Is polling dead (or dying)?
Posted January 4, 2018 1:02 p.m. EST
Updated January 5, 2018 11:18 a.m. EST
(CNN) — On Wednesday, the best known polling organization in the world made a big change. Gallup announced that it would no longer be doing a daily presidential approval tracking poll.
"Beginning in 2018, Gallup will start updating presidential job approval on a weekly basis, rather than on a daily basis," wrote Gallup chief Frank Newport by way of explanation.
I reached out to CNN's own polling guru, Jennifer Agiesta, for some more context about why Gallup made the decision they did. I also asked a few broader questions about the state of public opinion polling following the 2016 election. Our conversation, conducted via email and lightly edited for flow, is below.
Cillizza: Gallup is abandoning its daily presidential tracking poll. Why?
Agiesta: The economics of telephone polling strike again! Doing the kind of high quality, cell phone intensive, telephone interviews conducted by live people work that Gallup is doing here is just about the most expensive way to do survey research these days, especially when you consider the volume of calls necessary each night to have reliable daily data.
But they're not killing the project off entirely, instead, they are reducing the number of interviews they do each night from 500 to just over 200, for a total of 1,500 per week. This will enable them to continue releasing weekly averages on presidential job approval, with breakdowns on major demographic and political groups on a monthly basis. And it's the same overall number of interviews that went into each of the three-day averages that they released as the "daily" tracker. So it's not a total loss for those of us itchy for new numbers all the time.
The daily tracking numbers on the presidency have long been supported by other Gallup work. After publication, Gallup's editor-in-chief Frank Newport reached out to clarify that the demand for telephone interviewing generally has fallen, making it difficult to maintain the level of interviewing necessary to sustain a daily tracker. The most expensive part of any telephone poll is the cost of getting a respondent on the phone, not the length of the interview, so without a way to sort of split the costs for those initial calls, the approval tracking itself becomes exponentially more expensive.
It's also important to note that Gallup hasn't done daily tracking for every president. They have daily data for the entirety of Barack Obama's tenure, but not for George W. Bush or any president before him. Having weekly numbers on Donald Trump means we'll still have more frequent approval ratings than what we had for any president before Barack Obama.
Cillizza: Polling has taken a series of body blows in recent years, particularly after the 2016 election. Is Gallup's move at all influenced by all the negativity about polling these days?
Agiesta: It doesn't seem to be driven specifically by that negativity, more by a shift in priorities that Gallup set in motion well before the 2016 election. Back in 2012, you may recall, Gallup's daily tracking poll on the presidential race had a final national estimate of Romney 49% to Obama 48%, a bit different from the Obama 51% Romney 47% number the voters produced on November 6. Gallup revamped its methods after that result and stepped away from the horserace polling that's taken the most heat after 2012, so they've suffered a little less from the backlash than have other pollsters who are still in the horserace polling game.
Cillizza: With Gallup out of the daily tracking game, are there any big organizations left doing this sort of work? Why or why not?
Agiesta: Doing the exact same thing? Nope. And no one else was doing it before Gallup started, either. Gallup's daily tracking on presidential approval was the only one with these methods. There is one other pollster releasing a daily approval number for Trump, Rasmussen Reports, but as we've noted elsewhere, their methodology isn't exactly up to par with Gallup's.
That all comes back to the economics of polling. It's CRAZY expensive to do high-quality phone polling, even more expensive to do it every day. We've seen the effects of that as some news organizations have dialed back the amount of polling that they're doing in response to the increased costs, and in the groups that have switched from conducting regular phone polls to those using online interviews or other methods that are less expensive. That trend seems unlikely to change, especially as the share of Americans reachable only by cell phone rises. The latest data on that shows at least 52% of American adults live in homes that are reachable only by cell phones, and two-thirds or more under age 45 fall into that category.
Also, if you're reading, please pick up your phone when a pollster calls you. We really do want your opinions!
Cillizza: How useful -- and accurate -- is a daily poll? What can it tell us -- if anything?
Agiesta: There was value in Gallup's daily numbers, but there was also a lot of noise. If you look at the trend line for Trump's daily approval numbers here and then compare it to the same chart for his weekly numbers here, there's a whole lot more fluctuation in the daily numbers than the weekly ones.
Those two charts are both based on the same pool of interviews, the difference in the trend line is the difference between the stability of a seven-day average vs. a three-day average. By adding a fresh sample of 500 people every night, you get a fairly reliable read on where things stand in reaction to the events of the day. But one night of interviewing can only tell you so much: Your data are only as good as the people you can reach that night. On any given night, there will be people you can't get to -- because they're at a PTA meeting, or went to the movies, or had to work the night shift -- and you hope that it's a random subset of the population. But if it's not, then your data aren't so great until you get another shot at reaching those people.
Cillizza: Finish this sentence: "The decision by Gallup to end its daily poll tells us ________ about the polling industry more broadly." Now, explain.
Agiesta: It tells us that costs are driving the types of research being done by the polling industry more than almost any other factor. Methodological changes and targeted use of phone polling are likely to become more frequent, and for people who follow poll results, it's going to become more difficult to determine which polls to trust. There's some excellent advice on that front from experts you and I both trust here.
And if you, any of our dear readers, want many more sentences on this topic, here's the industry's take on the future of telephone polling in a report from the American Association for Public Opinion Research.
UPDATE: Gallup has clarified their reason for ending the daily poll and that change is reflected above.