Latent Partisans, or Why the Party ID Sample “Skews Democratic”
By now the Republican litany of complaints against pollsters has become a chorus of desperation. The old adage holds true – if all or most of the polls are against you, you are probably losing. The suggestion that pollsters are using incorrect samples is one that is employed by both sides in the last weeks of a losing campaign. And while there are occasional poor polls with shoddy methodologies or genuinely sloppy demographic screens, most pollsters know how to find a representative sample. There are ways to intentionally skew a poll one way or another – question wording, question order, dubious “likely voter” screens, etc. – but most polling outfits know how to construct a questionnaire that yields something reflecting actual voter sentiment.
That said, there is one area of the polling world that still comes in for controversy – Party ID. Quite simply, we do not know if Party identification really does bounce around or if it is a stable metric like income or religion. The reality is that it depends on how you define party ID and how you use it. Applying a rigid party ID screen is likely to produce some strange and inaccurate results, even though party identification doesn’t really bounce around all that much. Here’s why.
Consider the following types of voters who show up every four years. Assume for a moment that all are “likely voters”:
Partisans – These are Democrats who will support Obama and Republicans who will support Romney. These are the staples of American politics.
Independent Leaners – These are the self-described Independents who almost always vote as regularly for one party as do the partisans. Many times they are even more ideological than the partisans, e.g. Tea Partiers who support Romney but think the GOP spends too much, or Greens who support Obama but think he sold out to the coal industry, etc.
Swinging Independents – The stuff of legend, mostly, these are the voters who actually shift around from party to party each election cycle and often split their tickets depending upon particular issues and personalities. They are a small but important sample of the electorate.
Partisan crossovers – These are the “last proud conservative Democrat/liberal Republican” voters, dwindling every year. They hold on to old party identifications out of tradition or hope that their old party will return to older ways. Very common among old-school Vermont Republicans and white Southern Democrats. But dwindling as the parties become so ideologically aligned.
The above four groups represent most of the respondents in polls. Since their self-described party IDs don’t shift – though their vote preferences might – they will come out the same regardless of whether or not a pollster weighs for party ID or not. This includes the highly publicized self-described Indies voting for Obama or Romney, or proudly disaffected Republicans crossing to Obama or vice versa.
But there is a fifth groups that shows up in the non-weighted polls but gets eliminated and perversely squeezed out of the party ID-weighted polls – call them Latent Partisans
Latent Partisans – These are the folks who, during the long off-season, tell pollsters they are Independent or just “not aligned with a particular party.” In fact, they usually don’t think or talk about politics much at all. But they still vote regularly. And when the election season heats up, they not only gravitate toward one of the candidates. They also start to think of themselves as partisans of a sort. It may be revulsion against the other side, or some particular draw to one side. Whatever it is, these voters leave the Independent sub-sample of the polling screens and become partisans.
HOWEVER, if there is particularly strong movement of non-aligned, Latent Partisan voters to one of the two parties, you will often see a corresponding move that I’ll call: Embarrassed Independents.
Embarrassed Independents – These are the voters who typically think of themselves as belonging to one party, but are sufficiently repulsed and/or embarrassed by their party’s candidate and/or the partisan supporters that they tell pollsters that they are Independents. BUT – they do still plan to vote for their party’s candidate in spite of the embarrassing antics of the campaign – usually out of a “gotta keep the other side in check” sentiment. Or a “maybe the adults will take over again when the silly campaign is over” feeling. Think of longstanding moderate Republicans who like the Romney that ran Massachusetts but are turned off by the rightward tilt of the party. They call themselves Independents now though they will vote Republican. For now. In the future they may stay as Independent Leaners. Or maybe even become Swinging Independents. Or eventually cross to the other side. But that takes time, and for now they will quietly pull the lever for Romney in spite of what the GOP has become.
In non-weighed polls of late, these Latent Partisans show up as sudden Democrats. And these Embarrassed Independents are represented as declining Republicans. And so you can get a sample with a large Democratic party ID lean that accurately reflects the electorate as many mostly-apolitical Indies become Democrats and embarrassed Romney Republicans declare themselves to be Independents.
Now think about what happens when you specifically weigh Party ID, and peg it to some larger, static ratio. You not only lose the Latent Partisans. You actually end up overcompensating among Independents for the other side. As latent left-leaning Indies go to the Dem side, they get removed from the sample. As embarrassed Republicans move to the Indie category, they replace the lost Latents. But when you insist that Party ID is constant, you have to make up the difference by oversampling hard partisan Republicans so as to keep the D-I-R split at the constant ratio.
This is why Rasmussen’s tracker keeps leaning so Republican while every other pollster does not. Rasmussen uses a strict Party ID screen that they re-set only once a month. Leaning Partisans are thrown out. Embarrassed Romneyites replace them in the Indie pool. And the poll compensates by artificially including more Republican Partisans. Presumably, a re-set party ID screen in early October will capture the movement of Latents and Embarrasseds and so will correct this error.
But don’t fall for the cry of biased or unweighted Party ID samples. There’s a reason Party ID samples have skewed Democratic of late. Pretending otherwise is not scientific. It’s just a form of denial that will inevitably lead to disappointment.
LEGAL NOTICE ON CARTOON: The above copyrighted cartoon is licensed to run on TMV. Reproduction elsewhere without licensing is strictly prohibited. See great cartoons by all the top political cartoonists at http://cagle.com. To license this cartoon for your own site, visit http://politicalcartoons.com