As you probably know, I’ve been paying pretty close attention to the race for control of the Senate. You can see the six main models listed on the sidebar on the right. And recently, things have not been going so well, as I discussed in, Democrats Are Sad Not Delusional. So I continue to focus on Nate Silver’s model — currently giving Democrats a 36% chance of holding the Senate — and using Sam Wang’s model to remain hopeful.

Right now, Wang’s model gives the Democrats a 45±15% chance of holding the Senate. But it has been as low as 25±15% just a couple of days ago. It isn’t this number that gives me hope. Wang provides a more interesting statistic: the meta-margin. This is how far off the polls would have to be for the election to be a toss-up (50% chance that Democrats would keep control of the Senate). Currently, the value is R+0.4%. That is: Republicans are up by 0.4 percentage points and if the polls are off by 0.4 percentage points in the Republicans favor it would be a toss-up.

The reason this gives me hope is that polls usually *are* off by quite a lot more than that. Looking just at midterm elections (presidential election polling is better), the average magnitude of errors on elections since 1990 is 2.9%. Of course, that doesn’t mean that errors would be in the Republicans’ favor. In 1990, 1994, and 2002, the polls were off in the Democrats’ favor — meaning the Republicans did better than expected. But the last two midterms were off in the Republicans’ favor. And the last four elections total greatly favored the Republicans: +3.4%, +0.9%, -0.2%, +2.7%.

This doesn’t necessarily mean anything. The 1998 polls favored the Republicans by 4.9%. Then the 2000 polls favored the Republicans by 2.1%. But then it flipped and favored the Democrats by 4.0% in 2002. So maybe the polls are all making the Democrats look better than they will turn out to be. That would actually make sense, because the Democrats have been polling far better than anyone expected, given the fundamentals of this election. And if that’s the case, this could be a far worse election than I am expecting.

In the end, when I make my predictions going into the election, I’ll stick with the polls. But I hang onto the hope that the meta-margin provides. Of course, it is just a measure of the error that the models predict. Most of them claim that Democrats have about a one-in-three shot of keeping the Senate. And that means there is roughly a one-in-three chance that the Republicans will have a blow-out. So that’s not a lot to hold onto. But it’s something. As always, it will help if you vote.