“Covid-19 Is (Probably) Not an Exogenous Shock or Valid Instrument”, Jeffrey Clement2024-03-29 (economics, sociology, causality)⁠:

Empirical investigations of many information systems phenomena are complicated by endogeneity. The COVID-19 pandemic prompted a wide range of policy and societal changes that seem to present a natural experiment. Researchers have attempted to use these changes, such as mandated closures of non-essential businesses, as exogenous shocks or instrumental variables in causal inference, with the goal of evaluating a theory or phenomenon not related to the pandemic. However, the rationale that the COVID-19 response prompted changes (such as business and school closures) that were decided by an agent “outside the unit of analysis” is not sufficient to meet the criteria for exogeneity.

We concisely describe and demonstrate via simulation that the wide-ranging impacts of COVID-19—which were driven by politics, personality, and socioeconomics, and implemented in bundles—violate the parallel trends assumption for difference-in-differences analyses and the exclusion restriction for instrumental variable analyses.

Our hope is that this analysis helps IS researchers avoid such problems moving forward.

[Keywords: methods, difference-in-differences, instrumental variable, COVID-19, identification strategy, causal inference]