Many suggests that they are manipulating the number of Democrats and Republicans to boost Harris's numbers.
“So what they're doing is they're polling fewer Republicans. They're polling a disproportionate number of Biden 2020 voters in these states that were dead even,” pollster John McLaughlin said this week on the “Just the News, No Noise” television show.
“They're saying the Biden 2020 voters should be four or five points higher. It's ridiculous. So what they're doing is they're trying to pump Harris up. They're trying to suppress our vote. And this is, you know, there's smart people doing this, so I think it's intentional.”
I think this is happening to some degree, but it is not as widespread as years gone by. Sure, there are many pollsters who are hiding their demographics that could be fudging the numbers. In fact, I find that many of those pollsters showing large leads for Harris are not providing the numbers and a couple of who do show large leads and provided crosstabs seem to (in fact) be engaged in some over and undersampling to get there.
But if I look at the overall average of those pollsters who are providing their demographic crosstab information, I don't see it as I have in past years. Now that being said, of the pollsters who are actually providing their Demographics, the average toplines of those pollsters shows the race at 0.62% for Harris. This is considerably lower than the overall average you are seeing in places like 538, the Silver bulletin, or the Hill. This suggests that fudging may very well be happening with the pollsters who are both pushing these larger polling averages while hiding their crosstabs. Bad sampling is likely the reason they are hiding their crosstabs.
What is interesting and different from previous years is that when my spreadsheet recalculates for the turnout model it is actually pushing the number towards the Democrat. Harris's 0.62% polling average lead actually grows to 1.58% when I run it through modeling. When I push the undecideds into the mix, it grows a bit more to 1.71%. So according to what I can tell, the pollsters providing crosstabs appear (on average) to be sampling around a net one-point turnout advantage for Democrats, whereas I am predicting a two-point advantage. Perhaps I am being too generous to Democrats, given they lost the turnout battle in the 2022 midterm by three percent. I have debated whether the two-point advantage should be something more like a point and a half, but I have settled in at the two-point mark to keep it consistent with how I calculated previous spreadsheets.
The bottom line is that while fudging the demographic makeup of polling sample to help Harris poll better might be happening with the pollsters who are hiding the crosstab information, I don't see it happening to any large degree with those who are providing it. So far, the lion's share of pollsters seems to be fairly close to where they should be in my humble opinion.
Kommentare