Are natural hazards temporally random?

The paper employs statistical hypothesis tests to explore the question of whether natural hazards (hail and tornadoes being considered here) are or are not intertemporally random. The answer to this question, at least for these two hazards, is surprising and has important policy implications: hazards appear to be more likely in year t if an event was experienced in year t?1: hence, apparent ‘Gambler's Fallacy’ behaviour on the part of farmers and others may indeed be rational. Should this result hold for a full range of climate-related hazard types. as we suspect, hundreds of millions of dollars of crop loss and other damages may be preventable (e.g., rotating to frost-resistant crops if past events suggest greater than usual likelihood of early or late frost). The degree of positive serial correlation is also seen to vary a great deal regionally, with individuals in some states being unable to benefit from knowledge of prior hazard occurrence.