Abstract: We develop an agent-based model to reproduce the size distribution of R&D alliances of firms. Agents are uniformly selected to initiate an alliance and to invite collaboration partners. These decide about acceptance based on an individual threshold that is compared with the utility expected from joining the current alliance. The benefit of alliances results from the fitness of the agents involved. Fitness is obtained from an empirical distribution of agent's activities. The cost of an alliance reflects its coordination effort. Two free parameters $a_{c}$ and $a_{l}$ scale the costs and the individual threshold. If initiators receive $R$ rejections of invitations, the alliance formation stops and another initiator is selected. The three free parameters $(a_{c},a_{l},R)$ are calibrated against a large scale data set of about 15,000 firms engaging in about 15,000 R&D alliances over 26 years. For the validation of the model we compare the empirical size distribution with the theoretical one, using confidence bands, to find a very good agreement. As an asset of our agent-based model, we provide an analytical solution that allows to reduce the simulation effort considerably. The analytical solution applies to general forms of the utility of alliances. Hence, the model can be extended to other cases of alliance formation. While no information about the initiators of an alliance is available, our results indicate that mostly firms with high fitness are able to attract newcomers and to establish larger alliances.
Introduction
Innovation in the modern corporate landscape rarely occurs in isolation. Firms increasingly form R&D alliances to pool resources, share knowledge, and mitigate risks. While the topological structure of these collaboration networks has been extensively studied, the micro-dynamics governing the size of individual alliances remain poorly understood. Empirically, the vast majority of alliances (approximately 93%) consist of only two partners, yet a small number of "mega-alliances" involve dozens of firms. This paper presents an agent-based model (ABM) that treats alliance size as an emergent feature of strategic decision-making rather than an exogenous input.
The Agent-Based Model
The core of the model is the concept of "fitness," which serves as a proxy for a firm's attractiveness and activity level. In this framework, an agent's fitness is derived from its historical participation in alliances.
The alliance formation process follows a specific sequence:
- Initiation: A randomly selected initiator begins inviting partners.
- Invitation: Potential partners are invited with a probability proportional to their fitness.
- Decision Logic: An invited firm accepts if the expected Utility > Threshold. Utility is defined as the sum of the fitness of all current members minus a linear coordination cost (scaled by parameter $a_c$). The acceptance threshold is proportional to the invitee’s own fitness (scaled by parameter $a_l$), reflecting the higher opportunity costs for "elite" firms.
- Termination: The process stops once the initiator receives a predefined number of rejections ($R$).
A significant technical contribution of this work is the derivation of an analytical solution—a probabilistic recurrence relation—that allows for the mathematical calculation of the expected size distribution, bypassing the need for computationally intensive simulations.
Empirical Data and Calibration
The model was calibrated using the SDC Platinum database, covering 14,829 R&D alliances and 14,561 unique firms from 1984 to 2009. Alliances are represented as cliques in the network. The three free parameters $(a_c, a_l, R)$ were tuned to match the empirical data.
Results and Validation
The model successfully reproduces the heavy-tailed distribution of alliance sizes observed in the real world. The predicted size distribution falls almost entirely within the 95% confidence interval of the empirical data.
Key findings include:
- Parameter Stability: The cost-scaling ($a_c \approx 0.04$) and threshold-scaling ($a_l \approx 2$) parameters remained remarkably stable across different calibration runs.
- The Role of Incumbents: Large alliances are primarily anchored by high-fitness "incumbent" firms. Only these high-status agents possess the "gravitas" to attract enough partners to offset the linear coordination costs associated with larger groups.
- Robustness: The model is relatively insensitive to the exact value of the rejection limit ($R$), suggesting that the structural constraints of utility and threshold are the primary drivers of alliance size.
Discussion and Limitations
While the model provides a robust explanation for alliance size distributions, it has limitations. It currently assumes static fitness, whereas in reality, successful collaborations likely increase a firm's future fitness. Furthermore, the model does not explicitly account for "assortativity"—the tendency for highly connected firms to partner with each other—which may affect the fine-grained topology of the network core. Future work could extend this framework to include dynamic fitness and more complex initiation probabilities.
Summary updated