Monday, September 19, 2011

Colleen Chien: Predicting Patent Litigation

Patent litigation is an expensive headache, especially if the asserted patents seemingly come out of the blue. Patent clearance searches, used to identify patents that may potentially be asserted or infringed, can be daunting, due to the sheer number of patents that must be reviewed. A recent paper by Colleen Chien, Assistant Professor at the Santa Clara University School of Law (also covered at Patently-O), provides a way to narrow the field of potentially-litigated patents by analyzing the “acquired” characteristics of patents.

“Acquired” patent characteristics are those that accrue to a patent as a result of post-issuance activities. For example, acquired patent characteristics include changes in patent ownership, post-issuance investment by the patent owner (e.g., maintenance fees and reexamination), securitization, and number of citations by subsequent patents (“forward citations”). In contrast, “intrinsic” characteristics are those that resulting from prosecution, such as the number of claims, the number of prior art citations, and sometimes the number of patent family members.

In the paper, titled “Predicting Patent Litigation”, Chien analyzes a set of randomly selected litigated and unlitigated patents based on intrinsic and acquired characteristics. She finds that the acquired characteristics of litigated patents significantly differ from those of unlitigated patents. Specifically, litigated patents show more changes in patent ownership, are more likely to have had maintenance fees paid, undergo more reexamination, and tend to be more securitized and have more forward citations.  Chien further extends this analysis by performing a time-based analysis on post-issuance characteristics acquired prior to litigation, with similar results, indicating that these acquired patent characteristics can be used to help identify potentially-litigated patents ahead of time.

While the analysis certainly sheds some empirical light on the relationship between these acquired characteristics and potential litigation, the results (at least for some of the acquired characteristics) do not seem that surprising, as Chien acknowledges. For example, it seems obvious that a patent with up-to-date maintenance fee payments would be more likely to be litigated that a delinquent (and therefore expired and unenforceable) patent. Similarly, reexaminations seem to be rather obvious predictors of potential litigation, as described below.

One interesting extension to the time-series analysis (and somewhat related to a previous post regarding litigation timing) would be to examine the time difference between when a particular patent acquires certain characteristics and when it is eventually litigated. For example, how long before litigation does an eventually-litigated patent tend to undergo reexamination, or have its maintenance fee paid? This would provide valuable information, because even if the likelihood a patent is litigated can be predicted, the prediction is not that helpful if it only occurs immediately before litigation.

For example, take the case of reexamination, which unsurprisingly stands out as a particularly good indicator of potential litigation. Since the analyzed patents all issued in 1990, the inter partes reexamination process was not available, and only ex parte reexaminations were included in the analysis. While it would have been interesting to have patents eligible for inter partes reexamination (i.e., issued on or after November 29, 1999) included in the analysis, the value of doing so is of questionable value, since the very fact of inter partes reexamination with its adversarial nature (patent owners cannot file for inter partes reexams on their own patents; only third parties can) likely indicates that the patent is already in or very close to litigation. Similarly, since ex parte reexams can also be filed by third parties, presumably for reasons related to potential litigation, I would be curious to see how the analyzed patents break down in terms of reexam requester, and whether predictions based on the reexamination characteristic really provide useful information (i.e., allow the predicting party enough time to do something about it).

Finally, Chien suggests that mandating the public recordation of patent ownership transfers would assist in the identification of potentially-litigated patents. At first glance, this certainly seems plausible, simply because of the increased availability (and completeness) of ownership data. However, a party could simply record the transfer immediately before litigation, thus vitiating any predictive effect from the mandated recordation.

Also, is it really true that “if patents provide the right to exclude, the public is entitled to know who might do the excluding”? The purpose of a patent is to inform the public that some entity has the right to exclude others from performing the invention defined by the claims in that patent. Does it really matter who that entity is? Some think not.