Tuesday, March 1, 2011

Don't Believe Your Lying Polls

A number of recent polls finding support for workers' right to organize unions have drawn fire from the internet right wing. The PPP poll in Wisconsin found that a 2012 electorate would elect Governor Walker's former opponent with 52% of the vote. Democrat Barrett had lost in November by 7 points.

The general charge from the right is that the samples of the PPP and NYT/CBS polls are fundamentally flawed. Hot Air, leading the charge to discredit the two polls, rests on three arguments to attack the results.

Polling deniers are making a big deal out of the partisen identification of the sample. As in most long-form polls, participants were asked to identify themselves as "Democratic, indepdent, or Republican" but not necessarily in that order. The NYT/CBS polls is under attack for finding a 36/26/31 split for each of Democrats, Republicans, and Independents. Hot Air alleges that this finding proves that the sample is unreprentative of Americans:
an absurd sample for political polling. In December, Rasmussen’s general-population survey put Republicans ahead, 36.0% to 34.7% for Democrats. A recent poll by Gallup shows erosion in Democratic affiliation all through 2010. In 2008, Barack Obama won the popular vote by seven points nationwide, and the NYT/CBS poll assumes that the electorate has grown more Democratic in 2011.
Rasmussen has a well-documented house effect which leans towards Republicans and tends to push leaners, which is why independents typically register at lower levels in their polls than other firms' results. Many independent voters were leaning Republican following the election in November, and even then would have shown up as self-identified indpendents in other polling firms' reports. As even Ed Morissey noted previously, parties tend to gain in identification rates following victories, then quickly revert to a normal status. While Morrissey is determined to see this event as a damning inaccuracy in the NYT/CBS poll, even he would have told us in December that this partisan ID reversal was likely going to happen, not as an artifact of polling, but as a matter of reality. Ed Morrissey is comparing rotten apples to oranges when he reaches back to the Rasmussen polling in late November and puts it alongside February NYT polling.

Going on to compare party identification to the presidential popular vote is more specious. Indpendents don't often have a candidate to vote for at any level, and therefore have to choose between Republican and Democratic dandidates. Yet independents don't identify with the parties in polling. A 7 point win in 2008 is totally unrelated to partisan identification. It's not even clear whether Morrissey would say that we would expect to see a 7 point split between Republican and Democratic self-identification (with, presumably, independents splitting exactly evenly between McCain and Obama). In any event, if Morrissey had wanted to supply an apt comparison between the 2008 electorate and the NYT/CBS poll, he could have. Both exit polling and the NYT/CBS poll ask people to place themselves in ideological camps: liberal, moderate, or conservative. Exit polling in 2008 found a 22/44/34 split between those groups. This week's NYT/CBS poll finds a 19/37/36 split with 8% undecided. The comparison that Morrissey would have made if he were a serious commentator disproves his attack on the credibility of the poll.

Next, Morrissey conflates union membership with "living in a union household." As Nate Silver points out, a 2008 poll recently found that 9% of Americans are union members. 21% of Americans live in a household with at least one union member. That would explain Morrissey's inexplicable 'sample problem' here:
Next, 20% of the poll’s respondents claim to come from union households. However, only 11.9% of American workers belong to a union, according to a report published by the Bureau of Labor Statistics last month and noted by none other than the Times itself. Union membership fell to a 70-year low as a percentage of the workforce, which in itself is a rather damning statement about the view of collective bargaining by the vast majority of American workers. How exactly did the survey manage to comprise itself of almost twice as many union-household respondents for a poll on union policies as union members in the workforce? Interesting.
If you cut out all the trend blather in the middle, this is remarkably consistent with our intuitions about household membership. Not every union-member lives in a union-only household. The same is true for public-sector workers, which should be obvious to the author of this statement:
Government employment accounts for 17% of all workers, so a sample consisting of 25% public-sector households for a survey of adults (not registered voters) seems a little off.
Apparently Morrissey fails to account for the fact that most families are not single-worker entities. In a society in which only one member of a family worked, we would expect X employment status to match the number of households having X employment characteristic. Morrissey's assumption matches the 19th century ideal of domesticity, but neither reflects the modern upper class world nor any middle class historical period.

Rightwing blogs are making a bad statistical case (actually a nonsensical one in parts) in order to "question... the magnitude of the results." This is a very transparent effort to assure their politician friends that reality is not what it seems, that whatever science underlies polling can be discredited just like the science of global warming or evolution. All one has to do is close their eyes, plug their ears, and say, 'My people love me.'

While Gallup's partisan identification tracking poll is cited to attack the NYT/CBS poll's top line numbers of support for public union bargaining rights (60% to 33%), Gallup's own poll on the question revealed a nearly identical result (61% to 33%). The exalted Gallup's results show the same level of support for union bargaining rights as the NYT/CBS poll. While NYT/CBS might use a different method to 'push' identification from Gallup (i.e. categorize Independent-lean Republican differently), they find the same fundamental level of support for union bargaining rights.

The right wing blogosphere's emphasis on polling critiques is certainly a good sign for union supporters and public employees. The reality is settling in that collective bargaining is a popularly supported right of workers at the national level. The confirmation of this fact among mainstream polling firms is resounding. Pew finds plurality support for the WIunions when contrasted with Governor Walker (42%-31%). Gallup and NYT/CBS peg that national support for public sector unions' existence and bargaining rights at about 60%.

No comments:

Post a Comment