I really enjoyed this paper and how it clearly elucidated the theories of social network analysis, the idea of embededness and interplay between spatial and social attributes of geographical phenomena, and the need for a matrix-based model to describe and predict gang-based violence in Los Angeles.
The overarching and unspoken goal of the paper seemed to be to create a quantitative framework for visualizing social and spatial relationships. This distilling of observable social phenomena (gang rivalries and violence) into matrices and testable hypothesis, gave the paper a very logical flow and purpose. On page 311, the authors discuss how “identifying social positions as collections of actors with similar measures of equivalence allows…outcomes for similar actors to be operationalized and tested”. This is the first explicit mention of the predictive power of the outcomes of network analysis when applied to this specific issue of gang rivalries in Hollenbeck. Again, on pg. 315, the authors present their analysis as a testable hypothesis; “the hypothesis here is whether or not similarly embedded and positioned territorial spaces experience similar amounts of violence.” Their declaration of the relevant variables is very clear, as is the directionality of their hypothesis. I appreciate how succinctly they state the purpose of their network analysis and the predictive power of their results.
That being said, their methodology definitely glosses over some of the nuances of rivalry, resulting in binary positions on each matrix criteria, a criticism that they acknowledge on pg. 317. Several iterations of the calculations result in a grouping into either +1 or -1, allowing for dichotomized categorization which is very convenient for the purposes of data analysis, but are oversimplifications. They address this issue by expanding into 8 subcategories, but I would have liked an explanation as to why 8 was deemed sufficiently nuanced whereas 2 was two few and 30 too many.
I thought that the groundwork laid in the beginning of the paper (map with clear geographic and social boundaries outlined) provided a strong foundation for the discussion of the case study. Rather than apply their methodology to the crime data in a vacuum, the authors gave the reader background on why the observed relations among rival gangs might occur based on the geography of the space. I also appreciated the diagrams and thought they complemented the text nicely, although I did have a few lingering questions. Figure 3 represents each gang as a node and the portruding lines connecting one node to another indicate whether or not a rivalry exists between the two groups. But there is no discussion of the placement and relative distances between the nodes. Do the ones that are depicted closer to each other have contiguous turf areas? Are they organized according to a N/S/E/W axis that mirrors their real-world territoriality? These questions remained unanswered.
A further criticism would be the wording and use of survey from LAPD and *some* former gang members to identify the rivalries between each of the gangs. The use of the word “enemy” in particular is vague and seeing as perceptions of rivalry may arise from individual interactions and experiences, the accuracy of these designations is questionable, since alliances and feuds are dynamic, mistrust of outsiders is endemic, and their model is static. However, they did mention that the rivalries showed perfect symmetry between cops and ex-gang members alike.
But these are all minor criticisms of an otherwise excellent review and application of social network analysis to a contemporary geospatial issue. They suggest that these methods can be applied to all kinds of other scales and kinds of research in geography and I would like to see it applied to health epidemics, drug use, and homelessness, just to name a few.
-futureSpock