History says be skeptical of presidential polls
Published: October 5, 2012 at 9:06 am
Given my work and interests, it’s not surprising that I’m a big time political junkie. And when it comes to the presidential contest, and hot Senate, House of Representatives and gubernatorial races across the nation, I eat up the latest polls.
But when it comes to poll reading, one has to be cautious and discerning. It turns out that not all polls are created equal. Polling is not the straightforward statistical gauge of opinion and prediction that many assume.
So, when it comes to the latest take on who is ahead — President Barack Obama or former Gov. Mitt Romney — a big difference exists between surveys.
For example, a poll of adults might be interesting as a gauge of how the general populace views the two candidates. But since 1972, voter turnout has ranged between49 percent and 58 percent of the voting age population. Therefore, as a gauge for how the election is going, polls of adults are useless because a huge chunk of adults simply do not vote.
Then there are polls of registered voters. But this runs into a similar problem. That is, all registered voters don’t vote. Based on data from The American Presidency Project, during the 1996, 2000 and 2004 presidential elections, voter turnout as a share of registered voters ranged between 66 percent and 70 percent (data is incomplete, and no data was available for 2008, but the basic point is clear).
The most valuable polls tend to be surveys of likely voters. Still, challenges lurk here as well. For example, pollsters have to consider the history of turnout by party and by a wide array of demographics, as well as gauging the current atmosphere and attitudes, in order to get a sample of likely voters that hopefully will fit the estimated make up of actual voters on Election Day.
In the end, of course, polls are tested by actual election results.
Both Gallup and The American Presidency project provide handy looks at polling and results from previous presidential elections. Often, the final polls were pretty close to the final outcomes, but there have been far too many cases of polling being notably off the mark, as well as instances of elections changing rather dramatically in the final weeks.
In 2008, the final Gallup poll of likely voters put Obama at 53 percent and John McCain at 42 percent. Obama won 52.6 percent to 46.0 percent. Spot on with Obama, but underestimating McCain by four points.
Back in 1940, the last poll (of adults) on Oct. 31 had President Franklin Roosevelt leading Republican Wendell Willkie by a mere47 percent to 45 percent. FDR wound up winning big, by 54.7 percent to 44.8 percent. A similar phenomenon occurred in 1944, with a Nov. 3 poll putting FDR up by only 48 percent to 47 percent versus Thomas Dewey. FDR again won by a rather handy 53.4 percent to 45.9 percent.
The registered vs. likely voter polls lesson was on display in 2004.
Gallup’s last poll of registered voters had John Kerry at 48 percent and President George W. Bush at46 percent. But among likely voters, it was Bush 49 percent and Kerry47 percent. Bush won re-election by 50.7 percent to 48.3 percent.
Still, an exception crops up. In 1992, Gallup’s last poll of registered voters turned out to be a better estimate than its likely voter poll. Gallup had registered voters breaking 43 percent for Bill Clinton, 34 percent of President George H.W. Bush, and 16 percent for Ross Perot. The final tally was Clinton 43.0 percent, Bush 37.4 percent, and Perot 18.9 percent. The likely voter poll put Clinton up at 49 percent.
As for elections shifting in the late stages, in 2000, an Oct. 6 poll had George W. Bush at 48 percent and Al Gore at 41 percent, and as late as Oct. 26, Bush was up among likely voters in the Gallup poll by 52 percent to 39 percent. Again, the final vote was Gore 48.4 percent and Bush 47.9 percent.
Most dramatic, on Oct. 27, 1980, Gallup had President Jimmy Carter beating Ronald Reagan by 45 percent to39 percent. Of course, Reagan won by 50.7 percent to 41.0 percent.
And in 1976, the Gallup poll on Oct. 30 had President Gerald Ford at
47 percent and Jimmy Carter at 46 percent. Carter won 50.1 percent to 48.0 percent.
Of course, it also must be mentioned that from March onward in 1948, Republican Thomas Dewey never trailed President Harry Truman in the polls (of adults). Dewey lost the election 49.5 percent to 45.1 percent. At the same time, though, few historical observers point out that Dewey’s position in the polls had been steadily falling, from leading Truman 49 percent to 36 percent in late August to a lead of 45 percent to 41 percent in late October. It turned out to be the Truman surge at the end that surprised seemingly everyone, including headline writers at the Chicago Tribune declaring “Dewey Defeats Truman.”
The big lesson? Don’t be surprised if things diverge, perhaps quite markedly, on Election Day compared to what even the most recent polls communicate, never mind polls taken a month away from when people actually cast their ballots.
— Raymond J. Keating, chief economist, Small Business & Entrepreneurship Council.