All Polls Are Not Created Equal

Polling is a science, but not in the traditional sense of the systematic examination of a body of facts or truths. Rather the science of polling deals with the techniques you use to get the most accurate results possible.

There are many aspects to con-ducting a political poll, but the following are the three most critical to producing a sound product:

• accurate sample
• good voter screen
• carefully phrased questions

A good sample must include persons from all demographic and geographic sections. In recent years, most pollsters have been using a method called random digit dialing, in which the area code and phone exchange are selected. The computer then randomly makes up the last four digits to complete the phone number.

The size of the sample renders the margin of error. Most polls have a confidence level of 95 percent which means that if all persons in that demographic and geographic profile were surveyed, 19 out of 20 would have the same response.

Demographic and geographic characteristics are not enough. Interviewing just anyone who happens to answer the phone does not give an accurate picture of voter attitudes. A campaign only wants
to know what’s on the mind of those who are likely to vote. Those who don’t vote will not influence the outcome of the election so knowledge of their opinions is of no help in winning the election.

The methods for screening likely voters vary among pollsters and depend upon the specific election. For example, many voters will vote in a presidential general election only. So, if you are polling for a primary election in a non-presidential year, your voter screen for likely voters would not be simply, "Did you vote in the last election?" Rather, it might be something like, "Did you vote in the 1996 primary and the general election?"

Asking just one screening question is not adequate. Good pollsters continue with other questions to screen voters. They will ask the voter the likelihood of his voting in the specific election and perhaps a question or two to ascertain if the voter is an informed voter or not-- maybe who his representative is or where he votes. Whatever, a good pollster sets a criteria that must be met before proceeding with the survey.

We all know that a child can ask the same thing a dozen ways. It is the same with polling. Questions must be unbiased. Slanting a question in an improper way might get you the results you want to see,
but the results would be erroneous.
And it is imperative that the person asking the question does not reflect his own view when asking the question.

Complete accuracy in polling is not possible; you just try to get as close to the truth as you can. That’s why you see so many polls on the same topic that come up with such different results. Whether it’s for politics or the launch of a new product, following the rules is the key to getting accurate results.
They may not be the response you want; they will be the response you need.