Tracking the State and Behavior of People in Response to COVID-19 Through the Fusion of Multiple Longitudinal Data Streams

The changing nature of the COVID-19 pandemic has highlighted the importance of comprehensively considering its impacts and considering changes over time. Most COVID-19 related research addresses narrowly focused research questions and is therefore limited in addressing the complexities created by the interrelated impacts of the pandemic. Such research generally makes use of only one of either (1) actively collected data such as surveys, or (2) passively collected data from sources such as mobile phones or financial transactions. So far, only one other study collects both active and passive data, and does so longitudinally. Here we describe a rich panel dataset of active and passive data from US residents collected between August 2020 and September 2022. Active data includes a repeated survey measuring travel behavior, compliance with COVID-19 mandates and restrictions, physical health, economic well-being, vaccination status, and other factors. Passively collected data consists of Point of Interest (POI) check in data indicating all the locations visited by study participants. We also closely tracked COVID-19 policies across counties of residence of study participants throughout the study period. The combination of the longitudinal active and passive data helps overcome the limitations of active or passive data when used individually as well as the limitations posed by cross-sectional dataset and allows important research questions to be answered; for example, to determine the factors underlying the heterogeneous behavioral responses to COVID-19 restrictions imposed by local governments. Better information about such responses is critical to our ability to understand the societal and economic impacts of the COVID-19 pandemic and possible future pandemics. The development of this data infrastructure can also help researchers explore new frontiers in behavioral science. This article explains how this approach fills gaps in COVID-19 related data collection; describes the study design and data collection procedures; presents key demographic characteristics of study participants; and shows how fusing different data streams helps uncover behavioral insights often difficult to reveal from either data streams individually.