This study provides empirical support for recent theoretical models that allow for time-varying rare disaster risk. Using a unique database of 447 international political crises during the period 1918–2006, we create a crisis index that shows substantial variation over time. We show that...

We introduce a variation of Yu(2011)'s weighted bagging estimation method and show it substantially improves the predictability of the equity premium and other economic variables. This new machine learning method sharply improves equity premium predictability of many models with significant...