This morning, I pulled up 10,000 of the most recent patents to see what's going on with pendency and other issues.
Pendency: About 65% of the patents were original applications (i.e., not based on a prior US filing). The original applications issued after an average of 40.4 months delay (3 years, 4 months).* The remaining 35% represented continuations, CIP's, and applications based on prior provisional filings. Those non-original applications had a pendency of just under 3 years (35.3 months).** The shortest original application pendency was 4 months (No. 7,172,439) and the longest stretched for over 13 years (No. 7,169,572).*** Chart I below shows the histogram of pendency for the whole group in 6-month buckets.
Examiner: About 63% of the applications were examined by an Assistant Examiner. On average, the patented cases for assistant examiners took about four-months longer than examined only by a Primary Examiner.****
Foreign Priority: Unlike US Priority, a patent that claimed foreign priority did not seem to alter the expected time to issuance. About 41% of applications claim priority to a foreign patent document. The biggest foreign locale was Japan.
Citations to Prior Art: As the PTO has reported, most patents cite fewer than 20 references. In our sample of 10,000 patents, the median number of citations was 13-14. The average number citations, however, was over 25. That skewed average comes from a small number of patents that include a very large number of citations. Six patents issued to Shell (or a Shell shell) each cited over 1000 references and may have presented the entire history of oil drilling dating back to the 19th century. (See, No. 7,172,019). 15% of applications included 40 or more citations while 20% included 6 or fewer. The majority (64%) of the cited references were to US Patent Documents and over half of the patents cited no non-patent prior art.
Claimset Length: I also wanted to see whether longer claimsets increased the length of prosecution -- they do. Chart II below shows pendency as a function of claimset size percentile.***** This statistic is a bit odd because it is based on the number of characters in a claimset, but the character count appears to be a fairly-good single-variable proxy for the length and complexity of a claimset.
Technology Class: The greatest predictor of pendency is likely the technology classification. There are likely two primary factors driving the measured difference between technology areas: (1) Backlog; and (2) Heightened Scrutiny. Some art units have large backlogs of patent applications waiting to be examined while others have much smaller dockets. Since patents are examined roughly in order, a longer backlog results in longer pendency. In addition, some of those same units with large backlogs are also the ones who have instituted the second-pair-of-eyes review to ensure that the patents that issue should be issuing. This heightened scrutiny is expected to increase pendency as well. Table I shows a handful of classes and their average pendencies.
Recently issued Business Method patents (Class 705) took over five years to issue, and DNA-type patents (Class 435) were over four years.
Inventorship: Most patents list only one or two inventors. This is reflected in Chart III below. The number of inventors did not prove to have a significant impact on pendency. One patent did list 25 separate inventors. (No. 7,169,760). That patent was, however, quite far from the mean number of inventors calculated at 2.5.
- Cite as Dennis Crouch, Prosecution Data, Patently-O (February 15, 2007), http://www.patentlyo.com.
- * 95% CI 49.96 - 40.83 months.
- ** 95% CI 34.7 - 35.9 months. T-test shows difference between two sample averages at p<0.001. As can be seen from Chart I, the data is not exactly normal, but it is somewhat close. A more particular test would also factor in the skewed nature of the data.
- *** I excluded No. 7,177,516, which was filed over 19 years ago because it was held under a secrecy order.
- **** Average pendency for cases with assistant examiner was 41.4 months while the average pendency for cases signed only by a primary was 37.0 months. The difference between these averages is significant at p<0.001 using a t-test.
- ***** Error bars are shown at the 95% CI.