Fig. 1
From: Race and ethnicity data for first, middle, and surnames
ℙ(race = r|surname) values for each racial and ethnic group r, computed for linked names via the Census (y axis) and via the voter files (x axis). Data is coarsened such that each dot represents 2% of the range along each axis. The dots’ opacity corresponds to the total number of voter file appearances for names that fall into each range. For comparison, the unweighted correlations are 0.92, 0.93, 0.91, 0.93, and 0.26, respectively.
