UPDATE: There was a significant data coding error in this analysis. In some of the Insurance Information Institute tables, the position of "wind and hail" and "water" is flipped, so some of the values got put in the wrong columns when I copied the data into a spreadsheet.

After correcting this, "water damage" is still responsible for a large chunk of the increase in expected losses (around a third), but now wind and hail damage is responsible for the largest portion (around half).

This is still somewhat confusing, since loss ratios are down in most wind and hail-prone states, but it's now confusing in a different way.