Data in context: why some tightening in battleground polls shouldn't panic Democrats


Credit: Joe Biden for President, Flicker. License.

Slews of polls continue to be released daily, and every time there is a poll showing the race between Donald Trump and Joe Biden getting closer in key swing states, Democrats have nightmares about 2016 and our stomach churns. Believe me, I get it. It happens to me, too.

But when it happens, I like to crunch the data. When it comes to data, there are two main signs of optimism for Democrats: the makeup of those who are voting now, and Biden's history of overperforming the polls this year throughout the primary.

The most interesting phenomenon in this election is that certain polls are tightening just as people are early-voting and voting by mail in unprecedented numbers. Over 37 million people have already voted nationwide, and of those, the partisan split is 52% Democratic and 25% Republican, with the balance being carried by minor party and independent voters.

That statistics (compiled by the US Election Project), give or take a few points, holds pretty well across the board when comparing data for battleground states that release partisan voting statistics. In Florida, where over 3 million people have already voted, for example, Democratic ballots returned are outpacing Republican ballots returned 48% to 32%, and in North Carolina, by a 45% to 26% margin. Over a quarter of the 1.2 million votes cast in North Carolina have come from voters without a party affiliation. In Pennsylvania, Democratic ballots account for a whopping 72% of about 1 million ballots cast so far, in very early tallies that is not at all likely hold.

In another good sign for Democrats, Black voter turnout is high. In Georgia, 31% of the returned (and accepted) ballots came from Black voters, and in North Carolina, 23%. Black voters comprise 20% of registered voters in North Carolina, and 30% in Georgia. Even in South Carolina, where Democrat Jamie Harrison is taking on Lindsey Graham in the Senate race, Black voters comprise 32% of those who have already voted, 2 points over their share of registered voters. Keep in mind that most states do not report demographic data for those who have voted.

The other sign of optimism - which I addressed in less detail before - is that there is a case to be made that Joe Biden, not Donald Trump, is this year's turnout candidate, and it may well be the Biden-Harris ticket that outperforms polling, not Trump-Pence.

But let's put a fine point on it. The evidence is in, and Biden's show of strength is not anecdotal. From South Carolina on, Biden won almost every competitive race by a margin wider than he was expected to, and won some states he wasn't supposed to. Below is an accounting of states that were either competitive in the Democratic primary or is competitive in November, where the polling was going into the races, and where Biden came out.

I used Fivethirtyeight's polling averages. For Super Tuesday states, I used the latest date on or before the date of the primary that data was available on at least two candidates, and for other states, I used a reference point of where polling stood roughly 4 days before those primaries. Using data closer to the primary day for Super Tuesday was done for obvious reasons: a whole lot of developments happened between South Carolina and Super Tuesday that moved the numbers.

Here is the data on states and margins

  • South Carolina: Primary day 2/29. Polling average 2/25: Biden +7.8. Actual: Biden +29. 

Super Tuesday. Most recent polling average before the race because of changes immediately before Super Tuesday.

  • Texas: Primary day 3/3. Polling average 3/3: Biden -3. Actual: Biden +4.7.
  • Maine: Polling average: Biden -9.4. Actual: Biden +1.
  • Massachusetts: Polling average 3/3: Biden -6.3. Actual: Biden +7. 
  • Minnesota: Polling average 3/3: Biden -8. Actual: Biden +8.
  • North Carolina: Polling average 3/3: Biden +3. Actual: Biden +19.
  • Oklahoma: Polling average: 3/2: Biden +1.3. Actual: Biden +13.
  • Tennessee: Polling average: Biden +0.7. Actual: Biden +17.
  • Virginia: Polling average: Biden +16. Actual: Biden +30.
  • Colorado: Primary day 3/3. Polling average: Biden -21. Actual: Biden -12.

Mini-Super Tuesday: 3/10. Polling averages presented were 4 days out from the primary date.

  • Idaho: Insufficient polling data, but Bernie Sanders won the state in 2016 with almost 80% of the vote. In 2020, Biden beat Sanders by 6 points.
  • Michigan: Polling average: Biden +7.7. Actual: Biden +17.
  • Missouri: Polling average: Biden +20. Actual: Biden +25.
  • Washington state: polling average: Biden -3. Actual: Biden +1.5.

March 17 primaries. Polling averages presented were 4 days out from primary date.

  • Arizona: Polling average: Biden +24.5. Actual: Biden +11.
  • Florida: Polling average: Biden +43. Actual: Biden +40.

April 7 primaries. Polling averages presented were 4 days out from primary date.

  • Wisconsin: Polling average: Biden +15. Actual: Biden +31.

The only competitive state where Biden did significantly more poorly than the polling data indicated was Arizona, but primary polling data wasn't exactly great and the margin of 25 points was always seen as outlandish.

The point of presenting this statistics is that pollsters consistently underestimated Biden's strength (and overestimated that of a populist) for a simple reason: Voters were done, ready to put and end to the process, and weren't particularly eager to entertain pollsters once they made up their minds or pulled the lever.

Something very similar may well be happening. As we saw in the earlier part of this essay, Democrats are dwarfing Republican early voting numbers. More Democrats - and Democratic leaning voters - are done, doing their part, and are not greatly anticipating their next polling call. 

So are the numbers tightening? Yes. But we can't say whether at least part of that tightening is because more and more Biden voters are done voting and are self-selecting out of polling data, skewing those samples. If that is exactly happening we do not know, but it sure seems to have happened during the primary.

Moral of the story: don't panic.

I have as many butterflies in my stomach as anyone, but at the end of the day, worrying won't win this election for us. Working will.