Poll “Analysis” Shenanigans

UPDATE: Newsweek continues to lead with whatever data most favors Kerry. Like the previous post debate poll analysis piece they've decided to use registered voters without Nader this time referring to it euphamistically as the "national horse race". While the gap in likely voters has narrowed in this poll, characterizing this race as "getting even closer" when Bush leads or is tied in every other poll is a bit of a stretch.

UPDATE: CNN's decided to follow Newsweek's lead and fudge the numbers. In today's poll analyis piece they're calling the race "still tight" despite there being an 8 point gap in likely voters. This is because today, like the Newsweek piece mentioned below, they've sneakily decided to use "all voters" because this measure has Kerry closer. Compare this to their previous poll piece where they call the poll for Kerry despite only having a 1 point lead among likely voters. Apparently likely voters are good enough for Kerry, if they show him ahead.

UPDATE: Regression Analysis of polling data and an interesting interpretation.



UPDATE: Kaus and Patio Pundit also noticed Newsweek's curious poll analysis.

Even though it was anticipated weeks in advance of the debates, for the last few days even I, who was convinced the race was over after the CBS fiasco, was beginning to buy into the “Comeback Kerry” storyline actively forwarded by the mainstream media. Every news story I read before and after the debates had the same message that Kerry had closed the gap on Bush and had the “Big Mo” to eventually take the lead. Then I saw these charts HERE and HERE which hardly support the media’s conclusions. While polls are not perfect, the data doesn’t lie. Today I happened to come across the headline on Drudge that reports that the most recent Newsweek poll has Bush up 6 among likely voters – the measure universally accepted as the most accurate reflection of future voter turnout. Then I went to Newsweek and came across this headline “Too Close to Call: With the debates behind them, the contenders in the race for the White House remain locked in a dead heat in the latest NEWSWEEK poll”. Now it’s been a few years since I took stats, but with a margin of error of +/-4, 6 points is hardly “too close to call”.

What’s going on here?

Newsweek based the article on a measure “of all voters” instead of the more accurate likely voters that has the margin between Bush and Kerry 4 points closer. Why are they now using the “all voters” measure when it is less accurate, not emphasized by other news organization polls, and not used in every previous Newsweek poll analysis? The most innocent explanation is that the news media wants every election to be a close race to the finish and will report data that supports this. If both sides believe they have a chance of winning, they have an engaged audience on both sides. While this is probably a factor in this election as well, how likely is it that Newsweek and other mainstream media sources would be concerned about reporting it as a close race if Kerry was up 6 among likely voters? Not very.

Why is this a big deal?

Kerry’s base is substantially less committed to Kerry than Bush’s supporters are to him. Overwhelmingly, their vote is more an act of opposition to Bush than support of Kerry. This makes Kerry’s base particularly vulnerable to disinvestment. If at some point it became clear that a Kerry presidency was highly unlikely, voter turnout for Kerry would be extremely depressed. This appears to be CW within the mainstream media and they appear determined to do their part to make sure it doesn’t happen.