Insider: Pollsters Are Oversampling Anti-Trump Republicans


RUSH: Let's go to back to the polling for a moment, because in the polling it's almost identical. The polling of 2016 had Hillary winning by landslide margins. There were just a couple of polling units that had Trump, eh, holding steady, maybe even eking out a win. Same now. Every reputable polling unit you've ever heard of has Joe Biden winning in a landslide -- plus 10, plus 12, plus eight. Just a couple have Donald Trump barely hanging on and maybe winning.

One is Rasmussen. The other is Trafalgar. Which takes me to the polling and a theory. A friend of mine had a fascinating conversation with an off-the-record pollster. And this off-the-record pollster, meaning the pollster did not want to be known, did not want to be identified. He was content to share with my buddy how they are doing this, but he didn't want his name mentioned. And I don't know the name. My buddy has not furnished the name. But here's the theory. This is the theory to explain the polling discrepancies.

The polls are deliberately oversampling mushy, suburban Republicans to cover charges that they are shorting Republicans overall and deliberately under-sampling rural and small-town, conservative Democrat zip codes. All of this to prove that they are not oversampling Democrats. Then, they announce that they are scientifically fair on party affiliation, when they know in Trump's case he has far more support from Reagan Democrats than from Bush suburbanites.

And when I saw this theory, when this was explained to me -- and I got the note here during a commercial break about 20 minutes ago -- I beat myself upside the head. No wonder. This makes perfect sense. This is exactly how they're doing it. They're oversampling Republican suburbs. Who lives in Republican suburbs? White, educated women who hate Trump. White, college-educated Millennials who literally despise Trump. And the fact that they are Republican is all the better for the polling units because they can then say, "We're not shorting Republicans. Our sample is not short of Republicans."

In fact, in one sense, they're oversampling, to a degree, and they are shorting Republicans where the true Trump support is. So me run through this again and again explain to you. A friend of mine had a conversation with an off-the-record pollster who clarified how these popular nationwide known, these nationally known polling units are doing it and showing Biden plus eight and plus 10.

The first thing polls are doing is deliberately oversampling suburban Republican voters. That would be largely women. The word here is "mushy." So you could have some Republican men thrown in there, but you've heard the old saw here, that Trump is just doing horribly, horribly in the suburbs. Suburban women hate Trump. How many of you have heard that college-educated suburban women, they despise Trump.

Well, it turns out that these voters are being deliberately oversampled for two reasons. A, they're being oversampled because the pollsters love the fact they hate Trump, but also oversampling Republican suburbanites allows them to beat back allegations that they are shorting Republicans and deliberately under-sampling them. So who are they under-sampling? They are deliberately under-sampling people that live in small towns and rural conservative Democrat zip codes. And they're doing that to prove that they are not oversampling Democrats.

So they're doing two things here that are producing massive anti-Trump results, and they are ignoring voters who would be massively for Trump. And they're doing it in order to show that they're not shorting Democrats or Republicans, and they're not oversampling Democrats. In other words, they have found where to go to find Republicans who hate Trump -- suburbanite dwellers, suburban dwellers, mainly women.

They have also found where not to go to avoid Democrat support for Trump. And that would be rural, small town -- think swing districts. Think places Trump won in Wisconsin and Pennsylvania. They are not going to those places where conservative Democrats are. And they're purposely not going there so they can prove that they're not over-sampling Democrats. They love to be able to tell people questioning their work, "Oh, we're not over-sampling Democrats. In fact, we're over-sampling the Republicans in suburbs and we're under-sampling some of these rural Democrat areas."

And then after they do all of that, then they announce that they have been scientifically fair on party affiliation. They are doing it this way to refute allegations that they are under-sampling Republicans. Because, remember, in every poll that we have analyzed or looked at for the last number of years, you dig deep into the polling data, and you will find that the Democrats are massively oversampled. Thirty-eight percent Democrats, 24% Republican, 39% independent or whatever it is.

And so people always say, "See? See? No wonder they're getting all these favorable results for Democrats, because they're not sampling enough Republicans." Well, that's what this is meant to counter. They announce that they scientifically are a right down the middle in finding participants in the poll, proper party affiliation but they know in Trump's case that he has a lot more support from Reagan Democrats than he does from Republican suburbanites.

So if they oversample Republican suburbanites and undersample where conservative Democrats live, then they're gonna get results like they've got. And they can say all the while that their results are right on the money based on their sample. They can say they're not making anything up. They're not fudging anything. And so if Trump ends up winning again, they say, "You know, well, we must have picked the wrong sample."

But they're not gonna admit that anything was done on purpose. They're gonna stand by their sample as what they thought was the best cross-section of the voting public that they could find, poll after poll after poll. When in fact what they're doing is over-sampling Republicans that hate Trump. Can you imagine if you find that group -- if it's true, and it is, that suburban, white women, Republican women don't like Trump, if you can find a number of them to poll, you kill two birds with one stone. You go out and find a bunch of who hate Trump who happen to be Republicans. Oh, my God. We found a way to make sure we don't have to short shrift Republicans in the sample.

Then as an added bonus, you find out where Democrats who love Trump are and you don't sample them. You sample other Democrats where you know they hate Trump. And that's how you end up with what they've ended up with. Now, this all comes from a pollster admitting all this to a friend of mine, who is then passing it on to me for interpretation and analysis and eventual repetition to you.