Publications

How does collaboration with science change the direction of inventive activity? We study non-academic inventors' first co-invention with an academic partner using French patent data from 1976--2013 matched to scientific-author information. Measuring recombinant novelty as IPC4 combinations new to an inventor's prior technological history, we estimate dynamic effects with staggered difference-in-differences designs. A first science--industry collaboration generates a sharp but short-lived increase in exploratory recombination: inventors produce more new technological combinations in the collaboration year, while the share of novel combinations subsequently declines. We show that this pattern reflects a transition from domain opening to domain integration. The initial shock is driven by combinations involving new technological components, whereas later recombination increasingly combines newly accessed IPC classes with the inventor's prior technological portfolio. The domain opened by the first academic collaboration also continues to be reused in later patents, including outside the initial academic partnership and original co-invention team. These findings are consistent with science--industry co-invention operating as a learning-by-doing event that reshapes industrial inventors' search repertoires.

The Berry, Levinsohn, and Pakes (1995) model is a workhorse of empirical industrial organization, but GMM estimation is often bottlenecked by repeated share evaluation and fixed-point inversion on CPUs. We implement the demand-side BLP pipeline in JAX to exploit GPU execution with just-in-time compilation and automatic differentiation. On PyBLP-simulated benchmark economies, our implementation reproduces PyBLP estimates on identical markets and delivers large speedups in share computation that grow with the number of simulation draws, products, and markets.

This article examines the evolution of research topics published in top five economics journals using advanced textual analysis methods. We implemented Structural Topic Model (Roberts et al., 2014) on our database of articles from the top five journals, which we constructed using web scraping techniques from the RePEc platform. Our findings indicate that economics has evolved, with an increasing resurgence of empirical methods and a significant decline in theory, particularly in general equilibrium.

Professional esports teams increasingly rely on venture capital (VC) to fund their growth. However, the structural dynamics of this financing model and its financial viability remain underexplored. Anchored in signaling theory, this qualitative study based on seven interviews with esports team representatives and esports investors investigates the structuration of fundraising deals within professional esports teams, analyzing its stages and allocation of funds. It highlights that VC funding is often preceded by early-stage financing, such as business angels and Series A or B rounds, serving as legitimacy signals and used as a cyclical resource for specific projects like international expansion or player acquisitions. Interviews with team executives indicate that fundraising is critical for survival and project acceleration but does not lead to a viable income stream. Challenges include esports' fluctuating market appeal, limited investor understanding of the industry, and difficulties in monetization. However, emerging opportunities lie in leveraging fan communities, exploring cryptocurrency integration, and artificial intelligence applications. The findings deepen the understanding of VC structuration in esports, identifying adaptability and balancing current growth with financial viability as key success factors. Offering practical insights, this study underscores the need for innovative, viable business models to navigate the complexities of the esports ecosystem.

This exercise proposes a theoretical model inspired by and extending the recent approach of Bourreau et al. (2023), to analyze the effect of interoperability on innovation incentives in an asymmetric market of advertising-funded digital platforms (such as WhatsApp vs. Signal). Within a classic Hotelling framework with endogenous multi-homing, we explicitly incorporate strategic innovation choices in the presence of technological spillovers, allowing the dominant platform to copy the innovations developed by the minority platform. Our results show that the introduction or increase of interoperability reduces users' propensity to multi-home, thereby incentivizing the minority platform to intensify its investment in innovation to preserve its vertical differentiation. Conversely, the spillover discourages autonomous innovation by the leader. These theoretical results highlight the necessity of fine-tuned regulation of interoperability levels and the protection of innovations to effectively promote innovation dynamics.

This literature review aims to provide a synthesis of key themes in Industrial Organization, with a particular focus on markets characterized by network externalities. First, it explores the relationship between competition and innovation before examining the issue of compatibility and its implications, notably in the context of the Digital Markets Act.