In 2020's presidential election, it was shown that Joe Biden lost support from nonwhite voters, most of whom were ideologically conservative.

However, he not only still won this voting group, but he did better here than Kerry. Since 16 years is a fairly long time for demographic changes, I am wondering if there was a correlation between these two variables at the county level: the increase in minority population from 2000 to 2019 (the most recent data available now, and 2000 is a census year) and the change in two-party Democratic vote share, from 2004 to 2020 on the presidential level.

Notes: I am asking this question because I want statistical analysis for a project I am working on, and want to see how demographic change affected real-life election results, and not just projections. As an example, the coordinate for the national average is (10, 3.5) -- which means a 10 percentage point increase in nonwhite population in the US as a whole, and a 3.5 percentage point increase in Democratic vote share from 2000 to 2019.

  • I want this measured in percentage points, for both the population and the vote share. Apr 21, 2021 at 11:37
  • 4
    “Joe Biden lost support from nonwhite voters” do we actually know that? I’ve heard it a lot and it seems plausible, but the only evidence I’ve seen cited for that was exit polls, which are particularly flawed this year due to partisan differences in mail-in voting. Have you found better sources for this?
    – divibisan
    Apr 21, 2021 at 14:10
  • County and precinct level data, especially in non urban areas. Apr 21, 2021 at 14:36
  • 1
    Many minorities and young folks were adversely affected by covid stay-at-home measures, and simply voted their wallet; I wouldn't read too deep into the most unique election of our lifetime looking for patterns.
    – dandavis
    Apr 21, 2021 at 18:06
  • 2
    Why from 2004 to 2020 and not 2000 to 2020, to match up with the census dates?
    – CDJB
    Apr 22, 2021 at 8:13

1 Answer 1


No, there is little to no correlation. We can use decennial census data from 2000, demographic estimates from the American Community Survey, the MEDSL county data for the 2004 election, as well as Fox's feed for the 2020 election to investigate this. There is barely any correlation between the variables - a correlation coefficient of around 0.11, which indicates a very weak positive correlation.

As you can see from the graph below, most counties have seen a small increase in non-white population, but the change in two-party vote share varies significantly.

enter image description here

The code used in this answer can be found on GitHub.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .