Presumptive Democratic presidential nominee Joe Biden has a clear advantage in the national polls.
Perhaps not wanting to acknowledge Biden’s edge, some Republicans, including in President Donald Trump’s own circle, claim that the polls are oversampling Democrats and undersampling Republicans. Often, they point to the 2016 exit polls.
An in-depth look at the current high-quality national telephone polls reveals it’s quite unlikely the polls are inaccurate because of an undersampling of Republicans. The current polls have the same relative partisan makeup as earlier this year when the race was tighter and as the accurate final national 2016 polls.
Biden is leading not because the polls are undersampling Republicans, but because he simply has more support than Trump does nationally.
I looked at six high-profile national pollsters who polled after the third presidential debate in 2016 and have polled since June for the 2020 presidential race. These six pollsters included ABC News/Washington Post, CNN, Fox News, Marist College, Monmouth University and NBC News/Wall Street Journal.
View Trump and Biden head-to-head polling
On average, these pollsters had Biden up by 11 points over Trump among registered voters.
If one insisted on weighting these pollsters’ results to the party identification in the 2016 exit polls, Biden would still be up 10 points. (Weighting essentially means adjusting what percentage an individual or group makes up of the poll to what you believe the population as a whole looks like.)
In any analysis I do, however, I try to make an apples-to-apples comparison between pre-election polls from 2016 and pre-election polls from 2020. (See my analyses on Trump doing better among Black and Hispanic voters.) There is nothing special about the 2016 exit polls that pollsters should be weighting to them.
So let’s look again at the six aforementioned pollsters. Where I could, I took each pollster’s registered voter final result from 2016, and the final likely voter result when a registered voter result was not released.
Among these six pollsters, the average showed Democrats making up six points more of voters than Republicans in 2016. At the same time, they had Hillary Clinton defeating Trump by four points in the national vote. She, of course, won the popular vote by a little bit more than two points. That is, the polls were quite accurate.
An average of the latest surveys from each of these pollsters in 2020 shows that Democrats make up the same six points more of voters than Republicans. In other words, the polls have the same percentage of Republicans relative to Democrats as in 2016, and yet the polls have Biden doing seven points better than Clinton did in the final 2016 polls.
When you dig a little deeper you see why Biden is doing better than Clinton. He’s stronger among Democrats, Republicans and independents. It’s not that the percentage they’re making up in the polls is significantly different.
Take a look at the latest Marist College poll, which happened to be the most accurate of the bunch in 2016.
Biden has a 92-point lead among Democrats, a three-point lead among independents and trails by 84 points among Republicans. This is something we see consistently in the polls: Biden leads by a larger margin among Democrats than Trump does among Republicans, and he also carries independents.
Back in 2016, it was a different story. Marist’s final poll had Clinton up by 81 points among Democrats, trailing by 90 points among Republicans and trailing by 7 points among independents.
Likewise, Biden’s advantage has expanded over the last few months because he’s doing better among each group. In an average of these pollsters (save Marist who didn’t conduct a poll during that period that asked about the presidential race), Biden was up five points in late March and early April. The average of these polls had Democrats making up the same six points more of voters than Republicans as they do now.
The last thing I’ll note is that that party identification is an opinion, not a demographic. Voters can change their mind about which party they identify with, which is why most pollsters don’t weight by party identification like they do age, education, gender and race for example. There’s no magical figure to weight to. It’s conceivable that Trump’s poor job performance as rated by voters could have caused fewer voters to identify as Republicans now than four years ago.
Some pollsters, though, do weight their results by party registration or which party’s primaries a voter casts her or his ballots in. This statistic isn’t a current attitude but rather a measurable fact.
Last month’s New York Times/Siena College polls did exactly that. They found Biden up 14 points nationally and by nine points on average in the six closest states Trump won in 2016. Late last year, they had Biden by two points in these same battleground states.
Put another way, the poll that does weight for partisanship in a proper way was one of Trump’s worst polls this summer.
