Using movement data to analyze the effect of lockdowns and inform their implementation – the Italian case
While lockdowns are certainly effective in curbing the rise of infections, their imposition severely affects the life and health of citizens. For this reason, the extent of their deployment should be optimized both in time and space to minimize the number of people affected while guaranteeing the safety of the population. At the same time, contact-tracing initiatives can easily violate privacy laws, and are generally difficult to implement for a public administration. It is thus interesting to consider whether anonymized samples of social networks’ datasets still contain enough information to optimize the implementation of lockdowns. Starting from Facebook (FB) users’ movement data from META’s data for good program and publicly available data from the Italian Institute of Statistics (ISTAT), we show how a data-driven meta-population approach can be used to identify a spatial subdivision of a state that maximises movements within communities, while minimizing those outside of them. Specifically, we focus on the level of movements between provinces, administrative entities in between municipalities and regions. After verifying that FB movement data gives reasonable population density in each province, we show that a temporal clustering correctly identifies the first two national lockdowns without any prior information. Finally, by considering the most representative movement networks in both a lockdown and free-movement situation, we identify the optimal communities, i.e. macro regions that minimize the amount of traffic between them. Using two different approaches -modularity and resolution/relevance – gives largely comparable results, supporting the robustness of our findings.