Climate research like this quite new so this is most welcome.
I suspect that Miami looms large in driving these results. I noticed that coastal Southern Florida has a high hurricane count (by the measure used) and it was also at the epicenter of the subprime crisis, which is covered by the Fannie Mae data used in the study.
I’d like to know if the key result (that hurricanes and mortgage delinquencies are positively related) still holds if Florida data is excluded from the model. The other states were all relatively subdued during the housing boom that preceded the GFC.
I also noticed that LTV at origination is used as a driver. The authors did not include variables to proxy an updated LTV, like the change in house price since origination. I suspect that a variable like this would pick up the Miami effect. This would be tricky to model, though, because hurricanes will also impact prevailing local house prices.
There’s an interesting paper by Gallagher and Hartley (linked in comments) that looks at New Orleans data around Katrina. They found a much weaker, shorter term hurricane impact.
Anyway, it’s great to see some new research in this exciting and extremely challenging field.
https://www.garp.org/#!/risk-intelligence/all/all/a1Z1W000005xyqrUAA
Risk Weighted 