The collapse of opinion polling: Speculating on consequences

The New York Times asks "What's the matter with polling?" Public opinion polling is in a crisis, because few people are now willing to respond to phone surveys (participation rates have fallen from 80% forty years ago to 8% today) and pollsters may not use auto-dialers for mobile phones. This has made public opinion polling more expensive (so there is less of it) and less reliable, which in turn accounts for the big surprises -- to the media at least -- in the 2014 mid-terms in the United States, the Conservative Party romp in the United Kingdom, and Benjamin Netanyahu's strong re-election.

The Times, unsurprisingly, does not climb out of its echo chamber to notice that all of these were failures in forecasting right-wing victories, a circumstance that might lead to social rather than technical explanations for the failure of polling. It may be that conservatives no longer trust that pollsters will use their information honestly, so they just do not participate. The NYT's story gives no indication whether participation rates have declined similarly across the range of political beliefs. It would be surprising if they had.

We might well speculate about the consequence of all of this for our "national conversation," never mind presidential politics in the next 16 months.

In American presidential elections, ineffective polling will raise the cost of political campaigns even more, because politicians will not be able to target their spending so precisely. As the now unknown margin for error rises, there will be more battleground states. Since Florida taught us 15 years ago that every electoral vote is sacred, neither party may know enough to risk shifting money out of the "almost battlegrounds." While this will broaden American democracy and force candidates to the center, it will also be expensive and the agenda-setters in the media will decry it as another sad example of "money politics."

Political reporters will need to become more creative, because endless stories about the polling horse race will no longer be important or even slightly interesting. The only question will be whether the media admits this enough to move these stories off the front page.

Assertions by activists that this or that position in their favor is supported by the polling deserve essentially debilitating skepticism. The only question is how quickly the politicians who will have to decide such matters realize that issue polls are of rapidly declining importance to their own fortunes. The media, which funds a lot of these polls, will mock politicians who simply say that they do not believe the polls, even if the politicians are right.

The failure of issue-polling may eventually mean that we will again more reliant on the judgment of our elected officials and therefore more a republic than a democracy, at least insofar as we ordinary citizens are in a position to affect the regulatory state at all. Let us hope that this leads to a more thoughtful political class.

The media and the academic social scientists will try to claim that public opinion surveys are still useful, but the failure of election polls will make them look silly for saying so. Social scientists in particular will have to come up with new tools for their scholarship, such as it is. Expect to see a big push from academics to mandate even deeper questioning by the Census Bureau and other devices for extracting opinions from the citizenry.

Watch for new technologies that purport to measure public opinion more directly, including analytics from Google and Facebook, which are probably in a position to exploit an opportunity here, even if at great risk to the charge that they are actually affecting American political outcomes.

Release the hounds.