This is the repository containing the data and R scripts used to perform the analyses presented in the paper co-authored with Maria Rita D'Orsogna (Institute for Pure and Applied Mathematics - UCLA and California State University - Northridge)
The global COVID-19 pandemic has shocked the world with a wide array of societal consequences that go well beyond the dramatic public health sphere. Governmental interventions, along with the distress caused by the health-related threat, have fostered significant strain and tension across most of the world's population. As a consequence of these dynamics, social unrest has increased in multifold ways. As one of the results of social unrest, protests riots, and disorders events in general have emerged in many countries. In light of this, the present study aims at analyzing and characterizing the temporal nature and distribution of disorder events directly associated with the COVID-19 pandemic. It does so by using data retrieved from the ``COVID-19 Disorded'' Tracker (CDT) initiative developed by the `Armed Conflict Location & Event Data Project'' (ACLED) project. We specifically focus on the three countries that have exhibited the highest number of disorder events from January to October 2020, namely India, Israel, and Mexico, accounting for a total of 5,194 events. To investigate the temporal nature of these disorders, we employed a specialized class of point processes, named Hawkes process, that are characterized by the presence of temporal clustering and self-excitability components as two fundamental mathematical concepts for capturing inter-dependence among events. For each country, we investigated temporal clustering and self-excitability considering both the overall stream of protests, riots, battles, and violence against civilians and sub-national streams derived through the algorithmic generation of geographical clusters. We empirically show that temporal clustering and self-excitability emerge both at the national and the sub-national levels in all the three considered contexts, although with various magnitudes. The results indicate that this duality is critical in fully capturing the complex mechanics behind disorder events: national-level temporal clustering is not the outcome of random sub-national processes but, instead, is the realization of meso-level self-excitability patterns.