Why The Direction Of The Polls Should Worry Us

The surrounding horse-race narrative driven by the media around November is it will be a hard-fought nailbiter.  The polls on the other hand all show a single trend line – Joe Biden is winning and not looking back.  We can look at this in two ways.  The first would be that Joe Biden is actually dominating the contest today.  Or, we can look at it and see the polls are reflecting the mistakes of past elections, despite their protestations they are not.

The mono – directional leaning of the polls one way or the other is not a new phenomenon.  We saw this play out in multiple locations in 2018, and this was after pollsters said they had fixed the issues from 2016.  Take for instance Ohio.  That year, despite several polls taken by different companies from October to the election, not single poll showed now Governor Mike DeWine leading.  He ended up winning by four points.  Or my favorite: Florida.  In the gubernatorial race, out of 25 surveys taken from October to the election, only five showed Republicans winning.  What is so striking about these surveys is the lack of variance in the results.  Literally, almost every single poll showed then Democratic candidate Andrew Gillum up by a couple points.  One can say the narrowness of the margin shows the polls were only slightly off but if the that was the case and the election was close why did so, so many surveys show Gillum winning?

In polling circles it has become more common for pollsters to either purposefully or inadvertently follow a phenomenon known as herding.  In these instances, Polling Company A will, either in its methodology, sample size or weighting, follow what Polling Company B is and come out with the same result or close.  In some cases, pollsters will not even reveal survey results for fear of the backlash or ridicule.  Now, this is not to say this is happening in the run up to November but poll after poll nationally and in key swing states suggests the potential for one.

In recent years, pollsters have been fond of saying they have taken steps to ensure the accuracy of their surveys.  Many pollsters now sample more whites without college degrees and have made their samples older after weighting for demographics.  For example, Quinnipiac, well-known for its left leaning samples, now says it samples more white voters without college degrees after 2016.  Somehow, this tweak did not prevent them from missing the results in Florida by a whopping 8 points in 2018 nor posting a 13 point lead for Joe Biden in the quintessential swing state.

Now, to be fair, pollsters are dealing with a unique set of factors impacting their industry.  The first is that as technology has become cheaper and more available fewer voters are using landlines – the traditional method of contact for in-person interview surveys.  The second is fewer voters are simply willing to respond to pollsters.  The third is cost – the more complicated and advanced a survey is to reach voters the more it costs to create.  This is actually why so many surveys now are funded by partisan organizations; CNBC (D), PPP (D), Trafalgar (R), etc.  Most non-partisan surveys being run right now are funded out of universities which have the resources and academic inclinations to identify trends in public opinion,

Pollsters are not shy about confronting these criticisms.  For once, they readily admit their surveys could be flawed.  They also admit the media misrepresents their surveys and the fact they could be wrong (instead of every survey every time is right).  The problem, though, is they refuse to admit their own failings and their is a cottage industry of academic and polling institutions which hold their water for them.

Without getting too much into the weeds, nobody does this better than Fivethrityeight.com.  Set up by Nate Silver (who predicted a comfy win for Obama in 2012), the website routinely tells us how in fact, on average, the polling industry is doing pretty well.  I mean, they were only off a few points nationally in 2016 and were fairly close on the generic ballot in 2018.  Of course, this really only applies to about a month before the election.  Not the hundreds of polls prior, which historically, are not very accurate.

Redirecting back to 2020, state level and national polls show Trump clearly trailing.  But, despite their protestations, there is sizable number of voters indicating they are undecided (as I wrote here) and pollsters are not really pushing to find out who these voters will support.  Notably,  there is a significant gap in support for Republican Senators in NC and AZ compared to support for Trump.  Take for example a CNBC (D) survey which showed Senator Thom Tillis trailing 40-52 while Trump and Biden are tied at 47.  Does anybody think Tillis really is at 40 percent?

Despite their issues, PPP (D) actually surveyed the same race last month and found when they pushed undecided voters who support Trump but were undecided in the Senate race, Tillis trailed by a single point.  The same was found in Arizona.

This is not to singularly harp on pollsters.  The media misrepresents surveys and does not do a good job explaining how surveys could be wrong or the issues plaguing the modern day polling industry.  The media, unfortunately, also has carried this over to COVID for damaging repercussions for the nation.

For the average American, it is clear surveys show Trump is trailing and the GOP Senate majority is in danger.  IF THE POLLS ARE RIGHT.  And they are not repeating the mistakes of the past.  Until November we will not know.  But, until then, we should remain weary of buying into survey after survey repeating what others are saying about the election.

 

 

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s