Pick a variable — a CDC health measure, an ACS figure, or presidential margin —
and a geography. Every area with a value gets split into five equal-sized groups by that
variable; this compares everything else — health, wealth, and politics — population-weighted,
between the highest and lowest fifths.
Each group's figures are population-weighted (a place of two million does not
count the same as one of two thousand), and computed live from the same database the map itself
reads — nothing here is precomputed or cached ahead of time. Correlation, not causation: these are
county- or tract-level (ecological) statistics that describe places, not people, and a
place's political lean is itself bound up with its age, income and urban-or-rural mix. Census
tracts carry no election results of their own (elections are counted by precinct and roll up no
finer than county), so the Elections section is empty for tract-level reports.