Michael D. König

Abstract: We propose a general framework for studying how collaboration affects team production. Collaboration between agents is represented by a bipartite network, and equilibrium captures both complementarities among collaborators and substitutability across concurrent projects. We develop a Bayesian estimation procedure and illustrate the model using inventor collaboration networks in the semiconductor and pharmaceutical industries.

Abstract: We introduce an R&D network formation model in which firms choose both R&D effort and collaboration partners. Firms benefit from technology spillovers from collaborators while competition creates strategic substitutability in R&D. We characterize equilibrium, develop an estimation method that remains computationally feasible for large networks, and show that targeted subsidies to specific R&D collaborations can generate substantially larger welfare gains than uniform subsidies.

Abstract: We study the coevolution of knowledge creation, diffusion, and R&D collaboration networks. Firms collaborate with partners holding complementary technology portfolios, while innovation and spillovers change those portfolios over time. The resulting feedback can generate cyclical collaboration patterns observed in the data and yields policy implications for collaboration subsidies and competition.

Abstract: We develop and quantify an adaptive production-network model to study supply-chain resilience. Firms may exit after shocks, while surviving firms can replace lost suppliers subject to switching costs and search frictions. Using a large international firm-level network, we find that reshoring restrictions reduce output and increase volatility, and that network adaptation can amplify aggregate fluctuations.

Abstract: We build and estimate an endogenous growth model in which firms improve productivity through innovation or imitation and are subject to distortions. The model is disciplined by Chinese firm-level R&D and productivity data and compared with estimates for Taiwan. Counterfactuals show that misallocation of R&D resources has a large effect on aggregate productivity growth.

Abstract: We introduce a structural model for the joint evolution of network links and economic behavior. The underlying network game admits a tractable equilibrium representation, and a Bayesian Double Metropolis-Hastings procedure permits estimation with unobserved heterogeneity. Applied to R&D investment and collaboration in chemicals and pharmaceuticals, the model identifies positive knowledge spillovers and supports long-run key-player analysis.

Abstract: We analyze a model of R&D alliance networks in which collaborations reduce production costs while firms compete in product markets. We characterize equilibrium and the welfare-maximizing R&D subsidy program, then structurally estimate the model using R&D alliance data and company reports. The estimates are used to compare targeted with nondiscriminatory subsidies and to rank firms by their optimal subsidies.

Abstract: We study how networks of military alliances and enmities affect conflict intensity. The model combines network interactions with a contest framework and yields a closed-form Nash equilibrium. Using data from the Second Congo War, we estimate fighting externalities and evaluate interventions such as dismantling specific groups, weapon embargoes, and policies aimed at reducing hostility between groups.

Abstract: We develop a dynamic model in which firms improve productivity either through in-house R&D or by imitating more productive firms, subject to absorptive-capacity constraints. Firms near the technology frontier innovate more, while lagging firms rely more on imitation. The resulting balanced-growth equilibrium generates persistent productivity differences and a long-run productivity distribution with empirically realistic Pareto tails.

Abstract: We study network formation when agents value access to information held by others but observe only a limited neighborhood when choosing links. With little idiosyncratic noise, highly centralized networks emerge regardless of the observation radius; with more noise, local search can generate greater degree heterogeneity and higher aggregate payoffs. The model is estimated on co-inventor and scientific collaboration networks.

Abstract: We develop a dynamic network-formation model in which links depend on agents' centrality and have stochastic lifetimes. Using stochastic stability, we show that the long-run networks are nested split graphs and characterize their topology. The model matches key features of several real-world network datasets.

Abstract: We model the evolution of R&D collaboration networks when innovation results from recombining firms' knowledge. Firms form or sever partnerships according to network-dependent marginal benefits and costs, creating external effects on other firms in the same component. The decentralized process admits multiple equilibrium structures and can reproduce stylized features of observed R&D networks.

Abstract: We investigate efficiency and stability in R&D networks with network-dependent indirect spillovers. Efficient structures range from complete networks at low collaboration costs to asymmetric nested structures at higher costs. Stable and efficient networks need not coincide, and the divergence between stability and efficiency becomes especially pronounced in larger industries.

Abstract: We study network evolution when link creation and decay depend on node centrality. The dynamics generate nested network structures and can exhibit a discontinuous transition between hierarchical and homogeneous networks as link decay changes. The model also produces double power-law degree distributions with linked exponents.

Abstract: We consider a dynamic network-formation model in which agents form and sever links based on the centrality of potential partners. Capacity constraints on the number of links an agent can maintain generate a transition between dissortative and assortative network structures. This provides a simple mechanism for differences between technological and social networks.

Abstract: We analyze how the growth of firms' knowledge stocks through R&D collaboration depends on the algebraic properties of the collaboration network. Although efficient networks can range from complete graphs to quasi-stars, decentralized link formation typically produces sparse, clustered, heterogeneous networks. Longer evaluation horizons can move decentralized outcomes closer to efficient network structures.

Abstract: We study disease spread on a social network where individuals adjust contacts to avoid infection. Susceptible individuals rewire links away from infectious individuals, which reduces infections and raises the infection rate required for endemic disease. We formulate targeted lockdown policy as a computationally tractable semidefinite program and apply the model to the Netherlands using a population-level contact network. Accounting for rewiring allows substantially more intergroup contact under the optimal targeted policy.

Abstract: We introduce a Web-Based Affectedness Indicator (WAI) for real-time monitoring of economic disruptions. Using LLM-assisted classification and information extraction from more than five million company websites, WAI measures firms' responses to external shocks. In an application to COVID-19, the indicator closely tracks containment measures and predicts firm performance, providing timely firm-level information across industries and countries.

Abstract: We study a dynamic model of collective action in which agents interact and learn through a co-evolving social network. Comparing full information with local observation and DeGroot-style social learning, we show that social learning can generate a meaningful long-run probability of successful collective action. We estimate the model using large-scale Twitter data from the Egyptian Arab Spring and find an important role for network-based social learning.

Abstract: The share of R&D-active firms declined in Switzerland but increased in the Netherlands between 2000 and 2016. Using firm-level data and a structural growth model, we show that rising R&D costs help explain the Swiss decline, while innovation support sustains R&D activity in both countries. Counterfactuals indicate that improving innovation and imitation success can be more effective for productivity growth than reducing R&D costs.

Abstract: We analyze firms' choices between R&D collaboration and acquisition. Using text mining and transformer-based classification on a large corpus of news articles, we construct a novel dataset of R&D collaborations and estimate a structural model that jointly treats collaboration and M&A as endogenous. The results highlight substitution between R&D collaborations and acquisitions and its implications for technology spillovers and market competition.