Revisiting the Polls vs. Horse Race Narrative

Republicans are freaking out. Literally. Polls are showing Biden is set to win a bigger percentage of the popular vote than Clinton did in 1996 (and that was with a major third party candidate on the ballot). In Florida, Quinnipiac (these guys house effects bounce around more than CNN’s) finds Biden is set to outperform the the 2.5 percent margin the last seven Presidential elections in the state have been decided on… by 400%. Monmouth surveyed voters in Pennsylvania and found they were they were set to give Joe Biden a bigger margin than they gave Obama in 2008. Nationally, pollsters from Rasmussen to ABC to Fox to CNN have shown Trump has 90 percent plus approval ratings with Republicans but only getting 76 percent of their vote. For reference, he got 86 percent in 2016 and his approval ratings show he should improve on these numbers.

Yet, the horse-race analysis suggests something different. In Florida, Republicans have cut the Democratic voter registration in half. In Pennsylvania, the difference is even starker. Since 2016, Democrats have lost (not gained, net lost) 41,924 voters in the state. Republicans, in contrast, have gained 158,445 voters. That means Republicans have gained 200K voters in the state. These numbers follow earlier number pre-COVID numbers back in March the GOP was gaining on Democrats in Colorado and Iowa before Democrats called it quits on in-person voter registration canvassing. In North Carolina, Republicans added a net 83,785 voters between this March’s presidential primary and the final week of September, while Democrats added 38,137 and other voters jumped 100,256. During the same period in 2016, Republicans added 54,157 registrants, Democrats added 38,931 and others 140,868.

To put these numbers in perspective, Trump won Florida by over 112K votes and Pennsylvania by 78K votes in 2016. In both states, the GOP voter registration surge has eclipsed his total winning margins. In PA, by over 150 percent. In Minnesota, the Iron Range seemingly has firmly moved into the supporting Trump column with six Democratic Mayors endorsing the President last month.

So what gives? Well, this brings us back to the horse-race vs. polling narrative. Though FiveThirtyEight says differently (and of course they have no monetary incentive to do so), the polling industry to struggling to keep up with demographics, technology, voter response rates, and how elections have rapidly become far more cultural. Hence, pollsters have to make assumptions about turnout and the electorate upon which their polls are built. They then weight turnout on multiple lines. But, if that model is off, that determines the election.

By every polling measure Biden is on cruise control. But, if you look at the horse-race narrative you would assume Trump is winning. Voter registration numbers, the economy, unemployment, Gallup’s survey saying 56 percent of Americans feel better off than four years ago, the enthusiasm at Trump rallies, etc. Yet, this election is set to be a massive blow-out.

This, again, goes back to what do you believe. The horse-race or the polls? Both industries, and they are industries, have an incentive to pretend the race is close. The media which drives the horse-race narrative needs to keep interest in the contest so they tell us how close it is. Pollsters want their numbers to be right so they maintain public trust.

One has to be wrong and the other right. Either the assumptions the polls are built on are wrong (and it would be a massive, pervasive bias toward Democrats far surpassing 2014) or the horse-race narrative is off.

It probably is not that simple. Rather, for example, registration numbers could be continuations of trends we saw in 2016 whereby Republican registration gains in swing states are largely being driven by older, blue-collar white registered Democrats who have voted Republican for a long, long time. Or suburban Independents consistently now vote Democratic but just have not changed their affiliation.

On the polls, perhaps it is what Sean Trende described when he said “Who are these people that approve of the job he’s doing (in reference to Trump’s job approval ratings being higher than his numbers against Biden but aren’t going to vote for him?” Perhaps they are people who are going to end up voting for the president, no matter what they may say now. “There are people who like his policies, but hate his persona. But it could also be that these are voters who are saying they’re undecided, but they really aren’t.” In other words, the shy Trump voter. Or, that the polls simply are not hitting the geographic concentration of Trump’s support.

These polling misses or shy responses would not mean the entire polling model is wrong but rather the massive lead it seems Biden has is not as strong as it appears. For some perspective to prove the point, the same Rasmussen poll cited above gives Biden his twelve point lead largely because he carries 20 percent of Republicans. Take that way and the lead drops to 3 -5 points. or it could even be the bandwagon effect whereby a focus on polling based coverage tends to discourage supporters of one candidate while enthusing the others or driving some voters to adjust their response at the time to be part of the winning team (but once in the ballot box they return to form).

The albatross which cannot be colored over is COVID. Trump has bungled his response, at least from a PR perspective. Biden isn’t really proposing to do anything different than the President. Rather, he just is more empathetic and sounds nice. Which works for about six months – just ask Obama in 2009 and 2010.

So, which do you believe? A close race indicated by the horse-race analysis or the polling narrative of an FDR sized victory for Biden? Whatever you believe, none of us will probably even know by the end of the night of the election.


Leave a Reply

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

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

Facebook photo

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

Connecting to %s