It may be a little early for this, but probably not. As the political field begins to form for the 2012 general elections, we can expect to see more and more polling results released favoring one candidate or another for any particular office. And, it will continue to get worse in the weeks and months leading up to Election Day.
With that in mind, I think a short examination of how public-opinion polls can be manipulated to reflect whatever results the organization paying for the poll wants it to.
Over the years, three polling tactics have been proven to be particularly effective by providing misleading results; and, while the first of these could be attributed to sloppy preparation, the second and third can justifiably be considered deliberately deceptive.
The Unrepresentative Sample
The Loaded Question
The Push Poll
Let’s take them in order:
First, the Unrepresentative Sample
An unrepresentative sample is one that does not accurately reflect the population one wishes to survey. As implied by the illustration above, the man answering the door isn’t your typical, er, normal human being. So, if you wanted to know what normal humans (the ones who do not wear bird cages on their heads, for example) thought about something, this fellow would not be the one to ask.
The real-life example that is cited most often is a poll commissioned in 1936 by the Literary Digest regarding the presidential race between Franklin Roosevelt and Alf Landon. The sample names were selected from telephone directories and auto-registration lists. The results indicated that Landon would win easily, and events of the Landon presidency are discussed in civics classes to this day.
As we all know, Alf Landon actually lost to Roosevelt in a landslide, which is why this particular instance of polling error is so well known. Remember where the sample came from? Telephone directories and auto-registration lists. Well, in 1936 few American had either, and those who did were mainly upper-class. But the less-well-off could vote, too, and there were many, many more of them. Literary Digest‘s sample was unrepresentative of the population, and the election results proved it.