Article link: https://doi.org/10.1177/0001839219834011
1. [Positioning in broad research interests] We’d like to just start with a big picture first. Can you tell us about how this paper fits with your broader research interest?
Isin: My broad research interests are in the area of innovation strategy. I like to think about how organizational dynamics influence innovation strategies and performance of firms. By looking at how group dynamics influence syndicate evolution in venture capital firms, this paper brings us closer to understanding how innovation strategies are influenced by what happens inside organizations. I have another ASQ paper on venture capital resource allocation decisions published in 2007, which had the same flavor. It was also looking at political dynamics that happened within syndicates. That’s how I think it fits with my overall agenda.
Lei: I study social networks and firm collaboration in general. I’m especially interested in VC syndication. I have studied VC syndicate formation, its performance implication to start-up companies in terms of exit, and the evolution until that stage. This is one of the areas that I have been studying since my dissertation. This research paper is right in my domain of interest.
Could you tell us more about how you focused in on this particular research question, zooming in from your broader research interests?
Isin: It was mostly based on Lei’s own dissertation, where she started to think about group dynamics in the context of network tie formation, especially in multi-party partnerships. She partitioned her idea into several components: One was about the initial formation of multi-party syndicates. This project was about how they grow over time, and how members get added or deleted over time. Basically, she approached that larger problem, and then broke it up into smaller research questions. That’s how she was building that research agenda and I’m happy to be a part of it.
Lei: When I started this, I wanted to study the whole journey of VC syndication, from the very beginning – the initial formation – to the involvement of newcomers, which is the evolution of syndicates, and then to its impact on start-up companies’ performance end to end. This was where I had a bigger picture, but when I started looking at this research context, I realized that this was a well-studied dataset; this was a well-studied context. The key was to identify a unique perspective, and how I could find an interesting question that hadn’t been answered. Then I realized that there were very few papers in the network literature that talked about the group-level partnership formation, most are on dyad level. My AMJ paper focused on the initial formation of syndication from the group perspective. This paper was from a group perspective as well; so my angle was on how the group perspective can help us understand the formation and the evolution of VC syndicates. This had been overlooked for a long time, and I saw an opportunity to start from there.
Social network theories are one of the most well-established in our field and your work shows how to extend them, for example, by specifying two types of embeddedness, matched with the proper empirical setting. Could you tell us about how you came up with these two concepts of ‘depth and breadth’ of embeddedness? What was your motivation for approaching these two concepts?
Lei: When this project was started, actually we didn’t frame it using depth and breadth of embeddedness. Apparently, this is the result of evolution of our paper. When the paper was started, we focused on the overall familiarity of newcomers with incumbent members as well as the relational heterogeneity between the newcomer and the incumbent members in terms of tie strength. But during the review process, we were intrigued by the reviewers’ comments, and we figured that we needed to define the two kinds of concepts more clearly. We were stuck there for a while, but we tried to figure out how we can better demonstrate our idea. We realized that those two concepts that we initially used, the familiarity and the relational heterogeneity between newcomer and incumbent members, are more closely related than we thought and it was hard to tease them apart. When we reviewed the network literature more deeply, we came to the conclusion that there are two different ways of measuring embeddedness. Usually, when measuring embeddedness, researchers have used the average score of tie strength of dyads and the proportion of dyads that have prior ties. We realized that studying the evolution of syndicates provides a unique angle that may help us tease these two types of embeddedness apart, especially when we look at newcomer’s embeddedness with a group of incumbent members. It may seem the idea came in a moment, but it’s actually the accumulation of a long time of studying of this topic.
Isin: When we study networks, we look at a structure, and we envision the nodes in the structure as if they’re monolithic entities. However, most of the time they are not. When we are looking at alliances, for instance, we are really looking at firms, and there are units within firms, and then individuals within those units, and they have their relationships, preferences, and so on. Similarly, in this case, when we look at syndicates, it’s possible to think about them as nodes. A syndicate could be a single node, and then you can add another node. However, it’s important to realize that a syndicate is not a monolith, but it is composed of different firms with their own agendas. If it’s not the monolithic entity, then in which ways could an individual newcomer attach to that entity, the different parts of that entity? I think this is the key insight of this paper, and that’s how the idea of looking at the depth and breadth separately emerged. As Lei said, it was a journey. It didn’t just come. It took a while for us to refine that insight.
2. [Process of developing key arguments] You suggested that the new members’ experiences and the incumbent members’ status are at play. The complex dynamics of power among individuals that you suggested in your theory and the macro-level data that you used are fascinating. However, as those variables are well-known and thereby used quite frequently in many other papers, I think you must have taken a risk that might not add many new stories. Could you debrief some of the behind stories with these variables?
Isin: There was back and forth. As you have also identified very well, the challenge with this paper is that it talks about micro mechanisms, but uses macro level data, so it’s very hard to pin down those micro mechanisms that you can’t actually observe. That is where the reviewers helped us a lot by asking … about the mechanisms. We needed to bring in moderators that would help us distinguish the two mechanisms in the paper. One was about power, on how the power dynamic plays out in the case of conflict. The other was about mediation, on how mediation plays out in conflict situations. We went back to the literatures that I became very familiar with; micro-organization literatures on power dynamics, group dynamics, and exchange theory. We looked for moderators that would allow us to differentiate between those two stories. You are right that we were working within the constraints of a dataset that a lot of people may be familiar with. We had data on connections, but not the specifics of the content of those connections unfortunately. So, status and experience are the two moderators that helped us pin down those mechanisms as closely as possible. The other way we found was to do the qualitative study. I think you want to talk about that story separately. But Lei, is that fair?
Lei: Absolutely. When we started this paper, the moderators helped us pin down the mechanisms and strengthen our argument from hypothesis one to three. The mechanisms we laid out were power imbalance and mediation. We wanted to show in two ways that those are the two mechanisms that indeed play a role. First, we wanted to find moderators that make sense in the context, and second, we tried to identify them in the interviews. Interviews have been really helpful during the entire processes. They helped us to identify these two concepts of status and experience, because multiple interviewees mentioned the importance of them. At the same time, the interviews helped us illustrate those two mechanisms, and provide further evidence that those are really the mechanisms underneath our argument.
3. [Result interpretation] Sometimes researchers do not report insignificant results, and we rarely see insignificant results reported in papers as well. I was wondering what your reactions were when you saw the null effect over one of your hypotheses. Do you have particular stories that you recollect regarding the insignificant result? What would be your advice on such matter?
Isin: In general, my understanding and my philosophy is that it’s okay. It’s fine to have null results sometimes. You don’t always want to look for confirming evidence. As scientists, our job is to accumulate, collect, and report evidence that doesn’t necessarily show us significant results only. We are not searching for significance, but we are searching for truth. So, we learn from nonsignificant results as well. I do understand the nervousness around not finding significant results, especially when you’re working on a doctoral thesis. But as long as you are confident that there is truth in your story and that you haven’t made a fundamental mistake in design, then I think having a nonsignificant result is not the end of the world. It is a step in our learning process. It’s important to be transparent and truthful in reporting.
Lei: I agree. I also want to add that apparently you don’t want to have all insignificant results but rather (having) some insignificant results in your paper is not the end of the world. The key is to understand why that is the case and why you find those insignificant results. Can you explain it with an alternative explanation? When you have insignificant results, you definitely have to offer alternative explanations, helping audiences to understand why that happens. Also, you have to make sure that how this alternative explanation can fit in with your story. Why this alternative explanation is meaningful, and also won’t just throw out your entire framework or story out of the window… Ultimately, you need to have a good understanding of what is happening in reality to make sure that insignificance won’t be entirely a bad thing for your paper. For our case, we expected hypothesis three would suggest a stronger and significant effect of breadth embeddedness, but we also had to consider the motivation of newcomers when they know there are fractures or negative dynamics in the group, which can eventually make them not to prefer to join the group.
4. [Multi-method research] This is a paper which combines field evidence with quantitative analysis. Could you tell us about how you conducted the research? Did you start from qualitative observations or used them to confirm the validity of your constructs?
Lei: When I started this project, I initially did some interviews to understand the phenomena. For my entire dissertation, I interviewed quite a few venture capitalists to understand what the rationale is behind syndication behaviors from the beginning to the growth of the syndication. When I started doing analysis of the paper, however, I did not think qualitative interviews would be a part of the paper. Nevertheless, interviews had been very helpful for me to understand the phenomena from the start, to identify the holes in the literature, and to confirm what the focus should be in this paper. In the first submission, we did not include any qualitative data, maybe a couple of citations. We didn’t use qualitative evidence as heavily as we did in this current published version. But during the review process, we were asked by reviewers to clarify the mechanisms behind the argument. So, we went back and did more interviews, with greater depth. We realized that the interviews could be very helpful to strengthen our argument, if we include the direct evidence, citations, and remarks into our paper. So, thanks to the review process, we formalized the use of qualitative evidence. We need to give credit to the reviewers who helped us to include more evidence from our interviews.
Isin: I also remember conducting pilot interviews before we even submitted our paper to ASQ. We talked to several venture capitalists to see if the idea had legs. We wanted to have a reality check. If you are doing large sample work, you need to have that kind of reality check. I also remember reading a lot of blogs and practitioner books. I’m an avid consumer of that kind of material, from which you learn a lot about the process. Venture capital is a good context for that, because there’s a lot of written material that venture capitalists themselves have made available about their processes. But the formalization, or the increase in the number of interviews, came later. As I said, the big challenge with this paper was to establish the mechanisms. When reviewers asked us to do more, that’s when we said, “Okay, maybe we need more qualitative evidence.” That’s when it became more formal. It helped us break down the mechanisms for ourselves too, as the stories of the paper are fairly complex. It really helped us break down the mechanisms into their components, and what we thought was going on.
5. [Choice of dataset] Many scholars have used this VC setting for many research, and it is very interesting to see how a fresh perspective can make more contribution to our field even with well-known data. Can you describe how you came to work with it? How did you choose the context of research, namely the VC setting?
Lei: I totally understand the dilemma. One option is having your unique dataset, but it takes effort and resources to come up with a unique dataset. Another one is using an existing dataset. I ended up with the second option, because honestly, I did not have another option at that time. This was the dataset available to me at that time, and I did not have the resource to start something new. Even at that time, this dataset had been well-studied. You could find a lot of papers using it, and we all know the variables that you can come up with in this data have been well-explored. So my biggest challenge was to come up with something new, something that nobody had studied before. It was not in the very beginning that I came up with the group perspective. I had the dataset and I did interviews. I tried to come up with something interesting, fresh, and hadn’t been done by other researchers.
It took me a while to realize the gap in the literature. When I found that most of the studies were on the dyadic level, I asked “Why? What is the reason behind it? Is that entirely not doable (to use a group level)?”, because I was also new to the network literature at that time. I figured either few researchers had recognized it or something was not doable. It took me a while to realize an opportunity there; so I framed my entire dissertation on a group perspective. Instead of studying dyad formation, I studied group formation. Also, when you focus on dyadic tie formation, you cannot study the evolution of syndicates. But we all know that, in VC syndication context, venture capital investment almost always goes through multiple rounds until the start-up companies exit, which made it interesting since few had studied the evolution of collaboration before. And then I realized the limitation was on dyadic perspective and that this could become a fresh perspective I had been looking for. As I mentioned, the biggest challenge was to identify a unique and fresh perspective that nobody had used before. Ultimately, a dataset is not the most important; but it is important that you have a good and deep understanding about what’s going on in the phenomena, and what the mechanisms are in that phenomena. This may help you develop fresh perspective using a well-studied dataset.
Isin: I also used venture capital for my dissertation, except when I did, it was unique. There was literally nobody else who had the data. I was one of the first to get access, at least out of finance, and I was in fact a little scared that I would be framed as a finance scholar. Over time, after I wrote the dissertation and around the time when I published it at ASQ, the dataset became widely available. Even if you think you have a unique dataset, it may eventually not be so. I do agree that there is some value in thinking about fresh context, especially in your dissertation work. Because writing a dissertation is the one time when you have dedicated time, you can spend thinking about an ideal problem that you would like to solve, an ideal design to approach that problem. Over time, of course, path dependencies are set in place. You may get used to certain methods, certain datasets, certain approaches. But when you have a fresh and clean slate, it is good to be a little idealistic, and to try to bring a fresh perspective as much as you can.
However, I also see a lot of great papers written with datasets that have been around, so it is not just in the dataset. It is not sufficient to have a unique dataset, and sometimes it is not even necessary. I hope this paper can bring a fresh perspective to a well-studied area. I would agree with Lei’s assessment that, no matter how well-known or unique your dataset is, the question I would ask is what we can learn from it that we didn’t know before? What is an unexplored characteristic of this context that can help us learn more about what we are interested in? In our case, it’s turned out to be the temporal aspect of networks. The fact that partnerships grow over time was something that other studies had not picked up on in the network literature and we could observe. It is similar to my early work in the sense that, I looked at how investments evolve over time, as opposed to how syndicates evolve over time. This is clearly a unique aspect of this dataset that you cannot find very often in other datasets. That was the way I would approach it. It is about how well the data fits with the question that you want to answer, and about what unique opportunities it presents to answer the questions that you’re interested in.
6. [Future work] What kind of future works do you foresee around this topic? Could you give any advice, especially to PhD students who are interested in studying embeddedness, tie formation, or social networks in general?
Lei: I think a natural extension of this study would be to examine the consequence of the group dynamics, how that may influence the start-up company performance, in terms of innovation or success. At the same time, including qualitative interviews, fieldwork, and discussion of the process is kind of the trend in future research. Even when you use large quantitative dataset-based analysis, it would be always helpful to include some type of fieldwork evidence, and that will help you illustrate what is happening in the field and what is behind the mechanisms. Basically, a combination of the two would be a good way to provide strong evidence of your paper, and your argument.
Isin: I think there’s a lot of opportunity to think about power, politics and group dynamics, and how they influence innovation processes in organizations. For example, the original works in the Carnegie school had large emphasis on power dynamics, coalition building, and bargaining processes within organization, which have been a little bit overlooked; there has been a lot of focus on learning and adaptation, but not so much on the power and politics side. I sense that that is a field that will probably appear more in future research.
Hang-Jun Cho: Hangjun Cho is a Ph.D. candidate at Entrepreneurship and Family Enterprise department of INSEAD. His research explores how organizations utilize their various assets in directing strategic change, by which they advance their innovation, creativity, and long-term performance.
Ekin Ilseven: Ekin Ilseven is a Ph.D. candidate at Strategy department of INSEAD. Based on complex adaptive system perspective, his research investigates the temporal and multi-level strategic trade-offs involved in achieving organizational resilience.