Think back to last fall’s election, when many analysts – especially Republicans – were amazed that President Obama won the popular vote by almost 4 percent over Republican Mitt Romney, a 51-47 percent margin that was fully three percent higher than the final average of nine major polls, from Gallup to Public Policy Polling.
Even the most accurate of those surveys – the ABC/Washington Post poll and Pew Research – understated the Obama margin by a full percentage point.
The explanation has now arrived, courtesy of California Secretary of State Debra Bowen. It is that the vast majority of newly registered voters actually cast ballots. That’s something to think about as new polls begin to appear, now that the 2014 election season has informally begun.
The turnout among new voters confounded many pollsters, who often change their survey models in the final weeks before a major election to concentrate on what they call “likely voters.”
When some polls stuck to the registered model late last September and into early October, Republicans complained loudly of alleged bias among pollsters, saying that polls based on registered voters overstated the Democratic vote and undershot the number of Republicans who would turn out. For everyone knows, they said, that higher percentages of Republican registrants than Democrats come out for elections.
They even set up a Web site titled “Unskewedpolls.com” which took poll results and changed them to give Romney an eight-point edge one month before Election Day. (After the election, Unskewedpolls.com headlined that it was foiled by “revenge and racism.”)
Historically, the Republican presumptions are correct. They are more correct, though, in primary elections and special elections than in presidential votes, where the historical average turnout in California – which represents a decent cross-section of American voters – runs to about 79 percent of all registered voters, about 30 percent more than typically vote in primaries.
But something different went on last fall, and it suggests that pollsters will need to change.
Two factors altered things markedly: heavy registration of new Latino voters (26 percent more registered to vote last year than four years earlier) and online voter registration.
No major polling organization figured those new voters into sampling calculations when they began, late in the election cycle, to separate likely voters from the general pool of registrants.
Registrants have usually been judged as likely voters based on their prior turnout records and their levels of enthusiasm about current candidates.
But information gathered by Bowen, California’s top election official, suggests that new registrants in 2012 were more likely to vote than the average veteran voter.
In Orange County, for example, 82 percent of those who registered online after it became possible in mid-September actually voted. That’s a spectacular showing for persons who are not longtime, habitual voters that pollsters would usually classify as “likely.” Plus, 81 percent of new registrants who used the old-fashioned paper methods to sign up actually turned out. Equally spectacular and unexpected by the so-called experts. And much better than the overall performance of all registrants, only 67 percent of whom cast ballots.
In Santa Barbara County, the numbers were exactly equal among those newly signed up online and on paper, at 86 percent compared with an 81 percent overall turnout.
Those numbers were pretty much matched all around the state. And since almost two-thirds of those signing up for the first time last year registered Democratic, it turns out the polls were skewed to make the Republican turnout look significantly better than it actually was, not the other way around, as GOP activists and officials charged. So much for Unskewedpolls.com.
In a post-election talk at the Sacramento Press Club, California Field Poll director Mark DiCamillo, whose surveys turned out far more accurately than many of his national counterparts’, observed that the results square with human nature: If someone makes the effort to register in the waning weeks of a campaign, chances are it’s because he or she is deeply committed to some candidate or issue.
Polls disregarding those individuals, as every national measure of so-called likely voters did last year, will always be wrong, always behind the latest developments.
Which means that pollsters, like everyone else, must adjust to cope with today’s brave new online world – or else they will become obsolete.