I compiled a list of the over five hundred thousand reference citations in the almost eighteen thousand utility patents issued in the past two months. (Jan 1, 2009 – Feb 18, 2009, excluding reissues and reexaminations). In this sample, the average patent issued with 31 cited references. This is a skewed average – the median patent only cites 15 references. The vast majority of references (77%) are introduced by the applicant during prosecution while less than a quarter (23%) are cited by the examiner.
Patent documents (including US & foreign issued patents and published applications) represent 82% of the cited documents while non-patent literature represents 18% of the total. As was noted in the comments regarding PTO search tools, PTO searches are overwhelmingly focused on patent documents. Thus, 95% of references cited by PTO examiners are patent documents. (See pie chart aside).
Although – on average – applicants submit most of the prior art, Applicants submitted absolutely no references (zero) 15% of the patents in my study, and the median number of references cited by the applicant is only 7. Applicants cite over 200 references in only 2% of cases. The PTO's stalled IDS rules give the applicant 20 'free' references that can be cited without payment of a fee or further description. About 23% of the issued patents in my sample include more than 20 references cited by the applicant. The chart below is a histogram of applicant submitted references. About 12% of patents are "off the chart." In those cases the applicant cited more than 40 references.
Take-away: First, the data shows that applicant submitted references are substantially different from those cited by examiners. Non-patent literature is an important category of prior art that is largely missed by examiners, and the current information disclosure system does a pretty good job of getting more of that type of art in front of the examiner than he or she would otherwise see. A second point is that most of the time, examiners are not being swamped by reference submissions. It is surprising that 15% of patented cases include absolutely no applicant cited references. On the other hand, folks who submit over 200 references are deliberately choosing to do so. Generally, the very large reference volume comes in cases where the applicant knows that the innovation is important and is looking for more assured protection. In those cases, I don't see the harm in making the applicant somehow pay for the additional resources required to review those references. [Of course, in that case, the PTO should allow more examination time as well]
The next graph shows the average number of applicant-cited references for various top-patenting companies. The size of the blue bar reflects the number of applicant cited reference on average, for that assignee's Jan-Feb 2009 patents. (Scale at bottom). The graph is sorted so that companies with the greatest average number of cites are shown at the top. Although this is not strictly true, the difference between companies is largely a difference in areas of technology. Thus, companies patenting electronics generally cite less prior art than others. You may also notice in this list that US based companies tend to cite more prior art. The final chart in the series shows that – on average, US assignees submit over three times as many prior art references as do their foreign counterparts or patents without an assignee.