After the 2020 election there were a lot of declarations that polling was dead, permanently biased in favor of Democratic candidates, and especially obsolete for smaller district level races.
We never thought that was true. Our polling had been very good in 2017, 2018, and 2019 and it was good again in 2021. 2020 was the exception to the rule, not the new normal.
But in 2022 we were put to the test as one of the most prolific internal pollsters for Democratic candidates across the country. We polled in 58 competitive US House races and 11 Senate races.
Here’s how we did:
-Across all the House districts our polls came within an average of 2.7 points of the final outcomes. To put that accuracy in context, 538 has found the average error in House polls since 1998 is 6.3 points, so we cut that in less than half.
In the closest races our polls were even more accurate- in ones decided by 10 points or less our average error was 2.4 points and in ones decided by 5 points or less it was 2.1 points.
Contrary to the belief that polls will be biased in favor of Democrats forever our House polls actually had a slight- 0.8 points- bias in favor of Republicans this year. We had 27 districts where Democrats outperformed our polls, 21 where Republicans outperformed our polls, and 10 where we had the final margin exactly correct. We are comfortable with that being so well distributed rather than all off in one direction!
-Across all the Senate races we polled our average error was 2.3 points. 538 has found the average error in Senate polls over the last 25 years is 5.4 points so we cut that in less than half as well.
In 6 states Democrats outperformed our polls, in 4 states Republicans did, and in one we had the exact margin. On average our polls had a 1 point bias in favor of GOP candidates- almost half of that coming from Washington alone where we had Patty Murray up by 10 but she ended up winning by 15! (Finding her up 10 was enough to reassure our clients not to waste resources ‘saving her’ when she didn’t need saving!)
So we feel great about how things went this year and are optimistic for both PPP and the polling industry moving forward. We were very accurate at the district level, we didn’t overestimate Republicans, all these ‘truths’ about polling that emerged after 2020 were debunked.