For as long as we can remember, some political folks have subscribed to a theory, and it is this: Undecideds break for the challenger.
This theory holds that, as long as an incumbent is below 50 percent of the vote, he or she is in trouble, because undecided voters tend to favor the challenger. And it makes sense; after all, people know the incumbent well, and if they're not prepared to tell a pollster they'll vote for the incumbent, they're probably going to vote for the other guy/gal, right?
According to a Fix study of 25 competitive U.S. Senate races held over the last four elections, undecideds actually appear to break more for the incumbent than for the challenger. The below chart compares the final Real Clear Politics polling averages for all 25 races to the final results of those same races. Incumbents are in red, and challengers are in green.
The farther to the right that their bars stretch, they more undecided voters broke for that incumbent or challenger. If a bar stretches to the left, it means the candidate actually took less of the vote than late polls indicated.
A few takeaways:
1) The average incumbent saw his or her share of the vote rise 2.5 points between the late polls and Election Day. For challengers, their share rose just 1.6 points.
2) Even in states that matched the challenger's political lean (Democratic challengers in blue states and Republican challengers in red states), undecided voters weren't any friendlier. Incumbents beat their polling numbers by 2.6 points in these states, while challengers added only 1.5 points.
2) Only nine out of 25 challengers (36 percent) took more undecided voters than the incumbents they faced.
3) Nearly as many -- eight out of 25 -- actually took less of the overall vote than the late polls suggested.
4) Only three out of the 25 incumbents underperformed late polls when it came to their share of the vote. A strong majority (16 out of 25) saw their share of the vote exceed late polls by at least two full points.
5) Not one challenger who was trailing heading into Election Day was able to pull off the victory -- even as six of them faced an incumbent who was below 50 percent in the polls. Two incumbents who trailed, meanwhile, were able to pull off the upset.
So it appears the old adage doesn't really hold -- at least when it comes to the Senate.
Why does this all matter for 2014? It matters because there's already a huge amount of chatter about various Senate polls and what they really mean.
Many early polls have shown Democratic incumbents in tight races but well below 50 percent of the vote -- the most recent being the New York Times/Kasier Family Foundation polls that showed Sens. Kay Hagan (D-N.C.) and Mary Landrieu (D-La.) both in the low 40s. Others who have polled below 50 percent include Sen. Mark Pryor (D-Ark.) and, to a lesser extent, Sens. Mark Udall (D-Colo.), Mark Begich (D-Alaska) and Jeanne Shaheen (D-N.H.).
(On the GOP side, Senate Minority Leader Mitch McConnell of Kentucky has been in the mid-40s in almost every poll, too.)
So, if undecideds do indeed indeed break for the challenger, that would suggest Republicans are in an even stronger position to take the Senate than the polls show. That's not necessarily the case, though.
Now, the people who are undecided in early May might not be the same as those who will be undecided in early November, so it's not an apples-to-apples comparison. And it could be that people who are undecided today know less about the challengers right now but are more likely to gravitate toward them as Election Day approaches. (As Nate Cohn notes, undecided voters in the Times/Kaiser poll of Louisiana favored the GOP 56-18 and thus seem ripe to vote against Landrieu.)
It could also be the case that, as pollsters switch from measuring registered voters to "likely" voters in the coming months, the GOP will get a bump because their side is more motivated.
But it would also make sense that undecideds don't break for the challenger much these days. After all, in such a polarized political environment, the vast majority of people are pretty consistent about which party they favor, and early polls gauge partisanship as much as actual knowledge about the candidates themselves.
The numbers above suggest that, as Election Day approaches, it matters less how far below 50 percent an incumbent is and more whether they're beating their opponent. So if these incumbents can keep it close or maintain a small lead, they'll be in position to win.
And even if the incumbents are behind, undecideds often break for them.