We’ve written about polling before, but we’ve noticed that much of the mainstream press has resorted to questionable methods to downplay the likelihood that President Trump will be re-elected. Perhaps most egregious was a recent CNN Politics article that stated “Former Vice President Joe Biden holds a wide lead over President Donald Trump in the national race for the White House, according to a new CNN poll conducted by SSRS.Biden leads Trump 53% to 42% among registered voters, roughly steady from CNN’s poll in early March.” CNN trumpeted the polling to suggest that Americans were turning against President Trump. A less egregious, but still questionable poll from Monmouth University stated that “Joe Biden holds a 4 point lead over Donald Trump in the race for president.”
So what is wrong with these polls? Let’s start with the CNN Poll that was done by SSRS polling for CNN. The poll was based on 1002 telephone interviews. In the methodology section, SSRS states that of the sample, “35% described themselves as Democrats, 25% described themselves as Republicans, and 41% described themselves as independents or members of another party.” SSRS goes on to state that “the entire sample was weighted to reflect national Census figures for gender, race, age, education, region of country, and telephone usage.” The Monmouth University Poll uses sampling techniques that are quite similar to CNNs. There are two key points to highlight in these assumptions. First, CNN is using a sample that reflects the national census data for all Americans. This is not reflective of the electorate that usually turns out to vote. Second, CNN’s political affiliation results suggest that it’s polling is not representative of the nation.
If we compare the demographic data from the 2016 election with the CNN sampling, we see the wide discrepancies. According to the Pew Research Center, in 2016, 35% of the electorate described themselves as Democrats, 31% Republicans, and 34% Independents. Compare that to the CNN sampling: the percentage of Republicans has fallen by 6% from 31% to 25%. Monmouth uses a weighted sampling that is 26% Republican, 35% Democrats and 38% Independents. Again, the percentage of Republicans has fallen significantly. Our point is that the political affiliation in the CNN and Monmouth samples seem to underestimate the percentage of Republicans.
Another indication that something is amiss is to look at the breakdown by race. According to Monmouth, it used a sample that was 66% white, 12% black, 15% hispanic and 7% asian/other. So how does that compare to 2016? According to Pew, the 2016 electorate was 74% white, 10% black, 10% hispanic, and 6% asian/other. The variances between the 2016 electorate and the Monmouth sample are clearly significant. CNN/SSRS does not share the racial composition of their sample.
It’s certainly possible that the 2020 electorate will look different than the 2016 electorate. For example, in 2012 with Barack Obama on the ticket, the voting electorate was 13% black, 3% higher than in 2016 while the white % was approximately 72%, 2% lower than in 2016. Still, we think it highly unlikely that in 2020 the electorate will be dramatically different than the 2016 electorate. Clues can be seen in CNN’s own poll. 20% of CNN’s sample indicates that they are not enthusiastic or have no opinion about voting in the election, the highest it has ever been in CNN’s historical sample. Similarly, in Monmouth’s poll, they ask whether the sample is certain to vote. 82% indicate yes, but the share of Republicans saying yes is 89% while Democrats are at 84% and Independents are at 76%. To us, this indicates that turnout in 2020 is more likely to resemble the 2016 electorate than the samples being used by CNN and Monmouth.
The net result is that much of the public polling is simply unrealistic. We strongly doubt that CNN is correct that Biden is likely to win by 11 percent. We also suspect that CNN knows this is not the case, but simply wants to appease its viewership to give the impression that Biden is likely to defeat Trump easily. The reality is that 2020 is most likely going to be a repeat of 2016, a very close election that will come down to a small number of states like Wisconsin, Pennsylvania, Michigan, and Florida.
NewEnglandPolitics.com 