Election will test accuracy of polls
California has been one of the nation’s most heavily polled states in this election cycle. In the final two weeks of the campaign, at least eight different nonpartisan public polling organizations have surveyed the state -- some more than once -- along with uncounted numbers of polls commissioned by candidates, parties and interest groups.
The nonpartisan public polls all have shown Democrats Jerry Brown and Barbara Boxer leading in the races for governor and U.S. Senate, respectively. In the race for governor, the surveys have put Brown’s lead over Whitman at between six and 13 points among likely voters. In the Senate race, the various surveys have pegged Boxer’s lead over Fiorina at between two points and nine points.
But what if all the polls this year are wrong?
Normally, that doesn’t happen. For all the skepticism that some voters have about polls -- and the oft-expressed view by candidates that “the only poll that counts is the one on election day” -- polling in the United States has a very good track record for accuracy. That’s particularly the case when numerous polling organizations using different methodologies all come up with results that point in the same direction. Add to that the fact that “Election Day” in California started weeks ago, with about half the likely voters expected to cast their ballots by mail before Tuesday, meaning that late come-from-behind surges by any candidate are more difficult than they once were.
But some Republican analysts have argued that this year is different. Polls are underestimating the number of Republican “surge” voters -- people who don’t vote in every election but will vote this year -- and overestimating the number of Democrats, they say. Recently, two polling analysts, Nate Silver on the 538 blog and Sean Trende at Real Clear Politics, have analyzed that possibility.
The argument involves how pollsters estimate which voters really will turn out to cast ballots. Generally, about 90% of registered voters tell pollsters they plan to vote, but experience shows that many fewer do. In California, about 55% of registered voters have voted in recent midterm elections.
To screen for likely voters, pollsters ask a series of questions. Each polling organization handles that screening a bit differently. In the Los Angeles Times/USC poll, for example, registered voters were asked how certain they were to vote, how enthusiastic they were about voting and whether they had voted in the last two elections.
Currently, Democrats have a 13-point edge over Republicans among registered voters in California -- 44% of the state’s registered voters are Democrats, 31% Republicans. The screening questions in The Times/USC poll generated a pool of likely voters in which that edge shrank to four points -- 44% Democrats and 40% Republicans. The Field Poll, released late last week, assumed a five-point gap (44%-39%).
Those likely voter pools are both much more Republican than the electorate at large. Even so, both of those polls showed Boxer winning the Senate race by eight points. Those who believe that the polls are under-estimating Republican turnout argue that the likely voter pools are still not Republican-heavy enough. If that’s the case, the election would look more like 1994, when Republican turnout was noticeably higher than normal. One big difference between now and 1994: That year, Republicans started out with 37% of the registered voters, contrasted with today’s 31%.
So, is a Republican victory possible tonight? Of course. But it will require a GOP turnout considerably higher than pollsters have forecast.
-- David Lauter
PHOTO: A woman arrives to vote at a polling place on Venice Beach in Los Angeles, California, November 2, 2010. Credit: Reuters
Photos: California heads to the polls
Photos: The nation heads to the polls