Kevin Collins: But let's put that in context. Being 4 percentage points off makes a difference in close elections, to be sure, but it's also relatively small errors in the overall distribution of attitudes. More importantly, it's much smaller errors than the early days of polling which were far more inaccurate.

I, and probably most folks who post publicly about polling, get these kinds of responses periodically, ranging from uninformed through deeply ignorant all the way to "spreading disinformation." But let's take the question seriously. Does polling still work? (Spoiler: yes, yes it does).

My supposition is that polling data isn’t data in the same way horoscopes aren’t data ether. Broken clocks are right twice a day. Everyone needs to find another way of more accurately determining actual voting behavior. @chrislhayes.bsky.social @cnn.com

First, let's start with acknowledging the fact that 2020 had the worst polling errors on average in at least 2 decades, according to the survey research industry-group AAPOR task force's post mortem on 2020 pre-election polls (cite: aapor.org/wp-content/u...)

But let's put that in context. Being 4 percentage points off makes a difference in close elections, to be sure, but it's also relatively small errors in the overall distribution of attitudes. More importantly, it's much smaller errors than the early days of polling which were far more inaccurate.

You might also ask if presidential election polls are off by ~ 4 percentage points, does that mean all other survey research (often on topics other than elections) is broken? No, it does not. @pewresearch.org has put out some excellent research on this specific question

In 2021, they simulated how biased vs. balanced election weighting affects their estimates of various policy supports. The answer: not that much: www.pewresearch.org/methods/2021...

What 2020’s Election Poll Errors Tell Us About the Accuracy of Issue Pollingwww.pewresearch.org Given the errors in 2016 and 2020 election polling, how much should we trust polls that attempt to measure opinions on issues?

They have also compared results from their panel survey to benchmarks *besides* elections (yes, other benchmarks exist). Across a range of benchmarks, polling is still quite accurate: www.pewresearch.org/short-reads/....

Does public opinion polling about issues still work?www.pewresearch.org National polls like the Center’s come within a few percentage points, on average, of benchmarks from high response rate federal surveys.

The other question is "if not polls, then what?" A common answer -- frequently, lately, from the left -- is special election results. But special election results are simply not as good of predictors as polling is, at least for presidential elections.

On this point, I recommend you read @baseballot.bsky.social's excellent rundown here abcnews.go.com/538/2024-pre...

None of this is to say polling is perfect. But the answer to these imperfections is not to dismiss it entirely, but to (a) acknowledge the limitations , and (b) work on improving it.

If I’m reading this right, the 3 biggest polling misses in last 20 years underestimated the incumbent share in their reelection years.

Post