Election 2020: What About Those Polls?

We are about three weeks from Election Day and when I thought 2016 was a unique election, 2020 is a chaotic one.  Every week we have a new headline.  Over the past month, the news cycles has gone from President Trump accused of vituperating against dead soldiers, not paying his fair share of taxes, to being in Walter Reed hospital with Covid-19.    Now we're seeing Trump recovering quickly from Covid-19 with the help of experimental medication, but the next debate canceled.  These are interesting times.

Quickly, I still don't have any changes in how I see the electoral map in terms of electoral votes, but I've become more solid on FL as a state becoming safer for Trump.

I just want to make some comments about the public polling we're seeing currently in this race.  Since Trump's Covid-19 diagnosis as well as his public reactions to it, including a SUV ride to wave at supporters while still at Walter Reed, his public poll numbers have plummeted nationally.  Here are some of the changes we've seen:

• NBC\Wall Street Journal: Biden +8 Sep to Biden +14 Oct;
• Fox News: Biden +5 Sept to Biden +10 Oct;
• CNN: Biden +8 Sept to Biden +16 Oct;

On the face of it, it looks like the bottom is falling out on Trump's support.  In fact, at these spreads, Biden would probably be well on his way to 400 electoral votes.  However, while I can see that Trump's support has taken a hit over the past week through the combination of a negative news cycle of Trump catching Covid-19, his Administration botched communication surrounding it and that Covid-19 is Trump's weakest issue, I remain skeptical the change is that large.   There are several reasons I am skeptical.

• There are several polls that have not seen much of a change at all:
• The Hill/HarrisX: Biden +7 to Biden +5
• Economist/YouGov: Biden +8 to Biden +9
• IBD/TIPP: Biden +6 to Biden +3
• Additionally, the RealClearPolitics average in battleground states (PA/MI/WI/FL/NC/etc) Trump is running 0.5% ahead where he was in 2016.

• If Biden is so far ahead nationally, why is he still traveling to these battleground states and not typically safer GOP states such as Texas and Georgia?

•  In quite a few of these polls, they ask the question who they think will win the election/who their neighbors are voting for.  The results are quite striking:
• Fox News Oct: Trump 48%/Biden 39%
• Economist/YouGov: Biden 40%/Trump 37% (23% unsure)
• Gallup: Trump 56%/ Biden 40%

• Finally, one way to actually measure enthusiasm and trend with real data is to measure voter registration for states that register by party.  Voter registration by party shows the pool of available voters that a candidate should have to turn out (bigger pool=more available votes). Looking at several battleground states, The results are again quite revealing:
• North Carolina (Trump win 2016): 2016: Dems had 646K advantage/ 2020: Dems hold 400K advantage
• Florida (Trump win 2016): 2016 Dems had 330K advantage/2020: Dems hold 136K advantage
• Pennsylvania (Trump win 2016): 2016 Dems had 936K advantage/ 2020: Dems hold 717K advantage.

In summary, while national polls feed a media narrative of a widening Biden lead, voter sentiment of their thoughts of the race, voter registrations and battleground state polling suggest a very competitive race. 

Why would this be the case?  My thought is that the the polls have a wide variance because the methods they use and the distribution of respondents the pollsters receive from those methods.  

• Most polls use live callers as the way to receive responses.  They call folks and if the respondent agrees, have them answer 60+ questions.  While this method was fine 10-15 years ago when smartphone use wasn't as widespread, it is a relatively high cost for the respondent now.  Folks that are most likely to affirmatively participate in such polls generally pay higher attention to politics than the normal person, have time on his/her hands to answer all those questions and/or have adverse incentives to participate (i.e. paid survey taker).   

• Also, in an age where telemarketers are face great skepticism from respondents regarding the respondents privacy, it's not unreasonable to think that the responses would be what they think is the most socially acceptable, not what they really think.  For example, how socially acceptable was it in June to call Black Lives Matter a Marxist group?  Following the news not very acceptable.  Respondents therefore may tell the pollsters they approve. Hence, the poll responses are more of a reflection of what the respondent thinks the pollster wants to hear, not what they really think.  

• Live caller as the sole method also is very biased towards more white collar tilt in their distribution, which would tend to be more urban.  While a 30-ish old professional who keeps on top of the news is motivated to complete the survey, someone like my mother, who has no college education and works long hours, does not think too much of talking to a cold caller for 40 minutes on the phone.  Your truly has marked these cold calls as "Potential Spam" on my phone. Hence, a class bias is generally not accounted for in live caller polls.  

• In my professional life, I worked as a consultant that had to collect individual opinions on specific topics.  To gain those opinions, I used multiple methods of calls/emails/text to gain survey responses.  One way to address the respondent distribution and social intimation biases is to give respondents multiple options to respond with a shorter survey.  Reducing the cost of completing the poll survey will lead to a more inclusive respondent pool and more likely one that accurately reflects the electorate.

So there you have it, my thoughts on how to read the polls these days.  Just like in my professional life, I'm going to look at multiple metrics to assess this race, not just the polls.

Until next time, it is what it is.