Implications of Heterogeneous SIR Models for Analyses of COVID-19
This paper provides a quick survey of results on the classic SIR model and variants allowing for heterogeneity in contact rates. It notes that calibrating the classic model to data generated by a heterogeneous model can lead to forecasts that are biased in several ways and to understatement of the forecast uncertainty. Among the biases are that we may underestimate how quickly herd immunity might be reached, underestimate differences across regions, and have biased estimates of the impact of endogenous and policy-driven social distancing.
I thank Daron Acemoglu, Chris Avery, Victor Chernozhukov, Adam Clark, Jonathan Dushoff, Sara Fisher Ellison, Jim Stock, and Ivan Werning for helpful conversations and comments and Chris Ackerman and Bryan Kim for research assistance. The views expressed herein are those of the author and do not necessarily reflect the views of the National Bureau of Economic Research.