Using Household Rosters from Survey Data to Estimate All-cause Mortality during COVID in India
Official statistics on deaths in India during the COVID pandemic are either incomplete or are reported with a delay. To overcome this shortcoming, we estimate excess deaths in India using the household roster from a large panel data set, the Consumer Pyramids Household Survey, which reports attrition from death. We address the problem that the exact timing of death is not reported in two ways, via a moving average and differencing monthly deaths. We estimate roughly 4.5 million (95% CI: 2.8M to 6.2M) excess deaths over 16 months during the pandemic in India. While we cannot demonstrate causality between COVID and excess deaths, the pattern of excess deaths is consistent with COVID-associated mortality. Excess deaths peaked roughly during the two COVID waves in India; the age structure of excess deaths is right skewed relative to baseline, consistent with COVID infection fatality rates; and excess deaths are positively correlated with reported infections. Finally, we find that the incidence of excess deaths was disproportionately among the highest tercile of income-earners and was negatively associated with district-level mobility.
Malani acknowledges funding from the Becker Friedman Institute at the University of Chicago to purchase a subscription to the Consumer Pyramids Household Survey and the support of the Barbara J. and B. Mark Fried Fund at the University of Chicago Law School. We thank Mahesh Vyas, Kaushik Krishnan, Chinmay Tumbe, Shamika Ravi, Rukmini S, Prabhat Jha, Arvind Subramanian, Justin Sandefur, Abhshek Anand, Anmol Somanchi and seminar participants at the CMIE weekly webinar for helpful comments. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.