Corporate research spending spreads unevenly across business partnerships
A statistical method that untangles why companies form alliances reveals that peer influence on research budgets varies sharply from firm to firm.

When a corporation expands its budget for research and development, neighboring firms and strategic partners take notice. In corporate innovation networks, an increase in spending by one company often ripples across its collaborators, prompting competitors and partners alike to adjust their own investments. Yet measuring whether these shifts happen because firms copy their peers, or simply because similar companies choose to partner with each other in the first place, has long challenged economic measurement.
Untangling those forces matters because governments spend billions of dollars subsidizing private technology development. If policy makers misjudge how corporate decisions spill across an industry, public incentives can end up subsidizing investments that would have occurred anyway or failing to trigger the broader industry push they were designed to create.
Why is peer influence in corporate networks hard to measure?
Shared traits often disguise themselves as social influence when business partners make decisions. In corporate partnerships, a clear sequence unfolds: first, corporate executives evaluate potential allies and choose partners that possess complementary skills, similar technologies, or aligned strategic goals. Next, those linked firms observe one another's actions, learning from joint projects or adjusting their strategies to match partner spending. Finally, a company decides how many resources to allocate to its own laboratory research and product design.
A serious statistical complication enters that chain because many of the traits that draw two companies together remain hidden to outside analysts. A firm with aggressive leadership, an adaptable research culture, or an unrecorded technological breakthrough is both more likely to seek out high-performing research partners and more likely to invest heavily in innovation independently. When economists observe two connected firms spending heavily on research, standard statistical tools struggle to determine whether one firm inspired the other, or whether both were simply drawn together by invisible traits they already shared.
How does the new statistical framework separate partnerships from peer effects?
Jointly estimating how links form and how outcomes spread allows researchers to strip away the distortions caused by selective partnerships. In a working paper posted on June 1, 2026, to the arXiv preprint server, Duong Trinh and Santiago Montoya-Bland'on introduced an econometric method designed to solve this problem, which they named the Selection corrected Heterogeneous Spatial Autoregressive model.12

Their approach models the formation of collaborative ties and the resulting corporate investment decisions at the same time.23 The technique uses a finite mixture structure, a statistical grouping mechanism that sorts entities into distinct classes without requiring researchers to define those categories in advance.21 This structure captures heterogeneity, the reality that different firms react to their peers with varying intensity. Simultaneously, the framework accounts for unobserved, firm-specific characteristics that influence both the decision to form an alliance and the eventual choice of research budget, directly addressing network endogeneity, the confounding bias that arises when network connections are formed by choice rather than allocated at random.21
To estimate these parameters, Duong Trinh and Santiago Montoya-Bland'on developed a Bayesian data augmentation procedure, a computational strategy that treats hidden network factors as missing data and estimates them iteratively.21 The authors noted that this Bayesian formulation overcomes severe computational obstacles that prevent standard likelihood-based methods from solving large, interconnected systems.23 After confirming the reliability of the method in numerical simulation tests, the researchers applied the model to an empirical innovation network composed of United States firms.21
What did the model reveal about corporate research investments?
Corporate research and development budgets in the United States demonstrate positive spillover effects once the self-selection of alliances is taken into account.21 When partner firms increase their innovation investments, connected companies tend to raise their own budgets in response. However, the magnitude of that peer reaction is far from uniform.
The empirical findings revealed that peer responses vary considerably across individual corporations, demonstrating significant heterogeneity in how businesses react to external changes.23 By separating initial link selection from subsequent peer influence, the framework enabled the researchers to measure both direct effects, which reflect a company's immediate reaction to a policy change, and indirect spillover effects, which ripple through the network as partners respond to each other.21 The authors reported that these distinct measurements highlight how differently firms behave when exposed to outside research incentives.21
What are the boundaries of these findings?
The conclusions drawn from this framework describe specific modeled relationships rather than an exhaustive portrait of corporate decision making. The research relies on an econometric model that assumes unobserved factors can be effectively grouped through a finite mixture structure and corrected through specific distributional assumptions. Furthermore, because the study was posted as a preprint on arXiv, the findings represent preliminary working research that has not completed formal academic peer review.13 The empirical results also reflect the specific corporate alliance environment of United States firms, meaning the exact distribution of peer effects could differ in countries with different patent laws, corporate governance structures, or public subsidy rules.
Despite those boundaries, the framework provides an empirical tool for evaluating industrial subsidies. By quantifying firm-level spillovers, policy makers can better design targeted research policies that account for how an incentive given to a single firm propagates across its entire web of industry partners.21
The record
13 sentences trace to 3 sources.
- 1 Heterogeneous Peer Effects with Endogenous Network Formation See the source
- 2 Heterogeneous Peer Effects with Endogenous Network Formation Not published online
- 3 Heterogeneous Peer Effects with Endogenous Network Formation | Santiago Montoya-Blandón See the source