Fantastic Use of Data

A bottom-up study of the potential economic impact of climate change shows how a large scale analysis should be done

What Climate Change Will Cost Every U.S. County, 2080–2099(Kopp, Hsiang, et al. / Science)

I like to point out interesting uses of data, which the recent paper in Science, doing a bottom-up analysis of climate change’s economic impact on the US, most certainly is.

Unlike every other previous analysis, they started at a county level, assessing how each county would be affected by a 4°C increase in global temperatures. (Note that an average increase doesn’t mean that every location increases by precisely that amount. Some will increase by more, some by less. That’s the nature of averages.) The results and the data visualizations are all rather impressive and worth looking at.

This blog doesn’t dive into politics much, and I’m not going to debate climate issues here. The authors assumed a certain increase in temperatures. If you reject that assumption you’ll likely reject this study. But I find the method used for this approach to be interesting, and the paper well-worth reading for an understanding of how to tackle such a complex problem even if you find the underlying assumption to be wrong.

And some interesting conclusions

Agriculture is hugely impacted, but so is public health. In fact, regional outliers in Kern and San Diego counties, California, gain most of their improvements from better health. This is largely the result of eliminating many types of agriculture and the risky jobs that go along with them.

If you want to live in a place that’s doing well and plan to live much longer than I do, you really want to be north of the 40th parallel, or in a mountainous area — basically the Rockies or the heart of the Appalachians. There are only a handful of exceptions to this rule. While not studied, it is unlikely this effect stops at the border. Canada should do well.

If such simplistic analysis holds, then Europe mostly looks pretty good. The middle east looks to continue as a major shit-show and most of Africa will be a mess, as will India and much of southeast Asia. Russia looks pretty good too.

Optionally, you might want to live south of the 40th on the bottom of the world. Basically, New Zealand. Chile and Argentina both have pieces to the south of the 40th, but the bulk of both countries and their economies are in the “hot” zone. So is the rest of South America.

What’s missing

They assume a “normal” world, while also assuming a 4°C rise. This in itself is problematic. For example they don’t show as much damage as I would expect in places like Miami, New Orleans and even Houston, where a significant sea-level rise could lead to the abandonment of huge (and valuable) populated areas. They also cannot study the potential costs of dislocation of millions of people from places that are less viable. These kinds of “long tail” events are difficult to model, likely to happen (we can reasonably assume that some long-tail events will happen, we just don’t know which!), and likely to drive the largest regional and local impacts.

All in all, an interesting paper if you want to understand how complex data can be modeled.

If you’re interested The Atlantic has a less-academic take on the numbers.