Tribunal Theory

Patent.Law007The tri prefix of tribunal may evoke an image of three judges sitting jointly at the bench.  In practice, however, tribunal refers to any sitting judicial body – regardless of the number of judges.

The Court of Appeals for the Federal Circuit (CAFC) normally convenes in three judge tribunals to hear appeals. Motion practice, such as an emergency motion for a stay of an injunction, is often heard by a single judge.  According to statute, en banc panels usually consist of all (twelve) active members of the court. 28 USC 46(c).

A larger panel is thought to provide more stability and predictability, while a smaller panel theoretically provides a greater variance.

To see how panel size might alter litigation practice, lets take an example case where the lower court issues a permanent injunction that will be enforced within ten days. Question: Should the losing defendant hope for (1) a three judge panel to hear its emergency appeal or (2) a lone appellate judge? 

The answer, after several assumptions including independence, is that the appellant’s favored panel size varies with the strength of its argument.  If the appellant has a high (greater than 50%) chance of convincing each expected judge, then it is better off with a larger panel.  On the other hand, if the appellant has a low (less than 50%) chance of convincing each judge then he would likely get better results from a single judge.

This is easy to work out a computational example: For the first case, if we expect that the appellant has a 65% chance of convincing each judge of its case, the probability of convincing at least two judges rises to 72%. Thus, the appellant would be better off with a panel (72% chance of winning) rather than a single judge (65% chance of winning). On the other hand, if the appellant thought that its appeal had only a 35% chance of convincing each judge, the probability of convincing at least two judges drops to 28%.  Thus, the appellant who expected to lose would be worse off with a panel (28% chance of winning) rather than a single judge (35% chance of winning).  The larger panel also helps smooth over variance amongst the judges — tilting the scales further in favor of an expected winner.  A long shot case would hope for the squirrelly judge on a small (solo) panel.

In a May 6, 2007 distribution, Hal Wegner suggested a series of “Mini en banc” panels with seven (or five or nine) members — especially in the area of obviousness.  As the panel size increases, good cases theoretically win more often and should be losers lose more often.

Four Judge Panels: One simple step that could make a tremendous difference regarding stability of appellate determinations would be to shift the normal panel size from three judges to four judges.  The CAFC has power to make this change under 28 USC 46, which states that “the Federal Circuit may sit in panels of more than three judges if its rules so provide”.

On appeals, a split judgment is equivalent to an affirmance of the lower court decision. Thus, an appellant arguing to a four judge tribunal must convince three of the four judges of its case — creating a much larger, super-majority hurdle for the appealing party. The problem with the four judge approach is that it also creates an outcome skewed toward affirmance — something that could be seen as unfair. However, I expect that claim construction decisions reviewed de novo by a four member panel would help us return from our long vacation (or at least reduce the number of claim construction opinions being vacated).

The four judge approach would also violate Federal Circuit Rule 47.2, that requires “a panel consisting of an odd number of at least three judges.”

20 thoughts on “Tribunal Theory

  1. Getting back to the mathematics of the proposal, it is has some interesting aspects. I generated pictures of the solution “surface” based on the number of judges (1 to 12) and the probability of success with an individual judge (SIJ). The equation is at the bottom. (Dennis, I e-mailed you the pictures. Feel free to post them as commentators can’t post images).

    As expected, as the success with any individual judge increases (SIJ), so does the skew toward the expected winner as panel size increases. However, this occurs mostly in the linear region of SIJ (about 0.55 to 0.80). For an SIJ of 0.75, increasing the panel size from 3 to 5 judges skews the outcome 5.6% toward the expected winner. But with an SIJ of 0.85, going from a 3 to a 5-judge panel only skews the outcome by 3.4% toward the expected winner. The function is asymptotic with decreasing slope as the expectation of success increases.

    So Hal and Denis correctly state that a litigant more likely to win with a single judge is even more likely to win with an increased panel size, but the numbers are not that dramatic.

    The Far more interesting thing to note is the significantly steeper gradient with respect to SIJ than with the number of judges. Improving the merit of your argument has a substantially greater effect than increasing the number judges. That should be intuitive. But what is not so obvious is the significant increase in outcome that can be had by even modestly improving SIJ.

    For example, say you won with a 65% SIJ on summary at the district level and on appeal you improve your brief to an SIJ of 75%, you now have a 90% chance of victory with a 5 judge panel (84.4% with a 3 judge panel). Even more dramatic, increase the SIJ of your case to 85% and you virtually guarantee victory en banc (99.9%). This is a serious incentive to increase the quality of arguments on appeal which should serve as a strong inducement for the CAFC to adopt increased panel sizes. Litigants have to improve their arguments to win – a sort of brief arms race. Isn’t that what litigators do anyways?

    Litigants should also realize that it is in the steepest zone, between about 0.45 and 0.55 for SIJ, that the most significant improvement to outcome is available. The peak occurs at a SIJ of 0.5. Thus the often quoted 50/50 “crap shoot” at the CAFC is exactly where the low hanging fruit grows. For example, with a three judge panel, a one percent improvement in your SIJ from 50% to 51% results in a 1.5% gain in outcome. With a 5 judge panel the gain is 1.84%. One could imagine a game theory argument with a client advocating spending X dollars for improving a brief resulting a multiplier times X dollars improvement in expected outcome. For example, “investing” for a dramatic 15% improvement in the SIJ of a case from 0.55 to 0.70 increases successful outcome 20.9%, or from 57.5% to 78.4% for a 3-judge panel (in relative terms a whopping 36.3%)! The numbers are 24.4%, 26.6%, and 28% increase in outcome for 5, 7, and 9-judge panels respectively. The en banc numbers are even more compelling. Improve your SIJ just 10%, from 0.55 to 0.65, and your expected outcome goes from 64.4% to 87% when 12 judges preside!
    Of course, this rests on the massive assumption that SIJ can be significantly increased for a given case, that SIJ is uniform across judges in a panel (what convinces one judge often doesn’t convince another), as well as the overly simplified combinatoric model which ignores jurist interaction assuming jurist independence. There’s nothing like a well written 10 page brief with unassailable facts that can defeat a hundred page meandering drivel, or a lone dissenter convincing a panel of the error of their ways. History is full of such examples. Nevertheless, the numbers do point to the advantages of increased panel sizes and to aggressively improving arguments.

    Some data points:

    No. of judges______Success with Individual Judge(SIJ)______Outcome (to 3 digits)

    1________________ 0.55______________________________0.55
    __________________0.65______________________________0.65
    __________________0.75______________________________0.75
    __________________0.85______________________________0.85

    3_________________0.55______________________________0.575
    __________________0.65______________________________0.718
    __________________0.75______________________________0.844
    __________________0.85______________________________0.939

    5_________________0.55______________________________0.593
    __________________0.65______________________________0.765
    __________________0.75______________________________0.900
    __________________0.85______________________________0.973

    7_________________0.55______________________________0.608
    __________________0.65______________________________0.800
    __________________0.75______________________________0.929
    __________________0.85______________________________0.988

    9_________________0.55______________________________0.621
    __________________0.65______________________________0.828
    __________________0.75______________________________0.951
    __________________0.85______________________________0.994

    12________________0.55______________________________0.644
    __________________0.65______________________________0.871
    __________________0.75______________________________0.976
    __________________0.85______________________________0.999

    I didn’t read Hal’s paper (I have to get on his list), so apologies in advance to Hal if he already said any of the above.

    The solution equation is a summation of combinatorics (the standard combinatorial choose function = n!/(p! (n-p)!) ) over a majority or greater number of judges:

    Outcome(num_judges, SIJ) = Sum (for i=majority..num_judges, Choose(num_judges, i) * SIJ ^i * (1- SIJ) ^ (num_judges – i) ) ,

    Where majority = floor(num_judges / 2) + 1.

    (This ignores the even numbered panel split decision problem which is another flaw).

    Tautologies:

    Outcome(num_judges, SIJ) + Outcome(num_judges, 1- SIJ) = 1.

  2. Maybe have four judges where one of them is deemed a “lesser” judge whose vote counts for half of an ordinary judge’s vote.

  3. Dennis, I think this blog does us practitioners a great service, and I have a lot of respect for Hal Wegner, but getting back to the substance of the thread – have you and Hal flipped your lids this time? What’s unique about patent law (or appeals of veterans’ affairs or government contracts or compensation for government employees) that indicates that the way in which CAFC panels are composed should work differently than in any other circuit? Your argument for 4-member panels is applicable across the board, not just in patent law. But I’ve never heard of a tribunal anywhere having an even number of decisors, and I’m betting that if you look past the patent world and into the general literature of the workings of courts, you’ll find discussions of why panels are composed of an odd number of people.

  4. Linguist — tribunal is thought to be drived from tribus. Tribus, in turn, is thought to designate the three tribes of Rome.

  5. “Except that “tri” is not a prefix.

    This tells you what it takes to be a law professor these days.

    tribunal — tribune (as in Rome) — tribus (tribe)”

    Ha, ha, Linguist, you typed this in the wrong thread the first time! I guess this tells you what it takes to be an Internet pundit these days.

  6. Hey Linguist:
    With your searching skills I believe you qualify for a position examining patent applications at the USPTO. A quick look at wikipedia gets the definition of tri as being a prefix. link to en.wikipedia.org
    I guess you never rode a tri – cycle as opposed to a bi – cycle. I really should stop here but then there is the tri – angle, a shape defined by the number of angles formed by the sides of the shape. I know I shouldn’t do this so I will stop short of calling you a moron for that is trying enough.

  7. Linguist:

    Lay off the insults to Dennis. He provides a great service, and anyone who insults him is going to be instantly unpopular on this forum.

  8. Linguist,
    Maybe we can squabble over the definition of a prefix but the post simply suggests that tribunal “may evoke an image of three judges.” Furthermore, it seems reaonable to associate tribunal with three. Tribunal does evolve from tribune and thus originally from tribe. For tribe, the American Heritage Dictionary 3rd describes tribe as coming from the “Old French tribu, from Latin tribus, … or possibly from tri-, three.” From a linguistic point of view, I want to know if “three” in tribunal means three judges or does it refer to “third person” (ie an independent, unbiased third person brought in to adjudicate a two person conflict) ?

  9. Except that “tri” is not a prefix.

    This tells you what it takes to be a law professor these days.

    tribunal — tribune (as in Rome) — tribus (tribe)

  10. SJM: Thanks for the etymology. Good analysis.
    Joe Smith: Since obviousness is a question of law, the CAFC can review on appeal based on a JHOSITA– Judge etc.

  11. Mr. Hutz, I have the opposite view of Judge Newman’s opinions. She’s very pragmatic and understands the patent prosecuting bar very well, but as a legal matter I generally prefer her colleagues’ opinions to hers.

    Judges Gajarsa, Linn, and Lourie have all been patent attorneys.

    Judge Moore taught and litigated patents for a while before joining the bench.

    The other active judges have between 6 and 20 years of experience dealing with patent law as CAFC judges.

  12. The real problem is that many of the judges that now sit on the CAFC have no background in patent law. I believe only Newman and maybe one or two others were ever trained as patent attorneys. Funnily enough, Newman generally has the most thoughtful and logical opinions and I usually agree with her (not every single time, but much more than any other single CAFC judge).

    Simply making the panels 4 members may bring more consistency, but I am not sure it would bring better decisions.

    Joe

  13. The “tri” in tribunal may evoke an image of three judges, but it was never meant to indicate the number of judges. Tribunal is derived from the Latin word tribunus, a raised platform for a magistrate or tribune (tribunus). Tribune is derived from a term for the head of a tribe (tribus). Tribus was used in Rome to identify political divisions originally representing the three original tribes of ancient Rome, Tites (Latins), Rammes (Sabines) and Luceres (Etruscans). Tribus could be derived from tri- (three) and the root of be (-bhu), but has also been considered to be of Etruscan origin.

  14. M, one e-mails Hal Wegner and asks to be put on the list.

    Dennis, I wonder to what extent each judge’s decisionmaking process is influenced by the presence or absence of co-panelists. I imagine that a judge who has sole responsibility for an emergency appeal will feel more pressure to get the decision right, and may be more risk-averse in his decisionmaking process.

    If, on the other hand, a judge sits on a panel with 4 others, he might feel less pressure and exercise less independent judgment if he sees a consensus emerging among the other 4.

  15. Good lord, what an idealist. Isn’t the very nature of the law a sort of empirical imperfection? It’s built like an anthill, a lot of apparently random scuffling surprisingly resulting in a house with many rooms. As you point out there is no such thing as such a Judge or such an application, but even imagining such requires an objectivity not even the finest academic mind could be supposed to possess.

  16. Your post invites a question that I have been wondering about for a while: How close do Judicial decisions come to the decisions that would be made by an all seeing, all knowing Judge applying the law perfectly? Think of it as a quality control question: how often do the doors or wheels fall off the juridical product and why?

    Is there any literature out there about this issue? The best way to systematically improve the quality of the courts is to start by measuring the quality of the outcomes.

    In your new role as an academic this is an issue which might be worthy of a publication or two if you could find a methodology for attacking it.

Comments are closed.