RALEIGH – Early in the 2008 political cycle, I wrote about the proliferation of polling operations in North Carolina. Over the spring and summer, the numbers continued to proliferate. By some accounts, North Carolina was the most-polled battleground state in the nation, given its competitive races for president, U.S. Senate, and governor. By late October, 15 separate survey firms or organizations had polled the presidential race. Many had also polled down the ballot.
So, how did the pollsters do?
Pretty well, according to the aggregate data. The Pollster.com final average of surveys in the presidential race yielded a .4 percent margin for Barack Obama. That turned out to be a sound prediction of Obama’s slim margin of victory here. In the gubernatorial race, the final average yielded a 1.7 percent advantage for Beverly Perdue over Pat McCrory. The actual result was a 3-point Perdue victory. Only in the Senate race did the polling results differ substantially from the final outcome, by underestimating Kay Hagan’s 8.5 percent win with a final average of 4.1 percent.
These are simple averages. Obviously, some pollsters generated likely voter models that more closely matched the actual electorate than did their competitors, as you can see from the respective lists.
Going into the 2008 cycle, two Raleigh-based polling outfits, Public Policy Polling and the Civitas Institute, faced some important questions. Both were affiliated with a political party (PPP is explicitly Democratic) or ideology (Civitas is conservative). Would their affiliations bias their survey findings? PPP also made use of interactive voice response rather than live-person interviews, as did two other national pollsters with a significant presence in North Carolina (Survey USA and Scott Rasmussen, a former chairman of the John Locke Foundation). Would automated dialing sacrifice accuracy for economy, subverting the ability of a pollster to construct an accurate model of the electorate?
The answer to both questions, I’m happy to observe, is no. The final surveys from PPP and Civitas had Obama at +1, outperforming most other surveys. In the governor’s race, Civitas’ +2 Perdue was a bit closer to the final than PPP’s +1, but in the Senate race it was PPP’s +7 Hagan that nearly nailed the outcome, vs. Civitas’ +2. It’s worth noting that Survey USA also posted an excellent record this year, and while Scott’s state surveys missed the mark a bit, his national projection of the presidential popular vote was almost precisely the final outcome.
In short, IVR technology has proven its worth. And while you may disagree with the political leanings of the individuals involved, PPP on the left and Civitas on the right now have a proven record of clarity, accuracy, and honesty in measuring voter opinion in North Carolina.
Hood is president of the John Locke Foundation.