First of all, I would like to thank DaTechGuy for inviting me on his radio show. He blogs at DaTechGuyBlog.com and has been one of the few people who has been investigating the numbers behind the numbers in the Presidential polls.
Second of all, I would like to apologize for my Emily Latella moment (if you don’t understand that, google the name and you will find numerous video clips). I asked a candidate running for the U. S. House of Representatives about a state issue. Just for the record, the issue was forced unionization of family run daycare centers. This was on my mind because I had blogged about it earlier in the week (rightwinggranny.com).
Now to my point. One of the guests on DaTechGuyOnDaRadio this morning (if you missed the show, it will be up on DaTechGuy’s blog sometime this week) was Doctor Andrew Smith. Dr. Smith is the Director of the University of New Hampshire Survey Center, which specializes in election polling. Dr. Smith has been involved in polling since the mid 1980’s.
Dr. Smith pointed out that a poll is simply a specialized conversation. His organization randomly selects the people that they poll. There are two main aspects to taking a poll. First, will people talk to you? This can depend on the political situation at the time. If the candidate a person supports is not doing well, that person may be reluctant to talk to a pollster. Second, is the person you are polling being completely honest? Are they simply giving the pollster answers in order to make themselves look good? In many cases if a person is asked if they are likely to vote, they will answer yes because voting is considered a good thing–not because they actually intend to vote.
In polling the public, a pollster needs to ask questions people might actually have the answers to. If a pollster is asking very specific questions about a piece of legislation, the person answering may simply give an answer in order to avoid appearing ignorant.
The key to a successful poll is a random sample and interviewers that exactly follow the script they are given. The wording of questions and the order of questions can determine the answers that will be given.
When evaluating a poll, find out who is doing it and why. It is probably not a good idea to take polls conducted by candidates too seriously.
Most of the time a random sample will be accurate. One reason polls get more accurate as the election nears is that people begin to pay attention. In some cases, voters are not focused on the election until the week before and will give an answer to a pollster a month before that may change when the voter starts paying attention.
Dr. Smith described polling as an art with a scientific method behind it. I think that is a very accurate description.