A grim goodbye to Gallup’s presidential approval tracking
The end of an 88-year run
Gallup has been asking about Presidential approval since FDR was in office. You can, if so inclined, compare Trump’s current approval to that of Bush Senior, Nixon, or Obama. You can do things like investigate how early presidential approval correlates with midterm gains or losses, or how major events like 9/11 change public opinion of the president.
Or, you could, until yesterday. Gallup confirmed Wednesday morning that they will be ending the 88-year run of their presidential approval tracking, meaning that no future numbers will be produced. We may know that Trump’s approval at day 329 of his second term is historically poor, but we aren’t going to get those updates anymore.
Future decisions by the Trump administration will impact the president’s standing in other polls, but Gallup is out. For their part, Gallup says this is a “strategic shift”, reflecting “research goals and priorities”. They told the Hill that this had nothing to do with pressure from the Trump administration, and was purely an internal decision.
There are many polls in the world, so many that it is actually a problem to figure out which polls to trust. Gallup has been historically a gold-standard point of reference for scholars of the presidency, polling practitioners, and the general public. There’s no number of additional one-off polls run by different organizations with different methods that can make up for the loss of a time series like this. There’s no amount of money that can backfill an 88-year run of the same question.
Gallup has been asking this exact same question for almost as long as statistically significant survey research of the American people has existed. George Gallup conducted the first modern presidential poll using statistical methods in 1936, when he correctly predicted that FDR would beat Alf Landon and remain in the Oval Office. Gallup started asking about presidential approval just two years later.
This one question, “Do you approve or disapprove of the way [Name] is handling his job as president?” has seen survey research adapt from in-person interviews, to landline calls, to cell phone calls, and now to a time where most polling is done via online surveys. It wouldn’t be an overstatement to say that this question, and Gallup’s wealth of data on the answers, represent some of the very best public opinion data available.
For now, it seems like Gallup will continue to run their other public polling and serve as a resource for question wording and benchmarks, like general satisfaction with the country or economic confidence. We’re going to have to cross our fingers that it’s just presidential approval that no longer aligns with priorities (although, if it is, we should probably ask some follow up questions about why).
This isn’t the first polling resource we’ve lost to shifting priorities — some folks may remember the demise of HuffPost Pollster, and 538 was only recently shut down — but this is one of the most surprising and devastating decisions we’ve seen in recent memory.
As researchers, this loss feels very acute, particularly at a time when so many federal datasets are being lost. The erosion of irreplaceable resources leaves us all worse off.
 | A guest post by
|
I get the impression that Gallup is pivoting to more boutique and bespoke polling for specific verticals and, possibly, individual paying customers. I can imagine the money there is much greater for the work required. They have developed a large presence in workplace issues/polling like engagement, etc. It's almost like they are trying to out-Pew Pew.
Presidental approval and similar national-scale polling seems to have become a commodity, and almost a thankless one. And, like any commodity, quality is all over the place. And AI may push down the cost of entry as well as operational costs even further.
Speaking of 538, former head of 538 Nate Silver has been developing tools to try to suss out how high-quality and reliable various polls and pollsters are on these national issues, with mixed results. As a subscriber to his Silver Bulletin, trying to identify and adjust for bias/house effects, herding, sloppiness, and other polling maladies--and toss out really bad players--seems to be a case of Whack-A-Mole.
The other approach is what RealClearPolitics is doing--basically throwing all of their selected polls into a blender and letting the bias and junkiness average out. It sounds crude and has been accused of some right-wing bias, but it seems to work. Where it breaks down is in local races and politicians where the number of polls to throw into the blender is lacking, as is typical with aggregating insufficient data.