More on the Louisiana runoff

A couple days ago this article on the Fix at The Washington Post popped up in my Google Alerts. In it, Aaron Blake brings up something I had written about five days earlier--the very strong possibility that the Senate election in Louisiana will proceed to a runoff due to the inability of the two highest-polling candidates, incumbent Sen. Mary Landrieu and U.S. Rep. Bill Cassidy, to break 50%. 

Blake suggests a pretty simple (and pretty good, roughly) way to interpret the polls that don't ask about the head-to-head between Sen. Landrieu and Rep. Cassidy--just stack up the votes for Republican candidates--that includes state representative Paul Hollis and retired USAF Col. Rob Maness as well as Rep. Cassidy--and those are the votes Republicans get in the runoff.

It's a good approximation, and the best we can do from the polling available, which in the past year has been split approximately 2:1 in favor of the head-to-head polls. That method produces sensible results--Landrieu 48%, Republicans 44% doesn't seem too off. The problem I have with this method (and this is no fault of Mr. Blake's, but rather the fault of the polls we have to work with) is that in many of the four-candidate polls there are an obscene number of undecided or unsure votes, while in the head-to-head between Sen. Landrieu and Rep. Cassidy it's much lower, closer to the numbers in every other Senate poll. In the New York Times / Kaiser poll that had political junkies abuzz with Sen. Landrieu's 24-point lead over Rep. Cassidy, the excitement among Democrats should have been dampened by the sobering fact that 27% of respondents--over one in four--were unsure when given the choice between Sen. Landrieu, Rep. Cassidy, Hollis, and Col. Maness. There are so many ways that 27% could split--most of them in favor of the eventual Republican candidate--that simply stacking the Republican votes for Rep. Cassidy still doesn't say much.

That result is well documented by psychologists and sociologists, who have determined, in essence, that we humans suck at making choices. I propose that the Louisiana polling results (where undecided voters can make up as many as 28% of respondents in a jungle primary poll or as few as 4% in a head-to-head poll) can be explained like this: some respondents are undecided in the jungle primary but choose one candidate in the head-to-head because they find one of the candidates in the head-to-head decidedly less appealing than the other. However, those respondents may find one of the "extras" in the jungle primary--Mr. Hollis or Col. Maness, or both--equally appealing as one of the candidates in the head-to-head. As a result, while they picked one candidate in the head-to-head, they answered that they were undecided in the jungle primary.

Now here's the bad news for Sen. Landrieu: the two "extras" I mentioned just there are both Republicans. As a result, the reason some voters are undecided in the jungle primary is that they had to choose between Rep. Cassidy, Mr. Hollis, and Col. Maness. Most likely, the voters who are undecided between those three are going to be Republicans. The implication is that many of those voters who claim to be "undecided" are really just planning to vote against Sen. Landrieu. This theory can be summed up in the following scientific diagram:

It lines up with Mr. Blake's method of simply adding Republican votes together, and it means that Sen. Landrieu is not in good shape to win over most of those undecided votes. And if the head-to-head polls tell us anything, it's that if she doesn't win those, she doesn't stand a chance in the election.