Democratic candidate for U.S. Senate Michelle Nunn, right, participates in a rally before casting her own ballot in early voting at the Adamsville Recreation Center in Fulton County on  Oct. 15 in Atlanta. Nunn is running against Republican David Perdue. (David Tulis/AP)

The Senate race in Louisiana is headed for a runoff, and the race in Georgia is likely to join it. That means we could have two Senate races stretching into December and one of them into January — with control of the U.S. Senate very possibly in the balance.

And accordingly, we're starting to focus a little more on runoff polling.

These surveys suggest that Democrat Michelle Nunn has the runoff edge in Georgia (she led 51-47 in a CNN/ORC poll last week) and Republican Rep. Bill Cassidy has the edge in Louisiana (he's up 48-41 in a just-released USA Today/Suffolk poll).

But when it comes to these pre-Election Day polls of hypothetical runoffs, caution is the watchword.

The runoff polls, for instance, apply the same electorate from Election Day to the runoff. And especially in Georgia, where the runoff is still more than two months away and turnout will be lower, it's not a great measure. (Turnout in Georgia's last two Senate runoffs — in 1992 and 2008 — has been between 81 and 84 percent as high as turnout in the following midterm elections.)

In addition, that drop in turnout almost always comes at the expense of the Democrats. The last five statewide runoffs have seen Democrats lose an average of nine points from their Election Day margins. In other words, if Nunn's campaign were offered the same electorate Jan. 6 as Nov. 4, it would take it in a heartbeat. But that's probably not feasible — especially given that this will be the state's first post-New Year's Day runoff.

Turnout in Louisiana runoffs, by contrast, can be pretty similar to a midterm election. That's because the runoff is basically the general election in Louisiana, following the nonpartisan jungle primary on Election Day. In 2002, for instance, Sen. Mary Landrieu (D-La.) faced a runoff in which turnout was basically just as high (99 percent) as it was on Election Day.

But runoff turnout also varies significantly based on what kind of races are being held. In congressional runoffs, for instance, turnout can drop by as much as one-third if there isn't a big-ticket runoff at the top of the ballot getting voters back to the polls.

And while Landrieu has survived two runoffs, the odds in this one are stacked against her. The Suffolk poll shows Landrieu at 36 percent, Cassidy at 35 percent, tea party favorite Rob Maness (R) at 11 percent and 18 percent undecided.

Here, it's actually more instructive to look at the the Maness voters and the undecideds than the runoff poll. As Suffolk pollster David Paleologos notes, 89 percent of Maness supporters and 60 percent of undecided voters have an unfavorable opinion of Landrieu. These are the people Landrieu would need to win over in order to get to 50 percent plus one Dec. 6.

But while all of those voters appear likely to vote on Election Day, the question is who among them will turn out for the runoff. If Maness voters are disenchanted enough with Cassidy, do they stay home? And who among the undecideds comes to the polls? Landrieu is at 75 percent among the black vote, for instance, and if she can drive that number up, it would certainly help.

In both of these cases, though, the most important thing to remember is that there are still several weeks of the campaign to go, and especially in a state like Louisiana, both sides have effectively been biding their time for the runoff for a while now. The last few weeks is when stuff starts happening.

And lastly, consider the very real possibility that these races could be pivotal for control of the Senate. If that's the case, these two states will be inundated with money from all over the country, and nobody will be able to escape the importance of their state's runoff. And conversely, if the Senate majority isn't at stake, these could turn into pretty sleepy contests.

If could be a fun few weeks.