Let's start with the polling industry as a whole.
RCP tracks 18 different polling outfits that do multi-state polling. And of those 18, 14 of them over the course of the last decade have missed the actual results in favor of the Democrat candidate. Four missed in favor of the Republican. So if 78% of the polling consistently reports polling that overstates the final margin for Democrats, how much is that average miss? 4.6%. The least amount of error on behalf of the Democratic candidate is 2.2%, and the biggest miss is 6.1%, but the sweet spot for missing is just a touch over four-and-a-half.
As of Sunday, the RCP average of averages shows Kamala Harris with a 1.7% lead nationally. That's down from 1.9% before the debate, by the way. So what does that mean? Using 2020 as a baseline example, roughly 165 million Americans cast a ballot for someone for president. A lot for Trump, more for Biden, and almost 10 million for minor parties and independents. 1.7% of that 165 million is a 2.8 million vote lead nationally for Kamala Harris, giving the polling out there the benefit of the doubt that they're accurate. Now again, using 2020 data as the baseline, take a look at how California voted. It's as deep blue of a state as there is in the Union, and they voted overwhelmingly for Joe Biden. The final margin, once they finally finished counting a couple months after Biden was sworn in (I exaggerate, but not by much), was 63-42. In raw vote totals, Biden netted 5.1 million more votes just in the Golden State than Donald Trump. Going back to the RCP average as of Sunday, if Kamala allegedly has a 2.8 million lead nationally, and would have roughly the same results out of California, you could actually make the case that if you were to take California out of the equation, Donald Trump is actually leading the national vote in the remaining 49 states by a little under 3 million votes.
Now we get to the poll that was released over the weekend by AtlasIntel, the most reliable pollster based upon prediction versus reality over the last two elections - 2020 and 2022. In the former, AtlasIntel was only off by 2.2%. In the case of the latter, they slipped a little bit, missing by 2.3%. AtlasIntel by far has the best track record this decade, earning 2.7 out of 3 stars at out of 538's ranking system. They just released their new national numbers, entirely surveyed post-debate.
3.6% lead for Trump, eh? Let's say they're off by 2.3%. He's still up over a point. Remember that 1.7% RCP lead for Harris as of Sunday? Add in their average miss of 4.6%. That means Trump's probably up even more than that.
Nate Silver has been getting a ton of grief online by the fever swamp on the left because his election projection has been growing increasing in Trump's favor, currently standing at nearly 60-40 odds. Keep in mind, Silver is a Democrat and a Harris backer, but he's essentially become a pariah on the left because he's looking at all of these numbers and extrapolating the obvious. You can pile up all the raw votes in the blue states of California, New York, and Illinois all you want, but it's not going to equate to an Electoral College victory.
Polls do two things. The compilation of them show a snapshot of where the election is, and a trend of which direction the race is moving. And you can then extrapolate to what effect the popular vote shown in these polls would have on an Electoral College result.
hotair.com