By Authors: Mathijs De Vaan (University of California, Berkeley, Haas School of Business), Toby Stuart (University of California, Berkeley, Haas School of Business)
Interviewer: Domenika Uskova (Warwick Business School & Nova School of Business and Economics)
Article link: https://doi.org/10.1177/00018392241296140
Professor de Vaan, thank you for accepting the invitation!
Mathijs: Thank you for having me. I’m delighted to discuss our work on Referral Triads with the ASQ Blog readers!
First and foremost, how did the idea for studying referral triads emerge – was it theory-driven or data-driven?
Mathijs: For us, it was a mix. There is a large and very rich literature on social networks focused on the relationship between trust and social networks. It’s well-established that relationships, the longer they go on, usually lead to more trust being built. However, from our personal lives and from observations of how things work in organizations, we know a lot of instances in which there is significant trust, but you get the sense that this trust might lead to people having blind spots. They are not necessarily considering all the various options they have because they trust a person, and that person then becomes the default. My colleague and co-author, Toby Stuart, and I both had this feeling that there was something to this idea that trust essentially leads people to become defaults, and that there might be a downside to this. The way we started thinking about this paper was by combining this existing theory about the relationship between social networks and trust, and our intuition based on personal and organizational experiences. This brought the two ideas of trust and inefficiencies together.
Moving to the theoretical framing and conceptual contribution. You’ve described the referral triad as a distinctive social structure. What drew you to focus on the triadic level rather than the dyadic ties that most network research emphasizes?
Mathijs: I believe that many organizational and professional contexts are characterized by an intermediary that plays a critical role, perhaps more so than in our personal or social lives. There is almost no organization where intermediaries don’t play a very important role. We know this is true across the professions: we see it in healthcare, the legal profession, financial services, and consultancy. Essentially, any organization that relies on people working together and grows over a certain size depends on intermediaries. A lot of what happens internally in organizations relies on these people who know who to talk to to get a problem resolved, for example. These intermediaries play an incredibly important role; they fundamentally coordinate collaboration within and between organizations. This is an area that is ripe for more study because it leads to all these different collaboration and coordination problems, and I don’t think we know enough about these challenges.
“This is an area that is ripe for more study because it leads to all these different collaboration and coordination problems, and I don’t think we know enough about these challenges”
Your work utilizes two classic ideas – that strong ties create trust but can limit efficiency. How did you come to see this specific tension, between trust-building and optimal matching, as the defining mechanism in referral triads?
Mathijs: What’s interesting here is that you have to start thinking about the incentives of the people involved in this referral or intermediary challenge at both the individual and the organizational level. At the individual level, the incentives are almost always aligned with a trust-based mechanism. If we build trust with someone, we are going to continue to make referrals to them. The reason is that the search is really expensive. It requires a lot of time, and it’s hard to source the information needed to make referrals that are efficient but not based on trust. It’s a really thorny, challenging problem. Therefore, at the individual level, the incentives lead us to just go back to our defaults (our trusted ties). At the societal level, or even the organizational level, the incentives should be set differently; they should balance this trust-based mechanism with the need for efficiency. However, I don’t think organizations necessarily see this as a problem because they don’t really know the counterfactual – what happens when they don’t rely on the trust-based mechanism. Many organizations are not set up to address this challenge, and it’s not happening nearly as much as I think it should.
While individual professionals might be intuitively aware that default, trusted referrals aren’t always optimal, you suggest organizations lack the mechanism to address this. Is it fair to say that organizations remain unaware of the true cost of this inefficiency until they face an exogenous shock or internal crisis?
Mathijs: That could be the case, but I think the awareness is a bit nuanced. The real problem is that it’s really hard to think about how to get out of this cycle of default-based referrals. Therefore, most organizations don’t have a clear solution for this problem. They don’t have that counterfactual that would allow them to see what the alternative could be. Specifically, a non-trust-based referral mechanism that is efficient and works for all the individuals involved in the process.
What was the biggest challenge in establishing causal inference through the Instrumental Variable approach, while handling 890,121 records of sensitive data?
Mathijs: The challenges in establishing causal inference were twofold, starting with data access and its institutional complexity. The first hurdle was securing access to the 890,121 sensitive medical records, which are governed by numerous rules and restrictions. Beyond access, filtering this massive volume, which was only a small fraction of the total dataset, to create a useful analytic sample required profound institutional knowledge. It took a couple of years to fully master how the healthcare profession is organized and how its data are structured, illustrating the steep learning curve that acts as a significant barrier to entry for many organizational theorists. The second major challenge was achieving methodological rigor. Establishing a strong causal claim necessitated an IV strategy, requiring the instrument to satisfy several stringent conditions, like the exclusion restriction. This led to a highly iterative process: we tried and discarded multiple other instruments that showed flaws during workshopping before settling on the final, validated instrument, a necessary but time-consuming process for ensuring the highest quality in our causal claim.
The results suggest strong ties promote trust but reduce match quality – how do you think organizations can mitigate this trade-off? Which side of this trade-off do you think referrers should optimize for in practice, or is there a way to foster trust without sacrificing optimal matches?
Mathijs: That’s a great question. At a very high level, I think there are two main things that organizations can implement to mitigate this trade-off: incentives and information access. As noted earlier, individual incentives generally favor defaults because search costs are high. Organizations need to adjust these incentives to encourage intermediaries to consider a much broader sample of potential referral targets. One key finding in the paper is that being busy, or not having enough time, makes it more likely that intermediaries will resort to their strong ties. Therefore, an immediate action organization can take is to give intermediaries more time to make their decisions. If they are constantly under pressure to decide quickly, defaults are the more likely outcome. The second crucial component for mitigating the trade-off between trust and efficiency is reducing search costs by improving information access for intermediaries. Search costs are currently high because obtaining valuable data that accurately signals expertise is difficult. Organizations should proactively address this by providing structured data that gives referrers a clear sense of where the best expertise resides for specific problems, helping them identify effective partners both inside and outside their immediate network. Furthermore, AI presents a powerful solution: an AI system trained on vast organizational or medical data can assist intermediaries in searching for a massive space of potential partners. The AI should function as a supportive tool in partnership with the referrer, but it is vital that the model avoids simply reinforcing existing defaults; ideally, it should introduce randomness or non-obvious paths to actively encourage deviation from strong-tie referrals. Ultimately, the goal is to leverage these organizational systems to significantly reduce the friction involved in finding the best match, thereby allowing referrers to foster necessary trust without sacrificing optimal match quality.
That is intriguing! Regarding sacrificing optimal match quality, you found that lower match quality correlates with worse health outcomes – what might this mean for healthcare policy or referral systems?
Mathijs: I believe organizations should view the finding that lower match quality correlates with worse health outcomes with utmost seriousness, as their primary mission is improving patient well-being, making this a highly salient organizational outcome. The necessary policy and system changes are organizational, centered on the two key areas of incentives and information access. Specifically, when the stakes are high for patients, organizations must adjust incentives by ensuring intermediaries are given sufficient time to make decisions, and they must enhance information access to reduce the reliance on quick, default-based referrals. Ultimately, organizations must commit to implementing solutions that address these systemic issues, ensuring better match quality and measurably better patient outcomes.
That’s a real call to action regarding organizational policy. When it comes to the real-world pressures physicians face, we see the following finding in your paper: the post-hoc analysis shows that Primary Care Physicians rely on strong ties on busy days – what does this tell us about decision-making under time pressure?
Mathijs: That is a great question. The finding that people rely on defaults when they’re busy is not new; there have been several papers suggesting that people under time pressure rely more on defaults than they otherwise would. The reason we included this specific post-hoc analysis was to see if our theoretical argument, that people rely on defaults and should rely on them more when under time pressure, was consistent with our data. In terms of implications, it’s fairly straightforward: When decisions are important and high stakes, you must give people enough time. If you don’t, they will resort to defaults, which may not achieve the best possible outcome.
What was the most difficult or rewarding part of bringing this paper to publication?
Mathijs: One of the things I find really important when writing and reading papers is ensuring a strong connection between the theoretical framing and the empirics.
“One of the things I find really important when writing and reading papers is ensuring a strong connection between the theoretical framing and the empirics”
This was the part of the paper where my co-author, Toby Stuart, and I spent the most time, going back and forth on how to ensure these two things perfectly align. You start with your theoretical arguments, and then you have a complex dataset, ours had lots of different variables and fields that weren’t always clear to interpret. The challenge was making sure that everything we theorized about was measured in the data in the most appropriate way. There will always be some disconnect because that is the nature of working with observational data; it rarely aligns perfectly with what you have in mind. However, bringing the data as close as possible to the theoretical arguments was the most complex and time-consuming part of this paper and ultimately the most critical aspect of getting it published.
Finally, shifting to future research. Most network theories treat tie strength as static. Given the complexity of referral triads, what would a more dynamic theory of tie strength and trust look like over time as referrers and experts co-evolve?
Mathijs: One of the things we noticed is that the relationship between die strength and the outcomes is not linear, which suggests there are interesting dynamic patterns worth investigating. If you think about the relationship between tie strength and outcomes in a longitudinal way, the big questions for future research revolve around inflection points: Where does trust rise the most, where does it plateau, and are there levels where it ceases to grow despite continued interaction? Variables to explore include individual and contextual differences, such as those related to professional background, tenure, firm, and industry. A particularly interesting dynamic involves in-group versus out-group trust: while in-group members often build trust faster, organizational structure sometimes necessitates interaction with out-group members. In such cases, the trust built with those from different backgrounds, though potentially slower to form, might ultimately be more durable or stable than trust among similar peers. So, there is a significant opportunity to develop a more dynamic theory on the formation of trust in professional ties, reflecting the complexity seen in many real-world organizational settings. Professor de Vaan, thank you for taking us behind the scenes of ‘Referral Triads’ and sharing such insightful details on the challenges of balancing trust and efficiency in professional referrals
Interviewer Bio:
Domenika Uskova is a co-tutelle (double) fourth-year Ph.D. student in Strategy at Warwick Business School (United Kingdom) and Nova School of Business and Economics (Portugal). She is a quantitative researcher working with longitudinal archival panel data. Domenika investigates how data breaches shape the innovation and economic outcomes of R&D-intensive firms, focusing on their strategies for strengthening competitive advantage in response to such incidents. Her studies primarily span the U.S. and U.K. markets.