Page - (000324) - in Knowledge and Networks

322 a group (in this case the cluster); at −1, all ties are internal to the group.6 We can see in Table 15.2 that a large majority of the E-I indexes for the clusters are negative. These findings suggest that learning within clusters is much stronger than more global learning. Because the pattern of emergence of the global network works via a principle of local proximity, we might expect local clusters to cluster together on a geographical basis. In other words, we should expect the clusters identified above to cluster into larger regions. To examine this, we wanted a clustering technique that did not require us to assign the number of clusters. We selected Markov clustering, which uses a different strategy of community detection (van Dongen, 2008). The Girvan- Newman community detection procedure used above identifies community struc- ture by removing edges with high betweenness centrality until nonoverlapping groups appear. Markov clustering identifies community clusters by “walking around”; it identifies clusters as places where the algorithm spends a lot of time walking. This strategy intuitively captures the way information might circulate geographically. The Markov clustering identified 22 clusters, which at first glance might seem to approximate the county structure of Sweden. However, two of these clusters are very large and many others are quite small. Our interpretation is that the Markov clustering algorithm identifies the regional as opposed to the local clustering struc- ture of the network. These larger regional clusters attracted our attention because they suggest that one of the ways the national learning network might be integrated is through larger learning regions. These regional clusters could be significant in the circulation of knowledge among Swedish municipalities. A study of regional inno- vation and networks by Fleming, King, and Juda (2007), for example, found that such large components are positively correlated with innovation in patent co-author- ship networks. The two large clusters are indeed regions in a spatial sense. One of them (Fig. 15.1) represents the northern coast plus the Stockholm region (minus Stockholm itself). The second region (Fig. 15.2) runs spatially east to west in the southern part of Sweden and contains the Göteborg region. We also observe that a distinctive subregion can be detected in the Southern region (Fig. 15.2). This subregion is an extremely tight cluster of municipalities around the city of Göteborg. A possible explanation for this tight clustering is the formal creation of a metropolitan region. The formal association is called the Göteborg Region Association of Local Authorities, and the member municipalities are Ale, Alingsås, Göteborg, Härryda, Kungsbacka, Kungälv, Lerum, Lilla Edet, Mölndal, Partille, Stenungsund, Tjörn, and Öckerö. All these belong to the distinctive subregion visually detected in Fig. 15.2. The West Sweden region (Västra Götalandsregionen) that includes the Göteborg Region Association of Local Authorities has been studied in prior research. Gren (2002) notes that this is one of the best organized regions in Sweden, partly through 6 The E-I index for all the municipalities using this clustering was −.271 (the expected E-I index was .893, significant at <.05). C. Ansell et al.