The talk was part of the Organizational Networks session.
Blockmodeling collective action networks: Using direct and indirect approaches in tandem
Authors: Clare Saunders and Alejandro Ciordia Morandeira
Abstract: Over the past few decades an increasing number of social movement network analysts have made use of blockmodels to unveil general patterns of ties. Blockmodels significantly reduce the complexity of network data by classifying actors into discrete subsets, understood as differentiated positions or roles, and subsequently describing the ties between and within these positions in an idealised way. In order to find the most appropriate partition of nodes into different positions, analysts have been typically chosen one of two distinct methodological approaches: indirect or direct methods. The indirect approach is often inductively applied. It proceeds in two steps, first transforming the original network data into a single matrix displaying the similarities or dissimilarities between all pairs of actors’ profiles of ties to alters, and subsequently clustering these (dis)similarity matrices. The direct approach only works with matrices’ original data entries, rearranging actors until the computer finds the partition that best approximates a prespecified blockmodel structure – put differently, it is a more deductive approach. Each approach presents its own strengths and weaknesses but their inductive and deductive implications need not necessarily mean that one approach should be selected over the other. We argue that much can be gained from combining both approaches in the same analysis, even if one method will prevail over the other depending on the specific aims of the investigation. We suggest two ways in which this can be conducted. First, deductive designs, usually reliant on direct blockmodeling, can benefit from verification with indirect blockmodelling by providing an approximate benchmark for the assessment of the explanatory power of the resulting optimized partition. In other words, once the best solution for a prespecified blockmodel has been found, the analyst can assess its goodness-of-fit to the actual network and contrast it with the goodness-of-fit of the best possible atheoretical blockmodel with the same number of positions. Second, inductively generated explorative blockmodels could benefit from the optimization routines of direct approaches in order to avoid the misplacement of specific actors and decide on the positioning of dubious cases, thus ultimately refining the initial results obtained from indirect procedures. In order to illustrate our methodological proposal, we make use of two empirical interorganizational networks: one reflecting collaboration between political actors involved in climate policy in the UK , and the other, showing patterns of co-participation in environmental collective action in Spain’s Basque Country.
Concept, implementation and first results of a social network analysis to support cross-company workplace health management
Authors: Gabriele Fohr, Bert Droste-Franke, Carina Hoffmann and Andrea Schaller
Abstract: Since the Prevention Act from July 2015, workplace health promotion (WHP) has been a crucial setting for prevention in Germany. This is especially true for small- and medium-sized enterprises, which often lack the resources for WHP [Taylor 2016]. Within the KomRueBer project, funded by German Federal Ministry of Health (BMG), a cross-company network for promoting physical activity has been set up in a business park in North Rhine-Westfalia, Germany. With the BIG-Manual as a theoretical framework and step-by-step guidance for physical activity promotion, the cross-company network has been developed by a network manager. Its phases (A) finding, (B) preparation, and (C) cooperative planning process, are being evaluated by SNA methods. In line with the focus on the workplace, on the individual or node level we modelled a network of companies located at the business park and engaged in WHP, as well as other supporting organizations, that form the cross-company network of WHP. In addition, a second type of node was introduced: the sequenced events of the finding, preparation, and cooperative planning phase. We then got a two-mode network of WHP with events and participating organizations, aggregated for each time step (current time step plus earlier time steps). It was projected to one-mode to be able to calculate network-specific measures. The analysis covers first results in terms of visualisations (two-mode and one-mode, t1 to t5) as well as individual (company level) and community (cross-company level) network metrics. We were able to identify key actors and their roles within the network, and to evaluate the development of the complete network over time. Further work has to be done to evaluate following time steps, subgroups, non-responder, and a potentially conducive network composition for WHP.
Who steps up after a merger? The effects of boundary-spanning on post-merger taking charge behavior
Authors: Stefan Breet, Lotte Glaser and Justin Jansen
Abstract: Although prior research on mergers and acquisitions (M&As) has suggested that cross-legacy boundary-spanners serve as organizational change agents, an emerging line of research highlights the costs of developing and maintaining boundary-spanning ties. Building on the social networks and organizational identification literatures, we develop a social network perspective on cross-legacy boundary-spanning and post-merger taking charge behavior. More specifically, we argue that employees without boundary-spanning ties are more likely to engage in taking charge behavior when they are closely connected to the boundary-spanners of their legacy organizations. Our analysis of the social network of a post-merger organization shows that cross-legacy boundary-spanning has a negative effect on taking charge behavior, while proximity to boundary-spanners has a positive effect. Our study also reveals that the positive effect of proximity on taking charge behavior is strongest for employees who weakly identify with the new organization.
Organizational change and social network resilience among scientists over 20 years: A multilevel approach
Authors: Emmanuel Lazega, Avner Bar-Hen, Béatrice Milard and Antoine Descoubet
Abstract: Analyses of multilevel (“Fish/Pond”) networks of superposed and partially connected interdependencies (the first being inter-organizational, the second inter-individual) have shown the importance, as explanatory factors of scientists “performance”, of the size and centrality of researchers’ laboratories at the inter-organizational level relative to the importance of their own individual centrality in inter-individual networks. In this paper we take advantage of the fact that the period under examination in this study was a period of great reorganization of laboratories in French cancerology to look at the resilience of inter-individual networks over time as organizational contexts change or disappear. We track the institutional trajectory of 127 top level researchers and look at the extent to which co-authorship ties survive between these individuals over twenty years in spite of (or thanks to?) multiplication of affiliations, bifurcations and mobility from one organizational context to the other. We interpret the result in light of a combined theory of collegiality and its emphasis on the importance of personalized relationships for the coordination of collective agency in innovative activities such as scientific research. But also in light of public policy strategies using organizational “lego games” to drive scientific research and resistance to such strategies.
Establishing publication profiles of Russian universities: bipartite institution/journal network co-clustering approach
Authors: Angelika Tsivinskaya
Abstract: Over the last decade, research on bibliometric networks has received a lot of attention but it usually focused either on citation networks or on collaboration networks. These bibliometric networks are usually based on direct relations between publications or focused on the collaborations formed at the level of authors or institutions (Carusi et al. 2019). The aim of our analysis is to build a map of institutions and identify groups of universities similar in terms of publication profiles using a two-mode network between universities and journals (García et al. 2012). This research is grounded on the idea that researchers from different institutions prefer to publish their work in a different subset of journals in terms of not only disciplinary mixture but also quality and impact. The organisational culture can have a great impact on the choice of where to publish, especially if the cash-per-publication reward policy is present (Quan et al. 2017). The relationship between in! stitutions and journals is utilized to identify scientific community clusters employing a co-clustering detection on a bipartite network. The suggested approach has the potential of leading not only to the detection of academic institution communities based on the similarity of their research output but also to the clustering of scientific journals based on which are the most representative of each community. This methodology does not rely on predefined categories for publication profiles and purely based on similarity in term of publication strategies (Carusi et al. 2019). Our proposed approach is applied to the Russian universities using the data collected from the national bibliometric database (RISC). We chose this source over other alternatives as it has better coverage of journals compare to international databases which are skewed towards natural sciences and the English language. With our data, we have an opportunity to identify whether universities strive to publish in international journals or oriented towards the production of knowledge only for the local market.