Kopf: The Question of Sentencing Disparity, Part 2

In Part 1, I addressed intra-district sentencing disparity in the District of Nebraska.  I presented eight tables containing judge-specific sentencing statistics for fiscal years 2015-2016[1] and I asked for insights from readers of Simple Justice. As promised, I now set out some of my thoughts and respond to several of the interesting, and often funny, comments.

I start with two assumptions. The great bulk of the readers of SJ loathe the Sentencing Guidelines. Also, I assume that many of you don’t like Attorney General Sessions, and are particularly worried about his May 10, 2017 memo that, among other important aspects, requires federal prosecutors to charge “the most serious, readily provable offense.”

I don’t want to argue about the Sentencing Guidelines writ large, whether Sessions is a dolt or whether you prefer Attorney General Holders’ previous (and dainty) way of doing business. Indeed, if you are now a CDL in the federal trenches, you have no choice but to “Cowboy up” and deal with the legal realism of the moment.

By the way, I offer to pay the costs for our mean-ass host to tat his back like the one depicted in the foregoing photo. After all, as the mid-day cowboy, he enjoys wearing a Gerry Spence buckskin coat.

With these precatory remarks out of the way, I suspect that most would agree that offenders in the District of Nebraska and elsewhere[2] who are similarly situated should be sentenced in a similar fashion. Out of the “ink blot” like sentencing goals found in 18 U.S.C. § 3553(a) (factors to be considered in imposing sentence), we should concentrate on the one factor that is understandable on the face of it (plain meaning) and one that is relatively achievable and somewhat capable of measurement. That is, “avoid[ing] unwarranted sentence disparities among defendants with similar records who have been found guilty of similar conduct.” Id. at § 3553(a) (6).

I now turn to the meat of this post. Overall, Nebraska has a very heavy drug docket compared to the nation. About 45% of our cases involve drugs and more than 80% of the time we are dealing with meth freaks, far higher than what is seen nationally.[3]

There is one thing that jumps out at you right away when you view the judge-specific drug sentencing data. Compared to the Omaha judges, Judge Gerrard and I, who sit primarily in Lincoln, are much tougher when sentencing in drug trafficking cases. As Kirk Taylor wryly put it in commenting on Part 1: “Stay away from Lincoln…especially if you’re a drug dealer . . ..” Realize, please, that Omaha and Lincoln are only 60 miles apart, although it is true that the Platte River is between the two cities. (I threw in the note about the Platte River because, while it probably has no particular relevance to the subject at hand, I feel duty bound to provide a geography lesson as an extra bonus.)

When it comes to drugs, Judge Gerrard and I are way above the national sentencing numbers whether measured by a mean or a median. The national mean was 67 months and the national median was 50.  Gerrard’s mean was 103 and his median was 115. Mine were 91 and 72.  When it comes to drugs, the toughest sentencing judge in Omaha (Chief Judge Smith Camp) produced numbers of 74 and 60 for her mean and median and that was substantially lower than either John or me.

What gives? All I have is speculation.

Fact-bargaining goes on Omaha but not Lincoln, or so I suspect. I am guessing that Omaha does not draw a lot of task force cases relative to Lincoln. Omaha prosecutors rely upon “buy-bust” cases from the Omaha police department. Thus, historical conspiracies that produce large drug quantities due to the extended span of time and which we see all the time in Lincoln are not seen as much in Omaha. Even when they are brought, I suspect Omaha prosecutors are willing to drive down the Guideline range by bargaining about drug quantity. But, as I say, this is sheer speculation.

In any event, I am very happy that, under the leadership of Chief Judge Smith Camp and Judge Gerrard (our next Chief Judge), the District of Nebraska has decided to make a concerted effort to find an explanation by bringing in experts to help us. This will be a big task because it requires, or so I believe, a deep dive into each of the 446 drug cases (and perhaps the 110 firearms cases) set out in the most recent data run.

Next, I take a broader look. In the following bullet points, I examine all the data and not just the drug data.

  • There is an important limitation on the data that often goes unrecognized. If, as is the case in Nebraska, many drug cases involve cooperation that extends after sentencing (particularly in cases involving historical drug conspiracies) and those cooperators are rewarded by Rule 35(b) sentence reduction motions, then the point in time when the sentencing-length snapshot is taken becomes critical. The Sentencing Commission does not account for Rule 35(b) reductions. As we undertake to understand all of our sentencing practices, it will be important to see what the data look like when Rule 35(b) reductions are included assuming that our experts make a concerted effort to develop this new data.
  • Unsurprisingly, I am the nasty bastard in our bunch. With a mean of 71 months in prison and a median of 51 months for all cases, my sentencing numbers are way above the national numbers of a mean of 44 and a median of 21 and substantially above the Nebraska numbers of a mean of 50 and a median of 30. Moreover, I stay within the Guidelines nearly 51% of the time. This compliance number is not accidental. It is consistent with my view that the single best justification for relying heavily upon the Guidelines, even when those Guidelines are harsher than I would personally like, is the overarching need to treat similarly situated offenders similarly.
  • In cases where there is no government request, I vary downward only 20% of the time on my own. Yet when I do vary downward (whether at the request of the government or otherwise) I am lavish. I give a sentencing discount, measured from the low-end of the Guidelines, of a whopping 70%. Indeed, David Meyer-Lindenberg in his extremely thoughtful and detailed comment to Part I, was kind enough to point this out. Thanks David!
  • When it comes to kiddie porn, my numbers are like a poke of a stick in the eye of the reader. My mean was 157 months in prison and my median was 96 months. Those numbers should be compared to my Nebraska The composite for Nebraska was a mean of 111 and a median of 72. But for my numbers I have a likely but only partial explanation. I only sentenced 16 individuals for child porn during fiscal 2015-2016. I am pretty sure that several of those cases involved Rule 11(c)(1)(C) plea agreements involving the production of child pornography in extraordinarily perverse circumstances. That is, for example, mothers and fathers who actually used their own children to produce these materials. Knowing that I would, left to my own devices, max out such offenders, I suspect that defense counsel rightly elected to agree to very long sentences that were binding upon me in order to avoid letting me exercise the vaunted discretion that CDLs talk so much about.
  • I am completely perplexed by the firearm data. As Charles humorously questioned in response to Part 1, “Don’t traffic drugs in Lincoln, but if you do, be sure to carry a firearm?”[4] Judge Gerrard and I were substantially below our colleagues in Omaha and the nation when it came to guns whether measured by the mean or median. For example, my prison sentencing numbers were a mean of 44 months and a median of 17 months and Judge Gerrard’s numbers were 45 and 25. The composite numbers for Nebraska were a mean of 64 and a median of 37. My only guess is that our Omaha judges deal more frequently with hard core gang types when it comes to felon in possession charges while John and I deal more often with dip shits who, while being felons in possession, are not particularly dangerous. In this regard, John and I tend to draw cases from farther west than the Omaha judges. Since I practiced law out there for 13 years, I know that gun-toting but relatively harmless idiots abound in the wilds of our west.

To sum up, I suppose a wag (or a realist) could say that the Sentence-O-Matic 1000 produces reams of data that no one understands. But that doesn’t mean we shouldn’t try to examine what the hell we are doing. Is an empirical approach to understanding federal sentencing a fool’s errand? I hope not.

Richard G. Kopf
Senior United States District Judge (Nebraska)

[1] We have been collecting and publishing this data since 2007. See here for earlier data presentations.

[2] About 14 years ago, our court together with the federal courts in the District of Minnesota and the Southern District of Iowa participated in an in-depth study, funded by the National Science Foundation, of sentencing practices in these relatively homogenous districts. See Cassia Spohn, Sentencing Decisions in Three U.S. District Courts: Testing the Assumption of Uniformity in the Federal Sentencing Process, Vol. 7. No.2 Justice Research and Statistics Association (2005) (unfortunately the paper is behind a pay wall).  Among other things, and based upon sentencing statistics for fiscal years 1998-2000, Professor Spohn found that (1) the odds of incarceration did not vary among the three districts; (2) sentence lengths did not vary at all between Nebraska and Minnesota, but sentence length was 8 months higher in the Southern District of Iowa; and (3) there were significant inter-district differences in sentencing discounts for substantial assistance.


[4] There is an interesting methodological question unwittingly raised by Charles. Do the sentencing numbers from the Commission discussed here and in Part 1 take account of one case where there are separate offenses of conviction? For example, consider one case involving both (1) a drug conviction and (2) a gun conviction. While I laughed out loud when I read what Charles wrote, I am not sure that the data gets to the issue of correlating drugs with guns in the same case. In other words, I don’t think one case is counted twice for purposes of these numbers. I need to confirm my impression with the excellent data crunchers at the Commission.

21 thoughts on “Kopf: The Question of Sentencing Disparity, Part 2

  1. John Barleycorn

    About time! This whole tomayto, tomahto sentencing thing has got to end.

    But I am not sure calling a “feedlot” a concentrated animal feeding operation, as the federal government does, is gonna do the trick.

    Saddle Up Judge, saddle up…

  2. Jake

    Thanks for sharing all this data Judge Kopf. I’ve read both posts carefully and my only comment on the substance of your guest post is that I’m encouraged to know individual judges are making an additional effort to supplement the data provided by the annual Federal Sentencing Statistics report.

  3. Jim Tyre

    After all, as the mid-day cowboy, he enjoys wearing a Gerry Spence buckskin coat.

    Judge Kopf,

    Greenfield dumped the buckskin coat. With such a basic error, how can we trust anything you write?

    Seriously, an excellent pair of posts, thanks.

  4. Austin Collins

    Judge Kopf,

    It seems to me to be an eminently workable data set, particularly using some of the more modern ‘machine learning’ analyses as compared to the typical (even if advanced) statistical methods used in most studies of this sort. (for this data set, state vector machines and k-means clustering come immediately to mind)

    I’m a data scientist (doctoral work in theoretical and statistical physics), and am intrigued by which problems might be tractable within it.

    If you had 3 questions you most wished to have either answered or illuminated by the data set you linked to, what would they be? Seems like a fun side project to explore it, and given the size of the data set, wouldn’t take long.

  5. Richard Kopf


    What three questions do I want to be answered or illuminated? You get right to the heart of the matter. Thank you very much for your insightful comment. Here goes:

    1. With reference to section 3553(a)(6), for analytical and statistical purposes, how shall we define (a) “unwarranted sentencing disparity,” (b) “similar records” and (c) “similar conduct”?

    2. Are the apparent differences between Omaha and Lincoln for drug and gun cases real? In other words, for these cases, are Omaha and Lincoln judges actually sentencing differently in a statistically significant manner and, if so, why?

    3. If one factored into the inquiry the impact of Rule 35(b) sentence reductions (that are not present in the Commission’s dataset) and then one reran the “government below” numbers, what would be the result?

    All the best.


    1. SHG

      I’m required to drink hard liquor every time I see “unwarranted sentencing disparities.” I thought you should know.

      1. Richard Kopf


        That term is the last refuge of scoundrels like me. I am glad you find it as intoxicating as I do.

        All the best.


    2. Austin Collins

      Judge Kopf:

      (1) may be straightforward; if it is, I’ll get back to you with the results and some references to the methods used. At their heart, both of the techniques I mentioned are ‘classifiers’ — meant to predict what ‘group’ a new data point should fall in. However, one can also use them to discern and quantify the ‘distance’ between two complex data points.

      (2) Likely to very likely to be answerable from the data set, if you take out the ‘why?’. ML/AI doesn’t typically address causation, though many algorithms will tell you which ‘components’ for particular data points played the largest role — would that suffice?

      (3) If they are not present in the sentencing commission’s dataset, are the obtainable elsewhere?

      As I mentioned, I’m a data scientist, not a lawyer. So I’m assuming that rule citations will be googlable-enough for me to parse their relevance to the data set. (SHG: relax, not saying I can become an internet lawyer…. just see no need to ask for definitions from experts if they’re Google-able. 🙂 ) . If that’s not the case, any links to references would be appreciated.

      1. Austin Collins

        Apologies for the vagueness in (3) — by ‘they’, I meant ‘Rule 35(b) sentence reductions’ .

        1. Richard Kopf


          The data on Rule 35b reductions can be unearthed with a lot of work. This data will not be easily available to you. It could be done by you, but it is best done from inside our record keeping system called CM/ECF (Case Management/Electronic Case Filing).

          I am very excited by whatever you find regarding your points 1 and 2. I very much appreciate your offer and look forward to learning from you if you undertake the task.

          All the best.


  6. Elaine Mittleman

    If judges want to avoid unwarranted sentence disparities, should there be a chart of previous sentences in whatever category or specific descriptions – that the judge would then look at and try to fit into that chart? That would seem to be unworkable. Or should the judge stay within the guidelines on the assumption that practice would tend to diminish disparities?
    Or should the judge just give whatever sentence seemed appropriate – and then later do an analysis of his median and mean sentences?

    When I prepared a memorandum about cases discussing sentencing disparities, it seemed that the conclusion in cases was typically that there was no unwarranted disparity because the specific facts were different. As an example, one person may have pled guilty early in the process and another person waited until the day of trial to plead guilty. That difference in timing of the guilty plea provided an explanation as to why different sentences did not have unwarranted disparities.

    I guess that I agree with the principle that there should not be unwarranted sentence disparities, but I don’t have much confidence as to how that principle can work in the real world of sentencing by real judges.

          1. Richard Kopf


            Would it be unkind of me to say that, unlike you, the one sweltering and long afternoon I spent decades ago with Brother Spence convinced me that he was mostly all hat and no cattle?

            All the best.


          2. Allen

            That hat needs a different hat band. I would suggest something in venomous snake skin. Nothing says black hat like pit viper.

            1. SHG

              Much as I look quite dashing in it, I don’t wear it much. I wore it to bed the night of that pic, and Dr. SJ had strong words for me about it. I haven’t worn it again since.

Comments are closed.