« Multiple sections, multiple professors | Main | How many lives could PAS save? (PAS V) »

Thursday, October 29, 2015

Yes Virginia, there is a trial penalty, and it's four times larger than we thought

Last year David Abrams, Penn, stunned the criminal law world with a study concluding that defendants actually receive shorter sentences at trial than they do for pleading guilty. Rather than "penalizing" those who exercise their right to trial, we actually punish those who plead guilty more harshly, turning decades of plea bargaining debates on their head.

Abrams was wrong, and he’s not the only one.

My latest article reveals significant conceptual and statistical errors in the canonical methodology that cause most other studies, including those by the United States Sentencing Commission (USSC), to greatly underestimate the trial penalty. Underestimating the Trial Penalty: An Empirical Analysis of the Federal Trial Penalty and Critique of the Abrams Study.84 Miss. L. J. 1195 (2015) (Selected through peer review). Where leading researchers report that the federal trial penalty is only around 3-15%, I find that the average federal trial defendant receives sentences around 64% longer than if they had pled guilty instead. In other words, federal defendants cannot exercise their constitutional right to trial unless they are willing to risk a 64% longer sentence, a heavy "penalty" indeed. Where Abrams reports that Chicago defendants pay a similar penalty for pleading guilty, I reveal that his data actually suggests that plea defendants receive shorter sentences than those who go to trial. 

Applying these findings, I demonstrate that the federal trial penalty is so large that only a tiny fraction of defendants could ever rationally choose to go to trial. In such a system, the constitutional right to trial by jury becomes less of a “right” and more of a trap for fools.

Why are my findings so different than everyone else's? Glad you asked.

[More after the fold]

The biggest reason prior federal studies underestimate the trial penalty is that they fail to include the effects of the "acceptance of responsibility" discount. Under the Federal Sentencing Guidelines, defendants who “accept responsibility” by pleading guilty automatically receive a 2-3 point discount to their sentences, but lose this discount if they insist on trial. As such, it operates as a statutory plea discount, or "trial penalty," that sets the baseline for all plea negotiations. Because prior studies do not include the effects of acceptance of responsibility, they heavily underestimate the price defendants actually pay for going to trial. Acceptance of responsibility is written into the guidelines and the USSC's data itself, causing prior studies to miss these effects. Indeed, it is impossible to measure the effects of acceptance of responsibility without reverse engineering the data with several hundreds of lines of code, which I think I am the first to do.

I next explain that rather than measuring the "trial penalty" as that term is understood in crim law debates, Abrams asks whether a rational defendant would be better off going to trial. Defendants are better off going to trial if they face a negative "Abrams Trial Penalty," which Abrams claims to find. Abrams is quite upfront that he is not measuring the traditional trial penalty and, indeed, argues that crim scholars should focus on his new metric. The problem is that crim scholars generally do not discuss this metric because it is usually impossible to measure. A positive Abrams Trial Penalty would indicate that the average plea defendant receives a percentage discount that is larger than their percentage odds of acquittal if they went to trial. Because plea defendants do not go to trial, however, we cannot know what their odds of acquittal would have been without a highly expensive controlled experiment.  (FYI, I'd be open to grants to explore the question!). As I explain, Abrams' innovative methodologies cannot overcome this fundamental problem. By reanalyzing Abrams' findings, however, I show that the normal "trial penalty" in Abram's dataset is likely positive: defendants do pay a price to go to trial.

In addition, Abrams and many prominent scholars report average sentences as the average sentence excluding defendants that receive probation only. Because probation only is the lightest sentence you can receive, however, excluding those cases artificially inflates the average sentence and produces a metric that is quite misleading and largely irrelevant to crim law debates. Nonetheless, many well known sentencing scholars, including some commissioned by the USSC, persist in reporting the "average incarceration sentence" rather than, or in addition to, the actual average sentence defendants receive. As I explain, this peculiar metric appeared in the 1980's due to fundamental misunderstandings about the nature of censored data and selection effects, and confusion about the proper application of the Heckman 2-step correction factor and Tobit regression to control for censoring. (End stat technobabble). This and other common methodological errors are discussed further in my piece. (I put most of the technical stuff in footnotes).

One major limitation of my study is that, like virtually all sentencing studies, it cannot account for the effects of charge bargaining. Because charge bargaining works to increase the trial penalty, however, it does not affect my final conclusion that for the vast majority of federal defendants, trial by jury is not a "choice" or a "right." It is a "mistake."

Posted by Andrew Chongseh Kim on October 29, 2015 at 01:13 PM in Criminal Law, Judicial Process, Privilege or Punish | Permalink


There very well may be a trial penalty. But, is that necessarily a bad thing? Here is what I mean. If a case is plea bargained, the sentencing judge knows much less about the case and the defendant than if the case goes to trial. It is one thing for a sentencing judge to read some (oftentimes very brief) offense conduct facts written on a piece of paper that has usually been produced through negotiation of counsel. It is quite another for the judge to hear the victims testify live about how the defendant terrorized them, to see the defendant show no emotion, to see the defendant testify and lie, to see the defendant acting belligerent with his lawyer at counsel table during trial. All of those are things that I have witnessed during trials, and I can't help but think a judge who sits in a courtroom and sees such things during a trial is more inclined (perhaps rightly) to impose a higher sentence. Indeed, I know that some defense attorneys URGE their clients to accept deals in cases involving child victims because the defense attorneys know that when proof of the defendant's despicable conduct is displayed in a courtroom over the course of several days, the defendant's sentence is going to be much higher. It is not necessarily that that judge is punishing the defendant for going to trial, it is instead that the defendant is getting the punishment he deserves from a judge who is well aware of all of the facts. So, yes, those who go to trial may receive higher sentences. But, I think it is more a product of judges having a better sense of what the defendant did and who he is than it is a product of judges trying to penalize a defendant for choosing to try the case.

Posted by: Maybe | Oct 30, 2015 8:14:15 AM

Yes, the acceptance of responsibility gives a 2 to 3 point reduction to the sentencing level. The other element is that when there is a plea agreement charges are dropped. Defendants who go to trial have often been over charged. A defendant who accepts a plea almost always has charges dropped.

Nonviolent drug offender who are serving sentences of life without parole (at least those I have found - there is no data base - have two things in common, they were charged with conspiracy and went to trial. It took some time to realize this.

beth curtis
life for pot

Posted by: beth curtis | Oct 29, 2015 9:19:59 PM

It bears noting that Albert Alschuler has criticized David Abrams' study in his essay, "Lafler and Frye, Two Small Band-Aids for a Festering Wound." Alschuler reports that Abrams improperly coded many cases as "acquitted at trial" when charges were actually dismissed prior to trial. As such, Abrams' data artificially inflates the trial acquittal rate, making trial look much more attractive than it actually is. Alschuler's article gives sufficient reason to doubt Abrams' final conclusions about a negative trial penalty. My article reveals that Abrams' data likely shows a positive trial penalty even if we take his data as accurate.

Posted by: Andrew Kim | Oct 29, 2015 7:50:41 PM

The comments to this entry are closed.