I then conduct the following thought experiment. Given the spatial distribution of the population across counties in the year 1900, what was the average temperature exposure for the U.S population? So, I'm weighting the county average February and August Temperature by the year 1900 distribution of people across U.S counties. Intuitively, more people lived in Detroit in those days and fewer lived in Los Angeles. As you can see below, the average American lived in a place that was 34.1 degrees F. in winter and 73.8 in Summer and 22.8% of people lived in a coastal county.
summ tmean2 tmean8 coast [w=pop1900]
Variable | Obs Weight Mean Std. Dev. Min Max
-------------+-----------------------------------------------------------------
tmean2 | 2818 75438942 34.1235 9.595696 9.107881 69.19515
tmean8 | 2818 75438942 73.81865 4.546925 60.25134 91.89582
coast | 2818 75438942 .2284033 .4198783 0 1
Now, I repeat this exact same exercise using the same climate data but now I use the 2010 calendar year population distribution. We know the population has moved from the Rust Belt to the Sun Belt.
. summ tmean2 tmean8 coast [w=census2010pop]
Variable | Obs Weight Mean Std. Dev. Min Max
-------------+-----------------------------------------------------------------
tmean2 | 3110 304169234 40.06854 12.17356 9.107881 69.19515
tmean8 | 3110 304169234 75.0742 5.515168 60.25134 91.89582
coast | 3111 306665669 .3432595 .4748736 0 1
Note that the exposure to warmer winters has increased by 6 degrees! Exposure to warmer summers has gone up 1.2 degrees and coastal living has increased by over 10 percentage points. These composition shifts hint at what migration could do to "average climate exposure" if places such as Las Vegas and Phoenix are "too hot". But, note the great demand to be in warmer winter places! Perhaps in the future more households will telecommute and live Northern cities in summer and Southern Cities in winter. How is that for an adaptation strategy?
It looking at these calculations, it is important to note that there is no "climate change" here. I never shift the spatial distribution of climate conditions. Taking the climate distribution as fixed, I reweight these data using variation in how many people live in each of these counties. A more sophisticated exercise would also allow for the shifting climate distribution.
A long time ago, Mike Cragg and I talked about these issues in this 1999 RSUE paper.
