Why?

Toplines

• Sample skewing based on party self-identification undermines the idea polls can be accurate.
• A normal distribution with +/- margin of error doesn't work if the samples aren't representative.
• It creates two different bell curves that produce numbers that under-represent Trump support.
• The pollbusters index only includes polls that disclose their party samples.
• We average the last 5 polls and show the average skew.
• With a simple 1:1 50% unskew along party lines, the map changes significantly.

Thumbs on the Scale

2020 Exit Polling showed Republican and Democrat self-identification tied in Michigan and Nevada, and Republicans had self-identified turnout advantages in the remaining battlegrounds from R+1 in Pennsylvania to R+9 in Arizona.

Pollsters ask how voters identify with the Democratic and Republican Parties or as Independent/Other, and these samples rarely reflect the 2020 electorate.

Emerson College is the only poll that regularly (but not 100%) samples based on 2020 party identification and 2020 recalled vote.

Bad Skews Undermine Polling Trust

People mis-read polls all the time, but the bottom line is with a 95% confidence, the result should be within +/- the margin of error of the result. A Pennsylvana poll showing Trump and Harris tied at 47 with a +/-3.0 point margin of error could look like Trump 50 Harris 44.

If the Margin of Error = X, there is a 95% degree of confidence that the final result will fall within a range of 2X.

In 2020, the NY Times/Siena poll was off by a factor of 2X or more in Wisconsin (2.5x MOE), Pennsylvania (2.4x MOE) and Arizona (2x MOE). That is half of their final 6 battleground state polls, and we're to expect a 95% degree of confidence?!

Instead of the results happening within the range of a normal distribution bell curve, it's more like having two bell curves separated by the partisan skew.

Why This Matters

Polling is used to set expectations, which can over-inflate or under-represent candidate support.

The polls have a horrible track record of estimating President Trump's support.

The Pollbusters Index simply seeks to give users a peek under the hood, both indexing the polling toplines as well as the Dem partisan sample skew compared to the 2020 exit polls.

Is 2024 going to have the same partisan composition as 2020? Unlikely. But the indicators suggesting the electorate is shifting more Republican indicate the polls are understating President Trump's ballot performance.

Battleground States

See the averages of the last 5 polls and the built-in Democrat partisan sampling skews.

• Arizona Polls
• Georgia Polls
• Michigan Polls
• Nevada Polls
• North Carolina
• Pennsylvania
• Wisconsin
• Weighted Battleground Averages



What is Skew/Unskew?

How to read:

• Party Sample: Net R+ or D+ of a poll's sample
• Partisan Skew: difference between Party Sample and the 2020 exit poll party ID
• Simple Unskew: what would happen if 100% of Rs and Ds voted for their nominee and you unskewed the result.
• Unskew Chart Line: where Trump would be without the Dem skew
• Skew-Proof Chart Line: if Harris is below the Skew-Proof, she lags Trump without skew



This is not a true reweighted sample, it's just simple math. 100% of self-identified Rs and Ds won't vote for their nominee, but the skew metric and indicator lines on the charts visualize Dem bias in the polls.