Polling is a tricky field but a good poll can be a reliable indicator of public opinion. What is more, polls are all campaign professionals have by way of hard data that can be predictive and proscriptive (Campaign professionals are always on the lookout for both). So what do you do when the polls fail you? Not much.
Take it as a healthy, rejuvenating dose of skepticism and catalogue the event… So here goes:
Polls in July and early August showed the Democratic Senate primary closing, with Andrew Romanoff (D) getting a needed bounce against his opponent, Sen. Michael Bennet (D). Romanoff had erased Bennet’s double-digit lead and the feeling on the ground in Colorado was that Romanoff had the momentum he needed to score an upset victory. A last minute PPP poll had Bennet reclaiming his lead, but the margin of 6 points was so outside the trend line that it could be discounted as an outlier.
Furthering this outlier perception was that same PPP poll’s results in the GOP Senate and governor’s primaries that put Norton over Buck by two points and McInnis over Maes by one. Both results ran counter to months of other polling data that showed Maes and Buck leading. While both races were decided by wire thin margins, and the predictions that those races would be close were dead on, neither Norton nor McInnis had enough steam to beat their outsider challengers. The same could not be said for the Bennet race; he came back from an election week dive in the polls, but his win was no squeaker. Bennet won with an 8.4 percent margin, defying all expectations, including PPP’s.
A similar story held true in Connecticut. In the gubernatorial primaries, Ned Lamont (D) had held a lead over Dan Malloy (D) since Quinnipiac started polling this race in November, 2009. In fact, Lamont had almost a 20-point lead in May and June. That lead was gradually erased over the summer, but even a last minute poll released by Quinnipiac still had Lamont up by 3 points. In the end, the underfunded Malloy won that race by almost 16 points, a trouncing for the pollsters. Some may attribute this win to Malloy’s union support. Just over 20 percent of Connecticut’s 835,000 registered Democrats came out in this primary. In a low turnout race, organized union voters can make a big difference. However, the usual indicators that would show that level of support did not materialize in the Quinnipiac polls.
When the polls closed in Minnesota’s Democratic gubernatorial primary, it seemed as though this would be another story of data-tracking gone wrong. Mark Dayton (D) had never held less than a 10 point lead over both his opponents in this three way race. As the polls closed in Minnesota and returns began to trickle in, Margaret Kelliher (D) began to take the lead. An hour after the polls closed, she had a wide lead. Kelliher had the strongest turnout in the urban areas of the state and she buried Dayton, with almost 50 percent, in crucial Hennepin County. But as the night wore on, Dayton’s returns steadily increased and Kelliher’s stalled. By the end of the night, Dayton had won a narrow 1.2 percent victory over both his opponents.
There is a temptation in politics to cite polls as biblical truth, since there are so few other metrics to test a campaign’s success or failure. But it is easy to forget that there are so few hard, scientific indicators because campaigning is largely an art. And artists are fickle characters that create their own reality and care little for what was supposed to happen.
Noah Rothman is the online editor at C&E. Email him at firstname.lastname@example.org