CNN poll undersampled Black voters — that's how Trump made up 10 points over last month. THREAD.
Wondering how Trump closed his gap with Biden from 14 points in June to 4 points in August? So was I.
June cdn.cnn.com/cnn/2020/image…
August cdn.cnn.com/cnn/2020/image…
Wondering how Trump closed his gap with Biden from 14 points in June to 4 points in August? So was I.
June cdn.cnn.com/cnn/2020/image…
August cdn.cnn.com/cnn/2020/image…
There are 2 related reasons. The June poll included an oversample of Black voters, while the August poll did not sample enough Black voters to include them in crosstabs, which means fewer than 125 (see poll methodology). And the Aug poll oversampled "battleground states" voters. 
Pretty much all of Trump's gains and Biden's losses are attributable to white voters and independents. So this poll basically reflects and exaggerates the reaction of white people to the selection of Kamala Harris.
It's an interesting choice to undersample Black voters in their first poll after the selection of the first Black woman VP, and oversample from AZ, Florida,
Georgia, Iowa, Maine, Michigan, Minnesota, Nevada, NH, NM, North Carolina, Ohio,
Pennsylvania, Texas, and Wisconsin.
Georgia, Iowa, Maine, Michigan, Minnesota, Nevada, NH, NM, North Carolina, Ohio,
Pennsylvania, Texas, and Wisconsin.
This is not to say the poll should be disregarded, but CNN should make clear what their poll really says. That 10-point gap closure is very deceptive.
