In the not too-distant future, the PTO will likely re-invigorate its effort to force patent applicants to conduct pre-filing prior art searches. During his stint as IPO President, Marc Adler suggested that rational patent applicants should already be conducting pre-filing searches in order to assess patentability and define likely claim scope. Marc offered the then unsubstantiated comment that "if the initial search done by the applicant is as comprehensive as the one done by the patent examiner, it should follow that the pendency of the application will be shorter and the patent more likely will be valid." Despite Marc's suggestion, many applicants do not conduct pre-filing searches – largely because of the up-front cost of searching and analysis.
I wanted to test Marc's suggestion that pre-filing searches add value by leading to better patent applications and shorter application pendency. Although applicants do not ordinarily disclose whether a pre-filing search was conducted, this information can be teased-out by looking to see whether an information disclosure statement (IDS) was filed on the same day as the patent application itself. Of course, filing an IDS with the application does not necessarily mean that a pre-filing search was completed since the submitted art may have been obtained other ways. However, the IDS filing at least indicates that specific prior art was considered prior to filing, and I cautiously believe that such an early IDS filing does suggest the existence of a pre-filing search.
Yesterday, I created a set of 3000+ randomly selected patent applications filed in December 2004 and January 2005 and then divided those applications into three groups: Group 1 applications where no IDS was filed during prosecution; Group 2 applications where an IDS filing coincided with the filing of the patent application itself; and Group 3 applications where the first IDS was filed sometime after the application was filed. The data indicates the status of each application (abandoned, patented, or pending) as of June 26, 2009 – about 4 ½ years after filing. For this first pass, I simplified the data by excluding cases that (1) never published as applications (since data is kept secret for most of those cases); and (2) cases claiming priority to a "parent" patent application (since early IDS filing in the child cases provides little indication of whether a pre-filing search was conducted prior to the parent being filed).
Results: Looking 4 ½ years out from the application filing date, 25% of cases fell into Group 1 (No IDS); 26% into Group 2 (IDS filed with application); and 49% into Group 3 (First IDS filed after application). Table 1 shows these results.
Portion of All Cases | ||
Group 1 | No IDS Filed | 25% |
Group 2 | IDS Filed With Application | 26% |
Group 3 | IDS Filed After Application Filing Date | 49% |
Table 1 | ±3% margin of error at 99% CI |
In my sample, Group 1 applications were much more likely to be abandoned and less likely to be patented within 4 ½ years of filing than the Group 2 applications. Table 2 below shows this result. In particular, 40% of Group 1 (No IDS) applications were abandoned while only 26% of Group 2. Similarly, only 34% of Group 1 applications were patented compared with 46% of Group 2 applications. Pierson's chi-square test of independence results in P=0.000 – indicating that these differences as a whole are statistically significant.
| Abandoned | Patented | Pending | |
Group 1 | No IDS Filed | 40% | 34% | 26% |
Group 2 | IDS Filed With Application | 26% | 46% | 28% |
Group 3 | IDS Filed After Application Filing Date* | 24% | 40% | 36% |
Table 2 |
Table 2 shows that Group 3 (applications whose first IDS was filed after the application date) had an even lower average abandonment rate than Group 2. That number is likely skewed, however, because of the confounding relationship between keeping an application alive and filing a late IDS.
In looking at prosecution timing, I did not find a statistically significant difference in pendency between Groups 1 and 2. Group 3 applications did take significantly longer to patent/abandon. However, that result is likely skewed by the selection bias noted above.
Association Between Grant Rate and Pre-Filing Searches: These high-level results shown in Table 2 suggest a relationship between an applicant's success rate (patent vs. abandon) and the early filing of an information disclosure statement.
An alternative explanation for the relationship between higher grant rate and early IDS filing may be that both are indicators that applicants recognize higher-quality inventions and consequently work harder for valuable patent protection. In that scenario, there is no cause and effect relationship between a pre-filing search and a higher patent grant rate. If this quality-recognition theory were true, I would expect that applications filed by patent attorneys/agents (and larger law firms) to be more likely to include a pre-filing search. However to the contrary, my data comparing applications associated with one or more registered patent practitioners with applications not so associated shows very little difference in the rate of pre-filing searches. Rather, practitioners appear to focus on IDS filing more as a CYA burden as suggested by the greater likelihood of late-filed IDSs in practitioner filed cases. Practitioner-associated cases are also more likely to be patented and not abandoned.
Operation of the patent office as well as applicant behavior both vary depending upon the technology at issue. Table 3 shows IDS timing for each examination technology center. Applications in TC 1600 (Biotechnology and Organic Chemistry) are often seen as more valuable on a per-application basis. However, TC 1600's applications are less-likely than average to include an IDS filing with the original application. (χ2 P=0.000). Consistent with disclosure statements as merely CYA material, however is the fact that TC 1600 has the highest rate of later-filed IDS's.
IDS Timing | |||
Tech Center | Group 1 | Group 2 | Group 3 |
1600 | 11% | 24% | 64% |
1700 | 16% | 24% | 57% |
2100 | 29% | 26% | 40% |
2400 | 28% | 22% | 41% |
2600 | 20% | 33% | 40% |
2800 | 22% | 27% | 47% |
3600 | 35% | 27% | 38% |
3700 | 27% | 28% | 38% |
Table 3 | |||
Notes:
- Of course, this comes with the baggage of a retrospective study.
- This is a continuing project – suggestions are welcome.
