“The Welfare Effects of Reverse Payment Settlements in Pharmaceuticals” (with Colleen Cunningham and Matthew J. Higgins) (Slides)
Abstract:
According to the Federal Trade Commission, one of its top priorities in recent years has been opposing reverse payment settlements in the pharmaceutical industry. These “pay-for-delay” settlements prevent generic entry; the branded drug firm/patent holder pays the generic drug firm/patent challenger to abandon litigation and delay the introduction of its generic drug into the market. This paper seeks to analyze the welfare effects of reverse payment settlements in the U.S. antidepressant drug market, where three settlements with evidence of reverse payments occurred from 2002 – 2010. First, using details of the settlements derived from FTC cases, class action lawsuits, and other firm disclosures, a counterfactual entry date is posited, the date that generic entry may have occurred if not for the settlement. Next, a discrete choice, differentiated products model of demand for antidepressant drugs is estimated. The short-run change in welfare due to pay-for-delay can be estimated using the model together with the counterfactual entry date. In the short run, consumers are expected to gain from earlier entry of lower-price generics and producers lose as monopoly profits are dissipated. The effect on long-run consumer welfare is ambiguous, as lower profits for producers may lead to reduced investment in new drugs in the future.

“Deep Reinforcement Learning in Dynamic Games” (with Stephen P. Ryan)
Abstract:
We apply methods from deep reinforcement learning to overcome the curse of dimensionality in solving dynamic games. Economists using dynamic games to model endogenous changes in market structure are often restricted by the curse of dimensionality, where the size of the state space increases exponentially in both the number of agents and in the number of state variables, making solving such games using value function iteration intractable for many applications. We overcome this challenge using deep reinforcement learning, where neural networks provide flexible approximations to the value and policy functions that generalize over the state space and grow in complexity at a slower rate than value function iteration. Agents learn the optimal value and policy functions via simulated play of the game where they receive rewards for taking particular actions. We are the first to extend such methods to dynamic games in economics, where multiple agents interact strategically by best responding to each other. This method can potentially be extended to estimation via a nested fixed-point algorithm where the solution to the game is approximated using deep reinforcement learning for each guess of the structural parameters.

“Motives of Acquisition, Innovation, and Economic Growth” (with Leo Lam) (Slides)
Abstract:
In this paper, we consider three different motivations for acquisitions which we term incremental acquisitions, where the acquiror leverages the target's technology to develop new inventions within the same technological field; exploratory acquisitions, which involves the acquiror expanding into new and previously unexplored fields; and killer acquisitions, wherein the acquiror purchases the target firm primarily to halt its technological development, thereby preserving market power. Given these three types of acquisitions, we ask how different types of acquisition affect innovation, and subsequently economic growth and welfare. We seek to answer this question by developing a dynamic general equilibrium model of innovation and acquisition, explicitly incorporating the three acquisition motives. We also construct a novel dynamic panel of the patent portfolios of US public firms using M&A data, patent reassignment data, and patent licensing data, which we combine with patent citation and patent similarity data. This dataset serves two purposes: to categorize and empirically evaluate the effect of different types of acquisition on the innovation outcomes, and to estimate moments for calibrating our quantitative model for counterfactual exercise for policy analysis.

“Vertical or Horizontal Consolidation? A New Look at M&A in the Pharmaceutical Market” (with Jennifer Kao and Luca Maini)

“Structural Transformation, Human Capital, and Innovation” (with Terry Cheung and Leo Lam)

“The Role of Assets In Place: Loss of Market Exclusivity and Investment” (with Matthew J. Higgins, Mathias Krondlund, Ji Min Park, and Joshua Pollet)