Did The Polling Mess Up That Bad?

With 61 days to go before Inauguration, a look at what happened with polling as some of the post-election dust settles

a Lincoln Park Strategies feature

Special insights heard daily at #RealityCheck on WURD every Mon - Thur, 10am - 1pm ET LIVE at wurdradio.com or facebook.com/forwurd (11.18.20 edition). Wear your B|E. Support theBEnote w/ small & optionalmonthly/annual subscriptions(and get exclusives)!

306 electoral votes and a popular vote win by more than 6 million votes (current estimate) is not a bad night, so why did we feel so bad?

There are a few ideas. First: the way the numbers came in were not exactly showing a clear cut win for Biden right away. This was to be expected given how many states had to wait until Election Day to start counting. Also, as was noted in our last email before November 3rd, the most likely outcome was a close Biden win. Which is what happened. Regardless, it still didn't feel great. 

Second: even if the numbers didn't point to a blow out win for Biden, many of us were hoping it would happen. It didn't, and for many, this left a bad taste. 

Clearly the polling was not exactly spot on this year (again) - but the "polling is dead" reaction is probably a bit premature. It is going to take a while to really figure out what happened, but here is what we do know:

  1. The national polling was certainly more favorable to Biden than what ultimately happened, but not by a ton. On average Biden had a 8 point lead nationally and he is likely to end up ahead by around 5 points. Perfect? No. Wildly off? No.

  2. State surveys were a mixed bag. MN, AZ, GA, NC were all very accurate. WI, OH, IA not even close. 

  3. Biden did not over-perform in the polls in any competitive state. 

  4. Senate and House polling was pretty terrible.

  5. Sen. Susan Collins (R-ME) is indestructible.

If polling methodology was wrong, it would be wrong everywhere. This is not excusing the states that were missed, but it seems likely that the misses were due to unexpected turnout patterns more than methodology flaws. We will have to wait until the voter files are updated to know the answer, but as an example, in Michigan, it is estimated that 700,000 voters turned out who did not vote in 2018 or 2016 but were registered at the time. The fact that Congressional polling was so off also points to turnout as the main driver, but we will need to wait to find out. 

In the meantime, below are a few articles that do a good job of going through the numbers and sharing their views on what happened. You can also listen here to an initial wrap up on this subject Stefan Hankin did with Justin Wallen on WURD's Reality Check with Charles Ellison.

What Went Wrong With Polling? Some Early Theories

New York Times' pollster Nate Cohn says that the 2020 polling error was quite similar to the 2016 polling error despite pollsters' adjustments after 2016. To account for this new error, Cohn claims that either the initial diagnosis of the 2016 problem was misguided, or that there are new factors that have arisen since 2016 that make doing survey research even more difficult. Cohn favors the latter diagnosis, but acknowledges we must wait for more data and research untill we can know for sure. READ MORE …

One Pollster’s Explanation For Why The Polls Got It Wrong

After an interview with Democratic pollster David Shor, a Vox writer's main take away was that, "The kind of people who answer polls are really weird, and it's ruining polling." READ MORE …

‘A Black Eye.” Why Political Polling Missed The Mark. Again.

In depth read by the New York Times' David Leonhardt with a brief history of polling in the US and initial conclusions about this year's polling error. READ MORE …

powered by …