As many have noted on Daily Kos, the “state of the race” in terms of polling averages has been inordinately influenced by many partisan and low-quality polling agencies. In order to get an idea of what the polling averages might be if we remove all such pollsters from the mix, and use only well-respected, non-partisan pollsters, I have been doing my own calculations.
I’m not claiming it is the best method, or that it will prove accurate. My motivation for doing these calculations is because I can’t seem to completely ignore the polls, and I wanted to see what happens if all of the “junkier” polls are removed from the calculations.
Before I present the final numbers I came up, here is an outline of the basic method I used:
1. Starting with 538.com’s list of polls from the last several weeks (in almost all cases, only the last six weeks), I discarded all the polls done by entities that either: (a) have very low ratings, or are not listed at all, on the “Pollster Ratings” from 538.com and Nate Silver’s “Silver Bulletin; or (b) have known party connections (to either party) or other widely-recognized biases.
2. For each of the remaining pollsters, I assigned them a weight based on their average ratings and on how recent their latest poll was conducted.
3. For each state, I did the following. For the polls where I could find the crosstabs, I broke down their poll numbers by gender. I then used the percentage of 2020 voters that were women and the percentage that were men (which you can find HERE) to calculate the overall numbers for each of the two candidates. For example, in Pennsylvania, the 2020 voters were 52.75% women and 47.25% men. Meanwhile, based on the crosstabs of the 2024 Pennsylvania polls, using the weighted averages as mentioned above, the female vote is about 54.5% for Harris and 40.2% for Trump, while the male vote is about 41.6% for Harris and 53.9% for Trump. From this, for Harris, we get 52.75 x .545 percent (female vote) plus 47.25 x .416 (male vote) = 48.4 percent; and for Trump, we get 52.75 x .402 + 47.25 x .539 = 46.7 percent for Trump
4. For each state, for the polls for which I do not have access to the crosstabs, I just took the overall weighted averages for each candidate.
5. I then took the weighted average of steps (3) and (4) above. Note, since some pollsters put out polls rather frequently, I counted each pollster only once (even if they had multiple polls), taking a weighted average of that pollsters very recent polls.
Anyway, here are the numbers I came up with for each state, and nationally:
Pennsylvania: Harris + 1.9
Michigan: Harris +1.83
Wisconsin: Harris +2.93
Nevada: Harris +1.35
North Carolina: Harris -0.43
Georgia: Harris -1.17
Arizona: Harris: -2.14
National: Harris: +2.38
For your information, here is the list of the pollsters that I did use: Washington Post, Marist, Quinnipiac, Marquette, YouGov, Research Co., Suffolk, Beacon, New York Times/Siena, Fairleigh Dickinson U., Ipsos, Susquehanna, CNN, University of Mass. Lowell, University of Georgia, Muhlenberg College, East Carolina U., High Point University, Elon College, Data for Progress, Morning Consult, Florida Atlantic U., Mass INC., Hart Research Associates, Angus Reid Global, Data Orbital.
One unknown that we don’t know, and that the polls do not account for, is whether certain demographics will represent a larger portion of the voters than in the past. In particular, we don’t know whether women will comprise a larger percentage of voters than in recent elections. I am not sure how the early vote broke down by gender in previous elections, but looking at the early vote so far this time we have the following (women/men):
Georgia, 55%-45%
North Carolina, 54%-45%
Michigan, 56%-44%
Pennsylvania, 56%-43%
Wisconsin, 55%-39%
Nevada, 47% -43%
Arizona, 47%-46%
For Michigan, Pennsylvania, and Wisconsin, the percentages of early voters that are women are significantly higher than the percentages of all voters that voted in 2020 (and 2022 and 2016), so that looks good, but I don’t know if we can conclude much from these early women vote percentages.
Another big unknown, is how far off, and in what direction, the polls will be this year. In 2020, they overestimated Biden and Democrats in general by something like 4%. But since then, the polls underestimated the Democrats, including the 2022 Senate races, where they were off by more than 2% in most cases.
