Are Software Patents Different? Evidence from the Empirical Analysis of Litigation

Shawn P. Miller

Advisor: Alex T Tabarrok, PhD, Department of Economics

Committee Members: Robin D. Hanson, Joshua D. Wright

Hazel Hall, #348
November 08, 2012, 01:00 PM to 10:00 AM


Since software patents first proliferated during the 1990s, critics have argued they hinder, rather than further, innovation.  In the three essays that comprise this dissertation I obtain empirical evidence for two common criticisms of software patents—that they are more likely to lack innovation value and that they possess more uncertain property boundaries. 

The first essay is an empirical investigation of Jaffe and Lerner’s (2004) assertion that a swamped U.S. Patent Office has granted an inefficiently large number of patents with negligible innovation value.  I test this argument’s plausibility and determine the characteristics of patents without innovation by analyzing litigated patents subject to anticipation or obviousness decisions.  Using a selection corrected probit model, I obtain unconditional estimates of the likelihood patents with given characteristics lack innovation value.  I estimate a surprising 28 percent of all patents would be found at least partially invalid if litigated.  Software, business method and licensing firm-owned patents possess significantly higher innovation-based invalidity rates and I argue this may be due to these patents being more difficult for the Patent Office to review because their scope is more uncertain.

The second essay investigates Allison, Lemley & Walker’s (2011) finding that repeat patent plaintiffs overwhelmingly lose at trial despite economic theory predicting they should enjoy greater litigation success than owners who assert their patents in fewer lawsuits.  Counting consolidated judgments as one trial and analyzing a larger sample of lawsuits, I find that repeat patent plaintiffs are generally more likely to win trials.  However, this is not true of repeat software patent plaintiffs because these are not more likely to win infringement judgments.  I argue this fact is not inconsistent with rational software patent owner behavior but is best explained by the theory that software patents possess more uncertain boundaries.

The third essay seeks direct evidence of software patent scope uncertainty in extending the empirical literature on claim construction reversal rates to determine whether the Federal Circuit has been more likely to find error in district court construction of software patents.  Not only do I find that it has, I find that over the last decade software patents account for half the difference between the Federal Circuit’s high claim construction reversal rate and its lower average reversal rate on all other patent issues.  These results are cause for optimism in that in general the application of existing claim construction law is more predictable that has been feared.  However, this optimism does not extend to software claim construction, which consistent with Bessen and Meurer’s (2008) argument, is highly unpredictable.