Reply to “alive and kicking: Mortality of New Orleans medicare enrollees after hurricane katrina”

In the paper “Does When You Die Depend on Where You Live? Evidence from Hurricane Katrina,” published in the November 2020 issue of the American Economic Review (Deryugina and Molitor 2020-hereafter “DM”), we used administrative Medicare data to estimate the short- and long-run mortality effects of Hurricane Katrina on elderly and long-term disabled victims who were initially living in New Orleans. We found that despite a substantial mortality increase in the year of the hurricane, 2005, the cumulative probability of dying in the longer run was lower among Hurricane Katrina victims than among several comparison groups. This surprising result is apparent in plots of raw mortality rates, differencein- difference analyses using a variety of comparison groups, estimates from the synthetic control method, and survival model analyses. To explain why the mortality of Hurricane Katrina victims decreased in the long run, we compared the mortality of victims who moved to higher- versus lower-mortality destination regions. We showed that these movers’ ex ante predicted mortality was unrelated to the local mortality rate in the destination region, but their realized mortality was highly correlated with the destination mortality rate, demonstrating that place has a causal effect on life expectancy. Using a back-of-the-envelope calculation, we showed that the estimated place effect combined with the average decline in local mortality among the victims can account for most of the mortality decline among the Hurricane Katrina victims. Robert Kaestner (2021-hereafter “RK”) questions both our finding of a mortality decrease among Hurricane Katrina victims and our analysis of mortality among survivors who left New Orleans in the aftermath of the hurricane. While many of the concerns he raises are ex ante valid, most of them have testable implications that were already addressed in the published paper. Below, we respond to RK’s concerns, restating and elaborating on our original findings. Additionally, we show that the back-of-the-envelope exercise offered by RK to show that our estimates imply a decrease in mortality among New Orleans stayers is incorrect and does not fit the context.