem is that even when talking about protecting society and preventing recidivism, the court is tasked not with evaluating the expected future path of the defendant, but with punishing the defendant for a specific crime of which he has been convicted. The factors objected to (race, gender, family, and socioeconomic status as a proxy for race) are objected to because they are inherently unlinked from the crime of which the defendant has been convicted.

    To use those factors is to say that two people, both alike in moral turpitude, should be punished differently due to factors they could not possibly control. That is the essence of what is being banned by equal protection of the law. To the extent that predictors of reoffense are factors about the defendant that are within the defendant’s control, that’s an appropriate ground for punishment. To the extent they are not, punishment is unjust. 100% of a sentence is punishment, and the sentencing factors must therefore be limited to things which one can justly be punished for.

  7. Judge:
    Of the factors listed, as a sentencing judge you would likely still have a problem with gender for the same reasons that race would be problematic. However, once you accept the idea that a particular characteristic is, as you put it, a correlate rather than a direct causal factor, you are accepting that a particular factor can be a dependable proxy for the likelihood of a defendant reoffending. Of all the factors listed, above, prior criminal history appears to be the least objectionable in its usage because it is the most dependable as a proxy in this regard (gradually less so for the other factors, and, IMHO, not at all for immutable factors such as race or gender). But the question really becomes: are the other factors dependable proxies for a sentencing judge in selecting out the non-violent in order to lower prison populations while protecting the public? I’m guessing you would say “yes”, perhaps resoundingly. I am not so sure.

  8. I would use adult criminal history, substance abuse issues, level of stress (assuming that can be determined), sociopath (assessment based on much more than a screen) and are their adequate community support services for persons under community supervison.

    The sample sizes for criminal risk assement evaluations are too small to give reliable results if you include too many variables.

  9. Griff,

    As long as you recognize that your moral argument runs counter to the idea that we should reduce prison populations by trying hard to let out the non-violent least likely to reoffend, but keep in the violent and likely to offend, I understand completely. Unlike AG Holder, you realize that you can’t have your cake and eat it too.

    All the best.


  10. Justin,

    I believe I can, but why does AG Holder then complain about “basing sentencing decisions on static factors and immutable characteristics—like the defendant’s education level, socioeconomic background, or neighborhood” or “on the possibility of a future crime that has not taken place.”

    Now, understand Holder is talking about “a new generation of data-driven criminal justice programs . . . .” For example, there are data sets (sometimes massive) of the background of prisoners that can be used to show statistically what would happen if your defendant was similar to the output of the model. These are powerful tools.

    All the best.


  11. SHG,

    If you are looking to reduce prison populations through sentencing, and the metric becomes “likely to reoffend” or “likely to violently reoffend,” then if you are honest about it you can’t ignore or reject the social science data that is not directly keyed to race. I think Holder understands this, hates what the data shows, but wants to jump on the “let’s reduce prison population” band wagon–having really pretty cake and gulping it down too. His position is intellectually dishonest.

    All the best.


  12. Robert,

    First, you get a prize for your comment on supervised release. Large social science data sets may well help us to allocate our resources far better when it comes to supervised release. For example, instead of using 3 or 5 years as the typical length of supervised release, we might be able to shorten that period in a lot of cases if the reoffense metric was relatively low for a particular defendant. That would free up time for our probation officers to spend more time with tougher cases.

    Second, given the power of this data for supervised release purposes, I simply don’t understand why anyone would argue against use of this data to separate out the “likely to reoffend” from the rest when handing out prison sentences, particularly if your aim is reduce the prison population.

    All the best.


  13. This is a fair response; but I do share the goal of reducing our prison populations. To get concrete, here are a few strategies I can think of (just off the top of my head) that I would support:

    — Abolish all mandatory minimums
    — Reduce the maximum sentences for all nonviolent crimes
    — Decriminalize all misdemeanors that currently have maximum sentences of six months or less, changing them to civil infractions
    — Legalize the cultivation and possession of marijuana, and the sale of small amounts
    — Decriminalize simple possession (first offense) of every controlled substance; for subsequent offenses, send people only to minimum-security treatment facilities and/or release them to programs like 24-7 Sobriety
    — Make real, quality treatment and education programs available at all prisons
    — Establish a presumption of parole, so that when the individual becomes parole eligible he is automatically released (on supervision) unless the government opposes release and establishes by a preponderance that he should still be held

    These are all potential strategies for reducing prison populations and weighting those populations more toward violent offenders. I think they could all be implemented without having the kind of disparate impact on groups that are already disproportionately affected by our criminal justice policy that I worry about with “actuarial sentencing.”

  14. Peter,

    You write:

    “Ultimately, the problem is that even when talking about protecting society and preventing recidivism, the court is tasked not with evaluating the expected future path of the defendant, but with punishing the defendant for a specific crime of which he has been convicted. The factors objected to (race, gender, family, and socioeconomic status as a proxy for race) are objected to because they are inherently unlinked from the crime of which the defendant has been convicted.”

    That is brilliant Peter! Truly. But, where do find that limitation in 3553(a)? I can make the strictly utilitarian argument that once the government has established that you are a criminal the government has the right and is obligated to consider whether you will continue to be a criminal. There is nothing immoral or wrong about such an approach. It just has a “Clockwork Orange” sort of creepiness to it. But, with the all the “reduce prison population” chanting, that is where we are heading assuming we need a rational sorting mechanism.

    All the best.


  15. Also, it seems much less problematic for a probation department to use empirical risk-assessment tools to decide how much supervision someone needs once they’re on supervised release than it does for a judge to use them to decide how long the person should spend in jail.

  16. If you include gender(female/male), ethnicity (hispanic/other), race (black/other/white) and you need an average of 200 per subset the sample size should be 2,400 at least for that set of factors. If you include four age groups then you need about 10,000.

  17. We don’t find it in 3553(a), we find it in the 5th amendment. Due process of law, including the implicit (for federal) or explicit (for states) equal protection that goes along with it requires that the punishment for a crime be set in advance, be related specifically to that crime, and be the same for everyone. And due process requires that the exclusive precondition for punishment is conviction for an offense. Nothing else can lawfully cause punishment. And no offense under the law could have an innate characteristic as an element of the offense.

    Under due process of law therefore, punishment is, exclusively, for something you’ve done. It can relate to things you’ve done in your life that represent a proclivity to criminality, even if not themselves crimes. But punishment for innate characteristics is anathema to due process and the equal protection of the law, just as conviction of a crime for having an innate characteristic is anathema. It is not, and cannot be, something you’ve done that deserves punishment.

    Sec. 3553(a) can be interpreted fully consistently with this. It does not require any factors outside of the totality of the defendant’s actions to give a sentence in accord with the factors it outlines. Interpreting it in such a way as to include innate characteristics is, I argue, unconstitutional.

    The rule of thumb: if it couldn’t constitutionally be an element of a crime, then it can’t constitutionally be a valid factor in punishment.

  18. In light of your comment in reply, I looked a bit further to see if we do in fact find it in 3553(a). I would argue we find it in 3553(a)(4) and (5) These, as you know, direct the judge to consider the sentencing guidelines and policy statements of the sentencing commission.

    The sentencing guidelines, in turn, are governed by 28 USC 994(d) which states “The Commission shall assure that the guidelines and policy statements are entirely neutral as to the race, sex, national origin, creed, and socioeconomic status of offenders.”

  19. I’ve been working a bit on this. Would you mind if I emailed you my try at convincing you of the point?

  20. Does this assume a binary outcome (reoffend / not reoffend), and would grouping of outcomes (reoffend violent / reoffend nonviolent / parole violation / not reoffend) further multiply the sample size needed?

  21. Peter H.,

    Excellent point. But, now that they Guidelines are advisory, and one can easily read section 3553(a) the other way, one might argue that a judge is not required to ignore empirical evidence on the reoffense question even though it touches upon one or more of the categories mentioned in 994(d).

    Beside, I am not entirely sure empirical evidence of reoffense even if based upon things like socioeconomic factors is non-neutral. Put differently, if socioeconomic factors are treated alike–say we use various income levels as the metric but no income levels are weighted more favorably than some other–aren’t you treating socioeconomic factors in an entirely neutral fashion?

    Thanks for this. All the best.


  22. Even now being advisory, it’s still Congress’ intent that sentencing is to be done without regard to the 994(d) categories. A construction of 3553(a) which is not neutral to those categories is against the most commonsense way to make sense of the two statutes. After all, Congress thought they were mandating guidelines for sentencing neutral to those categories.

    I don’t quite follow your argument about non-neutrality though. If we break it into a bunch of income levels and then treat all of them equally, then it would not vary anyone’s sentence and is moot. If the statistical model for predicting recidivism uses income as a predictor, then it inherently has to weight some incomes as a higher chance of recidivism and an accordingly higher sentence.

    But under 3553(a) income and education of the defendant would be a fine factor to use, as it relates to the defendant’s productive contribution to society (or lack thereof) and more broadly to his history and characteristics. Oleson’s paper was talking about the offender’s parents’ income and educational attainment for socio-economic status, not the defendant’s income.

    It’s also ethically disturbing to use the fact of the defendant’s parents being poor as a reason to sentence more harshly, even if it is correlated with recidivism. Beating a man when he’s down comes to mind.

  23. Peter H.,

    See the post this morning where Dr. Oleson gets at some of the philosophical issues, albeit briefly.

    All the best.


  24. I agree with everything Griff has said (including that you are feeling better), but just wanted to make one other point. You, Judge, are correct that not basing sentencing decisions on immutable characteristics is a choice not to employ uncompromising utilitarianism.

    However, even most libertarians, for instance, reject pure utilitarianism. Robert Nozick, as I am sure you are aware, made an extremely compelling case for utilitarianism with “side constraints,” if one is going to be utilitarian. I think griff (and Holder, read with even a modicum of generosity) are making that point: yes, we want to imprison primarily, only if possible, those who pose a threat to the community, but the side constraints we advocate include not punishing people (I.e. not enhancing sentences) based on immutable characteristics. The reasons behind that side constraint are constitutional, moral, and practical. While it is beneficial to debate whether limiting that constraint to, say, race (as it seems everyone but Oleson agrees is both wise and required), it is not beneficial to stick our heads in the sand by pretending that the contour of those restraints is, quite precisely, inherent in any debate about the use of data-based metrics.

    Some data is okay to use, other data isn’t even if there are correlations precisely because the direction of causality, if any, is hugely consequential from any moral standpoint (including a utilitarian one). Holder drew the line, as did Monohan in the article you linked, at past conduct. That’s the gauntlet.

    Regards, Kerry

  25. Pingback: Better to be Sentenced by Ostrich or Pigeon? (Update) | Simple Justice

  26. It’s hard to add to Peter H’s work, but I would like to make two points:

    Disability, as much as race, is a factor protected from discrimination and unequal treatment under the law. If you want to give people more time for intellectual disability, you’re discriminating on account of disability. I know of no authority in the law to suggest that people should be locked up simply for being stupid.

    A second point is that poor people don’t commit more crimes. They commit more petty crimes and fewer big ones. They shoplift, while rich people throw pollutants into the world that kill thousands of people and severely injure countless more. They violate numerous statutes in the course of doing such things.

    Through accounting fraud and other devices, they cause people to be thrown out of their homes, sometimes causing them to die of exposure or other consequences of landing on the street.

    They corrupt our food supply for profit through countless violations of the law in their factory farms, thereby doing everything possible to promote antibiotic resistance along with periodic outbreaks of serious E coli infections – which when I was young was not even a pathogen! -and thereby kill significant numbers of people.

    And much more could be said. And these very consequential crimes are usually rewarded, not prosecuted, and when they are, which of these criminals is ever taken off the street, where they are doing vastly more harm than umpteen petty street criminals?

    It is as Anatole France said: “In her majesty, the Law forbids both rich and poor men to sleep under bridges, beg in the streets, or steal loaves of bread.” And that’s even when the statutes are even-handed.

    Finally, from a public safety standpoint, since the prisons are full of people with disabilities or other attributes that invite them to be classified as criminals, would it not make more sense to impose criminal penalties on school district administrators that knowingly violate special education law so as to endanger society by deliberately refusing to do their jobs so that they can dump these kids into the school-to-prison pipeline – rather than to punish their victims? In another instance, it is well-known that violent crime correlates closely to lead exposure, so that all industrial societies introducing lead into such places as paint and auto exhaust experienced violent crime spikes 20 years later, with corresponding fall-offs as the lead was taken out. Why not enforce the law with vigor on those whose polluting practices promote that violent crime?

%d bloggers like this: