Over the weekend, NBC News and Marist College dropped a bunch of new polls in key Senate races, just over a week before Election Day.
In isolation, there's not a lot surprising about the results; for the most part, Republicans have slight leads in mostly close races. But given that this isn't the first set of NBC/Marist polls in most of these contests, it gives us the opportunity to look at a somewhat meta issue about campaign polling: pollster trends. In short: prominent pollsters generally, but not always, agree about the trends in races. And, with that in mind, North Carolina may -- may! -- be getting interesting.
A number of sites aggregate polling averages over time, which is a good way -- a better way -- to track the state of a contest (see Real Clear Politics and Huffington Post, for example). What we're interested here is how the results from individual pollsters have evolved. This is itself a contentious issue; the experts at 538 have suggested that bad pollsters copy good ones to make their results seem more accurate.
But in the interest of seeing how polling firms have tracked, we took eight polling outfits (including two partisan ones, Rasmussen and Public Policy Polling), and graphed their polls in battleground states over time. Each dot below shows the margin between the Democrat and Republican in the race. It's the lines that are the interesting thing here, because they show which way each race has trended.
Pulling out some details:
Colorado -- The trend among pollsters here is consistent; Cory Gardner (R) has seen polls move in his direction pretty consistently. Except in Quinnipiac polling, which in September was an outlier and now has merged with the consensus.
Georgia -- A lot of the polling in this race has been done by firms not in the eight we identified above, which is interesting in and of itself. But the SurveyUSA trend (and those two other polls from CNN/ORC and PPP) are pretty much in line. More recent polls, though, suggest a trend back toward Republican David Perdue.
Iowa -- Polling consistently showed a pretty even race, with the pollsters starting to show movement toward Joni Ernst (R) earlier this month. Again, Quinnipiac's trend was different, moving toward weaker Ernst support, more in line with other polls. That this happened in a fashion similar to Gardner's shift suggests a change in turnout model.
North Carolina -- This chart looks sort of like Iowa's, with pollsters seeing a consistent result that has (at least based on one poll) suddenly shifted to the right. That NBC/Marist line is dramatic in comparison to the other recent polling firms which, even as of last week, continued to show a fairly flat level of support for Sen. Kay Hagan (D).
For the most part, the flow of the races captured by these pollsters reflects the polling average. Because, of course, the polling average includes these polls, and there aren't that many others to move things around. And for the most part, the polls are in agreement, a collection of watches telling basically the same time.
If other pollsters continue to largely agree with NBC/Marist's polling, as they have in other states, it suggests that the unusual result in North Carolina could make for a stressful last eight days for Hagan.