An Associated Press-GfK poll this week raised a lot of eyebrows when it showed the Presidential race at Obama 44 percent to 43 percent McCain. “How can that be?” many pondered, since so many other polls had Obama with a much bigger lead, upwards of ten points or more.
As some of my readers know, I am currently working on my Doctorate in Education Leadership at St. Mary’s College of California. A big focus has been on learning how to conduct quantitative and qualitative research, including surveys and polling.
One thing I’ve learned is that no two polls are alike. A lot depends on what questions you ask and on how you ask them. Other factors include how the sample of respondents is chosen.
I raise these issues because of polling discrepancies in the Presidential race, as well as some others. The Minnesota Senate race is a good case in point. With longtime “Saturday Night Live” comedian Al Franken running as a Democrat, it is probably the most talked about statewide race since Arnold Schwarzenegger became Governor of California. Celebrity attracts attention, and attention drives polling. In Minnesota, it’s been all over the map. Complicating matters is independent candidate Dean Barkley, who is consistently polling 15 to 19 percent.
A poll out today from St. Cloud State University has Republican incumbent Senator Norm Coleman up by nine points. But a “Big Ten Battleground” poll taken this past Friday has Franken leading by six points. A Rasmussen poll on Thursday has Franken ahead by four points, but a Survey USA poll from Monday has Coleman ahead by two points.
As I said, the polling is all over the place. Why? Let’s examine a few answers.
“Margin of error” – Look for this in any poll. It’s usually plus or minus 3 to 5 points. Let’s say a poll shows Franken 46 percent to Coleman 42 percent, with a margin of error of 4 points. At its widest, that means Franken could be ahead 50 to 38 percent or, at its closest, that Coleman is actually ahead 46 to 42 percent, given the four-point margin of error.
What questions are asked? – This is crucial. Some polls simply ask: “If the election was held today, for whom would you vote, Al Franken or Norm Coleman.” But if you add information such as “Democrat Al Franken or Republican Norm Coleman,” now you have introduced a potential bias based on political party. The better polls focus on what they call “likely” voters. For example, a person might be asked qualifying questions such as “Are you registered to vote?” and “Did you vote in the last Presidential race?” The idea is to narrow the poll to people who are actually planning to vote, rather than just having people express their candidate preference. With traditionally low voter turnout in the United States, this “narrowing” technique is crucial.
Who gets asked? - I’ve often heard people lament, “No one ever calls to poll me!” It’s a valid point. National surveys might make their prediction based on 1,000 responses in a country with 200 million eligible voters; but if you choose a representative cross section of the public, you can be very accurate. The problems with this are built-in biases. Polling companies often “over poll" in urban areas, leaving rural areas unrepresented. This is why Al Gore was at first declared the winner in Florida back in 2000. Exit polls were heavily skewed to urban areas. When the very rural and heavily Republican panhandle was counted, George Bush pulled ahead. Another big problem nowadays is that many homes have cell phones, but no land-line telephones. Most cell phone numbers are not listed in phone directories, so many of these people never have a chance to get polled.
So when you see a poll talked about on TV or in the newspaper, be a skeptical consumer. The three most important questions are: What is the margin or error? Who was polled? And what were people asked? Most reputable media and polling organizations will report those very issues up front. If they don’t, then doubt their results and call them to complain!
A new poll I conducted shows that 97 percent of Americans simply love the website www.MarkCurtisMedia.com. Just kidding, but check in often!