Working papers

With fewer than 10% of new drugs reaching the market, the drug development process is notorious for its high attrition rate. However, we rarely observe the reason for a drug’s discontinuation. It is known that pharmaceutical firms withdraw drugs after clinical failures, such as when trial results do not demonstrate adequate safety or efficacy according to FDA standards. At the same time, surveys suggest that firms also withdraw drugs for strategic reasons, such as when competition makes it unprofitable to continue development. Disentangling these two sources of attrition is necessary in order to predict the effects a government policy would have on the number of drugs that reach consumers. In this paper, I propose an empirical framework to separately identify the two components of attrition for each disease. To this end, I build a continuous-time dynamic model of the drug development process. In the model, firms take competitors’ R&D choices into account when they make exit decisions at different stages of the innovation process. To estimate the model, I use rich data on the development histories of experimental drugs, clinical trial outcomes, and disease-specific epidemiological characteristics. I find that, on average, strategic terminations account for 8.4% of all attrition, and as much as 35% for some diseases. Using these estimates in counterfactual simulations, I show that whereas large subsidies for clinical trials increase the rate at which new drugs reach consumers, the same effect can be achieved through any regulatory adjustment that can marginally help lower the probability of late-stage clinical failures. Furthermore, due in part to strategic responses by firms, changes that lead to fewer clinical failures in late-stage trials are 50% more effective for increasing new drug launches than equivalent changes associated with early-stage trials.

We estimate the effect of major airlines' use of the same regional airlines (as common-subcontractors) across different markets, on ticket prices. We consider the hypothesis that there is a complementarity between the anti-competitive role of multimarket contact and the role of having common vertical relationships, henceforth, common-subcontracting. To this end, we construct a measure of common-subcontracting, and find that, on average, common-subcontracting, when acting as a “multiplier” of multi-market contact, is associated with a 6.17% increase in prices. This positive effect on prices is in addition to the 2.66% increase in prices that is due only to the standard multimarket contact among major airlines.

Published

Global Concavity and Optimization in a Class of Dynamic Discrete Choice Models

with Yiding Feng (Northwestern University) and Denis Nekipelov (UVA), Proceedings of the 37th International Conference on Machine Learning, Online, PMLR 119, 2020

Discrete choice models with unobserved heterogeneity are commonly used Econometric models for dynamic Economic behavior which have been adopted in practice to predict behavior of individuals and firms from schooling and job choices to strategic decisions in market competition. These models feature optimizing agents who choose among a finite set of options in a sequence of periods and receive choice-specific payoffs that depend on both variables that are observed by the agent and recorded in the data and variables that are only observed by the agent but not recorded in the data. Existing work in Econometrics assumes that optimizing agents are fully rational and requires finding a functional fixed point to find the optimal policy. We show that in an important class of discrete choice models the value function is globally concave in the policy. That means that simple algorithms that do not require fixed point computation, such as the policy gradient algorithm, globally converge to the optimal policy. This finding can both be used to relax behavioral assumption regarding the optimizing agents and to facilitate Econometric analysis of dynamic behavior. In particular, we demonstrate significant computational advantages in using a simple implementation policy gradient algorithm over existing “nested fixed point” algorithms used in Econometrics.

Work in Progress

Competition, Wages and the Emergence of Computer Science Degree Programs in the US

with Emily Cook (Tulane University) and Devaki Ghose (World Bank)

This paper investigates the determinants of universities’ decisions to introduce computer science programs, and how these decisions affect equilibrium wages in computer science industries. Computer science (CS) departments first emerged in the 1960s and were rapidly introduced at U.S. universities in the subsequent decades. As of the Spring of 2017, 60% of U.S. public and private non-profit four-year universities offered an undergraduate CS major, and nearly 63,000 students graduated with a CS major from these institutions. The introduction of these programs—and the number of graduates they produced—was shaped by government subsidies, competition between institutions, and industry and faculty wages. The paper investigates to what extent these factors affected the supply of CS programs and graduates, and in turn, the equilibrium industry and academic wages for computer scientists. We model universities’ decisions about whether to offer a CS program and the size of their program. In the model, these decisions are based on equilibrium wages for faculty, which drive costs, and industry wages for CS graduates, which drive demand. We use historic data on CS program adoption at US universities supplemented by unique panel data on faculty compensation to estimate the model. We apply the estimated model to study the effects of counterfactual subsidies, competition, and industry demand on CS programs, graduation and wages over time.

Other Work in Progress

The Expected Herfindahl-Hirschman Index: A Concentration Measure for Innovation Pipelines, with Gaurab Aryal (UVA), Federico Ciliberto (UVA) and Margaret Kyle (MINES ParisTech)

Inferring the Values of Drugs from Stock Reactions to Changes in the R&D Status, with Gaurab Aryal (UVA), Federico Ciliberto (UVA), and Leland Farmer (UVA)