Zuzana Sasovova – VU University of Amsterdam
Ajay Mehra – University of Kentucky
Stephen P. Borgatti – University of Kentucky
Michaéla C. Schippers – Erasmus University Rotterdam
Julia Brennecke – University of Liverpool
Natalie David – University of Freiburg
Question 1. You investigate how self-monitoring determines brokerage in friendship networks in the workplace. However, you ground individuals’ motivation to become a broker on various instrumental benefits that arise from this position such as access to diverse information and control over information flow. Given that individuals experience different motivations to build affective (e.g., friendship) as compared to instrumental ties (e.g., advice), why did you not choose to investigate self-monitors’ brokerage behavior in instrumental networks, which intuitively seem to be more closely related to the motives you discuss?
It is worth noting at the outset that one can derive instrumental benefits from friendship networks (just as one can derive expressive benefits from an advice or workflow network). The nature of the benefit is not dictated by the kind of relation: the same relation can be put to different uses. That said, we focused on friendship because it is a discretionary relation and is therefore a good place to find evidence of the effects of personality on workplace networks. Instrumental ties at the hospital we examined were more likely to be shaped by new work demands related to the organizational change that the radiology department underwent. Moreover, note that our paper does not suggest that high self-monitors have exclusively instrumental motives when it comes to their social relations—that would be a grievous misrepresentation of self-monitoring theory. Our theory is focused more on how structures that are dispositionally congruent are attractive because they provide greater opportunities for the manifestation and reinforcement of dispositional preferences. Finally, our theory emphasizes not just what ego wants/is motivated to accomplish but also how others tend to react to ego—for example, laboratory based evidence that high self-monitors are relatively comfortable interacting with strangers is one reason we expected that, following an exogenous shock and the relationship rewiring it triggered, high self-monitors (relative to low self-monitors) would gain more new friends who are unconnected to their previous friends (i.e., new friends who are relative strangers). This, by the way, is why our measure of brokerage requires that alters see the broker as a friend (rather than focusing exclusively on the broker’s perceptions of the relations).
Question 2. Your paper leaves open whether high-self monitors consciously create and maintain structural holes or whether these are unconscious consequences resulting from their behavioral preferences. In fact, you acknowledge in the discussion that future research is needed to disentangle whether the network actions of high self-monitors are “deliberative and goal-oriented rather than routine and habitual”. What research design would you suggest to scholars aiming to approach this question?
One possibility would be to assign people social network goals (e.g., increase the number of structural holes in one’s network) and then examine the extent to which they succeed in achieving the goals as a function of self-monitoring personality and the goals assigned to them. The setting could be an informal social/networking event; or it could be something more extended, such as an executive development program. We need more insight into which kinds of network structures different kinds of people value—in preliminary work we have underway we are examining how self-monitoring differences are related to the value people attach to different networks (for examples of stylized network scales, see Mehra et al., 2014: http://ajaymehra.net/Documents/imaginaryworldsRSOtoshare.pdf). The kinds of networks people value are important to understand because people presumably seek to shape their networks in ways they, right or wrong, consider valuable/useful.
We also wanted to use this opportunity to share related analyses that we experimented with but that did not make it into the paper. The analyses had to do with changes in the triads that ego was involved in as a function of ego’s self-monitoring status and the alters’ self-monitoring status. The tools for these analyses were implemented in UCINET in the Network|Ego net|Longitudinal section of the menu system.
The simplest routine examined all triads that ego was involved in at T1 and determined what kind of triad each had morphed into at T2 (e.g., new hole, same hole, closed hole). For each ego, we could count how often, say, an open triad A1—Ego—A2 became closed due to the addition of a tie between A1 and A2. Using this routine we showed that, in line with our hypothesis, high self-monitors were more likely to form new structural holes by gaining new friends.
The more complicated routine, that used a randomization technique to assess the probability of obtaining differences as large as actually observed if hole closure were independent of the self-monitoring orientation of the broker, took the (dichotomized) self-monitoring status of both the ego and alters into account. For example, we were able to count how often an open triad a1—ego—a2, where ego, a1 and a2 could each be a high or a low self-monitor, changed into some other pattern, such as a closed triad, or a disconnected triad (e.g., the tie to a2 goes away). Using this routine, we showed, in support of our hypothesis, that a hole with a low self-monitor as a broker was more likely to close than would be expected by chance (this analysis was not included in the paper but was shared with the review team at ASQ). Because the Ucinet routine examines self-monitoring status of all three nodes in each triad, it also becomes possible to test whether a hole between two low self-monitors is more likely to stay open than a hole between two high self-monitors.
Question 3. Changes in network structure are not only induced by individual agency, but also by network self-organizing mechanisms (e.g., tendencies towards transitivity or reciprocity). Statistical methods such as stochastic actor oriented models or exponential random graph models allow accounting for these mechanisms while examining the effects of individual characteristics on network dynamics. Did you consider these methods? If so, why did you decide not to use them?
A premise of your question is that there are some changes in the network that we might ascribe to ego’s agency, and others that either “just happen” or, more sensibly, are the result of others’ tendencies. So in examining change in brokerage status, we’d like to take account of some of these unintended changes which some people regard as endogenous or “self-organizing”. For example, we might see high self-monitors accumulating many ties and bridging many holes. However the number of holes in an ego network changes not only as a function of ego’s actions, but of others as well (e.g., two alters forming a tie, reducing the number of ego’s holes). As a result, we cannot simply attribute the change in holes from T1 to T2 to the actions of ego. One approach to dealing with this (using the brokerage elasticity routine in UCINET) is to compute the number of holes that ego could have expected to result from making the changes in her network that we actually observe. This variable is essentially change in holes due solely to ego’s agency, filtering out the consequences of others’ actions. A related analysis we conducted in response to a reviewer’s comment who noted that “not all holes are at equal risk of closing over time” took the local structure around alters into account. Specifically, those holes embedded in multiple 3rd party relations may be more likely to close, whereas two alters with no friends in common besides the broker would be less likely to develop a tie. So, for each ego, we computed the number of friends that ego’s alters had in common at the previous time period and included this ego-level variable as a control. It did not change the reported effects of self-monitoring on brokerage dynamics.
Overall, our paper focuses on individual differences in self-monitoring orientation which lead to specific patterns of network churn and changes in brokerage positions. The approach we took differed from that taken by ERGMs and SAOMs. Our theory, and our hypotheses, are articulated at the individual level and because of the theoretical focus on impression management our analyses are mainly focusing on incoming ties: we ask how a person-level IV such as self-monitoring affects a person-level DV such as holes formed around the same individual. In contrast, ERGM focuses on ties, not on actors and the parameters are interpreted as tendencies in the graph instead of tendencies of individual actors. Although SAOMs do focus on actor tendencies, because these models were developed with an underlying modeling assumption of making changes in ego’s outgoing ties one at a time, it would be very difficult to test our hypotheses and conduct similar analyses with regard to brokerage dynamics. Fundamentally both ERGMs and SAOMs are models of tie formation/dissolution (see e.g., Block, Stadtfeld, and Snijders, 2016 for a more detailed comparison of applicability of ERGMs and SAOMs). They are not designed to regress a node-level outcome on a set of node-level characteristics or factors. They simply answer a different question than we were seeking to answer.
Question 4. Longitudinal data on informal organizational networks is typically difficult to get. How did you go about convincing the radiology department overall as well as the individual employees to participate in your study? Based on your experience, what advice would you give to junior scholars aiming to collect similar data? Moreover, did you provide the department with a report of your findings and, if yes, do you know what use they had for them?
Trust is key to obtaining high quality data. Respondents have to feel that you are not a stooge of management; and that you are genuinely interested in providing them with feedback that they will find useful/interesting. The first author (Zuzana) spent a lot of time building trust not only with the management of the department, but also with individual employees. She conducted a lot of face-to-face meetings with different functional groups, which were often followed up with one-on-one personal introductions when handing out the questionnaires. All this was very time consuming but it provided an opportunity to respond to employee concerns and questions (of which there were many). It is also important to give respondents what they value rather than just some summary of research. For example, in this research, which was part of Zuzana’s broader dissertation work, what respondents really valued were the maps of departmental workflow networks. They were less interested in the theoretical questions—about personality and network churn—that drove the ASQ paper. It is important to provide value to respondents, and value should be defined from their perspective not the researcher’s perspective.
Question 5. Finally, we are interested to know how you experienced the review process at ASQ. For instance, how many rounds of revision did you go through? How long did it take? Have the reviewers confronted you with contradictory requests and if yes how did you deal with them?
This paper benefited enormously from the helpful guidance provided by our editor, Phil Anderson, and our anonymous reviewers. We sincerely mean that. Many of the hypotheses in the final version of the paper were suggested to us by the editorial team as they helped us make sense of theory and results. The guidance was invaluable. The feedback on the first round was very clear but also super challenging. There were some differences in the counsel provided by the reviewers, but our editor did a masterful job of giving us clear guidance on how to proceed in such cases. We spent a full year working on the revision. However, it was very gratifying to receive a conditional acceptance on the second round—that only motivated us to make this the best paper we could. Linda Johanson provided excellent advice on how to restructure the paper and clarify the writing—we recommend that every prospective author read the excellent advice she offers in her piece “Sitting in your reader’s chair” (Johanson, 2007): http://jmi.sagepub.com/content/16/3/290.full.pdf
Block, P., Stadtfeld, C, and Snijders, T. A. B. 2016. “Forms of dependence: Comparing SAOMs and ERGMs from basic principles.” Sociological Methods & Research, In press.
Johanson, L.M. 2007. “Sitting in your reader’s chair: Attending to your academic sensemakers.” Journal of Management Inquiry, 16: 290–294.
Mehra, A., Borgatti, S. P., Soltis, S., Floyd, T., Halgin, D. S., Ofem, B., and Lopez-Kidwell V. 2014. “Imaginary worlds: Using visual network scales to capture perceptions of social networks.” In D.J. Brass, G. Labianca, A. Mehra, D.S. Halgin, & S.P. Borgatti (eds.), Research in the Sociology of Organizations. 40: 315–336. Bradford, UK: Emerald Publishing.