How Accurate Are The Polls?

If I weren’t so spooked by what happened in 2016, I would assume at this point that the 2020 election is over and done. But being Jewish first of all means that I suffer from a congenital conditions known as kayn ayin hara, which means beware of the evil eye.  I’m also something of a compulsive analyst of data, a combination of having been raised by a father who was an accountant and a college minor in stat.

Here are the numbers today: In the national polls, Joe has a 10.3% lead (52.2% versus 42%) in 538; in RCP his lead is 8.6% (51.1% versus 42.5%). A week ago Joe was at 52.4% in the 538 aggregate, and 51.6% in RCP’s aggregate poll. Slightly down in one, slightly up in the other. In other words, basically the same.

Where things get a little dicey is in the must-win states.  Here is where Joe sat on October 13, RCP in blue, 538 in red:

Here is where Joe stands today, again RCP in blue, 538 in red:

Note that a week ago, 538 had Joe above the magic 51% mark in 3 of 4 states. Note that today he’s at 51% in only 1 of 4 states. RCP had Joe at or above 50% in 3 of 4 states on October 13; now he’s hit that mark in only 1 of 4 states. So the race has ‘tightened,’ which is what all the media people would like you to believe. Because if you didn’t believe it, why bother to tune into their shows? Better watch the jewelry auction or how to cook fat-free food.

The good news is that the slight decline in Joe’s must-win states has not been matched by an increase for Trump. He has gone up in Pennsylvania from 43.8% to 45% in the 538 number, but in the other 3 crucial states, Trump’s either the same (MN) or slightly below where he was on October 13. As for the results from RCP, Trump’s down slightly in Michigan, and up a bit in MN, WI, and PA. If the statewide polls continue to move for the next two weeks the way they moved in the last week, Joe easily wins all 4 states.

Despite what is usually said about the lack of valid statewide polls in 2016, this belief is only partially true. In Michigan, for example, Hillary went into the last week with a 7-point lead. But the dog shit number was still above 7%. In Pennsylvania, statewide polls gave Hillary a 5 or 6-point lead, but the dog-shit number was between 8 and 10 percent. This is why I never discuss the polls without reminding my readers that what counts most of all are the respondents who haven’t made up their mind, if they have a mind, for how they are going to vote.

Pardon my sarcasm, but if you still haven’t figured out what a disaster Trump would be if he held onto the Oval Office for another four years, you probably don’t have a mind, or at least not a functioning mind.

The biggest problem with any attitudinal poll is that many respondents haven’t necessarily thought about the issue for the first time until the moment they pick up the phone. So to compensate for this problem, the pollsters analyze as much previous data as possible, tie the results to the various identifier categories they have on each voter (age, gender, race, income, etc.) and compare the most recent answers to the way this particular individual voted in the past.

This year, according to Nate Silver, most pollsters are paying more attention to educational attainment for voters because they believe that the non-college voting population was undercounted in 2016.

In the 2016 exit polls, voters who never went to college went for Trump by 51% to 45%. But voters whose annual income was less than $50,000 went for Hillary by 52% to 42%. So how do these numbers explain the fact that Trump won Pennsylvania and Michigan by a grand total of 65,000 out of 10 million votes cast in those 2 states? Sorry Nate, it doesn’t.

Can we trust the polls this time around? It really won’t matter what the polls say if our side makes sure to get out the vote.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.