An intensely pro-Trump sample: Why you should take the latest NBC poll with a huge grain of salt

NBC News released today the results of a survey purporting to show that in the aftermath of a Trump mob invading the Capitol of the United States, Donald Trump's approval rating has barely budged. According to the poll conducted by a Republican and a Democratic firm, Trump's approval rating is still at 43%, barely lower than the approximately 45% line he has held throughout his cursed presidency.

NBC's numbers stand in stark contrast with two other polls released over the last week. A Quinnipiac poll released on January 11 shows Trump's approval at 33%, and another survey by Pew Research showed it dipping to a low of 29%.

You should take NBC's numbers with an Olympus-sized grain of salt.

As always, the devil is hiding in the details. The poll, which stands in stark contrast with two other polls released over the last week, overrepresents white people by a few points and more importantly, those without a four-year college degree by double digits.

The NBC poll released today puts the share of the non-Hispanic white vote at 72%, a bit higher than the actual share of white voters in the November election. A Pew analysis of voter demographics just before the November elections showed that white voters were expected to make up 69% of the electorate. NBC's own exit polling of the November election put the share of white voters at an even lower 67%. A 3 to 5 point difference may not seem like a great deal, but a 5 point differential in favor of white voters can actually represent a net 10-point racial gap from the actual electorate.

According to both NBC and CNN exit polling, for example, Black and Latino voters each represented 13% of the November 3 electorate, while today's NBC polling puts them at 12% and 10%, respectively. At a total of 22%, that is four points lower than the actual electorate.

So, in addition to over-representing white voters by 5 points, NBC's pro-Trump survey underrepresented people of color by at least 4 points, netting at least a 9-point departure from the actual electorate.

But as we have learned over the past half-decade, a stratification of voters along racial lines is incomplete without an accompanying division along education lines. White voters certainly prefer Trump, but white voters without a college degree are his true base. So imagine my surprise at finding out that the NBC survey over-represents voters without a college degree and overrepresents those without.

According to the aforementioned NBC exit polling, 42% of voters in November had a bachelors' degree or higher, but in the new pro-Trump poll, that number stands at just 36%. That is a six-point underrepresentation of people with college degrees and a six-point overrepresentation of those without, for a net 12-point skew in Trump's favor.

This skew likely further means that whites weren't just over-represented in the sample, but that the white portion of the sample was much more intensely pro-Trump than the white electorate at large. That alone could have accounted for the entirety of the double-discrepancy that this outlier NBC poll appears to show in Trump's favor compared to the other polls that are much more likely to be representative of the current sentiments in the country and show Trump in a freefall.

No one should be surprised that a poll that over-represents non-college white voters by double digits shows more pro-Trump results, and no one should take such a poll seriously.