|
 |
 |
|
The flawed Gallup poll
Most political analysts would agree that if one poll gives different results than most others it is probably wrong. This is surely true of the recent Gallup Poll showing Mitt Romney with a seven point lead over Barack Obama.
You don’t have to compare the Gallup poll with any other survey to expose its flaws. A check of the internal consistency of this poll yields such implausible results that its findings are almost certain to be wrong. The key to this analysis is comparing results for registered voters and for likely voters.
Most poll readers assume that there is large falloff between registered voters and people who vote in presidential elections. In truth, most non-voters are not registered. The overwhelmingly majority of registered votes participate in presidential contests.
This close match between registered and actual voters suggests that all pollsters may be making a serious mistake in attempting to screen out the 10 percent of “unlikely voters.” This winnowing process may produce more errors than it resolves. A comparison between the two groups surveyed also provides a means for internally checking the plausibility of a poll.
According to the Gallup poll of October 18, Romney led by 1 percentage point among registered voters: 48% to 47%, but his lead ballooned to 7 percentage points among likely voters: 52% to 45%. By simple mathematics, under the assumption of continued 90 percent turnout, we can compute how large Obama’s lead must be among registered but unlikely voters to produce such results.
Simple algebra reveals Obama’s percentage among registered but unlikely voters according to Gallup results:
47% = 45% * .9 + % among unlikely voters *.1 = 65%
Thus, Obama’s support among registered but unlikely voters is 65%, 20 percentage points higher than his support among likely voters.
Simple algebra likewise shows Romney’s percentage among registered but unlikely voters.
48% = 52% * .9 + % among unlikely voters *.1 = 12%
Thus Romney’s support among unlikely but registered voters is an absurd 12%, 40 percentage points lower than his vote among registered but likely voters. These results also show that Obama’s lead among registered but unlikely voters is an equally implausible 53 percentage points: 65% to 12%.
Thus something is fundamentally wrong with the Gallup Poll’s screening for likely voters that produces internally inconsistent results, separate and apart from any sampling error. Unfortunately, even such flawed polls become part of the narrative of the campaign. Stories about polls dominate campaign coverage because they write so easily. Journalists can read polls and write stories about it without leaving their beds in the morning or even picking up a phone. And pundits can expound endlessly about polls since they can never be proven wrong or right. There is no external reality against which to measure polls and no one bothers to check their internal consistency.
Poll-driven journalism is not just meaningless. It is also pernicious because it detracts from coverage of what this election means for the American people. How many stories, for example, have we seen analyzing the consequences of overturning Roe v. Wade and returning abortion regulation to the states? This is a real possibility of Romney is elected. Such a fundamental change in social policy would profoundly change American life for both women and men, but you would never know it from today’s journalistic priorities.
Lichtman is a professor of History at American University.
Most Viewed RSS Feed »
