Taking polls with a grain of salt

[babylonsister]

Posted with permission.

http://maddowblog.msnbc.msn.com/_news/2012/03/14/10686541-taking-polls-with-a-grain-of-salt

Taking polls with a grain of salt
By Steve Benen
-
Wed Mar 14, 2012 1:00 PM EDT


Much of the political world gasped this week when the latest New York Times/CBS News poll found President Obama's approval rating down sharply over the last month, dropping from 50% to 41% for no apparent reason. Though the same poll still showed Obama leading the Republican field, the nine-point drop in support was one of the largest drops any president has seen in the poll in the last 15 years, and reversed a general positive trend for the president.

In many circles, Democrats had to resist the urge to panic, while Republicans started walking with a spring in their step. Both, however, were asking the same question: what happened to cause such a decline?

The answer, in all likelihood, is nothing.

As of this morning, I've seen seven national polls released this week: Pew Research Center, Reuters, Gallup, Daily Kos/SEIU, Bloomberg, Washington Post/ABC News, and the aforementioned New York Times/CBS News survey. I put together a chart showing Obama's approval/disapproval in each of the seven:



As you'll notice, there's a fair amount of variety here. The president's approval rating varies from 51% to 41%, while his disapproval rating goes from 43% to 50%.

Average these seven polls together, and what do you get? 47.3% approval, 47.1 disapproval.

As Ezra Klein noted this morning, when it comes to "the mystery of Obama's plummeting poll numbers, I'm not yet convinced there's any mystery here at all.... {U}ntil more polls show a significant drop for the president, I'm inclined to write this off as noise."

I agree. If all seven of this week's polls showed the same thing, it'd be far easier to take the results seriously, but that's just not the case. One poll found a sharp drop in the president's support, a few polls found a steady increase. Those looking for reasons to celebrate or mourn will have to look elsewhere.

I'd just add one more relevant detail: it's mid-March. I'd be lying if I said I ignored the polls at this stage, but the election is 237 days away. At various stages, the various candidates will go up and down in the coming months, trading leads and seeing peaks and valleys in their levels of support. There's just no reason for one poll to be taken too seriously, no matter what it says.

[elleng]

1. Agree AND

these polls are not comparing PrezO with ANYONE!!!

[gateley]

2. I LOVE polls and believe in them wholeheartedly if they support what I want to hear.

The negative ones, I can easily disregard.

[KharmaTrain]

3. National Polls = Beauty Pagent...

We don't elect presidents on a national basis but in 50 separate elections...thus these and any other national "popularity" poll are only fodder for the talking heads. I'll bet President Obama's numbers are awful in Alabama or Utah...but who cares! Inversely there are polls showing him doing strongly in Pennsylvania and Ohio...that does matter. So I'll stick with state-by-state numbers...

[progressoid]

4. "There's just no reason for one poll to be taken too seriously, no matter what it says."

But that won't stop them from being posted an a daily basis until election day on DU.

[Jennicut]

5. But the thing is, it wasn't really just noise. It was using an increased #of Repubs in one poll and

a decrease of Dems in another. The ABC and CBS polls had methodologies that were changed while the rest of the polls (Pew, Gallup, Reuters) continued to use what they have been using in recent months. Why ABC and CBS decided they had to change around the representations of Dems and Repubs is a weird, if not suspicious.

[DCBob]

6. I think polls are extremely valuable but one needs to understand their design and dynamics.

A good pollster can skew the results by the questions asked, the wording of the questions, the respondents selected, etc. I suspect in the polls that had the President numbers going down, they designed the survey to put emphasis on the issue of gas prices which could of course cause some people to react negatively to the President at that moment. I have not looked at the survey design but I suspect that was it because in the results they linked rising gas prices to the decline in his numbers.

[jenmito]

7. This morning on Morning Joe, Mark Halperin said the CBS poll had an over-sample of Repubs.

That was right after Mika cited the Reuters/Ipsos poll showing Obama at 50%. Finally, truth is told on Morning Joe.

[Drunken Irishman]

8. Mark Halperin said THAT?

Wow. Color me surprised.

[Drunken Irishman]

9. This is why you should always average the polls...

It helps weed out those polls that might be outliers, while not outright ignoring them.

RCP's average of polls from the '08 race had Obama's margin at 7.3 - only .3% less than his overall margin.

Average 'em out and you get a better idea where he stands. Currently, he's at 3.9% nationally. That includes Rasmussen and ABC, who show him trailing by two. That's probably closer to being right than, say, the polls that show him down.