The Selection Effects (and Lack Thereof) in Patent Litigation: Evidence from Trials
Abstract Using a selection corrected probit, I estimate the probability that patents will be found valid and infringed at trial. I combine for the first time detailed adjudication data with detailed patent data. I find that the selection effects for validity adjudications and infringement adjudications differ systematically. Additionally, infringement estimates do not appear to suffer from a substantial selection bias. The results highlight the importance of correctly specifying the selection mechanism in policy analysis. In contrast with previous studies, I find that the win rate for patents that go to trial is biased towards 50%. The bias is much more substantial for validity decisions, where I find unconditional win rates of 75% for adjudicated patents and 85% for matched patents. Win rates conditional on adjudication are below 60%.