Do the published applications of 2001 tell us about patent grant rate?
PATENTABILITY (17 Fordham Intell. Prop. Media & Ent. L. J. 875 (2007)), Professor John R. Thomas states:
Well, in response to that--it is a great question, but my response is what I think our group of academics is opposed to is the "teaching, suggestion, motivation" requirement. We are not opposed to that being relevant, we are not opposed to that being the epitome of obviousness, but we think that that blunt approach, which boils down to you have to have prior-art journal articles describing each and every step of the claimed invention and how to put them together, simply means that there are too many errors of patents being granted that should not be, because there are other sources of knowledge that will not necessarily be before that tribunal and there are, again, artisanal skills that simply cannot be accounted for.
So certainly, in terms of the finding of fact of ordinary skill, these will be questions that will be left to the trier of facts, as so many other similar style questions are--what is a reasonable person; how does a reasonable person behave? You cannot tell me that in tort cases that you have to have a journal article proving how a reasonable person would behave when she crosses the street. *904 I think, similarly, the level of proof has been so elevated, it is basically impossible for the Patent Office to reject an application.
As you know, right now the U.S. Patent Office claims a roughly seventy percent grant rate. [FN80] But, in fact, that is an Enron-style accounting statistic, because what that really counts is the number of final rejections that come out every year. Now, of course, thirty percent of those rejected will be re-filed the next year under the continuation strategy. [FN81] So the actual corrected grant rate we do not really know, because the Patent Office will not count them, but it is probably on the order of ninety-five percent. In fact, someone did a study of the first one hundred published applications after the 1999 amendments, and something like ninety-two of them had become issued patents, which was a little embarrassing. [FN82]
Footnote 82 is to Dennis Crouch, Statistical Interlude: First 100 Published Applications, Patently-O, Aug. 23, 2006, http://www.patentlyo.com/ patent/2006/08/interesting_sid.html. Although not noted by Professor Thomas, there were some caveats to Crouch's analysis that were pointed out by commenters to the blog post. One commenter correctly noted that the group of "published applications" is not representative of all applications, because patent applicants have the opportunity to opt out of the publishing procedure. Those who do opt out are likely to be most concerned about the consequences of the publication of an ultimately rejected application. This would be an issue of underinclusion. Another commenter pointed out that the proper focus should be on the final state of the allowed claims, rather than on the percentage of allowed applications. This is related to a theme raised by LBE in 88 JPTOS 1068 (2006) and papers cited therein.
There is another issue. The first 100 or so published applications may be systematically non-representative. Although Professor Thomas incorrectly stated Now, of course, thirty percent of those rejected will be re-filed the next year under the continuation strategy, it is correct that approximately 30% of patent applications are CONTINUING applications which include cons, divs, cips, and RCEs, with RCEs (which merely continue prosecution) the most abundant. Of the first twenty applications published on March 15, 2001, three did not lead to issued patents. Curiously, of the remaining 17 published applications, 13 were divisional applications related to an already issued patent and 4 were continuation applications related to an already issued patent. Not one of the 17 published applications that yielded an issued patent was a first-filed case. Given that 70% of U.S. applications are NOT continuing applications, one can see that the first 100 or so published applications in the year 2001 were not statistically random. Unless perhaps you think it is likely that something with a 30% probability is going to come up 17 times in a row.
The idea that there is a 92% patent grant rate based upon the published applications of 2001 (as suggested by Thomas' statement: In fact, someone did a study of the first one hundred published applications after the 1999 amendments, and something like ninety-two of them had become issued patents, which was a little embarrassing.) is without merit.
This would not be the first time that Professor Thomas has misrepresented patent grant rate statistics. As pointed out in an IPBiz post on April 15:
In turn, footnote 202 states: John R. Thomas, The Responsibility of the Rulemaker: Comparative Approaches to Patent Administration Reform, 17 Berkeley Tech. L.J. 727, 754 (2002).
The interesting part is that Thomas says something about the 97% patent grant rate that Quillen and Webster didn't really say. As seen in the 2006 paper by Quillen and Webster, the paper in 11 Fed. Cir. B.J. 1 established upper and lower bounds for patent grant rate. By 2002, Quillen and Webster had adjusted these bounds; by 2003, Clarke had criticized the formalism of 11 Fed. Cir. B.J. 1 and by 2004, LBE had criticized the adjusted formalism. The viability of the 97% number had been strongly questioned before the year 2007. Further, given the information within footnote 17 of 11 Fed. Cir. B.J. 1, the lack of validity of the 97% number was apparent at the time of the writing of 11 Fed. Cir. B.J. 1.
Statistical Interlude: First 100 Published Applications
The problem with the use by Thomas of the work by Crouch is the following: The first 100 published applications like all the applications published in the first 18 months had to have earlier filed applications to be eligible for publication. Every application filed on or after November 29, 2000, that had a benefit/priority claim which went back more than 18 months was immediately eligible for publication.
The first 100 published applications were by no means statistically random. There were earlier filed applications in their "patent family history." This is not true of patent applications generally (only about 30% of applications are continuing applications). Thus, the statement by Thomas: In fact, someone did a study of the first one hundred published applications after the 1999 amendments, and something like ninety-two of them had become issued patents, which was a little embarrassing. is misleading to the extent it suggests that 92% of all patent applications lead to issued patents.