Thursday, July 31, 2008

Gallup's Poll Controversy

In the past few days, many Obama supporters seem to have been a little down. The most recent Gallup poll shows McCain leading Obama among likely voters. I have written about this before and cited how registered voters are the most dependant voters who we can poll. Likely voters very rarely turn up, are less excited and are also very likely to be more fickle on their voting decision. Luckily, those of us who read polls constantly know that Gallup is one of the worst pollsters out there and this recent poll is a glaring example of pollsters trying to falsely tighten the race.

Gallup was asked how they determined if a voters was a "likely voter".

"As for how "likely voters" were identified, USA Today reports that respondents were asked "how much thought they had given the election, how often they voted in the past and whether they plan to vote this fall." Fair enough. But the very next sentence raises even more questions about whether USA Today's effort is actually a snapshot of the electorate, as its website claims, or enters the realm of forward-looking hypothesizing. Buried in the ninth paragraph of USA Today's own write-up, they reveal that "McCain's gains came because there was an even number of likely voters from each party. Last month, the Democrats had an 11-point edge."
Abramowitz says this contradiction is the equivalent of polling malpractice. "It is simply not plausible that there would be an 11-point swing in party ID among likely voters or that there is now an even split in the likely electorate between Republicans and Democrats," he wrote in an email to the Huffington Post."

So, Gallup claims that voter's are identifying with the GOP more this month than last. Maybe, but as Abramowitz points out, 11 points is impossible for the beleaguered party. There hasn't been a single poll out this season that suggests party ID is anywhere close to even. So, what Gallup did was search and search until they found the number that created an even split.

From fivethirtyeight,

"How do you get from a 47-44 Obama lead among RVs to a 49-45 McCain
lead among LVs?A few quick calculations shows how. You have 900 RVs and 791 LVs, so that means that among your 109 UVs (that's unlikely voters according to
Gallup) Obama leads McCain by a whopping 61% to 7%. Putting it another way,
according to Gallup 16% of registered Obama supporters are unlikely to vote
compared with only 2% of registered McCain supporters."


Keeping the race close is what the media and pollsters want. I don't think that they're trying to "get" Obama, I just think that elections are the a tremendous source of revenue, especially for pollsters like Gallup who usually don't get any attention from anyone in the off season. It is a bit scary though. The polls drive the media narratives and those narratives can set the tone for the campaigns.

Gallup Chief Frank Newport admits that this was a little fishy.

So sure, "under a scenario" where McCain's voters are energized at a level equal to Obama's and the national distribution of party ID is equal between Democrats and Republicans, perhaps it would make sense to see McCain with a four-point lead in a poll with a plus/minus 4 percent margin of error. But engineering coverage of a poll with metrics contrived to show results under a certain "scenario" sounds more prospective and hypothetical than the paper's stated mission of covering polls as momentary snapshots and "not forecasts of far-off election days." As Newport said on MSNBC this morning: "The likely voters simply tell us that turnout could make a difference."



Yeah, but polls are not supposed to be based on a series of "what-if" scenarios. I'm done with Gallup until they apologize or run a story admitting their errors.

My Gallup-esque poll question: If McCain was 46 and Obama was 71, who would you vote for?

Ridiculous, but not too far from what Gallup did.

1 comments:

Anonymous said...

I think the Gallup organization is in trouble. There is always the Quinnipiac poll where Grandpa Hogan went to college.