The Strength of Varying Tie Strength

The Strength of Weak Ties" argument (Granovetter 1973) says that the most valuable information is best collected through bridging ties with other social circles than one's own, and that those ties tend to be weak. Aral and Van Alstyne (2011) added that to access complex information, actors need strong ties (high bandwidth") instead. These insights I integrate and generalize by pointing at actors’ benefits and costs. Weak ties are well-suited for relatively simple information at low costs, whereas for complex information, the best outcomes are expected for those actors who vary their bandwidths along with the value of information accessed. To support my claim I use all patents in the USA (two million) over the period 1975—1999.

💡 Research Summary

The paper revisits two seminal strands of social‑network theory—Granovetter’s “strength of weak ties” (1973) and Aral & Van Alstyne’s “high‑bandwidth” argument (2011)—and unifies them under a cost‑benefit framework that conditions tie strength on the complexity and value of the information being exchanged. Granovetter showed that weak, bridging ties are especially effective at transmitting novel, simple information across otherwise disconnected clusters. Aral and Van Alstyne, by contrast, argued that the acquisition of complex, tacit knowledge requires strong, high‑bandwidth relationships that provide trust, repeated interaction, and deep coordination. The author proposes that both claims are correct, but that the optimal network strategy depends on the nature of the information: simple, low‑value data are best accessed through many weak ties, whereas complex, high‑value data demand strong ties, and the most successful actors are those who can dynamically adjust their bandwidth to match the information’s value.

To test this hypothesis, the author exploits the entire United States patent record from 1975 to 1999 (approximately two million patents). Patents serve as a rich proxy for both information complexity (through technological classification, citation depth, and novelty measures) and economic value (via forward citations, licensing revenue, and market impact). The author constructs a co‑inventor network for each inventor, quantifies tie strength by the frequency and duration of joint patenting, and classifies patents into “simple” (incremental improvements, low citation counts) and “complex” (highly interdisciplinary, high citation counts, extensive prior art). Using regression models that control for inventor experience, firm size, and field‑specific trends, the study finds three robust patterns:

1. Complex, high‑value patents: Inventors who maintain strong, repeated collaborations (high bandwidth) with a small core set of partners experience significantly higher forward citation rates and greater commercial success than those who rely on many weak collaborators. The interaction term between tie strength and patent complexity is positive and highly significant, confirming that strong ties are a prerequisite for the diffusion of sophisticated knowledge.

2. Simple, low‑value patents: In contrast, for patents that are largely incremental, a larger number of weak ties correlates with higher citation counts. Weak ties provide diverse, low‑cost channels for spreading straightforward information, confirming Granovetter’s original claim in a modern, technology‑driven context.

3. Dynamic bandwidth adjustment: Inventors who vary their collaboration intensity across projects—using strong ties for complex inventions and weak ties for simple ones—outperform peers who adopt a uniform strategy. This “bandwidth‑matching” effect yields the highest overall citation impact and the greatest probability of licensing or acquisition events.

The cost analysis underscores that strong ties entail higher maintenance costs (time, coordination, trust building), but these costs are justified when the expected payoff from complex knowledge is large. Weak ties, being cheap to sustain, are optimal when the marginal benefit of additional information is low.

Policy implications follow. Firms should institutionalize a “bandwidth management” function within R&D, allowing project managers to allocate resources to either deep, high‑bandwidth collaborations or broad, low‑cost outreach depending on the project’s knowledge requirements. Moreover, the findings suggest that innovation ecosystems (e.g., science parks, industry consortia) should facilitate both types of connections: platforms for rapid, low‑cost networking alongside mechanisms that support long‑term, trust‑based partnerships.

In sum, the paper extends classic network theory by demonstrating that the value of a tie is not static but contingent on the informational context. By empirically validating the “varying tie strength” hypothesis with a massive patent dataset, it provides a nuanced roadmap for individuals, firms, and policymakers seeking to maximize knowledge acquisition and diffusion in an increasingly complex technological landscape.