brokerage is a process not a structure; repeat after me: brokerage is a process not a…


The best conceptual presentation that I saw at this year’s Sunbelts was a paper by UCI’s David Obstfeld and Steve Borgatti entitled “Brokerage is a Process, Not a Structure: A Clarification of Social Network Language.” Their basic argument is simple but very provocative: it is time to do away with “purely structural” definitions of brokerage and to consider the qualitative content and context in which the ties are made.

The problem? The now hegemonic and widely accepted structural operationalizations of brokerage (i.e. Burt 2005; Fernandez and Gould 1989) as implying a state whereby an actor stands between two actors who are themselves not directly connected to one another misses a lot of things that are intuitively “brokerage” but because of allegiance to the standard definition are currently under-theorized and under-studied.

Prototypical examples of some of this brokerage that does not conform to the standard model are a “matchmaker” in a sexual/marriage market or a “talent scout” (what Hirsch 1972 referred to as “boundary personnel” in culture industries and DiMaggio 1977 explicitly referred to as brokers). Other examples come from Obstfeld’s own research on productive exchange networks, where a broker brings together two other people with complementary talents to work on a creative project that otherwise wouldn’t have worked together.

It is clear in these cases that the structural definition of brokerage as resulting from control benefits due to the bridging of structural holes does not work in any of these cases: here the whole (no pun intended) point (and explicit goal) of the broker is to close the structural hole and have the partners meet. In fact, there is no benefit to the broker if the “forbidden” triad does not turn into a plain old fully-connected triad (the prospective partners don’t like one another, the artist and the label are unable to reach a deal, etc.), in fact the broker in this bizarro world makes a living by getting rid of forbidden triads.

But of course as Obstfeld and Borgatti point out, this is not such a bizarro scenario. In fact most brokerage systems are probably based on this model of productive exchange (or the “third who joins”) rather than on the divide-and-conquer tertius gaudens imagery, which may only be the tip of the brokerage iceberg. In this sense, by defining brokerage narrowly in traditional structural terms, we may be missing most of the brokerage processes that are going on right under our noses. For instance, in the aforementioned types of context you may want closure and clustering, and in fact lots of structural holes may decrease performance. That is, those who join also benefit.

The solution? An expansion of the idea of brokerage to make the gaudens case a special case of a more inclusive set of processes. Obstfeld and Borgatti propose something like a 2 x 2 table cross-classifying gaudens and iungens types of brokerage with dense and sparse network structures. So that you may have forms of advantageous brokerage that both occur in dense networks and that require joining rather than segregating action on the part of the broker.

I think that the Obstfeld/Borgatti analysis of brokerage is a great example of the theoretical limits of “pure structural” concept-definition and the benefits that come from bringing in a qualitative sense of culture and process into network theory that Paul DiMaggio talked about in his “Nadel’s paradox” paper.

Written by Omar

January 29, 2008 at 9:06 pm

19 Responses

Subscribe to comments with RSS.

  1. Amen to that!

    Taking structure for granted is almost always a mistake — structure masks more important underlying drivers.



    January 29, 2008 at 9:10 pm

  2. Hmmm… are you fighting straw men? I always interpreted structural holes as a sort of potential. The gap in the network is a potential that somebody might exploit. Thus brokerage = action + structural possibility. Who interprets this in any other way? How can a hole, by itself, make anything happen?



    January 29, 2008 at 9:18 pm

  3. I think the key issue is not that accruing benefits is being conceived in a way that is separate from action or as pure “structure”, but that a lot of brokerage involves a different type of structure altogether than the one that has traditionally been thought of as necessary. In fact the type of structure that has been defined as the “opposite” of brokerage (i.e. closure). In both types of structure action is necessary to get the benefits, but the kind of action (splitting versus joining) is qualitatively very different. So the structural definition, in fact “defines away” a kind of action that indeed emerges as brokerage when this is seen as not tied to any particular structure but as a type of qualitative process.



    January 29, 2008 at 9:22 pm

  4. I think part of this is a matter of how research questions and associated understanding evolves and progresses over time.

    So, someone comes in and notes the critical importance of structural holes as sources of information, etc (insert favorite DV), and then argues that the structures (or holes) are the locus of activity (that is, the IV of choice). Then, someone else comes in and says, wait a second, it can’t just be about the structure, the node itself has some underlying properties and these need to be understood (process or certain characteristics), and so forth. The IV then in effect becomes the DV to be explained. There is a tendency then toward reduction in the process — unless someone steps in with something like ‘culture’ to trump everyone else (Omar?) — and overall I think there is progress in explaining phenomenon. The structure then in effect falls out of the equation over time (in some ways its a mere correlation now), as more fundamental drivers and processes are sought.



    January 29, 2008 at 9:25 pm

  5. Oh, forgot to add: thereafter a Giddensian voice comes in and says that we’ve all been using false dichotomies, its all about interaction.



    January 29, 2008 at 9:38 pm

  6. Teppo – I don’t think (having never read the paper or seen the presentation) they’re saying that structure falls out of the equation. In fact, if structural holes facilitate brokerage processes, then you couldn’t make this argument. More opportunities for brokerage would be present in structural holes. Instead I think you could say that structural holes just aren’t sufficient (and probably not even necessary) to produce the kinds of returns associated with brokerage.



    January 30, 2008 at 12:21 am

  7. I haven’t read the paper either (link anyone?) — but simply saying that structural issues end up quite often being epiphenomenal to, and artefacts of, more fundamental causal drivers that are nested.



    January 30, 2008 at 1:52 am

  8. Since I had to miss Sunbelt, it was great to get reports from your blog. Thanks Omar.

    And thanks for the report on this paper. A couple of things to note though:

    1. The title of this paper is worrisome. Having arguments about what brokerage ‘is’ is bound to be unproductive. Brokerage is a concept we use to to organize our thinking about relationships or processes in the real world. So, to say, that it really ‘is’ one thing or another is essentially meaningless. It’s a concept. Similarly, it is a strange objective to ‘clarify social network language.’ If I come up with my private language for analyzing social networks, can anyone tell me that I’m wrong? No. Language cannot be wrong. The question is whether one concept or another is useful for organizing our thinking about the real world.

    2. The coordinator role, which seems pretty much what Obstfeld is trying to capture with his ‘tertius iungens’ idea, is one of the forms of brokerage in the Fernandez//Gould scheme.

    3. Are matchmakers really about eliminating structural holes? In a short-term sense, of course! They succeed when formerly-disconnected people get together and stay together. And sure, it seems awkward to say that they are exerting ‘control’ over the happy couple. But note that the matchmaker is trying to make a living and presumably not just cover her costs. She charges a price, after all, and the price will be increasing in her reputation for being able to make matches that no one else can make. This is how control is exercised in many cases: extracting profit. Think ebay as matchmaker. Very similar. And quite profitable, no? And in both cases, the reason for the profits is that they have identities that allow them to facilitate matches between disconnected parties that they couldn’t do on their own or through an alternative matchmaker. And ebay leaves enough on the table so that they keep on coming back. So the structural holes persist despite the tertius iungens!


    ezra zuckerman

    January 30, 2008 at 11:19 am

  9. If I come up with my private language for analyzing social networks, can anyone tell me that I’m wrong? No. Language cannot be wrong. The question is whether one concept or another is useful for organizing our thinking about the real world.

    The true voice of Chicago pragmatism :)



    January 30, 2008 at 1:26 pm

  10. Thanks Omar for capturing our presentation so well and starting this chain. I found all the posts interesting and will focus on one that touches on several issues.

    1. I agree 100% with Ezra that “brokerage” (and any other concept) should be used to “organize our thinking about relationships and processes in the world.” I would also contend, however, that any dialogue that clarifies those concepts so that they better organize our thinking is – quite the opposite of Ezra’s contention – bound to be productive. It is my right to fashion whatever private language I please (or ratify an existing public language in use) but when the use of a private or public language is the source of confusion or blocks advancement, we are well-served to seek out a clarification, especially if such a clarification is parsimonious and has a certain face validity. In this case, Steve and I are suggesting that the term brokerage, as it is currently used, conflates structure and process and that clarification of our language (and theory) in use can help us disentangle these two important dimensions of the phenomena.

    2. Ezra hasn’t seen the paper – no fault of his own – but the tertius iungens (“third who connects”) approach to brokerage – that accounts for two of the four types of brokerage we discuss in the paper (see below) is simply not addressed by Fernandez and Gould. Fernandez & Gould (1994:1457) offer a conception of brokerage that “…does not permit the endpoints of the brokerage relation to be directly connected.” (I’ve talked with Roberto Fernandez about this explicitly and believe this is a fair characterization but could be wrong.)

    3. The paper (which is not yet ready for primetime) takes issue with the use of term “brokerage” to refer to social network structure and argues that such an approach conflates structure and process preventing us from studying how brokerage processes influence structure, or vice versa. We suggest as the solution: Disentangle treatments of social network structure from the term “brokerage” which is defined as a social process.

    The typology of brokerage (as a process) we offer (taken from my 2005 ASQ paper) is as follows:

    1. Coordinate action or information between parties who have no immediate prospect for direct introduction or connection. (Conduit)
    2. Actively maintain and exploit separation between parties (tertius gaudens)
    3. Introduce or facilitate ties between parties where a continuing coordinative role is unnecessary, diminishes in importance, or simply not offered – the tertius iungens (brief tertius iungens)
    4. Introduce or facilitate interaction between parties while maintaining an essential coordinative role (sustained tertius iugens)

    Key point: We further suggest that each of these types of brokerage may take place in sparse networks where alters stand on either side of a structural hole, dense networks where all nodes share a strong tie, and mixed networks consisting of various patterns of weak, strong, and absent ties.

    4. Ezra raises an interesting matchmaker vignette at the end though I see brokerage as referring to a widely and increasingly dispersed behavior of which formal or quasi-formal roles are only one case. If I introduce Joe and Julie (who have no preexisting tie to each other) based on a shared interest they have, I am acting as a broker. If I call a meeting between Rick and Raylene who already have a tie with each other, for the purposes of starting a new project, I am also acting as a broker. So brokerage is not restricted to activity around structural holes – something we can start to think about if we separate structure from process. BTW, whether the broker expects or extracts a profit is a separate and important question that is also worthy of extensive further consideration, but one reason brokerage and/or entrepreneurial action is often a high risk/high reward activity is that there is often little control over whether there will be an immediate or delayed profit. The broker’s expectation for profit and whether they get one or not ought not to be assumed.


    David Obstfeld

    January 31, 2008 at 2:36 pm

  11. Hi David. Great that you saw this. I had been meaning to send you a link and ask for your paper when it’s ready.

    I don’t have time at the moment for a thorough response, but I wanted to quickly concede that you’re right about Fernandez and Gould. I should have checked the source. Conceptually, there is a great deal of overlap with the coordinator role, but I concede that they do not focus on on the broker’s closing the triad.

    More soon…



    January 31, 2008 at 4:37 pm

  12. Here is a bit more of my reaction to David’s post:

    1. What is the difference between structure and process? In David’s post, “ties” seem to be an element of structure. But when we say that two actors are tied, we are presumably just saying something about how they have interacted in the past and perhaps something about their predisposition to interact in the future. It’s just interaction (or perhaps just directed orientation, if we think that I have a ‘tie’ to Scarlett Johansson) captured over time. In a word: ‘process.’ So, I’m not sure what it means to say that brokerage is an aspect of process rather than structure. All we seem to mean by structure is process captured over time.

    2. I think it’s good that David (in his ASQ article, and now again) has focused our attention on cases where actors benefit from actively facilitating interaction between parties who might otherwise not have interacted. But as I have told David in the past, there are disadvantages of focusing so heavily on ‘local’ patterns of interaction– i.e., among members of a triad– because it potentially misses big differences in the implications of those local patterns depending on where they are situated in larger structures (i.e., enduring patterns of interaction among many more nodes). So sure, I’m fine with calling both those cases David mentions ‘brokerage’; for cetain purposes (e.g., outcomes we might be interested in for them as individuals or collectively) they might be identical. But for other purposes, — e.g., which one is likely to be more profitable– they are quite different and it’s therefore not clear what is gained by giving them the same label. And relatedly, my point about the matchmaker is that you can potentially miss something important by just focusing on the closing of those triads, since the act of facilitating exchange often helps to reinforce the matchmaker’s position in the larger system (e.g., by growing its reputation for making matches no one else can make).

    Another way to make the point about the importance of the larger structure is to consider two cases that meet David’s definition of tertius iungens: (a) a firm that facilitates a meeting among its rivals to coordinate on prices (or perhaps more benignly, to coordinate on an industry-wide marketing campaign or marketing effort); and (b) a firm that facilitates a meeting between one of its rivals and one of its important customers. The second firm is a pretty dumb firm, no? And the first firm is potentially quite smart (if ethically or legally suspect). The larger point is that what looks the same locally can look very different once you situate it in the larger structure.



    February 1, 2008 at 5:08 am

  13. From the perspective of computer science and engineering, there’s a rich amount of publications and theory on process algebras and their relations to structure. It’s not an easy subject, it mixes with distributed systems. Much of the recent work deals with mobility, i.e., actions which combine sending and receiving data with the alteration of links and sites.

    A good attempt to unify different process algebras is Robin Milner’s notion of bigraphs. Bigraphs combine a specification of nested sites and linking nets between sites. Process algebras set up mappings from bigraphs to bigraphs, via reactions, transitions and equivalences. Which alas probably sounds like gibberish till one starts to work with it all.

    On Milner’s home page, he’s looking for reviews and comments on a first work teaching bigraphs at a grad school level.

    I think most of the examples I’ve read here can be modeled usefully using bigraphs and process algebras.



    February 1, 2008 at 7:47 am

  14. From reading the comments (rather than the source–a new shortcut) I am missing any reference to the current literature on networks from complexity science/self-organizing systems. In that literature OF COURSE brokerage is a process. The network consists of the flows. This has been applied to everything from protein structure, bio-chemistry, and baboon colonies to collaborative filtering. Holes are opportunities and in Watt’s thinking you don’t want density of network everywhere. That is what brokers are for, to make a bridge that is not an overload and creates network robustness. Psychology has recently reinvented ethnography, don’t tell management science is reinventing current network theory. Barabasi, Watts, Bonabeau, Holland–is this literature not part of the conversation?



    February 6, 2008 at 11:33 pm

  15. Eleanor,

    Management science draws mostly on sociological network theory (see for instance Burt 1982) which was there before the first physicist decided that networks were cool (unless you count Harrison White). Hence most of what you refer to as “current network theory” is itself a recent reinvention (of the wheels and all) of sociological network theory (excepting Holland if that’s a reference to the work of Holland and Leinhardt). So management science through its connection to sociology has actually been ahead of the curve before complexity and self-organization became trendy.

    For a post as to why I think molecules shouldn’t be mixed with people see here.

    For a post on why E. F. Keller thinks that a lot of network physics is not applicable to biology see here.

    For a post on why social networks really are different from molecules and traffic see here.



    February 6, 2008 at 11:43 pm

  16. This is why I can’t stand the “new” science of networks.



    February 7, 2008 at 12:45 am

  17. […] and Management Theory) Newsletter. Among other things, we learn that Ron is interested in brokerage as process, that being a real-life broker is quite stressful, and that Ron has a Second Life avatar. JK: I’m […]


  18. […] have been thinking about these issues for a book I’m writing on brokerage. Like David Obsfeldt and Steve Borgatti, I find the meat of the issue to be in the *process*, not the structure—in figuring out why […]


  19. […] can read about him here. Also, you may find this online discussion of a presentation he gave at Sunbelt interesting. The presentation comes out of a paper that he and […]


Comments are closed.

%d bloggers like this: