Commentary Magazine


Topic: partisan identification

Skewed Presidential Polls Should Be Trashed, Not Published

With the public and the pundits hungry for more information about the election, the focus on polling seems to be greater than ever. Unfortunately for the pollsters, so has skepticism about their results. Part of that lies in the natural unwillingness of partisans to accept that their side is losing. Thus, Republicans take polls that show their side winning as truthful while scoffing at those that show Democrats ahead; Democrats play the same game. We’ve seen a lot of this during this election cycle. But as much as we should guard against the partisan knee-jerk when reacting to certain polls, that doesn’t mean that they must all be taken at face value. Case in point is the latest Washington Post/ABC News poll of the presidential race published today. It gives President Obama some much-needed good news by showing that he leads Mitt Romney 49-46 percent. That three-point margin is an improvement by one point over the last Post poll taken two weeks ago.

But the problem with the Post poll is revealed in the paper’s story about its findings:

Partisan identification fluctuates from poll to poll as basic orientations shift and with the sampling variability that accompanies each randomly selected sample of voters. In the current poll, Democrats outnumber Republicans by nine percentage points among likely voters; the previous three Post-ABC polls had three-, six- and five-percentage-point edges for Democrats. The presidential contest would now be neck and neck nationally with any of these margins.

In other words, the pollsters know this is a bad poll but went ahead and published it anyway.

Read More

With the public and the pundits hungry for more information about the election, the focus on polling seems to be greater than ever. Unfortunately for the pollsters, so has skepticism about their results. Part of that lies in the natural unwillingness of partisans to accept that their side is losing. Thus, Republicans take polls that show their side winning as truthful while scoffing at those that show Democrats ahead; Democrats play the same game. We’ve seen a lot of this during this election cycle. But as much as we should guard against the partisan knee-jerk when reacting to certain polls, that doesn’t mean that they must all be taken at face value. Case in point is the latest Washington Post/ABC News poll of the presidential race published today. It gives President Obama some much-needed good news by showing that he leads Mitt Romney 49-46 percent. That three-point margin is an improvement by one point over the last Post poll taken two weeks ago.

But the problem with the Post poll is revealed in the paper’s story about its findings:

Partisan identification fluctuates from poll to poll as basic orientations shift and with the sampling variability that accompanies each randomly selected sample of voters. In the current poll, Democrats outnumber Republicans by nine percentage points among likely voters; the previous three Post-ABC polls had three-, six- and five-percentage-point edges for Democrats. The presidential contest would now be neck and neck nationally with any of these margins.

In other words, the pollsters know this is a bad poll but went ahead and published it anyway.

It’s true that partisan identification isn’t set in stone. But do the pollsters or the editors at the Post who were presented with this survey for publication really believe the electorate is that heavily skewed in favor of the Democrats? If that were true, that would mean America is leaning even more heavily toward President Obama’s party that it did in November 2008 when his “hope and change” fever was at its height. Getting such a result at a moment when every other poll indicates that Romney has made up the ground he lost in September to tie up the race, if not go ahead, should have alerted the pollsters that their sample was badly skewed. Adjust the figures to the level where other polls show party affiliation and the result would have been a lead for Romney, not Obama. It should have told the Post and ABC that this poll was not worth publishing.

We won’t know just how much those who vote in this election will lean toward one party or the other until after November. But the notion that an election as close as this one will produce a plus-nine result for the Democrats is ludicrous.

There will be those who will simply charge the pollsters — and their sponsors — with political bias and claim that they deliberately sought to cook the poll so as to give a win to President Obama at a time when other polls and the public’s mood has shifted against him and toward Romney. In reply, the pollsters will simply say that their sample was random and that they merely transcribed the choices of their respondents. I’ll take them at the word about that. Random is random, and perhaps that’s just the numbers they got. But not all random samples are kosher. The party identification numbers should have made it clear to them that this was a bad survey and that they needed to try again, if only as a control to see that they didn’t produce a glaringly inaccurate survey. That they didn’t do that is an indication of a lack of seriousness, if not bias. Just noting a margin of error (in this case of 3.5 percent) isn’t enough.

The skewed sample means that the Post/ABC poll is an outlier and will be dismissed as such by serious observers. But it raises serious questions about the willingness of major news organizations to publish material that they already know is tainted and almost certainly inaccurate. Stories like this make it clear why the public views journalists with such disdain.

Read Less