Jay Cost Breaks Down the Party ID Conundrum

This blog has hammered the party ID problem in polling from the beginning.  The incomparable Jay Cost puts the pieces together in his megaphone at The Weekly Standard lays out how the polls split on voter turnout and how that drives whether they find the race a blowout in favor of Obama or a dead heat:

Republicans, by and large, are frustrated with recent polls of the presidential election because they think Democrats are being oversampled. Many pollsters respond by saying that “weighing” the polls for partisan identification creates its own problems and might end up skewing the polls in the wrong direction.

A lack of weighing creates its own problems, which many pollsters often fail to acknowledge. Specifically, many polls have, in my judgment, overestimated the Democrats’ standing right now. I base this conclusion not on a secret, black box statistical methodology or some crystal ball, but rather on a read of American electoral history going back to 1972. If I am right, then some of the polls are giving a false sense of the true state of the race, and will likely correct themselves at some point or another.

One important “tell” in my opinion, is this president’s continued weak position with independent voters, who remain the true swing vote. Obama’s average overall margin over Romney in these same polls is roughly 4 percent.

Bottom line: You do not get a four-point lead overall with a tie among independents, unless you are squeezing substantially more votes out of your base than your opponent is. And more generally, you are not “winning” an election in any meaningful sense of the word when 3/5ths of unaffiliated voters are either undecided or against you.




Cost sees “two ways the polls are tilted in favor of the president.”

First, many of the polls are guessing that Democrats are set to turn out at levels that match or sometimes exceed 2008. Take two examples – recent polls in Ohio and Florida. I’ve included the 2008 and 2004 exit polls as a baseline for consideration.

The midpoint between 2004 and 2008 is D+1.5. You’ll notice that Gravis, Washington Post, and Fox basically see a replay of 2008 while Rasmussen and the Purple Poll see roughly something in between 2004 and 2008. Relatedly, the polls on the high end for Democrats see a 5-point lead or better for the president (with Gravis being a strange exception), and Obama at or near 50 percent. The polls that see a tighter partisan split basically see a toss-up.



Cost then proves out the same point with Florida and points out the trend holds for Virginia, Colorado and likely Nevada (a lack of polling impacts the analysis for NV). He concludes his party ID discussion with the following:

All told, we see a statistically significant relationship between Obama’s margin and the Democratic advantage in partisan identification. In other words, there appears to be a bimodal distribution of the polls. They are not converging around a single point. Instead, some (notably Rasmussen, Purple Strategies, Survey USA, and Mason-Dixon) see Obama ahead by just 1 to 3 points in the key swing states, while others (notably the Washington Post, Fox News, PPP, and NBC News/Marist) see an Obama lead that ranges between 4 and 8 points. And the difference looks to be built around how many Democrats are included in the polling samples.

If it comes down to whether or not this will be a repeat of 2008 — which is basically what the latter camp of pollsters is suggesting — then my money is on no. Of course, it is possible that I am wrong. I have no crystal ball looking forward. All I can do is look back through history, where I see on average a nationwide Democratic identification edge of about 3 points, which is also roughly the midpoint between 2004 and 2008. That is my guess about 2012. It is an informed guess, but it is still a guess. If I’m right, then Rasmussen, Purple Poll, Mason-Dixon, and Survey USA are closer to the mark. But I could be wrong, in which case Fox, PPP andWashington Post are closer to the mark.

Importantly, the pollsters are guessing, too. They are guessing via the myriad of choices they make about when to poll, whom to poll, and how to poll. By Election Day, polling will be much more “scientific” than it is today; but now there is quite a bit of “art.” That’s how we wind up with two points of convergence, instead of just one.

As I mentioned earlier, a big “tell” here is that Obama cannot build any kind of a lead among independent voters. That suggests to me that his advantage is built entirely on Democratic enthusiasm, which right now is above its historical trends and clearly on a post-DNC bump. Nobody in the postwar era has won the presidency by carrying less than 49 percent of independents, and Obama is quite a ways below that mark, even if some polls show him at or above 50 percent nationwide and in the key swing states.


  1. wholefoodsrepublican
    Posted September 26, 2012 at 10:52 am | Permalink | Reply

    suggest using a MEDIAN, not a mean, that will take care of outliers better.

  2. Mike DeVault
    Posted September 26, 2012 at 11:08 am | Permalink | Reply

    These polls are driving me NUTS!!!! I want to believe what I am reading here but the contrarian viewpoint is overwhelming.

  3. Posted September 26, 2012 at 2:16 pm | Permalink | Reply

    Sorry guys, you are grasping at straws. First, party’ identification is fluid and can’t be compared across pollsters, which measure it in different ways. More importantly there is an increase in independents that when prodded indicate they really lean Republicans. To by itself probably explains the numbers you are seeing.

    As always avoid looking at any one poll and focus on averages. Obama is winning.

  4. Posted September 26, 2012 at 7:15 pm | Permalink | Reply

    Turnout is not “fluid” though, you can go back decades and predict that a normal-looking election will be something like D+2-4. Even with all the ducks lined up in a row in 2008, the best the Dems could manage was a +7.

    So if a poll oversamples Dems by 9-10 points you can with confidence say it’s rubbish. No ifs, no buts, it’s rubbish.

    And you CAN compare across pollsters by weighting everyone by the same reasonable D+ turnout (say, D+3). Do that, and virtually all the polls show it as tied with Romney possibly slightly ahead.

    Take the mid-point of turnout from 2004 and 2008 in the swing states and do the same…..you end up with the race very close there as well, at most a couple of points either way.

    Independents “leaning Repub” suggest to me they have broken from Obama but haven’t quite convinced themselves to vote FOR Romney yet.

One Trackback

  1. […] to Jay Cost, there is a disparity between polls which Cost calls a “bimodal distribution”, http://battlegroundwatch.com/2012/09/26/jay-cost-breaks-down-the-party-id-conundrum/ which Cost says results in a difference in poll results which “looks to be built around how […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: