Q&A of the Day – Are polls being used to attempt to suppress Trump voters?
Each day I’ll feature a listener question that’s been submitted by one of these methods.
Email: brianmudd@iheartmedia.com
Twitter: @brianmuddradio
Facebook: Brian Mudd https://www.facebook.com/brian.mudd1
Today’s entry: I’m starting to believe polls really are being used to attempt to suppress voters. No one in their right mind would think Joe Biden is up by double-digits nationally let alone in states Trump won! My question is how easy is to conduct these polls and get them reported? How much money are media companies putting into these bogus polls?
Bottom Line: Pollsters. For four years they’ve been the most loathed element of the political reporting process for most Trump supporters. And for generally earned reasons based on 2016’s election outcome, however as I’ve pointed out – the national polls were more or less correct. Context is always key. Yes, most pollsters' oversampled support for Hillary Clinton but not to the extent most think. The RealClearPolitics average of national polls showed Hillary Clinton winning by 3.3%. She won the popular vote by 2.1% - thus the average national pollster was only off by around 1.2%. The bigger issues came with the state polling. Here are the states where the average polling incorrectly showed Clinton winning in 2016:
- Michigan
- Pennsylvania
- Wisconsin
They should be ground zero for where errors in polling may take place again. The average pollster missed those states by 4.5%. And guess what, the polling averages once again show Biden with leads in all three. Recent history would seem to suggest that if pollsters are set to blow it again – those three states would be good places to start. But does it really rise to the level of conspiracy? One of the oversimplifications is that the media entity sponsored polls, a la ABC/Washington Post are actually conducted by them. They’re not, they’re the ones who pay to commission the polls. They hire accredited firms to do so. There are numerous accredited firms. Some which poll nationally, many more which operate regionally or locally. As a result, the cost of accredited polls has a huge variance in cost. Here’s an estimate of the cost of a poll:
- Local political polls: $2,500
- Statewide polls: $10,000
- National polls: $40,000
The cheapest possible accredited poll costs $600. That would be a single question local poll. It’s also possible to spend over $100,000 on a national poll, using callout as opposed to automated services. In the grand scheme of the money thrown around during election cycles, the barrier to entry for media outlets is low. That’s led to the proliferation of polling over the years. Now, having addressed part of your question – back to the theory of voter suppression through polls...
There’s no evidence of the intent to engage in voter suppression through polling. Additionally, since accredited firms are hired by those sponsoring most polls, were any evidence of this to surface, it’d literally put the accredited firms out of business. That’s not to say it isn’t possible, just that the risk for those who’d go along with the scheme would be about as high as it gets. The most likely explanation is how poor the samples and modeling are within many of these polls. Take yesterday’s NBC/Marist Florida poll showing Biden with a 4-point lead. Here’s their methodology:
Samples were combined and balanced to reflect the 2017 American Community Survey 5-year estimates for age, gender, income, race, and region. Results are statistically significant within ±3.6 percentage points. There are 1,001 registered voters. The results for this subset are statistically significant within ±3.9 percentage points. There are 743 likely voters defined by a probability turnout model which determines the likelihood respondents will participate in the November 2020 election based upon their chance of vote, interest in the election, and past election.
Notice any potential points of failure? Here are three:
- Balancing samples against 2017 estimates
- Use of registered votes as part of the sample as opposed to exclusively likely voters
- Assigning a 3.9% margin of error when only 743 likely voters were sampled in a state in which over 7 million have already voted
Each potential point of failure introduces increased odds for the poll to be meaningfully wrong. Margin of error rates are predicated on the notion that they’ve essentially gotten everything else right in the poll. As I say, if the premise of anything is false, anything built on it will be too. All of this is to say that just about all of the polling non-sense is explainable without conspiracy. I do think it’s worth questioning the validity of the accreditation agencies for polling but that’s for a different day. Now could there be an effort by some in the news media to attempt to use polling numbers themselves to suppress? Absolutely and that’s the most likely dynamic in play here. Consider this from just the past two days.
Range in national polls:
- Rasmussen: Trump +1 (margin of error 2.5%)
- CNN: Biden +12 (margin of error 3.8%)
Which direction has news media steered the conversation? Have you been made to feel by most news media that Donald Trump might be set to perform better nationally than four years ago, or that he’s potentially going to lose in a landslide? Also, the example helps illustrate my point about the absurdity of polling samples and margin of errors. In CNN’s poll the best possible performance by Trump is a greater than 8-point loss. In Rasmussen, the worst possible performance for Trump would be a loss of 1.5% - which is still better than the 2.1% loss by Trump in the national popular vote four years ago. They can’t both be right. So, which is it? That’s where news media discretion often kicks into play. It’s understandable when a news network cites their own paid for polling but it’s interesting when they commonly lean on the polling of even competing media outlets over those like Rasmussen without media affiliation. That’s not a conspiracy, that’s just reality.
Photo Credit: Getty Images