The Des Moines Register’s Iowa Poll always makes news and with good reason: The pollster that conducts it, Selzer & Company, is among the best in the country, according to FiveThirtyEight’s pollster ratings. On Saturday evening, the poll had an especially interesting result in Iowa’s Senate race. It put the Republican candidate Joni Ernst six points ahead of the Democrat, Representative Bruce Braley. Most other recent polls of the state had shown a roughly tied race.
Consider the implications. Republicans need to pick up six seats to win the Senate. Right now, they’re favored to win the Democratic-held seats in Alaska, Arkansas, Louisiana, Montana, South Dakota and West Virginia, according to the FiveThirtyEight forecast. That’s six seats right there. In Kansas, however, the independent candidate Greg Orman is a slight favorite to defeat the Republican incumbent Pat Roberts — and Orman could caucus with Democrats if he wins. If he does, Republicans would need to pick up one more seat somewhere.
That’s where Iowa comes into play. If Republicans are favored there also, they have a path to a Senate majority without having to worry about the crazy race in Kansas. Nor is Iowa their only option. Polls have also moved toward Republicans in Colorado, where their candidate Cory Gardner is now a slight favorite.
This is an awfully flexible set of outcomes for Republicans. Win the six “path of least resistance” states that I mentioned before, avoid surprises in races like Kentucky, and all Republicans need to do is win either Iowa or Colorado to guarantee a Senate majority. Or they could have Roberts hold on in Kansas. Or Orman could win that race, but the GOP could persuade him to caucus with them.
Sounds like it’s time for Democrats to panic?
Not quite, at least according to the FiveThirtyEight model. Republicans are favored to take control of the Senate but the race is close; essentially the same conditions have held all year. As of Sunday morning, the GOP’s odds of winning the Senate are 60 percent in the forecast, only half a percentage point better than where they were after our previous update on Friday.
What’s the flaw with the narrative I described above? It conceals too much of the uncertainty in the outlook. Republicans, for instance, are almost certain to win the Democrat-held seats in Montana, South Dakota and West Virginia. But Alaska, Arkansas and Louisiana are closer calls. Republicans have between a 70 and a 75 percent chance of winning each state, according to the FiveThirtyEight model. It’s proper to describe the GOP candidates as favored, but that’s much different than they’re being guaranteed to win. (Democrats were aided slightly by a CNN poll of Louisiana, also out on Sunday, which had their incumbent Senator Mary Landrieu behind but by a closer margin than other recent surveys have shown.) It’s also not certain that Republicans will hold all their own seats apart from Kansas. Georgia, where our forecast gives the Democrat Michelle Nunn a 27 percent chance, remains somewhat competitive.
As for Iowa itself, the Des Moines Register’s poll may be a great one — our forecast model weighs it more heavily than any other in the state, and Ernst’s chances improved to 56 percent from 48 percent as a result — but it’s still just one poll. As I described last week, it’s usually a mistake to bank on any one poll as opposed to the average or consensus. There are intrinsic limits on how accurate one poll can be, especially if it has a small sample size as the Register’s poll did (546 likely voters).
Finally, we still have more than five weeks to go until the election. If I reprogram the model, telling it the election will be held today (Sunday, Sept. 28) — I hope you can get to the ballot booth in between watching football — Republicans would be somewhat heavier favorites, about 70 percent to take the Senate.
But there’s still a lot of campaigning to do, and one should be careful about concluding that Republicans have the “momentum” (a concept that is constantly misused and misunderstood by other media outlets). Just two weeks ago, it was Democrats who’d gotten a string of strong polls. The FiveThirtyEight model is pretty conservative compared to most others out there. It didn’t show as large a swing toward Democrats as others did two weeks ago — they never quite pulled even in the forecast — and it’s not showing quite as large a swing back toward Republicans now.
So what conditions would merit outright panic from Democrats?
They should keep a close eye on North Carolina and Kansas. These states have been moving toward Democrats in our forecast, helping them offset Republican gains elsewhere.
But these are also races in which the Democrat is doing better than the “fundamentals” of the states might suggest. The Democratic incumbent in North Carolina, Kay Hagan, is pretty clearly ahead in the polls today (including in a CNN survey that was released on Sunday). However, two other states with vulnerable Democratic incumbents, Colorado and Alaska, have shifted toward Republicans. Perhaps if the Republican challenger Thom Tillis can equalize the ad spending in the Tar Heel State, the polls will show a more even race there as well.
And the Kansas race is still in its formative stages. No one has yet polled the race after the Democratic candidate, Chad Taylor, was officially allowed to remove his name from the ballot on Sept. 18. Since then, the ad spending has become more even after nearly a month in which Orman had a pronounced advantage over Roberts. In about 7 percent of our forecast model’s simulations, Democrats held the Senate solely because they won Kansas and Orman elected to caucus with them; without it, Republicans would already be 2-to-1 favorites to take the Senate.
Democrats should also monitor the polls in Louisiana, Arkansas and Alaska. As I mentioned, if these go from being probable GOP pickups to near-certain ones, it will make a lot of difference in the model.
A more macro-level concern for Democrats is that some of the highest-quality polls, like the Des Moines Register poll, tend to show the worst results for them. Quinnipiac University polls, which also have a good track record, have recently shown clear Republican leads in Iowa and Colorado. And the highest-rated polls of the generic Congressional ballot tend to show a Republican lead. This pattern is the reverse of 2012, when Democrats tended to do better in more highly-rated surveys. It may be that some of the mediocre polls will converge toward the stronger polls in states like Iowa and Colorado; a Public Policy Polling survey of Iowa to be released later this week is also likely to show Ernst ahead, for instance.
Still, the whole advantage of having a statistical model like ours is that it provides for some discipline — a rule-driven approach that doesn’t flinch just because the media narrative does. Democrats may have had a restless Saturday night, but no one poll ought to change your perception of the campaign all that much.
Check out FiveThirtyEight’s latest Senate forecast.