This is some simple analysis to show the relationship between the partisanship of a state and error in the polling margin. Because the lines generally slope down that means that polls tend to over-estimate the Democrats in Red states, and the same of GOP in Blue states. The intercept of a 0pp mean error is generally around a tied (D+0/R+0) state.
MIT Election Data and Science Lab, 2017, "U.S. President 1976–2020", https://doi.org/10.7910/DVN/42MVDX, Harvard Dataverse.
FiveThirtyEight Pollster Ratings: https://github.com/fivethirtyeight/data/tree/master/pollster-ratings