It is often mentioned that the media has an interest in making the Presidential campaign closer than it is, or at least closer than it was. If a five point lead is accurately described as auguring a massive landslide, it could make things seem inevitable and the public would lose interest. Thus we get storylines about how Obama’s lead was “only” x points. It is hard to tell how much effect this differential between reality and media framing shapes subsequent reality.
However, what is not mentioned is how pollsters may have an interest in making the Presidential campaign closer than it is. When polls show a landslide, there is less interest from those sources (media or campaign) that fund polling activity. Thus if you are a polling company, you also have an incentive for closer results (granted, offset by the need to maintain credibility).
In this campaign, there are a few areas where polling decisions would tend to bias results in favor of McCain. The first is the use of likely voter samples rather than random samples of all eligible voters. These likely voter models are complex and allow a lot of subjective decision making enter into the process with fairly significant shifts in outcome. At present, many of these models favor types of voters more likely to vote for John McCain.
A second area is based on type of phone connection. Many pollsters only call land-lines or underpoll cell phones. Given the increasing trend of people who only have cell-phones, and the different characteristics of those that do, this also can greatly effect polling results.
Pollsters often make efforts to address all of these issues, but these efforts necessarily add in ambiguity relative to simple, large random samples. Given that there may be an incentive to make this election seem closer than it is, and the issues that would allow for sleight of hand in producing polling results, is it out of the question to suspect that the polling trends that seem to differ with the underlying fundamentals and the shocking Democratic results in Congressional races is the result of malfeasance?
Update: I should also add that campaign ground organization is something that is also typically unrepresented in polling this far out, is typically very effective, and is massively in Obama’s favor.
- VWI
