Out of court settlements are ubiquitous in patent litigation, due to the significant costs of pursuing a trial and uncertainty over its outcome. This is believed to encourage "frivolous" litigation, however the secretive nature of settlement agreements poses a challenge to evaluating the quality of cases filed and the cost they impose on innovators. I develop a dynamic model of litigation in which the defendant screens out plaintiffs through a series of sequential out-of-court settlement offers, exploiting heterogeneity in the likelihood of cases to be dismissed or successful. Employing a newly constructed dataset that combines granular information on all patent infringement cases filed in the US from 2007-2021 and data on the parties involved, I estimate my model. This allows me to retrieve (i) the distribution of the ex-ante probability of victory for plaintiffs, (ii) the unobserved distribution of settlement amounts and (iii) the relative bias of courts towards plaintiffs and defendants. I find that patent litigation costs over $48.8 billion a year to listed defendants, $24.7  billion of which are from settlement transfers. Furthermore, cases brought to court by patent assertion entities are smaller in size, lower in quality and more likely to settle for smaller amounts. I then consider two counterfactuals. First, I consider several fee-shifting rules, finding that conditional fee-shifting is effective in discouraging frivolous litigation, however it can induce costly delays before settlements. Then, I turn to forum shopping. I find that restricting plaintiffs' discretionality in the choice of venue would dramatically reduce the incentives to file low-quality cases, especially in the E.D. of Texas.

We study how fee shifting affects patent enforcement and innovation. Our identification relies on the U.S. Supreme Court's 2014 ruling in Octane Fitness v. ICON Health & Fitness, which broadened courts' discretion to award attorneys' fees against exceptionally weak infringement claims. Using a quasi-regression discontinuity design comparing cases pending at Octane's release with those filed before or after, we find that the decision increased defendants' willingness to contest weak claims and prompted plaintiffs, especially Patent Assertion Entities (PAEs), to file stronger ones. Pending cases saw higher fee awards and lower settlement and plaintiff success rates; subsequent PAE cases involved stronger patents and higher success rates. A difference-in-differences analysis further shows that Octane boosted innovation: public firms that were particularly exposed to PAE assertions prior to Octane increased R&D and patenting in response, and private startup firms performed better in venture capital markets.

Understanding how many granted patents are actually valid, and why some examiners or technologies produce more invalid patents than others, is central to evaluating the functioning of the patent system. Yet existing data only reveal validity for the small fraction of patents that are later challenged, creating a severe selection problem. This paper develops a framework that combines PTAB challenge data, litigation records, and USPTO examination outcomes to recover the joint distribution of patent validity and examiner skill. We model both the examiner’s grant decision and the petitioner’s challenge decision within a unified latent-variable structure and derive conditions, akin to a Heckman correction, that allow us to identify the distribution of examiner-specific false positive rates and signal precision despite heterogeneous examiners and endogenous selection into review. The resulting estimates provide the first systemwide view of patent quality and examiner heterogeneity corrected for strategic selection into validity challenges. 

The strategy of innovation by acquisition is commonly discussed in MBAs classes and featured in the news, however, little research exists on its effects on startups entry and innovation. This paper evaluates the effect of systematic acquisition of startups from oligopolistic incumbents on the incentives for the former to invest in R&D. I propose a general "Musical Chairs'' model of competition, where demand depends substantially on the ordinal quality ranking of incumbents and only a fixed number of firms can operate with profits. Several common demand system fall within this framework. I assume that startups hold asymmetric information about the quality of their assets, however they can be acquired before they enter the market. Two main motives drives the acquisitions: preemption and technological race. Market leaders are strongly motivated to buy startups to prevent followers to catch up, while laggard are mostly motivated by the risk of being pushed out of the market from new entrants. Depending on the industry structure, I find that either force can prevail, determining which incumbent acquires the startup and the acquisition price. Young and competitive industries lead to high incentives for innovation, while established industries with significant technological gaps between the leaders and the laggards provide very weak incentives to invest. I enrich the model introducing dynamics and allowing incumbents to be forward-looking, and I discuss key indicators for antitrust concerns on the dynamics effect of these acquisitions.