Would someone please explain why we have polls on the upcoming election that say Obama's lead is as high as 13% while two very reliable polls, released in the last 24 hours, that say his lead is 1%? (A statistical dead heat!) Is this wide range of numbers as odd to you as it is to me? I am not into conspiracy theories, as I have said numerous times, but something seems amiss when the range is so vast. And why do we have so many poll anyway? What purpose do they really serve? Every one of them is based on various bits of data directly rooted in the questions that are being asked and when they are being asked. None can say who will finally vote.
Does anyone remember that the polls in 2004 said that John Kerry was up by 10% in most of the battleground states only to have him loose these states and the nation, 51% to 48%. I am not an expert on polling but I do know that these numbers are so widely divergent that I would still not be totally surprised if they are proved wrong. The race seems very volatile right now. Remember, they all say that there is a 2.5 to 3% variation. If this is true then the race is much closer than many are leading us to believe. I do not know the answer to this question so I ask anyone who has a credible response or a serious thought: What do these polls really mean right now?
Comments are closed.
My Latest Book!
Use Promo code UNITY for 40% discount!
The value in a ton of polls is balance. Any on the extremes are probably wrong. If you have 10 polls and one says 1% and one says 13% then both are probably outliers and the real value is somewhere in the middle. If we only have one poll then we don’t know if it is an outlier or if it is closer to reality.
When interpreting poll results, keep in mind that there are two types of error. Sampling error, which is reflected by the “margin of error” usually reported with poll results, reflects the variation that arises from random sampling. This is easy to assess and diminishes as the sample size grows. The other kind, which is called “nonsampling error,” lumps together all other sources of variation and bias –inadequacy of the sampling frame (the master list of telephone exchanges from which phone numbers are drawn), peoples’ unwillingness to respond or unavailability when calls are made, biases introduced by the way questions are asked, etc. Most surveys have no measures of nonsampling error. So the usual “margin of error” grossly underestimates the actual uncertainty.
In pre-election polling, the big source of uncertainty is, “Who will actually vote on election day?” Many who say that they intend to vote never do. “Registered voters” are very different from “likely voters.” And there are different ways to assess who the “likely voters” are. We can build models for turnout based on what happened in previous elections, but it’s fair to say that this year’s election is so different that we really have no idea how accurate these preelection polls are. Even the exit polls on election day, which were quite accurate before 2004, are now suspect. So, the short answer is, we really don’t know much.
Much of the discrepancy has to do with how many Democrats and how many Republicans you are polling. Rasmussen for example polls a certain percent more Democrats than Republicans based upon earlier predictions of elections.
The crazy thing in some of these polls is the difference gap between McCain and Obama is exactly the difference betweeen the number of Democrats and Republicans one is polling.
If this is close to right and I think it is it seems like Gallup and a few others are intentionally polling a much greater number of Democrats than Republicans and I do have to ask why?
In the end, the polls mean very little since they are skewed in several different ways. One also wonders how much the Dow Jones and the market is being skewed by several different factors at this time? Do people really still believe the country we live in is really a level playing field?
JLS hit the nail on the head. “Registered voters are very different from likely voters.” Everyone who has commented agrees the polls mean very little. The liberal press might actually fear seeing Obama’s numbers too high for fear the Democrats will believe they “have it in the bag” and stay home. Who is likely to show up at the polls? Traditionally, Republicans have a better percentage of turnout than Democrats. Also, what else is on states’ ballots? Here in California there is a huge proposition to define marriage as one man to one woman. No matter how hard the liberals work to get out the vote, this proposition will bring out an even higher percentage of conservatives than usual. Will it be enough to turn the state to McCain? I doubt it. But I think the point is that you just can’t be sure who’s going to turn out on election day.