Supplemental Material
A full replication package, including all code and data, for the visualization reported here is available on Github.
Data Sources
Intergenerational Economic Mobility
I use intergenerational economic mobility measures as prepared and described in Chetty et. al (2014). A full replication package containing data and code for this study is available through the Harvard Dataverse here.
Estimates of intergenerational mobility are generated at the commuting zone (CZ) level. Estimates are reported only for CZs with more than 250 children in the relevant samples. This is the reason that a small number (29) of CZs in Figure 1 are missing data. Relative mobility (RM) is the slope from an OLS regression of child income rank on parent income rank within each CZ. Relative mobility can be multiplied by the difference in parent ranks (on a 0-100 scale) to obtain the expected difference in child’s rank (on a 0-100 scale). Absolute upward mobility (which is the measure of economic mobility utilized in my visualization) is the expected rank of children whose parents are at the 25th percentile of the national income distribution.
Estimates of Surface PM 2.5
I use monthly measures of North American, ground-level, fine particulate matter (PM 2.5) for 2010-2021, as estimated by van Donkelaar et. al (2021). These estimates are generated through a process that utilizes geographically weighted regression to calibrate satellite measures of PM 2.5 with ground-based observations. A more detailed description can be found here. The resulting measures of PM 2.5 are high resolution raster files (0.01° x 0.01°) covering most of North America, but not including Alaska or Hawaii.
Data Preparation
Intergenerational Economic Mobility
The measures of economic mobility do not require any further preparation beyond their original format, as described above.
Estimates of Surface PM 2.5
The first step in aggregating the monthly estimates of ground-level PM 2.5 is to summarize each 0.01° x 0.01° grid as its average monthly PM 2.5 value across the full time period. The resulting file is a high-resolution (0.01° x 0.01°) raster file of average monthly PM 2.5 levels from 2010-2021 across North America. Finally, I roll up these high-resolution measures to the CZ level, resulting in average monthly PM 2.5 levels from 2010-2021 at the CZ level across the contiguous U.S.
Figure 1
I calculate the quartiles of both the economic mobility measure as well as PM 2.5 levels (dividing economic mobility measure and PM 2.5 levels into four equal bins, respectively). I then assign each CZ into one of 16 bins (four mobility bins x four pollution bins) determined by which economic mobility quartile and PM 2.5 quartile it lies within.