Essays in empirical industrial organization

The field of empirical industrial organization uses data to analyze the structure of industries in the economy by measuring the parameters that drive the behaviors of firms and consumers in these industries. Part of the literature focuses on markets in which firms interact in an imperfectly competitive setting. Research in this field heavily relies on models with a game-theoretic foundation. Many market structure models endogenize the number of firms entering a market. Not only industries but also other types of organizations operate through interactions among their members and can be analyzed with analogous game-theoretic models. The underlying hypothesis is that agents making a certain decision receive a non-negative payoff, conditional on the expectations or actions of other (potential and actual) agents acting in the same environment. These considerations have been crucial in shaping the first two self-contained chapters of this dissertation. In the first chapter, I study the workers' decisions of joining teams within an important scientific experiment. In the second chapter, joint with Laura Grigolon, we provide empirical evidence of the link between common ownership and firms' decisions of entering markets in the Ontarian cancer drug industry. The fact that decision-makers operating in a strategic environment have in expectation non-negative payoffs is parallel to revealed preference arguments at the basis of discrete choice models of consumer behavior. As in the market entry literature, consumers' choices are interpreted as revealing something about an underlying latent utility. By observing how consumers’ decisions change, as their choice sets and market conditions change, one can gain insight into the underlying determinants of consumers' preferences. In the third chapter, joint with Liana Jacobi and Michelle Sovinsky, we analyze the potential complementaritarities in consumption of the so-called sin goods (marijuana, alcohol and tobacco) taking into account persistence in behavior. For the development of this dissertation, I rely on rigorous descriptive analyses and the development and estimation of structural models. With these approaches, it is possible to give informed assessments to policy-makers and in the case of structural models to quantify the impact of feasible policy changes. In Chapter 1, I present an empirical structural model that quantifies the main drivers of endogenous team formation and team performance when the allocation of individuals to teams is decentralized. Many companies currently adopt decentralized approaches to production. These arrangements are widespread in scientific institutions, as fellow researchers typically collaborate on a voluntary basis. The emergence of such arrangements poses several challenges to an economist. First, it is important to understand which elements drive the decision to join projects. Second, it becomes critical to develop tools to correctly measure the performance of teams when the decision to participate in projects is endogenous. These steps are fundamental to assess if decentralization is desirable to obtain successful outcomes with a larger probability. To address these challenges, I use unique data from Virgo, an international collaborative experiment in science. Researchers involved in Virgo choose which project(s) to work on. For the analysis, I use the information on projects' characteristics, outcomes, and participants. I develop and estimate an entry game with incomplete information where heterogeneous agents decide simultaneously whether to join a project. The payoff of joining depends on exogenous project characteristics, including a measure of ex-ante quality, and the expectation on the actions of potential project-mates. Strategic complementarities and substitutabilities can arise in this setting as workers might find beneficial or detrimental to work with others. I measure project outcome in terms of probability of project completion. I find that the pool of expected project-mates drives the decision to join a project while project quality is less important. The larger the pool, the lower the probability of joining a project, as a consequence of the congestion effect due to increasing coordination and communication costs. Heterogeneity in researchers' characteristics explains the selection into projects. I show that controlling for both projects’ ex-ante quality and endogenous project participation matters for obtaining unbiased estimates of teams’ performance. Finally, I consider a counterfactual centralized mechanism in which strategic interactions have no value. I find that this alternative allocation leads to excessive project participation and decreases the probability of project completion. Hence, adopting a decentralized mechanism of project allocation within a firm can be more efficient because workers internalize the costs and benefits of working with each other. In Chapter 2, joint with Laura Grigolon, we document the features of a highly innovative industry characterized by a concentrated ownership structure, the Ontarian cancer drug industry. The analysis has the objective of studying the effect of common ownership on the decision of generic producers to enter the market. Common ownership, namely the practice for large institutional investors of owning stakes in competing firms, has raised the attention of antitrust scholars because the degree of common ownership grew in recent years. Some empirical studies show that it has a large effect on the strategic behavior of companies held by institutional shareholders. Common ownership linkages are a well-established feature of many industries, including the cancer drug industry, for which hospital and public drug program spending in Ontario is dramatically increasing over time. These factors make it an appealing setting to understand the consequences of the common ownership phenomenon. We use unique data on the timing of cancer drug entry in the market (branded and generics) and collect information on patents, drug approvals, and drug indications. We complement our dataset by gathering ownership data mainly from 13F filings. With these data, we empirically assess the presence of common ownership and quantify which components mainly drive the link between common ownership and market entry. In particular, we show that investors' concentration plays an important role in defining common ownership in the years before the entry of a generic in the market. Common ownership may have anticompetitive effects and be harmful to welfare. With the results of this paper, we make the first important step in identifying the target of eventual policy interventions to reduce this practice, for this industry as well as for other innovative industries characterized by a high level of concentration. In Chapter 3, joint with Liana Jacobi and Michelle Sovinsky, we analyze the potential complementarities in use when individuals choose to consume bundles containing marijuana, alcohol, or cigarettes (sin goods), taking into account persistence in consumption for these substances. Two-thirds of Americans are in favor of marijuana legalization. This substance, however, might be consumed in combination with other substances, such as alcohol and tobacco. Moreover, past use of one of the substances might have consequences for the consumption of that substance and other sin goods, especially if one considers complementarity in consumption. Therefore, it is important to understand whether consuming marijuana affects the consumption of other substances and what changes when one considers the potentially addictive nature of these products. We develop and estimate a dynamic model of multi-substance use allowing for persistence in behavior. For the empirical analysis, we uniquely combine data from two primary sources. The first are individual-level panel data from the Panel Study of Income Dynamics (PSID) survey, which contains information on demographics and consumption behaviors of young adolescents in the US. The second source are pricing data for marijuana, alcohol and cigarettes collected from administrative tax data and transaction data. Our parameter estimates show that it is important to account for correlation across unobservables and persistence in behavior when analyzing the decision of using the sin goods in combination. Moreover, we find that the past use of a substance influences not only its current use but also the decision of using the substance together with other substances. Our results provide insightful information on the long-run effect of marijuana legalization for the concurrent and future consumption of potentially substitutable products.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.