To provide a quick recap of the hypotheses under investigation, we first predict that the PTO will respond to negative shocks to the Agency's resources by (Hypothesis 1) granting patents at relatively higher rates to applicants in those technologies with stronger proclivities to continue the application process upon rejection in comparison with applicants in those technologies with weaker such proclivities, and (Hypothesis 2) allowing a greater number of claims within each allowed application in those technologies prone to high repeat filing relative to technologies not prone to high repeat filing. By responding along both the extensive and intensive margins of allowance in this manner--by both issuing patents more frequently on an application-by-application basis and allowing more claims per application--the Agency may find itself better able to diminish repeat filings and concomitantly decrease (or at least slow the growth of) its backlog of patent applications.
Our findings are consistent with both of the above hypotheses. That is, our results indicate that during times of resource constraint the PTO grants patents in high-repeat-filing technologies, which include, among others, information and communication and health-related technologies, at a relatively higher rate than patents in low-repeat-filing technologies, which include, among others, furniture, house fixtures, and apparel textiles. Additionally, our results indicate that upon deterioration in resources, the PTO increases the number of allowed claims in patents in technologies with stronger proclivities to continue the application process upon rejection compared to those technologies with weaker such proclivities. Thus, our findings provide compelling empirical evidence that the PTO is indeed, in certain circumstances, biased toward granting patents. Moreover, our findings suggest that the Agency is targeting its granting proclivities in certain fields over others.
Primary Results: Graphical Analysis
Description of graphical approach
In the Online Appendix, we present tabular regression results for the basic difference-in-difference approach mapped out above. (96) In the main text, however, we aim to confront this analysis in a more visible manner. As such, we focus on a pictorial depiction of our key findings. If anything, such graphical analyses are more comprehensive insofar as they allow us to observe how the story materializes dynamically on a year-by-year basis.
More specifically, in Figure 1, we plot a year-to-year time trend in the differential grant rate between technologies that are highly prone to repeat a filing upon rejection--which include, among others, software, business methods, medical and surgical devices, and genetics--and technologies that are not especially prone to file such repeat applications, which include, among others, furniture, house fixtures, and apparel textiles. (97) We plot this trend over the 1991 to 2010 period. Each value in this time trend captures the degree to which grant rates in the high-continuation-prone groups exceed those in the low-continuation-prone groups at that time. However, rather than presenting a trend in the absolute differences in the grant rates between such groups, we normalize such differences such that they equal zero in the base period of 1991 and thereafter plot how the normalized differences evolve over each subsequent year. For instance, assume that the grant rate of the high-continuation-prone group equals 70% in 1991 and the grant rate of the low-continuation-prone group at that time equals 65%. We nonetheless plot a difference equal to 0 in 1991. Our interest is in learning how this baseline differential itself trends over the subsequent time period. As such, hypothetically, if the low-continuation-prone group's grant rate stays at 65% in 1992, but the high-continuation-prone group's rate increases to 72%, then we would plot a value of 2 for 1992 because the difference in those rates would have grown by two percentage points over that one-year period. (98) Similarly, if the high-continuation-prone group's rate decreases to 67%, then we would hypothetically register a value of -3 for 1992.
By normalizing the differential grant rate to zero in the base period, we effectively account for and disregard any fixed, time-invariant differences in granting tendencies that apply to applicants across technologies. (99) We stress that our methodological framework is not simply capturing any inherent disparities in grant rates between high- and low-continuation-prone technologies. Such disparities are, if anything, irrelevant sources of information for our analysis. What is relevant for our purposes is how those differences (whatever level they may be at) change over time in connection with an alteration in the resources of the PTO. That is, when the Agency experiences a negative shock to its resources, do we see the grant-rate difference between high- and low-continuation-prone groups growl This graphical and methodological framework allows us to hone in on that association of interest. (100) To facilitate a visual representation demonstrating how trends in the differential grant rate between high- and low-continuation-prone applicants correlate with corresponding trends in the resource health of the Agency, we simply overlay the differential grant-rate time trend (marked by the solid line with triangles in Figure 1) with a time trend in the annual "sustainability scores" (marked by the dashed line with circles).
Note that this approach of plotting the time trend in the difference in grant rates across our "treatment" groups and our "control" groups also allows us to neutralize any across-the-board time trends in grant rates. As will be discussed momentarily, the differential grant rate between the treatment and control groups plotted in Figure 1 begins to rise when the Agency's resource state deteriorates. This rise is not merely a reflection of a general upward trend in grant rates across the whole Agency over those years--for example, a trend in which the grant rates rise by five percentage points for both the high-continuation-prone group and the low-continuation-prone group. Because in each year we are taking the difference between the grant rates of these respective groups, we are effectively netting out any national trends in PTO practices.
Ultimately, one can see that this graphical approach, by accounting both for fixed differences in granting tendencies across technologies and for general, across-the-board differences in granting tendencies across years, effectuates the dual layers of differencing characteristic of the "difference-in-difference" experimental design discussed in Part IV. Differencing out across-the-board trends in granting practices (by making comparisons across technologies) and inherent differences in granting practices across technologies (by making comparisons with respect to the baseline reference period), we are able to isolate the theoretical prediction of interest. In other words, this experimental design allows us to rule out a number of potentially confounding explanations for the observed patterns, leaving us with greater confidence that we are identifying the causal effect of the PTO's ability to meet its expected examination demand on its tendencies to grant excessively. (101)
Before interpreting the graphical findings depicted in Figure 1, we note that Figure 2 replicates the grant-rate analysis from Figure 1 for the case of our alternative dependent variable--the average number of allowed claims in each technology-by-year group. That is, Figure 2 plots a year-by-year trend in the difference between the average number of claims in allowed applications of the high-continuation-prone technologies and the average number of claims in allowed applications of the low-continuation-prone technologies. This difference is normalized to zero in the 1991 base period (as above). We likewise overlay this differential claims-number trend with the year-by-year trend in the Agency's sustainability score, allowing us to visualize whether the PTO will begin to allow more claims precisely when the Agency's resource status begins to deteriorate.
Description of findings
With this experimental framework set out, what do Figures 1 and 2 tell us? Overall, these visuals are remarkably consistent with the theoretical predictions of our analysis. That is, our findings are consistent with the PTO responding to negative shocks to the Agency's resources by (Hypothesis 1) granting patents at relatively higher rates to applicants in those technologies with stronger proclivities to continue the application process upon rejection in comparison to applicants in those technologies with weaker such proclivities, and (Hypothesis 2) allowing a greater number of claims within each allowed application in those technologies prone to high repeat filing relative to technologies not prone to high repeat filing. To begin, consider the early years in the sample--the early 1990s. At this time, the sustainability score rises, suggesting that the Agency's resources (and hence its ability to process the applications demanded of it) may have been improving. In part, this may have been due to the fact that the PTO began to collect the substantial twelve-year maintenance fees for the first time during these years. (102) With such favorable conditions, one might predict that the Agency would find itself in little need of taking any distortionary actions--namely, granting excessively--in order to diminish repeat filings. Consistent with these predictions, over these early years, the Agency does not appear to have altered the manner in which it treated the high-continuation-prone groups relative to the low-continuation-prone groups either in terms of whether to allow applications at all or in terms of how many claims to allow.
The fact that the differential granting outcomes...
Does the U.S. Patent and Trademark Office grant too many bad patents? Evidence from a quasi-experiment.
|Author:||Frakes, Michael D.|
|Position:||V. Results through Conclusion, with footnotes, p. 644-676|
To continue readingFREE SIGN UP
COPYRIGHT TV Trade Media, Inc.
COPYRIGHT GALE, Cengage Learning. All rights reserved.
COPYRIGHT GALE, Cengage Learning. All rights reserved.