Are Modern US Polls Equipped To Detect Electoral Changes?

A number of recent surveys and reports of 2020 have raised eyebrows. While national and state level polls show Joe Biden with a steady lead over President Trump less than two months from Election Day, registration figures from a number of states and specific demographic surveys show the topline numbers of surveys might be hiding true voter sentiment.

The first head-raising survey came out of Florida last week when an NBC/Marist survey found Biden and Trump tied. The survey actually showed the President ahead among Latinos in the Sunshine State. For reference, he lost Latinos in 2016, per exit polls (see there issues here) by almost twenty points. Another survey of Latinos in blue Miami-Dade County found Biden leading among the demographic 55 percent to 38 percent, significantly under-performing Clinton.

Biden’s struggles have raised alarms in Democratic circles, not just because of his struggles among the conservative Latino electorate in Florida composed of anti-communist Cuban and Columbian Americans, but because his numbers nationally among Latinos lag Clinton.

It shows in other surveys. A UC Berkeley survey taken in August showed Trump winning 30 percent of the Latino vote in California. While nobody expects Trump to win the state the fact he is running so well among a group considered hostile to him means polls might be missing how deep his support is among the group in other battleground states like Arizona., Texas, Georgia, and North Carolina, where there cumulative votes could determine the winner.

Most national surveys show Biden doing just fine among this group but they often under-sample this demographic in much the same way they under-poll blue-collar and rural voters. For example, in Nevada, a state notoriously hard to poll, a NYT/Siena survey found Trump down by a mere four points (within the MOE). In New Hampshire, a state which has been under-polled this cycle but most analysts assumed Trump would lose he is down by three.

Similarly, as I have written extensively about, virtually every poll shows Joe Biden ahead in the battleground states by varying margins. On the one hand, this is a good sign for the Biden camp it also reveals a warning sign. If a candidate is winning a state by significant margins then yes, every survey should show a lead. But, leads within five points from different pollsters should be a worrying sign the surveys are herding each other and might be missing key unfavorable groups in their cross-tabs.

The media doesn’t hep by buying into the hype of every survey and failing to understand the under-pinning of the results. Virtually every survey has a failing (some worse than others) from election modelling based on 2016 exit polls (inaccurate) to being opt-in surveys (Yougov/CBS) to being overly reliant on over or under-sampling certain demographic groups.

All which might explain why polling results are mirroring each other and not events on the ground. In recent weeks, from Pennsylvania, to Iowa, to Florida, there have been reports of increased GOP voter registrations. All the while polls have shown comfortable leads for Joe Biden. While the media focuses on the minutia of the day to day actions of the campaigns or the tabloid stories of the day (Trump called soldiers “losers” or “downplayed” the virus) they often have little impact on the results of the election. The strategies of the campaigns do.

The GOP has made a significant effort to register wayward Democrats and turn out voters in Minnesota and elsewhere who did not vote in 2016. In Pennsylvania, more voters ID as Republicans in the former Democratic bastion of Cambria County. Sure, the county won’t outweigh urban Philly, but it does not need to combined with other rural counties. These subtle shifts often are undercounted in polling in much the same way late breaking deciders are in shifting a race.

Which begs the question is modern polling even able to accurately capture the varied and shifting opinions of the American electorate. Especially one increasingly divided by class and less by income or race. Indeed, the Trump campaign’s concerted effort to focus on working class issues in its contacts with voters is likely an under-reported reason why an immigration hardline President is doing better with the demographic since GW Bush. This is compounded by the fact the better-off and more urban/suburban voters are more likely to opt-in and respond to surveys.

It does lead one to discount many surveys and question their results, especially in swing states like Iowa, Ohio, and Florida, where pre-election polling in 2018 told us the states would welcome Democratic Governors for the first time in over a decade. Just look how that turned out. Numerous questions remain about whether pollsters have solved their issues. Simply over-sampling blue-collar voters doesn’t solve other issues with regards to divisions along class. In the end, modern day polling is far less trustworthy than the past leading to elections being far more uncertain than in the past. For losing candidates anywhere, that should be music to their ears and a warning to all winning candidates until the actual votes are tallied.


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