Differential polling error is calculated for the 2020 election as the relative difference in polling error for two candidates (Biden and Trump).

Relative error is calculated as (observed−expected)/observed, where the “observed” value represents actual election results, and the “expected” value denotes the predicted results from polls. Differential polling error is actuated as the relative error for candidate Trump minus the relative error for candidate Biden.

This statistics combines the relative polling under and over performance for each candidate. For example, in the state of Utah, if Joe Biden is under-performs his polls by 4% (-4% relative error), and Trump out-performs his polls by 8% (8% relative error), the differential polling error for this state would be (8%) - (-4%), or 12%, meaning that for this state, candidate Trump had a 12% advantage relative to polling.

I omitted 3 voting states or districts for this analysis: Hawaii, Alaska, and District of Columbia.

Hawaii and Alaska would have made the state by state visualization tiny, and made a not very attractive map. The District of Columbia can be considered an outlier in this study in that Trump signficantly under-performed his polling there (but his polling in DC was incredibly low). Trump was polling at around 12% in DC prior to election day, and gathered around 5% of the vote on election day. His relative performance would be about 120%, which would have thrown the color scheme way out of wack.