Don’t Swear Off Polls, Help Fix The Problems

Defunding polls, reducing personnel and refusing to ask horserace questions won’t solve the problems that election polls face.

Polls got a lot wrong in 2016 ― Brexit and the U.S. presidential election, for example ― and now some people are ready to swear off polling altogether.

A French newspaper says it will no longer pay for public opinion polls as part of its reporting on the country’s upcoming presidential election. The venerable Field Poll in California closed down in late 2016 after nearly 70 years in operation. The future of the long-running poll collaboration between CBS News and The New York Times is up in the air after key departures from the Times. Even before the U.S. presidential election, other media outlets, such as the Columbus Dispatch, drastically reduced the number of polls they conduct.

Polling’s biggest problem might not be errors in election estimates. Errors can be fixed with enough investment and research. The greatest threat to polling is that after 2016, there seems to be less willingness than ever to spend money to improve the state of election polling.

Polls were partially right on the U.S. presidential election ― Democratic nominee Hillary Clinton did win the popular vote ― but across a few critical states, there was a wave of Republican voting that turned the Electoral College to Donald Trump. Polls missed certain types of voters who voted heavily for Trump, which made a Clinton win seem more likely.

Similarly, polls showed that there was a possibility that “leave” could win the Brexit vote in the U.K., but most showed “remain” with the advantage.

Yet there’s nothing that can offer a broad view of a population like a public opinion poll. Le Parisien’s plan to talk to people in bars and factories will provide some public opinion, but what about voters outside of bars and factories? Polls remain the best way gauge how people think everywhere. Certainly polls have their problems and have failed to perfectly measure opinion in some recent elections, but talking only to certain groups won’t measure an electorate accurately, either.

To get an accurate idea of the public’s views, we need more empiricism, not less. Reporters will always have biases, and only certain types of people will be willing to talk to reporters. People are also unwilling to respond to polls, but pollsters who offer anonymity stand a much better chance of overcoming response problems and biases than reporters looking for on-the-record comments.  

To get an accurate idea of the public’s views, we need more empiricism, not less.

Polls won’t disappear if media outlets stop investing in high-quality work. There are plenty of polls. The number of election polls has grown continually over the past few cycles. That’s actually part of the problem: More data isn’t necessarily better. We need more high-quality data, not just more data. It made matters worse when Gallup didn’t release horserace polls for 2016 after missing the outcome in 2012, and when the Pew Research Center refrained from releasing final likely voter horserace estimates in October and November.   

If high-quality pollsters, such as Gallup, Pew, CBS-Times, Columbus Dispatch and, formerly, Field, aren’t in the picture, lower-quality pollsters will be right there waiting to provide numbers and information, hoping to make a name for themselves and generate business in the process. But their polls will be more prone to error. And when there are hundreds of polls all saying the same thing ― as most polls did when they indicated Clinton would win ― it’s easy to develop a false sense of certainty and safety in concluding that that’s what will happen.

Even some high-quality polls missed in the 2016 election. But those pollsters will fix their problems. Fly-by-night pollsters looking for a little press won’t invest in testing and improvement. We need the better polls to stay in business and keep polling elections. And if members of the media want reliable information, they need to invest in polling, not swear it off. Outlets that can’t afford to spend money on high-quality polling can help by reporting on high-quality data instead of amplifying anything and everything regardless of quality.

Election polls aren’t going anywhere. It’s embedded in human nature to want to know what the status of a contest is, and politicians will always need that information to help run their campaigns.

That information is often flawed. But it’s the best we have — and we should keep trying to make it better.