[Andy Leipold (guest-blogging), June 27, 2006 at 10:03am] Trackbacks
Judge vs. Jury Acquittals, Day 2

Wow. Some very thoughtful and perceptive comments were posted yesterday on the basic question of why federal judges are more likely to acquit than federal juries. This is fun.

Let me pick up on a few of the points that people raised in their posts, to show what the data revealed. These figures are based on a study of about 77,000 federal criminal trials completed between 1989 and 2002. Those who read yesterday's post will recall that I am trying to figure out both why judges are more likely to acquit than juries, and why the judicial conviction rate dropped sharply between the late 1980s and the early 2000s.

Several folks hypothesized that the type of crime involved -- violent, property, drug, etc. -- might explain the disparity. Perhaps certain types of crimes are both steered toward a particular factfinder and are particularly likely to end with an acquittal. For example (I speculated), financial and regulatory crimes might be directed toward judges because both sides worry that jurors will misunderstand the evidence, and perhaps it is also the complexity of these cases that make them hard for the government to win. Or perhaps defense counsel avoid juries in violent crime cases, and so on.

There may be something to this, but the figures don't show much of an effect by case type. I looked at six crime categories: violent, property, drug, immigration, regulatory (crimes involving customs, social security, the mail, etc.) and public order offenses (a group that includes primarily traffic offenses -- really -- and guns). Two things became clear: (1) juries always convict more than judges, no matter what category of crime, usually by 15 to 30 percentage points, and (2) defense counsel prefer juries to judges in all types of cases. In only one category did the percentage of bench trials come close to half: 45% of public order trials are tried by the court. Interestingly, judges convict in only 47% of the public order cases, compared to an 83% conviction rate by juries.

Another line of posts offered the insight that the "conviction gap" might be explained by the strength of the evidence. To paraphrase Judge Posner, perhaps innocent defendants want judges because they worry about a mistake being made, while guilty defendants want a jury trial, hoping a mistake will be made. I think this explanation has great intuitive appeal.

But two brief points on this. First, it is darn hard to measure the strength of the evidence, at least on a large scale. The best indicator we have of the case strength is the outcome, but that leaves us chasing our tail if we are trying to evaluate the outcomes. This doesn't make this explanation wrong, it is just hard to know if it is right.

Second, if strength of the evidence explains a big piece of the conviction gap, we still need an explanation for why the jury conviction rate has remained steady while the bench rate fell significantly over the period studied. So while I like this explanation very much, I worry it is still speculative and incomplete.

I have left out a lot of detail, much of which I hope is provided in the article (on SSRN, at and perhaps in other posts later this week As always, questions and comments welcome at

I still say it's the asymmetry. Bench trials happen only in the cases when defendants want them, Fed. R. Crim. P. 23(a), and prosecutors don't use their blocking position. So, unless defense counsel are deluded as to the kinds of cases that would benefit from a bench trial, bench trials by definition are trials in which defendants have availed themselves of a significant procedural advantage. Not surprising to see more acquittals.
6.27.2006 12:12pm
Keith Wollenberg (mail):
I think the asymmetry is part of the explanation, but that the inflection point occurred just as the Federal Sentencing guidelines came on line is unlikely to be coincidental, in my opinion.

Given fixed sentencing, which many judges do not like, is it much of a stretch to hypothesize that in a close case, a judge might acquit, rather than impose a draconian sentence? Even if the bias were unconscious, it seems plausible that this could have an effect.

Even if we reject this idea, it is likley that the very existence of the guidelines skews which defendants plead, which choose jury trials, and which choose bench trials.
6.27.2006 12:18pm
Richard Bellamy (mail):

Second, if strength of the evidence explains a big piece of the conviction gap, we still need an explanation for why the jury conviction rate has remained steady while the bench rate fell significantly over the period studied. So while I like this explanation very much, I worry it is still speculative and incomplete.

Over the period studied, I am guessing that mandatory sentence lengths have generally increased. Judges always know when their conviction for stealing the loaf of bread will require a 20 year conviction.

Juries, not so much.
6.27.2006 12:27pm
Maniakes (mail) (www):
Perhaps there was some cultural shift in law schools, such that their graduates became more skeptical of prosecutor, or otherwise more likely to acquit. Do you have access to data on where the judges went to school and which year each of them graduated?
6.27.2006 12:39pm
Yes on both counts.
Yes, the guidelines definitely influenced defendants' calculus, as sentencing outcomes became (somewhat) more predictable. Going to trial was still a bet, but at least you're permitted to see the denomination of the currency.
Yes, a judge's verdict may be influenced by his/her distaste for the likely sentencing outcome. I have no idea if the number of cases is statistically significant, but I know for certain of one case of mine in which it occurred (based on a later conversation with the judge).
6.27.2006 12:41pm
P.S. There's no inconsistency between our positions, by the way. "Asymmetry" is kind of a procedural meta-explanation; it really only explains how it is that defendants can effectuate their preferences. As such itapplies irrespective of why defendants perceive a judge as more hospitable in a particular case.
6.27.2006 12:45pm
byomtov (mail):
Do defendants know who the judge will be when requesting a bench trial?

If so, is it possible that they will opt for bench trials if the judge is thought to lean pro-defense? If that's the case then we would expect to see a relatively high rate of acquittals, but also bench trials unevenly distributed across judges.

It wouldn't be the case that judges are more likely to acquit across the board, but that some judges are more likely to acquit, and they get a disproportionate share of bench trials.
6.27.2006 12:45pm
Bob Bobstein (mail):
Totally substance-free comment: Terrific work engaging the commentors' points right out of the box, Prof. Leipold. Not all guest-bloggers have had a great time here, it's great to see you enjoying and benefiting from the experience.
6.27.2006 12:47pm
Bruce McCullough (mail):
If the assignation of each case to either judge or jury trial was random, then such a statistical comparison might shed light on the "conviction rates" of judges and juries. But the assignation is not random. In fact, what may be driving these results is what statisticians call a "sample election" problem.

People who are innocent and who think that a judge can see this are more likely to choose to be tried by a judge. People who are guilty and who think that they can confuse a gullible jury are more likely to choose a jury trial.

Hence, the discrepancy between judge and jury conviction
rates may well have nothing to to do with the judges
and the juries, and everything to do with the defendants
selecting themselves to be tried by judges or juries.

I gave a quick scan to the paper, and I did not see that
it considered the sample selection bias possibility.

Heckman won the Nobel for his work on this problem,
and it's well known in the literature.
6.27.2006 12:54pm
Shelby (mail):
Has anyone with a working knowledge of game theory looked at this issue? It's a classic situation, with one party (the defendant) adjusting its actions based on its perception of how other parties (judge, jury) will act in response to anticipated events (presentation of evidence). This is largely a more formal way of putting the "strength of evidence" argument; apologies if it was advanced in the previous post's comments.

Maybe Slate's economist, having evaluated penalty kicks under game theory, is willing to have a go at this.
6.27.2006 1:36pm
Maybe Judges and Juries have different standards of reasonable doubt?
6.27.2006 1:45pm
I think that Judges take a more critical look of the actions of the police then juries do.
6.27.2006 1:49pm
byomtov: Generally, the assignment of a judge is random, but you do know who your judge is early on. I wonder if requests for bench trials depend on the judge you get?
6.27.2006 1:54pm
CJColucci: Don't wonder. Of course a defense lawyer will consider the judge. In the district where I practiced, there were judges to whom no defense lawyer would entrust the fact-finding function, irrespective of other considerations; they'd take their chances with 12 people off the street.
BMcCullough: sample selection is of course an underlying flaw. Are judges more likely to acquit? Are acquittals more likely in the kinds of cases where defendants opt for a bench trial? Too many confounding factors to permit a conclusion.
6.27.2006 2:01pm
Londo (mail):
Could it be that Judges convict on a curve, while juries tend to get one and only one shot at it? A guy who steals $50 from his neighbor may seem like a non-crime to a judge who just sat through 10 attempted rapes, a murder, and a kidnapping. But to ordinary people, the theft of $50 is a much bigger deal.
6.27.2006 2:16pm
Bruce McCullough (mail):
KevinM: It is not obviously true that there are too many
confounding factors to reach some reasonable conclusions.
This is just a special case of the general problem of causal
analysis with observational data. There is a wide variety
of statistical methods available to handle the problem,
beginning with the pioneering work of Holland and Rubin.

What we would really like to know in this situation
is what would have happened to the defendants who
chose a jury, had they instead chosen a judge, and
conversely for those who chose a judge. Only then
can we decide whether judges underconvict, etc.

While we cannot know these counterfactuals, we
nonetheless may be able to place bounds on the
probabilities that judges and juries convict, and
these bounds may be strict enough to answer
the question of whether judges or juries overconvict.
6.27.2006 2:46pm
The strength of the evidence issue is probably very strong. I don't think the issue revolves around sensing the strength, but rather, only bringing a case before a judge, not because the person is innocent on the facts, but innocent on the law. there are times when you want a judge to try your case, i.e., when there is barely any evidence of knowledge, if at all, in, say a fraud case (possession of a forged instrument). a jury would likely still convict because "he should have known better".
in essence, the judge is only used when he would be your best juror, and you've basically perempted an entire panel.
6.27.2006 3:23pm
Christopher Cooke (mail):
I largely agree with jutblogger, but would add one wrinkle: in my experience, bench trials almost always involved defendants who were likely guilty, but maybe not guilty beyond a reasonable doubt, or defendants who were not sympathetic people at all, but likely not guilty. I think, in such circumstances, the defendant and counsel are betting that the judge will apply the law to the facts and acquit based upon the failure of the prosecution to meet the burden of proof.

However, if the defendant is likely guilty beyond a reasonable doubt, but has some very sympathetic facts on his side, in my experience the defendant will usually insist on the jury trial (hoping for jury nullification), unless he or she also thinks the judge strongly favors his side for the same reasons as the jury.

Of course, no defendant in a federal case will waive the jury trial when the judge is viewed as pro-prosecution.
6.27.2006 3:37pm
Patrick Wright (mail):
A cynical question - are there any interesting results when factoring in the political party of the President that appointed the Judge?
6.27.2006 5:11pm
David Lewis (mail):
I rushed over to say, "The Guidelines!, The Guidelines!" But Keith Wollenberg beat me to it. I really think that in the old days, in the kinds of cases likely to go to a judge trial, the judge would end up thinking, or feeling, something like, "Well, he's probably guilty, but he's not such a bad fellow, I can just give him probation." And, until 1987, he could. I'm strongly inclined to think the loss of discretion in sentencing made judges apply the reasonable doubt standard more stringently than they had been.

But this explains only why judges are convicting less, not why juries convict more. I would have thought that the guidelines would have produced fewer jury convictions, since it's my anecdotal experience (as a defense lawyer) over the last 20 years that being able to quantify the sentencing difference between going to trial and pleading has made lots of defendants decide to plead, when under pre-1987 law they might have gone to trial. And since the cases where people decided to plead were probably the less triable cases, that should have increased the percentage of acquittals in tried cases. But not so.

Has the percentage of tried cases decreased since 1987? My guess would be yes. How about the number of prosecutors per tried case? Yes? That would do it. More prosecutorial time per trial, caused by more pleas, would in all likelihood raise the conviction rate. (Aside: There is nothing more dangerous, I have always thought, than a prosecutor with time on her hands.)
6.27.2006 5:42pm
Hattio (mail):
David Lewis
Has it occurred to you that the guidelines have given people more reason to go to trial and hope for a miracle. I mean, if I'm in my thirties and looking at a max of 35 years, and the prosecutor offers me only 25, I figure I'm an old fart by the time I get out anyway, might as well roll the dice.

And I noticed that you said nothing more dangerous than a prosecutor with time on her hands. Any importance to the gender of the pronoun you chose?
6.27.2006 5:50pm
David Lewis (mail):
In my last paragraph, that should be "How about the number of prosecutors per tried case. Increased? Yes? That would do it."
6.27.2006 5:51pm
nick (mail):
Fascinating post.
6.27.2006 5:58pm
David Lewis (mail):
Nah, Hattio. No gender point intended. I just alternate pronouns to be fair, no one having invented a neutral pronoun (except maybe "their" which I can't help feeling is plural).

There is certainly some of the "roll the dice" phenomenon when big sentences are in question. The guideline benefit for pleading guilty is not so enticing in those cases. But I think that was always so, even before the guidelines. The guys looking at long sentences in those days knew it. On the other hand, there was parole in the old days, so a 20-year sentence didn't necessarily mean more than 7 or even fewer years.

Such long sentences are not the most common, of course. The BOP website says about 60% of prisoners at any given moment are serving 0-10-year sentences and only 20% or so are serving over 20 years. Of course, over time, the total number of prisoners sentenced to lower sentences is much greater even than this, since the BOP snapshot does not account for the fact that, for example, the 3% of prisoners serving a one-year term or less does a complete turnover in a year, while the 20-year guys turn over only every 20 years or so.

I think the impact of the guideline benefit for pleading is much greater at the lower end of the sentencing scale, where most of the cases are, and it would not have been offset much by the effect of the guidelines (and mandatory minimums) in making long-sentence types decide to roll the dice and go to trial.
6.27.2006 6:15pm
Snacktime (mail):
It strikes me that during the same period, guilty pleas have also become vastly more common. According to the 2004 USSC sourcebook, roughly 96% of federal defendants pleaded guilty since 2000. before 2000 the rate was lower, however; in 1997, it was roughly 93%. I am having trouble finding earlier data, but my understanding is the percentage of guilty pleas rose steadily since the late 1980s (i.e. after the guidelines). Thus, during the same period in which the acquital rate of judge trials was going up, so was the percentage of guilty pleas.

Now it could be that the same cause explains both: i.e. the sentencing guidelines (strict mandatory sentences) motivate more guilty pleas and also lead judges (aware of the strict sentences) to acquit more often. But maybe the acquittal trend has to do with the kinds of defendants foregoing trial altogether?

Let's assume that a defendant is more likely to seek a judge trial if he thinks he's got a relatively strong case. As the stakes go up with mandatory minimums, pleas become relatively more attractive even to these folks. so even some of the defendants with strong defenses forego trial and plead guilty instead. this might mean that nowadays, only the defendants with really, really strong defenses decide to go to bench trial rather than plead guilty.
6.27.2006 6:51pm
Alan H. Martin:
Consider the phenomenon in jury trials of judicial reversal of a conviction notwithstanding the verdict. Consider also the phenomena of dismissals in any trials, both with and without prejudice.

Might analysis of these events' rates and trends allow you to test any hypotheses about strength of prosecution cases or, defense strategies; particularly across different types of crimes?
6.28.2006 10:36am
David Lewis (mail):
You know, the phenomenon of acquittals in judge trials may be even more pronounced than it appears. In the districts in which I work, one common reason for going to a bench trial is to preserve constitutional claims without bothering with a jury trial (when the government refuses to agree to a conditional guilty plea, preserving the right to litigate the constitutional claim on appeal). In these cases, the defendant essentially does not contest his guilt before the judge based on the evidence at a suppression hearing or the like. So the judge invariably finds her guilty. These "procedural" bench trials count as convictions, even though the defendant did not really present a factual defense.
6.28.2006 11:42am
David Lewis (mail):
Snacktimes data suggest support for my hypothesis about the reason for the increased conviction rate in jury trials: more prosecutors concentrating on fewer trials.
6.28.2006 11:44am