Claim Construction Reversal Rates II – District Court Judge Experience

    By David Schwartz

How do district court judges with varying levels of experience perform on claim construction? Are judges more likely to have their decisions affirmed when they have previous claim construction experience? Yesterday I provided some background on the large database I compiled and some overall results. [LINK] Today’s post provides an analysis of the data based upon the number of previous appeals of claim construction. A draft of the paper, Practice Makes Perfect? An Empirical Study of Claim Construction Reversal Rates in Patent Cases, can be downloaded here.

If district court judges improve after appellate review of claim construction, one would expect that the reversal rate would decrease as the number of appeals increases. More specifically, a district court judge with more prior Federal Circuit feedback should have a lower reversal rate than a district court judge with less experience. Assuming that the judge decided the second case after learning of the decision of the Federal Circuit in the prior case, the second time a particular judge is before the Federal Circuit, he or she should be more likely to have his or her claim construction affirmed. Figure A below illustrates the reversal rates of district court judges broken down by the number of prior claim construction appeals.


The left-most set of bars represents the results from every judge’s first appeal, with the blue bar indicating the percentage of cases that had to be reversed or vacated due to an incorrect claim construction, and the red bar indicating the percentage of cases with any erroneous claim construction. Moving to the right, the pairs of bars represent the results from each judge’s subsequent appeals.

These results suggest that district court judges do not improve as the Federal Circuit reviews their cases. Contrary to conventional wisdom, the reversal rate does not appear to decrease when a district court judge has multiple decisions reviewed by the Federal Circuit on claim construction. In fact, the highest reversal rate is for judges with four prior claim construction appeals. Other than judges who have been appealed exactly four times, the range is very narrow, between 26.8% and 30.5% for reversals and 31.0% and 40.4% for errors. Thus, there does not appear to be significant expertise gained by district court judges via direct Federal Circuit review that causes the claim construction reversal rate to decrease. (As discussed in more detail in the paper, the study is subject to several limitations inherent in studying appellate cases, including for example, a potential selection bias and a potential distortion if the cases are not distributed evenly across the district court judges. Ideally, for research purposes, cases would be randomly assigned to a judicial district (i.e., no forum shopping), and a random subset of those cases would be appealed.)

Up next: Reversal rates based performance by district court judges after a first reversal by the Federal Circuit.

I welcome comments from the readers of Patently-O.

33 thoughts on “Claim Construction Reversal Rates II – District Court Judge Experience

  1. 33

    Let me provide a bit more detail on the data behind Figure A. Starting from the left, in the first column, I determined whether the first appealed case for each judge was Affirmed or Reversed/Vacated. Of these cases, 29.6% were vacated or reversed due to an erroneous claim construction. (In the paper, I provide the number of observations in each category.) In the second column, I determined whether the second appealed case for each judge was Affirmed or Reversed/Vacated. Fewer judges fit within this category since not every judge had two claim construction cases appealed. I did not use the cumulative record — including both the first and the second case — for the judge here. Instead, I only included the result in the second case. Of these cases, 28.9% were vacated or reversed due to an erroneous claim construction. And so on for the remainder of the categories. One may expect that after a judge has had three, four, five or more cases reviewed by the Federal Circuit, his or her reversal rate on subsequent appeals would decrease.

    As mentioned in the blog post and the paper, there is a potential selection bias. Accordingly, I mainly presented descriptive statistics in the paper. For those who are interested in measures of statistical significance, I calculated a chi-square test p-value of 0.285, which is greater than the expected p-value for a 95% confidence level. This calculation is provided in the paper.

  2. 32

    Yikes, that previous comment won’t quite apply to the 0 previous cases bins. I see what you mean there. The reversal for each judge is either 0 either 1. (I don’t think that should be called a “rate.” That’s the confusing part.) You could sum those rates for n judges, divide by n and get the group average, which should also be the 29.6 figure for that first group.

    The next column the possible “rates” for individual judges (having exactly 2 appeals) are 0, .5 and 1. Sum those and divide by n for the group average, which may or may not be 28.9.

    As I understand it, you are looking at the very next case following the criterion number of prior appealed cases. I’m saying, add up the individual reversal rates and take a mean for each group. The 10/15 example I used is extreme because probably none of your judges have that many cases under their belt. But the math is the same regardless of how many total appeals for each judge. Get the individual reversal rates for all the judges, then ask whether the mean a function of the total number of appeals.

    I guess I really question the validity of looking at just the last case for each judge and saying of those, X% were reversed, etc. I guess what you’re asking is what is the probability that a judge who has Z appealed cases under his belt will be reversed on the next appeal. And is that probability a function of Z. I’m not convinced that your data answer that question. But, as you can see, I’m not entirely out of the fog here.


  3. 31

    Sorry, Dave, you’ve lost me. Or I’ve lost m’self.

    Let’s look at the first blue bar. It has a value of 29.6% and you’ve called that a “reversal rate” and it’s the “rate” at which judges w/ 0 appeals have been reversed on claim construction. You don’t tell us how many judges so we don’t know the n, but let’s say for argument it’s 100.

    I presume you took the individual reversal rates of all 100 judges, added them together, and divided by 100 to get a mean of 29.6%.

    The individual reversal rates are not 0 or 1, they are some discrete value between, and including, 0 and 1. If, for instance, Judge Learned Handjob had 15 cases and got reversed on 10, then his value is — hang on a sec’, let me find that calculator — 10/15 = 0.66666. If his buddy, Judge Anthony Swastika’s value was 0.2500, then their mean is 0.4583. Add in the other 98 judges and you have an n of 100 with (n-1) 99 degrees of freedom. I guess you can do a t-test, multi-variant analysis, or whatever you want to do, including SD and SE. Of course, unless you know the data are normally distributed, which you don’t, you would prefer a non-parametric test.

    I admit I didn’t read your formal paper, so my interpretation of your technique might be pretty well buggered (this is a test to see if “buggered” gets past Dennis’ censor), but the foregoing is the way I interpreted your bar-graph. How else to you get a reversal “rate?”


  4. 30

    I’m a bit unclear for which variables you would like to see the standard deviation. Standard deviation measures how widely spread values are in a dataset. As an example, take one dataset including the numbers 2, 4, 6 and 8, and a second dataset including the numbers 5, 5, 5, 5. The average of each dataset is 5, and the standard deviation would tell us that the first dataset has a larger spread than the second dataset. As a second example, consider a dataset including information on the gender of the judge. Gender information is a nominal variable and lacks any numerical value. If a study reports the percentage of judges that are women, calculating a standard deviation would have little or no meaning.

    The data illustrated in Figure A above is similar to the gender example. There are only two potential values. We could call them “Affirm” and “Reverse,” “0” and “1” or even “0” and “55.” The names assigned to this variable are arbitrary. For the blue bar in the first category (judges whose claim construction appeals had never been appealed), there are only two possible results: 1. the judgment was reversed or vacated due to an erroneous claim construction; or 2. the judgment was not reversed or vacated due to an erroneous claim construction. Because there are only two potential values for each observation and these values lack any numerical meaning, I don’t believe that reporting the standard deviation would be useful.

  5. 29

    Way up there at the top, above all this patent troll crpa that has nothing to do with this post, M. Samardzija asked this question:

    Do these data support Congressman Ira’s push for specialized courts?

    Does anyone know who and what this refers to? Is there a proposal in Congress for a parallel IP court?

    Also, Mr. Schwartz, if you’re still with us, put some standard of error or standard deviation indicators on your data. Data mean nothing without some indication of the variance, and the author comes across as un-professional or hiding something. It’s like deleting all the case cites from a brief. I know you, like all lawyers, are scared of numbers with more than 2 digits — just enter the data and push the “S.D.” button on your calculator.


  6. 28

    All judges operate under the backwards completion principle, regardless of whether they admit it to others or themselves. To the extent they render decisions that go against their personal beliefs, it’s because its because the case law or statutory law is too plainly in opposition to their beliefs (or more likely they are not imaginative enough or capable enough to overcome the accepted conclusions of prior case law and statutory construction – cf Cardozo).

  7. 27

    Mr. rat,

    I hope the irony of my thought below doesn’t fly over your head, but, in your comment above at 1:01 PM, when you wrote:

    “Gee, the way I learned it was ‘power corrupts and absolute power is even nicer’. I guess I went to a different school.”

    you were the one suggesting that Orwellian-type oppression was nice. I guess your memory is short. I never suggested anything like that. I just don’t like pejorative slur terms.

  8. 26

    “the term carries the connotation of someone who unjustly is demanding a tariff of innocent passers by.”

    Thank-you Mr. Bloom.

    It may be a fair question to discuss whether the term is properly applied to one entity or another but the attempt by JAOI(TM) and others to entirely stamp out the expression reminds me of Big Brother in Orwell’s 1984 where the oppressor sought to control debate by controlling the vocabulary.

  9. 25

    “Let’s face it, the world would be hell of a better place if everyone were a bit more quixotic.”

    The world would be a better place if fewer people occupied themselves with tilting at windmills and instead spent more time grounded in reality.

  10. 24

    Dear Mr. Bloom,

    I appreciate your honesty.

    I also appreciate your compliment – THANK YOU.

    Let’s face it, the world would be hell of a better place if everyone were a bit more quixotic.

    I’d be honored to buy the beer.

    I think you would be pleasantly surprised at my quixotic accomplishments over my 40 years of inventing et al. endeavors.

  11. 23

    The term “patent troll” evokes for me the image of the troll under the bridge in “The Three Billy Goats Gruff” (link to, who jumps out and eats anyone who tries to cross his bridge. I believe this was the intent of whoever coined the phrase. So yes, the term carries the connotation of someone who unjustly is demanding a tariff of innocent passers by. I think it’s a particularly colorful image, and that it fits a few cases that I’ve encountered.

    A better comparison than “shyster lawyer”, which is way too general, would be “ambulance chaser.” I’m sure that term offends a great number of tort lawyers. Like “patent troll,” it at least superficially appears to condemn a particular type of behavior. Like “patent troll,” it’s a colorful phrase that’s unlikely to die out simply because some people don’t like it.

    For what it’s worth, I’m not sanctioning the use of either term to refer to a specific entity. But your attempt to stamp it out is quixotic, at best.

  12. 22

    Dear Mr. Bloom,

    By your picking the particular meaning of troll you did from the dictionary, were you suggesting the term “patent troll” is not pejorative? Or would you be constrained to agree with me that the use of “troll” in relation to IP matters is closer to the dictionary’s definition of “demon”?

    Come on now, you can say it, we all know it anyway, be honest, just like in “My Cousin Vinny.”

  13. 21

    “You, of course, can see the difference between trolling for fish and NBC advertising its TV show as, ‘Adult predators troll Internet chat rooms in pursuit of children.'”

    Yes, of course I can. And hopefully you can now see the difference between a patent troll (which term was derived from the mythical Scandinavian creature) and trolling Internet chat rooms. I meant no disrespect in pointing out that NBC’s use of the term “troll” is completely unrelated to the use of the term in the patent domain. It was your example, and it doesn’t support your argument, because you’re talking about two completely different words.

    You can use the term “shyster lawyer” or make lawyer jokes all you like. I don’t think anybody is seriously suggesting that we should try to stamp out all use of derogatory comments about lawyers. Our efforts are better spent trying to make sure there is no substantial basis for the comments…

  14. 20

    Dear ugly rat and Mr. Bloom,

    With all due respect, your remarks are shallow and or are vainly designed to obfuscate and the obvious truths I’ve discussed above. Oh, it is a verb, not a noun, and it is a fishing term of art blah, blah, blah. No one ever suggested anyone here or anywhere else for that matter was calling a patent troll a child molester – please, gimme a break. You, of course, can see the difference between trolling for fish and NBC advertising its TV show as, “Adult predators troll Internet chat rooms in pursuit of children.”

    I have come to expect better from IP professionals. Read my words again and try to find some meaningful criticism. If you find meaningful criticism, I’ll buy your all the cheese and beer you can consume over the weekend.

    Seriously, for the reasons I described, the pejorative term “patent troll” is akin to the pejorative term “shyster lawyer” and neither term is suitable for constant use on this or any other IP blog.

    Do you or anybody else find fault with that conclusion?

    I will have more meaningful things to say about what has been happening to our IP community – you’ll begin to see the destructive toll the slur term “patent troll” has contributed to our IP profession.

  15. 19

    “Troll is a terrible term. That is why an MSNBC TV News Special was advertised as follows: ‘Adult predators troll Internet chat rooms in pursuit of children.’ Of course, in this example, troll is not pejorative enough, but I think you get my point.”

    JAOI (TM), your example of why “troll” is a terrible term is way off. MSNBC is using troll as a verb, in this case meaning “to look, or search”. This is related to the verb “troll” as in “fishing, by pulling a lure through the water.” COMPLETELY unrelated to “troll” as a noun, which refers to “a dwarf or giant in Scandinavian folklore inhabiting caves or hills.”

    Maybe I’m being insensitive when I refer to a person (or more likely, an entity) as a patent troll, but I’m definitely NOT calling him (or it) a child molester.

  16. 18

    “Power tends to corrupt, and absolute power corrupts absolutely.”

    Gee, the way I learned it was “power corrupts and absolute power is even nicer”. I guess I went to a different school.

    As to your obsession with ‘patent troll” – “a rose by any other name would smell as sweet” and a “patent troll” by any other name would smell as bad. The problem people have with “patent troll” is not because of the word but the concept it names and using some other name for the concept will not change the fact that it is, is intended to be and will always necessarily be negative.

  17. 17


    As child, we were taught, “Sticks and stones can break my bones, but words (names) can never hurt me.” As intelligent adults we can see that that is not true, but it still has an effective subconscious impact – that is why name calling in not uncommon, e.g., in politics.

    And, as children, we were taught from children’s books over and over to h a t e and fear trolls.
    Thus, it is human nature to h a t e and fear trolls.

    If asked, neither Hill & Knowlton nor Burston Marstella could have come up with a more effective pejorative means of attacking self-employed inventors than to use the term “patent trolls.” Troll is a terrible term. That is why an MSNBC TV News Special was advertised as follows: “Adult predators troll Internet chat rooms in pursuit of children.” Of course, in this example, troll is not pejorative enough, but I think you get my point.

    Using the term “patent troll” is akin to using racial slur terms. For example, how would Patently-O readers feel if the slur term “SHYSTER LAWYER” was used constantly to describe attorneys practicing as lone independent professionals? I’d bet most readers would be upset, and the IP community would rightly object and not let it continue.

    So, I beseech you; please refrain from using slur terms. You’ll be a better person for it and you’ll want to thank me in the morning.

    I’m sure everybody would like to know that there is a more accurate and politically correct way to describe the slang slur “patent troll.”

    It is more sensitive and preferable to use the phrase “patent abuser,” or simply refer to a patent abuser as someone who “abuses patent rights” lest you offend the honorable efforts of inventors who are gainfully self-employed in their chosen profession promoting progress in America. Some inventors may abuse their patents just as some lawyers may abuse their license to practice law.

    The term “troll” should be reserved for insidious abusers; for example, when discussing pedophilia.

    The damage caused by the slur term “Patent Troll” should stop.

    Those who have used this slur in deceitful ways in the past should be sanctioned.

    More to come…

  18. 16

    Barack Hussein Obama is a brilliant, passionate, charismatic and great young American. And if he is elected to the most powerful position in the world, there is a chance that power may corrupt him as power has corrupted so many others:

    link to
    “Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men.”
    Lord Acton expressed this opinion to Bishop Creighton in 1887.

    However, it appears that Obama has yet to be corrupted while some might argue that the Clintons have been long ago corrupted, that the Clintons have taken oaths of office with a grain of salt.

    Conservative Republicans have tried to bias voters against Obama for reasons including his middle name, HUSSEIN. Both Mr. and Mrs. Obama were upset and fought back. I am confident all Patently-O readers would agree that such attacks are reprehensible, to say the least.

    John McCain, the presumptive Republican nominee (who, incidentally, has been tried, tested and passed), has forcfully rebuked members of his own party after name calling attacks against Barack.

    Which is to prove that not all proverbs are true; for example, this one link to
    “Sticks and stones can break my bones, but words (names) can never hurt me”.

    This proverb is simply not true, else why would a-hole republicans resort to using “Hussein” in an attempt to disparage Barack Obama?, and why would American presidential candidates rebuke such tactics if they did no harm?

    Where am I going with this on the world’s most popular patent blog? Anywhere?, Anywhere?, Bueller?, Malcolm?, Anyone?

  19. 15

    Dear Dave Schwartz,

    I hope this isn’t too far off topic, but, as a self-employed inventor, who has practiced many of his own inventions over the past 40 years but not all, I am interested to know if there is any perceptible bias favoring bigger businesses over independent inventors at the District Court level, and if so which venues, the CAFC level or the Supreme Court.

    My reason for asking this will become even more apparent when I post a more in depth comment later today as time may permit.

  20. 14

    Thanks again for everyone’s comments. Some quick responses are below.

    1. The data presented in this post are based upon appealed cases. For an analysis of the reversal rate based upon number of patent lawsuits handled (instead of appealed cases), please see Figure X on page 36 of the paper.

    2. The comment stating that district court judges may develop a reputation at the Federal Circuit is correct and cannot be excluded by the data.

    3. I did code the Federal Circuit cases for whether the district court judges construed the claims too broadly or too narrowly. I’ll include this information in my second article analyzing the data.

    4. Ignoring for a moment the possibility of a selection bias, the samples sizes (i.e., 57 judges with 4 cases appealed, 105 judges with 3 cases appealed, etc.) are large enough to generate potentially statistically significant results. But the sample may be too small to generate “significant” results, absent large differences.

    5. Last I want to address the potential selection bias. One commenter above stated that “a lower quality judge will be appealed more often….” Others have provided similar comments. These are fair concerns. For those who are interested, I attempted to analyze this potential selection bias using appeal rate data of various judges in Section D of Appendix B on pages 64-66 of the paper.

  21. 13

    The judges with the biggest patent dockets will get appealed the most. The more appeals filed, the more often the judge will get reversed. As mentioned by others, the sample size is too small for the inferences drawn. And the sample is too small to gain much of any kinds of inferences, except maybe trends. I just don’t see how you can try to pull these kinds of detailed conclusions when only about 100 patent cases per year even go to trial. (And that’s a real statistic that is useful. It has also remained pretty constant from what I recall, over the CAFC’s tenure.)

  22. 12

    “Result Oriented Decisions are the Antithesis of Blind Justice”

    as you posted, I was preparing my comment.
    Take a step further, your second, number 2) theory is consistent with an ominous conclusion that follows my comment. My comment:

    “Result Oriented Decisions are the Antithesis of Blind Justice”

    In simple terms as I understand it, one result of establishing the Federal Circuit Court of Appeals in 1982 was for the public to have the ability to consistently gauge the scope of a patent and its claims and whether or not a particular thing infringes a particular patent.

    If the CAFC’s panel opinions had been consistent with each other, a body of law would have been built up to accomplish that result, but CAFC panels’ opinions varied so as to create conflicts — consistency was not realized. And then came en banc Phillips, which, upon reading, gave one hope that finally there would be a fairly comprehensive guide to difficult claim construction issues so consistency would become a reality.

    As was suggested by “Public left sea sick” to Mr. David Schwartz above, “If you follow the link to link in the [“Public left sea sick”] comment above, you will find an Appendix, “Listing the 16 cases,” a post Phillips v AWH analysis of file history disclaimer claim construction issues.

    This analysis suggests that, since Phillips, most remarkably, the consistency rate seems to have worsened!

    What’s up with that?? All I can figure is that “metoo” and “pds” put their fingers on it as follows:

    Above “metoo” wrote:
    Re: “… how can a judge learn from the past when the CAFC is constantly changing the rules to suit its whim of the day?”

    Yesterday, “pds” wrote:
    “… the FC construes the claim language to get the end result they want. There is so much apparently conflicting law regarding claim construction that it isn’t difficult to come up with a rationale that fits the result one wants.”

    All of which point to the obvious: It is time for the Supreme Court to address this growing conflict and grant cert to the pending Petition “Public left sea sick” wrote about. An important case, but not a hard case for the Supreme Court to deal with – Provide clear and unmistakable instructions that District Court Judges and CAFC Judges construe patent claims in accordance with the Phillips’ guidelines to a tee. As Phillips is reviewed by the Supreme Court, any blanks that may be found in Phillips can also be addressed.

    Many might think that the Supreme Court cannot and will not pass on the opportunity to once and for all set the CAFC straight on the course it was initially established to follow – the CAFC obviously has run amuck – issuing RESULT ORIENTED DECISIONS is a vice that if left unchecked will grow like a cancer. Where will it stop? I mean, how can one reconcile en banc Phillips with a GROWING numbers of inconsistencies in controlling claim construction matters?

  23. 11

    I think you have a correlation-causation problem here. The fact that the reversal rate does not go down after multiple appeals might be:

    1. Only perceived “bad” claim constructions get appealed in the first place. So although a district court judge improves, the rate of appealing his claim constructions goes down while the rate of reversal remains constant.

    2. The district judge develops a reputation at the Federal Circuit, so more prior reverals means more likelihood of future reversal, even if the district judge improves.

    Not saying that either is necessarily the case. But the stats here do not exclude either theory, I think.

  24. 10

    My thought is that it will be hard for judges to “learn” from previous appeals. I would expect the judges to “learn the field” if all claim construction cases were the same (i.e. same terms to be construed in same technology). However, given that there is no standard patent case, how could we expect judges to “learn” the correct claim construction in an electrical engineering case after they’ve been reversed three times, once in a biotech case, once in a mechanical engineering case, and once in a software case? We might not.

  25. 9

    “In fact, the highest reversal rate is for judges with four prior claim construction appeals.” So judges continue to merely “practice” law and learn very little? I’m confused.

  26. 8

    Mr. David Schwartz,

    If you follow the link to link in the comment above, you will find an Appendix, “Listing the 16 cases,” a post Phillips v AWH analysis.

  27. 7

    To “metoo”
    Re: “Perhaps what the data evidences is that trial judges have a hard time learning from previous appeals simply because the CAFC is constantly contradicting itself. In other words, how can a judge learn from the past when the CAFC is constantly changing the rules to suit its whim of the day?”

    To “pds”
    Re (in yesterday’s part I thread) “… the FC construes the claim language to get the end result they want.
    There is so much apparently conflicting law regarding claim construction that it isn’t difficult to come up with a rationale that fits the result one wants.”

    + +

    The thread link below has an extreme example of “metoo’s” and “pds’” concerns.
    It discusses a 3-0 CAFC decision in which the panel IGNORED the CAFC’s own strong “CLEAR and UNMISTAKABLE” precedents required for a patentee to disavow claim scope, and INSTEAD affirmed a District Court’s claim limitation which was found (only) in the file history in the patentee’s Request for Reexamination (not found in the specification, and no hint of limitation found in the claims) — that claim limitation was based solely on an “IMPLICATION”!

    In other words, the CAFC affirmed a claim scope disavowal solely by file history “implication” which contradicts the CAFC’s own “clear and unmistakable” precedents:

    link to

    The above link includes these links:

    • File Attachment: CIAS_Petition.pdf (190 KB).
    • File Attachment: CIAS_Petition_Appendix_H.pdf (733 KB) (Listing the 16 cases).

    The extreme example I’m referring to is described in this quote from pages 6 & 7 of the Petition that was linked-to in the article link (emphasis added):

    “… The result is that the public is left at sea as to what a patent’s claims mean unless and until one or another panel of the Federal Circuit decides how it is going to view these prosecution history statements.

    “In this case, for example, the District Court held that during reexamination of the patent-in-suit, CIAS, in describing a prior art patent system, had “IMPLIED” a particular operation of that system and that the “IMPLICATION” of this description was a limitation to the CIAS patent’s claims. Without explaining how such an “IMPLIED” description and “IMPLICATION” of a limitation could satisfy its own “clear and unmistakable disavowal of claim scope” standard, the Federal Circuit affirmed the District Court’s limiting claim construction and consequently judgment of noninfringement. If all that is required to create a limiting claim construction is an “IMPLICATION” based on a description of prior art, then patent applicants and the public as a whole are placed in an untenable position. VIRTUALLY EVERYTHING AN APPLICANT MIGHT SAY ABOUT PRIOR ART COULD GIVE RISE TO AN “IMPLICATION.” Yet applicants routinely must describe or discuss prior art cited to or by the patent examiner. No one will know whether that discussion will limit the patent’s claims unless and until there is litigation and the Federal Circuit speaks.”

  28. 6

    There are far too many variables here for this analysis to be useful.

    The mere fact that you are only looking at cases which are appealed and not all cases, dooms the analysis.

    Further, a lower quality judge will be appealed more often and is also likely to be reversed more often. Of course, there is always the possibility that the lower court judge got it right and the CAFC got it wrong or that the lower court judge correctly followed the established law and the CAFC decided to write new law.

    Dennis does some of the same types of attempts at statistical inference and with all respect I am left with the impression that you are not having the work reviewed by a qualified statistician.

  29. 5

    Perhaps what the data evidences is that trial judges have a hard time learning from previous appeals simply because the CAFC is constantly contradicting itself. In other words, how can a judge learn from the past when the CAFC is constantly changing the rules to suit its whim of the day? And certainly one could argue that de novo review, combined with frequent 2-1 decisions, makes it impossible to glean anything meaningful from reversal statistics.

  30. 4

    Prof. Schwartz, thanks for sharing. I have 2 questions and a comment (which I realize you might not necessarily respond to):

    1. Are these results in this post statistically significant? I note that in Table III of your article, only 57 trial judges have had their claim constructions appealed 4 or more times, and only 30 have had 5 or more. I wonder whether the sample size is large enough to tell us anything.

    2. Is the number of appealed claim decisions the best measure of judicial claim construction experience among trial court judges? I wonder if a better measure might be each judge’s overall experience with patent cases.

    For example, suppose that Judge A has done 5 patent cases, all 5 were appealed, and in all 5 appeals the parties disputed claim construction questions.

    Suppose that Judge B has done 30 patent cases, but only 15 of them have been appealed, and claim construction issues were only raised in 5 of the appeals.

    Overall, I would guess that Judge B would be reversed fewer times on claim construction than Judge A, yet they appear to be treated as having equal “experience” in some parts of your study.

    * Comment: There are many things that CAFC reversals can “teach.” One is to do a better job with claim construction. Another is to write your orders in a way that makes them difficult to review or reverse. Some judges may get reversed more often than others because they write more lucidly and any “errors” they make are easier to discern.

    Very interesting stuff, and thanks for sharing.

  31. 3

    Do these data support Congressman Ira’s push for specialized courts? Do these data demonstrate that the District Court judges are poor at interpreting patent law or that the Federal Circuit keeps moving the target on the District Court judges?

  32. 2

    Maybe this should be looked at for the firms arguing the cases. Do less exprerienced firms lead judges down the wrong path, or do more experienced firms overargue the terms, and win, to the detriment of the client?

  33. 1

    Just a suggestion you may have already taken into account:

    Many who look at reversal rates might assume that the judge construed claims too broadly and was reversed. Perhaps you may want to differentiate between reversals for “too broad” and “too narrow”.

    Is there a statistically significant difference between various technical arts?

Comments are closed.