Project Details
This investigation aimed to increase understanding of Gray Zone conflicts. In particular, it explored the role of violent non-state actors within these dynamics. In doing so, this project combined detailed qualitative research with large-n quantitative testing across a diverse set of cases: Colombia, Libya and Ukraine. This research was conducted at START as a Strategic Multi-Layer Assessment Initiative. It was part of a larger endeavor involving additional sub-awardees and deliverables.
First, despite the commonly accepted belief that the role of military versus other instruments of power on the DIMEFIL spectrum is relatively minor, our analysis shows that kinetic events have an outsized effect in shaping actions across the rest of the spectrum. Second, both VNSA proxies and state forces seemed to be concerned about domestic and/or international legitimacy. As such, if one side de-escalated (for example decreased Black Zone activities in favor of more Gray actions), the other side was likely to follow suit, so as not to be viewed as the aggressor. Third, and relatedly, a serious caveat is in order to the second point. Specifically, we found that VNSAs were less adept at correctly interpreting state aggression or conciliatory action. Generally, the tended to interpret even conciliatory actions as aggressive. Fourth, in seeking to quickly identify potential partners, this research shows that practitioners can chooses to aggregate actors of the same type without substantial loss of fidelity, insofar as they behave similarly. Moreover, further simplification is possible by examining dyadic pairs of conflicting belligerents (as opposed to all parties at once). Doing so limits the number of instruments of power that need to be considered, while identifying the Zones of conflict that predominate in a given dyad.
This project developed and employed semi-automated quantitative data assessment, recoding and deduplication procedures in order to undergird both frequentist and Bayesian statistical analysis. Specifically exploratory multivariate regression analysis was employed alongside the development of a Bayesian Belief Network (BBN) model. Qualitative research, including both thick-description and process tracing, was also utilized.