Sept. 24, 2024, 1:18 p.m.
We’re into that fun part of fall when temperatures are slowly dropping, apples are ripening, pumpkin spice everything is everywhere, and football season has hit its stride. And public polls are piling up faster than the leaves in your driveway.
The volume of polls is truly astounding. At one point last week, we had four new Pennsylvania polls released on the same day. FiveThirtyEight's polling database shows 18 presidential polls (national and statewide) on Sept. 18. Just that one day. From my own experience working in poll aggregation, the volume will continue to increase through the end of October.
It wasn’t always this way. As recently as the mid-1990s, public polling was dominated by the major media polls (which are still around) and maybe a few independent polling companies—a few universities, Gallup, and Pew, for example.
But times have changed—we have so many more polls now, due to the confluence of three media and technological advancements all emerging in roughly the same time frame: the 24-hour news cycle, the internet, and cell phones, quickly followed by smartphones.
The 24-hour news cycle kicked in when cable news networks like CNN and Fox came on air in the early and mid-1990s and created a demand for constant feeds of new content, a dramatic change from network news programs that ran for half an hour each night. Not long after that, the internet created the opportunity for even more content. Consumers demonstrated a demand for even more news and opinion, and there was no reason for these new outlets to put a cap on supply the way the morning paper did. Digital media became a whole new branch of the news industry.
Around the same time period, cell phones became cheap enough that more people began using them. Until the 2000s, polling had relied on calling household landlines—you could easily take a random sample of phone numbers, which were tied to geography by area code and exchange, and talk to a relatively representative segment of Americans. People answered their household phones for the most part, too, even though the introduction of caller ID slowed down responses by allowing people to screen unknown callers.
As people began dropping their landlines, however, a complication arose: While pollsters could use automatic dialers to dial a bunch of landline telephone numbers, which allowed them to only use human labor if someone answered, the Federal Communications Commission banned using auto dialers with cell phones. Originally, this was to prevent users from paying for extra minutes of cell phone talk time due to spam calling, but the rule remains in place.
This makes calling cell phones more expensive because you need more interviewers to dial each number by hand. To pile insult on injury, texting and smartphones led to people no longer answering phone calls on their cell phones. That gets us to where we are now with telephone polls: They are expensive, and response rates are low.
Entrepreneurial pollsters worked to leverage the internet’s connectivity and cheap access to millions of people for gathering opinions. Gathering opinions from people on the internet requires some effort to find those people and get them into “panels” (groups of people willing to take polls) that you can survey, but the process was (and is) much cheaper than a telephone poll. That lowered the threshold for new pollsters to get in the game.
On the media side, the demand for constantly refreshed content meant that new pollsters suddenly had an outlet—the gatekeeping of the major networks was gone. There was not even a need for an official media partnership to get attention, as social media became an easy way to transmit information to millions of users.
As the number of polls increased, so did the need to find ways to make sense of them. CNN had long run a “poll of polls” as part of its coverage, but in the early 2000s, RealClearPolitics started averaging polls as a standalone function on its website. And Pollster.com became the first widely cited poll aggregator that used a statistical model to compile the polls.
From there, it was only a short step to Nate Silver’s debut forecast in 2008, which combined a model-based poll aggregate with election “fundamentals” like economic indicators. The success of his model in 2008, and of more models in 2012, fueled the rise of media election forecasting.
And that brings us to where we are today. We have dozens of polls per week, around a dozen poll aggregators, and several forecasts running around the internet, in our pockets, and accessible 24 hours a day. The higher volume doesn’t necessarily mean we know more. It just means there’s more noise.
Contributing editor Natalie Jackson is a vice president at GQR Research.
