Flood Risk Stress Testing: Lessons from the Fort Lauderdale Extreme Rain in April 2023
Zip Code Risk Data Offers New Clues
The New York Times reports that Fort Lauderdale has faced a record rain deluge. A “silver lining” of this shock is that it offers a test of the predictions made by the emerging field of climate science’s ability to predict the geography of risk. Back in 2021, I wrote a piece about the emerging competition between different entities predicting local climate risks. If such spatial risk rating entities can predict relative risk exposure of different areas, then loan lenders, insurers and urban planners can nudge our economy to adapt to increased climate risks by introducing incentives to push economic activity to relatively safer locations (i.e areas that the risk raters rate as relatively safe). Loan lenders can offer more generous lending terms in safer areas, insurers can design contracts that offer cheaper premiums in safer areas and urban planners can up zone in these safer areas. Together this suite of policies moves economic activity to “higher ground” and this creates a more resilient economy that suffers less when future disasters take place. This net dynamic would mean that future empiricists who estimate “climate damage functions” of the form
Economic damage in $ = a + b*Recent disaster will estimate a smaller “b” . This would be statistical evidence that more recent shocks cause less economic damage because the economy is using microeconomic incentives to adapt to anticipates shocks. This is the main theme of my 2021 book.
Returning to the Fort Lauderdale deluge. We will soon know the extent of the flooding. This can be mapped by zip code to calculate for the 11 zip codes in the city of Fort Lauderdale how much flood damage occurred.
This realization of the damage can be compared to the First Street Foundation (FSF) ex-ante prediction of flood risk. FSF provides their data by zip code for free and here are the data;
A 10 indicates very high flood risk score
a 1 indicates very low flood risk score
zip Code First Street Foundation Average Flood Score
1. 33301 7.784037
2. 33304 7.054352
3. 33305 7.745545
4. 33306 7.953757
5. 33308 6.718842
6. 33309 3.092274
7. 33311 3.932072
8. 33312 3.306016
9. 33315 6.506026
10. 33316 7.008132
11. 33334 6.330962
I see that First Street Foundation predicts that zip codes 33309, 33311 and 33312 are the safe zip codes. Does the data created by the April 2023 shock support this claim? Have homes and properties in these zip codes suffered less flooding? This natural experiment (the heavy rain) offers a test of the model predictions.
How do applied climate scientists use natural experiments to update and validate their pinpoint spatial models? If these pinpoint climate science models prove to be accurate then they will help us to adapt to climate change.
Let me close with a point about Type 1 versus Type 2 errors in decision making.
Law Students are taught that it is more costly to society to execute an innocent man (a Type 1 error) than to release a guilty man from prison (a type 2 error).
This is an example of asymmetric loss function. We must make a decision on a man’s prison sentence without knowing for sure the truth about his innocence. In the case of flood risk and modeling a property’s flood risk; do we prefer to make type 1 or type 2 errors?
In a nation filled with risk averse people, I believe that we prefer to label a safe property as flood risky rather than label properties that are dangerous as safe.
If we overly trust an incorrect model of spatial climate risk, we can increase our risk exposure!