More on QWIII

The approach to grant rate taken in the third paper by Quillen and Webster (QWIII) is a bit different from that in the second paper by Quillen and Webster (QWII). To give an example, consider the year 1993.

In my 2004 paper in JPTOS, I had worked through their corrected grant rate for 1993 to illustrate how they were obtaining their numbers:

The 85% number in QWII is an average of data for the six years from 1993 to 1998, and is presented as a grant rate adjusted for patents granted on a continuing application and its parent. In general, the objective is to obtain a grant rate based on (patents granted)/(applications submitted). In the particular approach of Appendix V, one attempts to use data only from one year. The number "applications submitted" is measured by the sum of "applications allowed" and "applications abandoned" in one particular year. To correct for the effect of continuing applications, QW2 have a two step approach to calculate an "adjusted applications abandoned." The grant rate presented in QWII become (patents granted)/(patents granted + adjusted applications abandoned), so that the accuracy of the approach depends critically on the term "adjusted applications abandoned." To measure this term, QWII take the number of applications abandoned and subtract from that number 0.69 X the number of continuing applications in that year. The multiplier 0.69 is derived from the sampling of the 1000 patents and numerically is (1-97/297), purportedly the fraction of continuing applications on which the parent application does not issue as a patent. To illustrate the methodology, consider the data for the year 1993. There were 104,351 patents granted, 60,763 applications abandoned, and 50,896 continuing applications. QWII multiplies 50,896 by 0.69 to obtained 35,130 "adjusted continuing applications" . The 35,130 number is subtracted from 60,763 to obtain the number of "adjusted applications abandoned," 25,633. The grant rate becomes 104,351/(104,351 +25,633) or 80%.

In the current variation (QWIII), the new grant rate numbers for 1993 are

corrected for continuation applications 76%
corrected for continuation and CIP applications 84%
corrected for all continuing applications 91%

In QWIII, the term "corrected for _____ applications" means the number of "_____ applications" is subtracted from abandonments. Thus, correcting for "all continuing applications" means a larger number is subtracted from abandonments, and the grant rate is higher than correcting only for "continuation applications."

In QWII, an attempt was made to correct for multiple patent families, because, if this were not done, one would be counting the patent grant, but not the application, therefore, obviously creating a grant rate in excess of 100%. In QWIII, this problem still exists, as one notes grant rates in excess of 100% for "corrected for all continuing applications" for the years 1995 (116% !!!), 1999 (101%), 2000 (107%), 2001 (111%), 2002 (103%), 2003 (103%), 2004 (100%), and average 1995-2004 = 103%.

Even in QWII, the correction was not done properly:

To correct for inaccuracies in their model arising from situations wherein both a parent and a continuing application issued as patents, QWII relied on data from a sampling of 1000 patents which sampling had been published by John R. Allison and Mark A. Lemley [denoted hereafter as AL], 53 Vand. L. Rev. 2099, 2100 (2000) [see 12 Fed. Cir. Bar J. at 37-38]. In the AL data, of the sample of 1000 randomly sampled issued patents, 297 represented patents resulting from continuing applications, which includes continuations, divisionals, and continuations-in-part. For 92 of the 297 patents [31%], a patent issued on the parent. Of the 297 patents, 141 of the patents arose from continuation applications (as distinct from divisionals or continuations-in-part), and of these 141 patents, there were 19 in which a patent issued on the parent, which corresponds to 6.4% (19/297).

QWII did NOT even ATTEMPT correction for multiple patent families wherein NO PATENT ISSUED ON THE PARENT.