Some countries seems to be very poorly located, as they are in the path of all sorts of natural disasters. But some do better than others in coping with their perilous situation. In particular, death tolls from cataclysms seems to be, in general, of an order of magnitude larger in developing countries. What else could influence such numbers?
Eiji Yamamura, from a rich and homogeneous country that does quite well with earthquakes, tsunamis and typhoons, studies ethnic heterogeneity in two different ways in this regard. The first is ethnic polarization, which describes how close the distribution of ethnic group is from a fifty-fifty one, and ethnic fractionalization, which can be interpreted as the probability that two random people are from the same ethnic group. Once one adds the miracle instrument of cross-country regressions, legal origins, the first indicator has shows that heterogeneity has a positive impact on natural disaster deaths (meaning more of them), while the second has none.
Now Yamamamura takes this as a sign that ethnic polarization is a better indicator of ethnic heterogeneity than ethnic fractionalization. This looks like some seriously flawed reasoning here, which is repeated several times in the papers: the fact that some indicator tests favorably some hypothesis does not necessarily mean that it measures what the hypothesis says. What if in really the hypothesis is false? And in any case, on what theory would this hypothesis be based? I can easily imagine good reasons why homogeneity would lead to fewer deaths, a better social cohesion that leads to better institutions coping with disasters, like in Japan.