A sharp reader caught a statistical problem with that much-cited NBC/WSJ poll that shows Republicans are getting whooped in the shutdown:
@texasbryanp In today’s NBC/WSJ poll, 1 in 5 respondents work for govt. Only 8% of general pop works for govt. See last page.
— Tampa (@S1CT) October 11, 2013
It’s true. Here is the complete poll with the sample, the questions, etc. On the very last page we get the breakdown of where the 800 respondents work.
The wording of the question is broad enough to sweep in spouses of government workers and other situations in which the respondent themselves may not work for the government in some way, but someone closely related to them does. Oversampling government workers may not always lead to a skewed result — government workers include the bureaucrats we all have in mind here, but also the wife of the local sheriff who’s as conservative as can be. Nevertheless, we do have government workers heavily sampled, by more than twice the actual percentage of Americans who work for the government. Then let’s factor in the poll’s partisan sample, which was 43-32 Democrat, the percent that admits voting for Obama, which was 44 versus just 35 for Romney, and this looks like a left-heavy sample. I’m not screaming UNSKEWED POLLS!!!1!!!1! here, but, it’s questionable. Then, factor in NBC’s role. That network is not to be trusted.