American novelist Samuel Clemens (a/k/a Mark Twain) wrote, “There are three kinds of lies: lies, damned lies, and statistics.” Opinion polls are a variety of statistics and some recent mis-uses of Presidential race opinion polls demonstrate how they can be employed to mislead and misinform. The Newsweek polls have gained substantial notoriety, but there are others, such an invalid Zogby poll touted on that font of enlightenment, the Huffington Post.

First, a review of the Newsweek polls. On June 20, Newsweek released a Presidential-race poll which had Barack Obama leading John McCain by a hefty 15 points. This poll was treated to a veritable festival of exuberance and delight by Newsweek and by the talking heads at MSNBC, among others. Then, on July 11, Newsweek released another effort at polling and to the obvious dismay of Newsweek, MSNBC, et al., it showed Obama leading McCain by only 3 points. The Newsweek/MSNBC folks were caught . . . either the Newsweek polls were flawed, or their favorite candidate was suffering a major drop in popularity in a relatively short time. Do I need to say that the July 11 coverage was much more restrained than the June love-in?

Judge for yourself. The Newsweek story on the June poll, authored by East-coast elitist Michael Hirsch, featured a photo of a beaming, happy Barack Obama; began with the news that “Barack finally has his bounce;” predicted that “Obama may enjoy more than one bounce;” and announced, “Obama is running much stronger at this point in the race than his two most recent Democratic predecessors, [Kerry and Gore], who both failed in their bids to win the White House.” Acknowledging that other national polls had Obama with a 4 to 5 point lead over McCain, Hirsch simply blew past the adverse information, opining that “the latest evidence of [Obama's] gaining ground goes well beyond” the margin of polling error. Heady stuff.

The Newsweek story on the July poll was much more sombre. Jonathan Darman’s piece featured a dark, backlit profile of Obama and began with the admission that “some of Barack Obama’s glow may be fading.” Darman, seemingly befuddled by the turn of events, wrote that “Obama’s overall decline from the last NEWSWEEK Poll, published June 20, is hard to explain.” He continued, “But perhaps most puzzling is how McCain could have gained traction in the past month.” Sigh. Darman did not mention, of course, that the new poll put Obama behind the leads enjoyed by losing candidates Kerry and Gore.

The truth is not really difficult to see. The June poll was an anomaly, probably due to sampling error. Some polls are “outliers,” that is, they lie outside the normal range of results. Statistics for public opinion surveys are calculated to the “95% confidence interval” – - meaning that 95% of the time, the sample’s mean values will fall within the poll’s margin of error. But 5% of the time, those mean values can lie outside this margin and have no relation to what the “true” means are. In addition, the June Newsweek sample may not have been a good one – - Newsweek may have oversampled Democrats or other types likely to be supporting Obama in larger numbers. Both of these factors may have come together to produce an off-the-wall result in the June Newsweek poll.

Unlike the Newsweek writers, the good folks who post at Huffington’s are undaunted by either facts or contradictions. Thus we arrive at Dawn Teo’s dewey-eyed and mildly incoherent piece, Zogby: Obama Leads Arizona; McCain Support Still Dropping. The first word of the title – - Zogby – - gives away the ruse. The Zogby polls, once somewhat respected, are no longer valid. They have been so far off the mark that I predict another colorful usage will emerge from this campaign as “Zogby” gets verbed, thus: “Did you see where McCain got Zogbied on Huffpost?” To “Zogby” meaning to be attacked with an obviously invalid poll result.

To understand why most Zogby polls are invalid and thus aren’t included in the Real Clear Politics tallies, all that is necessary is to read Zogby’s own description of its polling methodology. Zogby takes its numbers by sampling its “online panel” . . . in other words, Zogby uses volunteers who want to be polled! There is even a link at the site where you, too, can volunteer! This is NOT generally considered to be a valid method of taking such polls, which should be based on random samples. The reliability of Zogby polls falls somewhere between the Newsweek variety and the AOL “Straw Poll” which, as I write this, has McCain up by a 73% to 27% score.

While the Zogby poll is demographically weighted, from the sparse description it also appears to be an “All-adult-voter” poll. If you peruse the poll listings on Real Clear Politics, you will notice that there are parenthetical notations beside the poll names; for example, Rasmussen Tracking (LV), AP Ipsos (A), Gallup Tracking (RV). “A” means “Adults;” “RV” means “Registered Voters;” “LV” means “Likely Voters.”
Political scientists usually consider “A”-sample polls to be less reliable. Only about half of the eligible voters actually go to the polls; the participation figures in recent Presidential elections are: 1988 – 50.1%; 1992 – 55.1%; 1996 – 49.1%; 2000 – 51.3%; 2004 – 55.3%. Obviously, it doesn’t matter who non-voters favor when it comes down to election day. The best predictor of future behavior is past behavior and experience has demonstrated that non-voters simply don’t vote, even if they say they will.

There is no clear preference between registered-voter polls and likely-voter polls. Registered-voter polls tend to include some respondents who are non-voters; on the other hand, likely-voter polls tend to exclude recently registered voters. Of the “tracking” polls – - Gallup and Rasmussen – - one uses an LV sample, the other an RV sample. Looking at these two polls together, tempered by reference to other polls, probably gives us the best current indicator of polling trends.

So, if they have so much confidence in Obama, why do the left-leaning media rely so heavily on questionable poll results? Perhaps because the polls reported on Real Clear Politics show a close race beginning to get closer?

As I noted in a previous post, the Presidential preference polls for the first half of July do not accurately predict who will win in November. Today is July 15. The race for the White House begins tomorrow.

My thanks to Paul Zanucci and L Brown for their educational comments on my previous posts.

If you want energy independence . . . JOIN US! Visit American Sentinel

Related posts:

The Polls: Obama’s Lead, McCain’s Advantage

Newsweek Poll – - McCain Closing Fast On Obama?!?

The Newsweek Poll – - Is Someone Cooking The Numbers?

LIES, #@%*$# LIES, AND OPINION POLLS

Recommended links:

Faith Of My Fathers – John McCain’s Service

Wait For The Wind – Expect A September Surprise?