The good, bad and indifferent in last week's polls
Posted May 5, 2014
A new low?
Last week, The Washington Post declared, "Post-ABC News poll shows Democrats at risk in November as Obama’s approval rating falls."
The problem with the story – not the headline – is that it emphasizes "a new low" for Obama. It does so because most polls conducted primarily for and by mass media tend to reference only their own prior polls. If you read only The Washington Post, it's as if all other polls on the same topic fail to exist.
Yet, if we go to Pollster, which collects and aggregates all relevant polls, we can compare. At Pollster, click on, "Obama approval." While you can select certain polls, or types of poll, I included all polls and find the average is 44 percent approval. Also, the trend line for Obama approval is ticking slightly upwards. Clearly, The Washington Post measures one of the worst ratings for Obama. Other polls are similar, to be sure, but they on average are not as bad news for Democrats.
Republicans are from Twitter?
Do political preferences correlate with social media use? A survey by the Harvard Institute of Politics suggests so. A New York Times blog on the Internet thinks it’s a big deal, but I had the opposite reaction. Ever since Godfather’s pizza started to taste differently after people learned a Republican running for the presidency owns the company, I’ve come to expect bigger differences between partisans. The differences here seem rather tame to me.
Where are the kids?
Here is an interesting "big data" approach for thinking about the difference between midterm voters versus habitual voters, and why this matters for 2014, from the New Republic. Note to Democrats: if accurate, it's depressing news.
A subset of Democrats’ problems in 2014 is that "the kids" don’t vote in midterms. If youth are disproportionately likely to vote Democratic, yet less likely to vote in midterms especially, then team D has a turnout problem.
Where are the Dems?
More good news for North Carolina Republicans: The two basic considerations for thinking about likely outcomes in elections are the distribution of preferences – about even so far between U.S. Sen. Kay Hagan and whoever is the Republican candidate – and the turnout rates of partisans. The New York Times' The Upshot blog says turnout is going to be a big problem from Democrats in North Carolina come November. I don’t disagree, but it is possible state-level frustration among Democrats with Republican rule in the General Assembly might mobilize their turnout above expectations.
How much weight?
Last, a new polling firm based in Georgia has come under scrutiny for its sampling methods. Insider Advantage is using social media to collect opinions and then weighting these responses in a seemingly post-hoc manner. To be fair, it is increasingly difficult to conduct true probability sampling for national or state-level samples. Yet, probability sampling is the only method that permits pollsters to generalize from the sample to the larger population with known parameters. Anything else is educated guessing, as good as the guessing might be, about whether the poll results match the population’s opinions.