Are Mobile Phones Skewing 2012 Presidential Polls?
With just 42 days until the 2012 election, Americans are paying closer attention to presidential polling results. Not much has changed since the Republican and Democratic conventions: if you look at a rolling average of the last several months, President Obama has maintained a 3-4 point lead over Gov. Romney (within the statistical margin of error).
Traditionally, polling firms call landlines to get respondents’ voting preferences, but 33% of US households are now mobile-only. Research indicates that mobile-only households are considered to be more likely to vote Democratic, so more pollsters are making sure to include specific percentages of mobile users in each survey.
Before putting together a survey, pollsters must think in multiple dimensions so they can reach a representative sample of the electorate. Each poll includes select percentages of different population groups, such as:
- likely voters vs. registered voters
- Republicans vs. Democrats
- survey response rate
- live interviews vs. robocalls
“Landline vs. mobile phone” is a critical consideration, but because federal law prohibits robocalls to mobile numbers, pollsters must call these numbers to conduct live interviews. Historically, pollsters have faced similar challenges; in the 1936 election between President Franklin Roosevelt and Gov. Alf Landon of Kansas, the largest polling operation was Literary Digest, a weekly newsmagazine. In its election poll, Literary Digest mailed out 10 million postcards to people who were:
- members of specific clubs and organizations
- previously registered to vote
- listed in telephone directories
- registered car owners
After counting the surveys that were returned, Literary Digest reported that Landon would beat Roosevelt 66% to 33%. As it turned out, the 1936 election was one of the most lopsided ever — Roosevelt won 98% of the Electoral College and 60.8% of the popular vote! Did the magazine get it so wrong simply because they oversampled people with telephones (40% of the population)?
Not at all. Literary Digest sent out 10 million surveys, but only 2.3 million were returned, a dramatic nonresponse bias that wasn’t recognized until after Election Day. Additionally, the survey’s sample revealed a selection bias (magazine subscribers, car owners and telephone customers) that helped skew the results toward a more affluent crowd more likely to vote Republican.
Literary Digest folded not long after releasing its 1936 survey; more scientific polling outfits like Gallup, Roper and others soon became leaders in measuring public opinion. As we become more connected and data becomes easier to gather and crunch, it’s a safe bet that polling methodology will grow even more complex; for all we know, SMS could become the next major tool for measuring public opinion!
Comments
Powered by Facebook Comments