The result is that output per capita at a 40 year horizon would be
20 percent below baseline (recall that in our base case scenario, output per capita at a horizon of
40 years is 2 percent above baseline). By contrast, Acemoglu and Johnson’s coefficient from an
instrumented regression of log GDP per capita on log life expectancy is -1.32 (Table 9, column 1),
implying that the mortality improvement we consider would lower GDP per capita by 41 percent.