In June, the Government Accounting Office (GAO) issued a performance audit of the U.S. Patent and Trademark Office (USPTO), “Patent Office Should Define Quality, Reassess Incentives, and Improve Clarity.”  The report was prepared at the request of the House Judiciary Committee, which asked the GAO to examine trends in patent litigation and identify opportunities to improve patent quality. To address the latter, the GAO attempted to identify the factors that contributed to “low quality patents.” The GAO’s analysis of these factors was based primarily on a lengthy survey of USPTO patent examiners on various factors that affected the quality of the applications they examined.
The GAO report identified Requests for Continued Examination (RCEs) as a key issue impacting patent quality:
Requests for continued examination: Applicants are currently allowed to file an unlimited number of requests for continued examination, which is a request by an applicant to reopen examination of the patent application after the prosecution of the application has been closed. Applicants request continued examination most often after final rejection of an application, according to USPTO officials. Such requests provide applicants with virtually unlimited attempts to secure a patent, which is problematic for patent quality, according to some stakeholders. Some stakeholders also told us that such unlimited requests can wear down examiners, making them more likely to eventually grant the patent. While USPTO has reached out to patent applicants to learn why they were using such continued examinations, the number of requests for continued examination continues to pose a burden for the agency’s examiners. 
In its introductory paragraph the GAO report expressly linked “low quality patents” to “an increase in patent infringement suits,” and in the above quotation it links RCEs to “low quality patents,” stating “Such requests provide applicants with virtually unlimited attempts to secure a patent, which is problematic for patent quality, according to some stakeholders.” Other than the “stakeholders” the GAO offers no evidence for this assertion.
Then in October 2016, two authors of the GAO report published a blog in which they echoed these conclusions, stating that the “USPTO is also not able to make terminal rejections. Applicants can re-apply an unlimited number of times, and this contributes to the sizeable backlog of unexamined applications and subsequent time pressures at the agency.” Some commentators propose that the USPTO should abolish RCEs and close prosecution after a fixed number of office actions.
The problem of “unlimited” RCEs is a myth. While it is true that the patent statute does not limit the number of RCEs, the reality of patent examination and prosecution results in a negligible number of high count RCE patents, and they have no impact on the patent litigation landscape.
To bust this myth, I analyzed the frequencies of RCEs for both litigated patents and a random sample of issued patents. The litigated patents included 12,753 utility patents drawn from over 24,400 patent litigations filed between January 2010 and October 2016, with filing dates after November 29, 2000. The control group was a random sample of 38,265 patents, also filed after November 29, 2000. Using data from Patent Advisor, I obtained the number of RCEs in each patent, as well as data identifying Tech Center, Art Unit, classification and so forth. Figure 1 summarizes the analysis.
Click here to view table.
The first column shows that the majority of the issued patents, about 72%, issue with no RCEs at all. Second, in both the Control and Litigated groups, more than 99% of the patents issued with three RCEs or less. There were only 14 patents (0.00037%) in the Control Group, and only 6 (0.00047%) in the Litigated Group, that had seven or more RCEs. Patent examiners are made of stronger stuff than to be worn down by these thimblefuls of high-count RCEs. More importantly, it demonstrates that high-count RCE patents do not contribute to the current spate of patent litigation.
I also looked at the RCE patterns in different Tech Centers ("TC"). The average number of RCEs per patent is shown in Figure 2, for both the Control Group and the Litigated Group.
Click here to view table.
I break TC 3600 into two groups, “3600BM” which comprises the “business methods” art units 3620, 3680 and 3690, and “3600” which covers the remainder (e.g., transportation, construction, apparel, structures, engineering). Tech Center 2400 (Computers) and TC 3600BM have the highest average RCE rates, but still far less than one RCE per patent on average. In addition, the average number of RCEs as between the Control and Litigation Group in each Tech Center is very close overall, with the most pronounced differences in TC 1600 (Biology) and TC 3600. This suggests that litigated patents do not have significantly different RCE characteristics as compared to non-litigated patents—contrary to the suggestions of the stakeholders interviewed by the GAO.
Different Tech Centers do experience different RCE distributions. Table 1 shows the percentages of patents in each Tech Center by the number of RCEs:
Click here to view table.
In each column, the highest values are shaded red and the lowest values are light green. Here, we see that TC 2800 (Semiconductors, Memory, Optics) has the highest percentage of patents without any RCEs at all (79.77%), while TC 2400 (Computers) has the lowest (57.68%). Even so, about 41% of patents in TC 2400 issued with 1-3 RCEs. The differences between Tech Centers may be at least in part a result of differences between examiners. Considering just the 338 examiners in the study with more than 20 patents each, the average number of RCEs per examiner per patent ranged from zero to about 1.2, a significant variance.
Because the GAO suggests that RCEs are linked to highly litigated patents, I looked at the RCEs patterns of the most frequently litigated patents. Table 2 summarizes all patents that had been in at least 80 patent cases since 2010:
Click here to view table.
There is only a single RCE in these 13 patents (which account for over 2,000 litigations). The top 100 most litigated patents, accounting for about 20% all patent cases since 2010, have only 37 RCEs between them. This average (0.37 RCEs per patent) is exactly in line with the average figures in Figure 2. On the other end of the spectrum, 89% of the patents in the Litigated Group were involved in three or less lawsuits, and for these patents, the average number of RCEs per patent was 0.39, a statistically insignificant difference. In short, high-count RCE patents themselves do not contribute to patent litigation.
Indeed, anyone who knows prosecution understands that all things being equal, the more RCEs, the less likely the patent is to be litigated. This is because more RCEs typically means more office actions, which in turn implies more amendments and statements in the file history as the applicant attempts to narrow the claims to reach allowable subject matter. Thus, higher RCE counts typically result in narrower, not broader claims. Further, multiple amendments provide fertile ground for defendants to find a basis for further narrowing claim constructions. Table 2 above shows that in the most frequently litigated patents, the average number of office actions (and hence amendments) is less than 1, exactly in line with expectations. Figure 3 shows the average number of office actions per RCE for the Control and Litigated Groups:
Click here to view table.
Here, the values below five RCEs are most reliable, as the number of patents with six or more RCE is very low, as noted above, and so the averages are very sensitive to outliers. This graph confirms that on average the number of office actions is just about two times the number of RCEs, precisely what we would expect under the current rules of compact prosecution which make the second office action final in most cases. At the upper end the number of actions drops off sharply; a review of the patents at this upper end of the RCE count shows these tend to be serial RCEs with information disclosure statements filed after an earlier allowance. In other words, the applicants were not wearing down the examiner, but paying to submit additional prior art, perhaps to bolster the patent or otherwise delay issuance.
Let’s return then to the GAO report. First, the GAO’s repeated use of the term unlimited introduces an implicit argument and bias. While it is true that the statute does not put a limit on the number of RCEs, the word itself implies there is a problem. Those of us who have, had, or were children know that too many requests for anything can overwhelm. Unlimited suggests the desired conclusion in the last sentence: “the number of requests for continued examination continues to pose a burden for the agency’s examiners.” Providing the basis for this assertion are the all-knowing “stakeholders” who suggest that unlimited RCEs can “wear down” the examiner. I can only surmise that these stakeholders do not, in fact, prosecute patents because if they did, they would know this is not true, as I have shown. Most examiners I have spoken to since the GAO report told me that RCEs are easy to handle because they already know the details of the application and the prior art. So, far from being worn down, most RCEs are easy work for examiners and typically move the claims towards a narrower scope. Perhaps this is why comments from stakeholders are mentioned here, but the comments of examiners were not.
The GAO survey asked the examiners to evaluate which factors impacted their ability to conduct a prior art search and examination of the patent application, and the degree to which each factor made their work more or less difficult. Overall, there were 47 factors that could impact the difficulty of search and examination.
Significantly, the GAO did not ask the examiners whether “unlimited RCEs” impacted their work. Instead, the GAO asked:
How much easier or more difficult do the following factors make it to complete a thorough examination: The ability of applicants to file RCEs.
This was the proper question, because it did not implicitly bias the examiners’ views of RCEs. Accordingly, any conclusion that the GAO made regarding the alleged "unlimited RCE" problem could not be supported by the survey responses.
Most importantly, the GAO report did not attempt to compare the differences between the various factors, but simply reported the percentages of responses for some of the factors without any context. This made it impossible for the reader to figure out which factors were really most important to examiners. To overcome that problem I obtained the tabulated survey data from the GAO’s website, and re-analyzed the results to obtain comparative rankings.
Out of the 47 factors that make examination and search much more difficult for examiners—the factors that would seem to contribute significantly to patent quality— the “ability of applicants to file RCEs” ranked near the bottom, 40th out of the 47, with just 1.6% of examiners indicating that this factor made their work “much more difficult.”
By comparison, the PTO’s misclassification of patent applications was much more significant, ranking 5th overall, with 43.6% of examiners indicating that misclassification made examination much more difficult. “New patentability related case law (e.g. Alice)” ranked 17th (23.2% of examiners), pro se applications ranked 19th (21%) and poor quality drawings ranked 25th (18%). Even the USPTO’s own IT systems were considered a greater impediment to examination than RCE, ranking 32nd (8.1%). Why the GAO chose to focus on RCEs rather than these more significant factors is unknown.
Further, the GAO's own data on RCEs is essentially absent from its analysis. In a footnote the GAO acknowledges that only about 10% of all RCEs filed in 2015 were for a third RCE. While the GAO provided the numerical details of the examiners’ responses in their discussion of other quality factors, they provided none in support of their RCE analysis.  This may also explain why the GAO relied only on some “stakeholders” for their conclusions, rather than the actual data from the survey.
Finally, examiners already have the tools to combat overly aggressive applicants: if an applicant files an RCE without a substantial amendment, the examiner can issue a first-action final rejection under MPEP 706.07(b). Given that the fees for RCEs escalate and are higher than typically new application filing fees, it would be wasteful for an applicant to file multiple RCEs with the same or trivially amended claims to “wear down” the examiner. This costly strategy, if is used at all, is extremely rare: in my study, I found only 45 patents (0.0008%) that had three or more RCEs and more final rejections than non-final rejections (which happens only if there are multiple first-action finals). In a number of these cases, one or more RCEs were after allowance, and used to file supplemental IDS's. This data suggests that only a tiny fraction of applicants are filing RCEs without making significant amendments.
That said, while there is no evidence that frequency of RCEs itself impacts patent quality, proposals to change the RCE rules are worth considering. Figure 4 shows that in both the Control and Litigated Group, about 97% of the patents issued with five or fewer office actions (counting both non-final and final rejections):
Click here to view table.
This is consistent with the common observation that two office actions are simply not enough for the examiner and applicant to identify whether there is allowable subject matter. Thus, instead of routinely making the second office action final, the fourth action can be final, at which point the first RCE would be available. Subsequent RCEs would be necessary in only about 5% of cases and could be priced accordingly. Adjustments to the examiner count system and the number of hours allocated per application would likewise need to be made.
It’s time that the myth of “unlimited RCE” be terminally rejected.