Presidential elections are decided by a candidate's ability to garner 270 electoral votes on Election Day. The popular vote total on November 4 is literally useless regarding America's selection of its next President, as the 2000 Election so ably demonstrated. Based on that indisputable fact, in theory the presence of national polls delivered by numerous sources should have little bearing on the candidates' standing as the election winds down.
However, the reality of national polls is quite different than their electoral significance. In many respects, perception is crucial to a political election race. It affects the way swing voters view each candidate. It drives the media narrative regarding which direction each campaign is headed. In other words, it provides a snapshot in time that defines the national position each candidate occupies as the election moves forward.
There are three major poll result providers that most political pundits, media sources, and astute observers follow to obtain an effective gauge as to the state of the 2008 Election: Nate Silver's Fivethirtyeight.com, Pollster.com, and Realclearpolitics.com. All three websites have traditionally been very fair and non-partisan in their reporting and their poll presentations. Their analyses have been based on conclusions drawn from trendlines, polls (and their internals), current narratives, and a host of other salient factors.
Why, then, has the latter organization chosen to remove and disqualify national polls and samples favorable to Democratic Party Presidential Candidate Barack Obama in today's Sunday poll average? In this article, I will examine three factors that are casting a serious shadow upon Realclearpolitics' (which will be referred to as RCP throughout the article) ability to claim a non-biased status in its poll result reporting: poll sample type, poll inclusion/disqualification, and poll favorability.
Poll Sample Type
This section really has more to do with the Gallup Daily Tracking Poll than anything - as it greatly affects the poll numbers being reported by RCP.
Throughout the primary and general election season, Gallup has been reporting results based on samples taken from "Registered Voters". This means the sample is taken from voters who are eligible to vote, with no preference expressed with regards to turnout based on "traditional" or "expanded" likely voter patterns. In other words, this sample type assumes that equal parts of those who have been mobilized to register to vote will show up on Election Day - and that those voters are sampled with that assumption in mind.
Given the overwhelming primary turnout for Barack Obama and the paltry turnout for Republican John McCain, this sample type seemed appropriate and fair - as it gave both men the benefit of the doubt with regards to voter turnout. If popular support for both candidates follows the primary turnout, then the "Likely Voter" samples would be virtually useless because both "expanded" and "traditional" models both make the assumption that turnout will follow one model (2004) or the other (projected 2008). In this case, McCain's turnout during the primaries was so miniscule that an effective gauge of his general election turnout could not be projected with any degree of certainty.
Instead, what Gallup has done (and RCP picked up and run with) is release two "Likely voter" samples that project results based on both the "traditional" and "expanded" models along with the "Registered Voter" sample. As Obama's ground and mobilization forces have overwhelmed McCain's (particularly in battleground states), it is no surprise to an astute observer that Obama has led the "Registered Voter" sample type by 6-11 points during the past several weeks. However, RCP has chosen to include the more McCain-friendly, "traditional" voter sample (Fivethirtyeight and Pollster have not) to water down Obama's perceived support in its average - even as the traditional model assumes a level of support, registration, and turnout that is not consistent with the organization, ground game, and registration models of each campaign.
Poll Inclusion/Disqualification
This component of my analysis relates to RCP's decision to include polling firms like Zogby in its poll averages - while at the same time refusing to include a firm with a lower pollster-introduced error like Research 2000. According to Fivethirtyeight.com, the pollster-introduced error for each firm is as follows:
Zogby +2.16
Research 2000 +1.73
Each firm generally uses representative samples of each political affiliation: Republican, Democratic, and Independent. Each firm uses about 1100-1200 "Likely Voters" in its daily tracking poll. Why, then, does RCP deliberately exclude Research 2000 and include Zogby in its averages?
In short, RCP is cherry-picking the poll that is more favorable to McCain and leaving the poll favorable to Obama out of the average. There is also the matter of the R2K poll being displayed on the Daily Kos home page every day, as though that in and of itself would mean the poll is biased. The fact is, Research 2000 Daily Tracking Poll is conducted by experienced and professional pollsters - and is not influenced by the sponsor in any way. The only plausible reason that R2K is disqualified from RCP's polling average centers on conservative bias against its more Obama-favorable results.
Poll Favorability
This refers to RCP's decision to drop several reputable, Obama-favorable polls released less than seven days ago from its averages. This morning, RCP excluded the following national polls from its posted average:
Pew Research Obama +7
CBS News/NY Times +14
LA Times/Bloomberg +9
Ipsos/McClatchy +9
As a result, the RCP average dropped from 6.5% to 5.0%. To the best of my knowledge, new polls have not been released to replace these polls - which, again, are less than a week old and even include samples taken from days congruent to some of the daily tracking polls. I suspect that these polls would have found their way into the polling averages if they were more favorable to McCain. However, as they show large leads for Obama, they were pushed to the background as though they didn't exist to portray a tightening of the race (even though Fox News pollster Scott Rasmussen shows the race actually widening to a six-point lead for Obama this morning).
As always, these considerations are provided for your digestion and assimilation. They are promulgated as my own observations based on months of analysis of RCP, Fivethirtyeight, and Pollster as related to their polling averages and accompanying analyses. My analysis of RCP is that it is displaying a vested interest in pushing national poll samples and sample types that favor the Republican Candidate for President, John McCain, rather than the non-partisan site it claims to be - and as such, should be viewed with an eye of skepticism as the 2008 Election draws to a close.
http://democratictribune.com/...