I analyze damage from hurricane strikes on the United States since 1955. Using machine learning methods to select the most important drivers for damage, I show that large errors in a hurricane's predicted landfall location result in higher damage. This relationship holds across a wide range of...

Can forecasts of natural disasters alter their destructiveness? Poor forecasts increase damages when individuals do not mitigate risks based on the false belief that they will be unaffected. We test this hypothesis by examining the impact of 12-hour-ahead forecasts on hurricane damages and find...