Harry Enten notices that there are many fewer polls than there were during the first 8 months of the Trump administration and analyzes how this may be distorting our view of Biden’s success and failure:
The polling industry has taken its shots following the 2016 and 2020 election when former President Donald Trump was underestimated.It’s not surprising. Polling can help shape a narrative about where the country is and where it is going. When polling is wrong, people can be misled and the stories being told about the electorate may have been flat out incorrect.
But therein lies why it is so important that we have good polling: it’s one of the few ways we know whether anecdotes people hear are a true representation of the public at large. It shapes not just election coverage, but also strategies for dealing with areas such as getting more people vaccinated against Covid-19.
Unfortunately, it seems that we do have fewer polls right now to help us do that during President Joe Biden‘s administration compared with the Trump administration.
We can see this by looking at the number of polls asking about Biden’s job approval rating. There have been about 750 of those polls published, so it’s not like we don’t have any data.
A closer look reveals, however, how inflated that number is. More than 60% of those are from Morning Consult and Rasmussen Reports alone. The latter has a record of overestimating Republican candidates, and neither meets CNN polling standards.
Four years ago, there were about 1,070 polls released at this point on Trump’s approval rating. That’s 300 more approval polls than we’ve had about Biden at this point. The two pollsters who made up the most number of polls in this average were Ipsos and YouGov, and they accounted for just a little bit more than 40% of all the surveys about Trump’s approval rating. This is significantly less than the two top pollsters are making up of all the polls today.
The fact that we have a smaller number of pollsters making up the lion’s share of polls coming out puts them in a powerful position. When that occurs, they can drive the narrative of where the country is standing on issues to a greater degree. If they’re accurate, this isn’t a big problem. If these pollsters are not accurate, the public and the press may be misled.Judging whether a poll is accurate or not isn’t so easy.
One way to do this is to see how close a pollster’s final surveys are to the election results like FiveThirtyEight does. Using FiveThirtyEight’s polling scores (one of many ways we can judge accuracy), we can see that there are fewer pollsters who have done well in the past that are publishing polls now.
By my count, there have been a little more than 40 polls published about Biden’s approval rating that have come from companies rated some form of an A (i.e. A+, A, A- or A/B). At this point four years ago, there were a little less than 70 polls rated in this top tier.When you expand out to pollsters rated a B+, the difference is even more stark. There were about 540 polls rated in the A or B+ category at this point in the Trump administration. There have been only about 140 at this point in the Biden administration. In 2017, this was driven a lot by the non-partisan Gallup tracking poll and YouGov. Gallup no longer publishes a tracking poll.
Of course for some people, the most important thing about polling is it tells you who is up and who is down in the horse race. The media often makes judgments about races from the polls, and plenty of people decide who and how to donate to candidates.
The first tea leaves of what 2022 will look like come from the generic congressional ballot. It was one of the first signs in 2017 that Trump’s unpopularity was transferring to House Republicans.There were nearly 200 generic congressional ballot polls published at this point in the 2018 cycle. Many of those did belong to the Ipsos tracking poll. Eliminating those, you still have about 90 generic ballot polls.This year there have been only a little more than 60 congressional ballot polls published.
Again though, it’s not just the number. It’s the quality of the poll.Using the same FiveThirtyEight system, there were 29 polls through early September 2017 that were rated somewhere in the A range. Through early September 2021, there have been only 4 polls in the A range.
A lot of the generic ballot polls produced also come from pollsters who poll for candidates, such as Change Research and McLaughlin & Associates. While most of these polls released aren’t for candidates and may very well be accurate representations of the electorate at this point, it’s something to keep in mind for both this data and for what types of polls may be published in the lead up to next year’s midterms.
In total, there have only been about 35 generic ballot polls published by non-partisan groups this cycle. The majority of those come from YouGov. Last cycle, excluding the Ipsos tracking poll, there were double the number of generic ballot polls from non-partisan groups at this point.
I assume that the polls are correctly showing trends. But Enten makes a good point that the least reliable polls are doing the most polling right now which isn’t good at all. And I have to wonder why. They polled Trump constantly … and the polling never really changed. Now we’re getting sporadic polling and Biden looks like he’s cratering in the polls. Huh…