On the Weekly Standard blog, Bill Kristol quite rightly notes that the New York Times/Kaiser Family Foundation polls in the Arkansas, Kentucky, Louisiana and North Carolina Senate races are suspect, because in all four states the retrospective question about voting in 2012 shows Barack Obama doing much better in the race against Mitt Romney than he actually did in each of the states.

There's one other reason for suspicion about these samples of registered votes. The NYT/KFF poll also gave respondents an option of saying they did not vote in 2012. In response, 21 to 32 percent say they did not vote in that presidential year.

Since turnout in off-year elections is always significantly lower than in general elections, what is the likelihood that these 2012 nonvoters will become 2014 voters? Not very high, I'd say. But the polls still count their choices in 2014.