A news release touting the most recent poll results probably focuses the attention of the presidential campaign participants like nothing else. The poll results are used with great skill by one side of the other as evidence of their candidate's strengths or the other guy's weaknesses. But it's not confined to the campaigns. Anyone who follows the campaigns with any regularity is glued to it, too.
But polls aren't always accurate. Walter Hickey in Almost Everyone Misinterprets The Polling On The Presidential Election — Here's Why gives us a peek into the making of the sausage with an explanation of some of the impairments to poll accuracy.
Briefly, they are as follows:
- Margin of Error - All polls have a margin of error, and it can be significant. For example a politician polling 48% with a 4% margin of error places the real result somewhere within 44% and 52%. However, Mr. Hickey notes, the error is more likely to be on the small end of that 4% rather than the large end. What's missing from Mr. Hickey's discussion is how the pollster arrives at the stated margin of error. It always seems to be 4%.
- Political bias -- The pollster may have a partisan slant which gets built into the poll. While those poll results might be suspect, they can often be wrong so consistently that they can provide useful information to those willing to factor that in.
- Statistical bias -- The pollster may pull a set number of phone numbers from a directory, and the political leaning of the people answering the phone is random. But as everyone who has looked at a red/blue map knows, some regions contain more residents who lean in a particular political direction. So maybe the locale has more people committed to one or the other of the parties. Or maybe the person answering the phone prefers to call him/herself an independent. Maybe the sample wasn't large enough. Or my favorite, more people inclined to vote Republican look at the caller ID and don't answer the phone.
- Sampling -- The pollees may be contacted by land line or cell phone. And each of those methods can yield results that skew toward a particular political party. So the good pollsters will use a hybrid and include in their results only those responders who say they intend to vote.
Finally, Mr. Hickey touches on but doesn't expand on a topic that should be very significant. The main stream media had an ongoing agenda to minimize the Tea Party effect, and Tea Party rallies don't occur that much any more. Many still see only Democrats and Republican with the assumption that Tea Party fans have been absorbed into the Republican party. To some degree, that's right. And to that degree, there are two elements in the Republican party.
Many of the old school Republicans can be distinguished from Democrats only by the targets of their largess. However, there are millions of voters out there who are very concerned about the monstrous national debt and the huge annual deficits. While many of those people will vote for a Republican as the only viable alternative, they wouldn't necessarily identify themselves as Republicans, especially to pollsters.