The poll has Mary Landrieu at 36, Bill Cassidy at 32, Rob Maness at 6 and 23 percent are undecided, and in the runoff Cassidy is ahead of Landrieu 43-40.

Those numbers are terrible for Landrieu, and they confirm what a lot of other polls show – namely, that she struggles to get to 40, much less climb over it, in the primary, and cannot get out of the low 40’s in the runoff. In that respect, the poll, commissioned by Raycom for distribution among all its Louisiana TV stations (WAFB in Baton Rouge, WVUE-TV in New Orleans, KSLA-TV in Shreveport and KPLC-TV in Lake Charles), isn’t much different than most of the others out there.

For Maness, the poll is horrible and his camp will likely be screaming about it. With 23 percent undecided and his vote at just six percent, he’s probably demanding to know how the pollster presented him as an option.

And for Cassidy, the numbers are acceptable, though if he’s anywhere near 32 percent in the primary he’s not going to be particularly satisfied with that result. At 43-40 going into a runoff with 17 percent undecided, though, he’d be your next U.S. Senator from Louisiana.

But what’s interesting about this poll, as the headline says, isn’t the results. What’s interesting is that this is a poll of registered voters rather than a poll of likely voters.

Which is unusual. Particularly this late in the game; the poll was conducted Oct. 14-19, meaning three weeks out from Election Day and less. After all, early voting starts tomorrow.

It’s not typical polling practice to do a survey of registered voters this late. Almost every poll you see at this point in the game will screen for likely voters so as to get the most accurate numbers possible.

And Democrats are the ones usually doing registered-voter surveys, because they make the numbers look better – and also because Democrats think they can dragoon all kinds of people to the polls who aren’t particularly engaged in the political process…or know who the incumbent is.

But this poll isn’t a Democrat poll. The company doing it is Multi-Quest International out of Metairie, which is owned by Loyola University professor John Grimm. Grimm’s bio at Loyola’s web site doesn’t exactly mark him as Stan Greenberg or Ron Faucheux…

Mr. Grimm has been involved with the presidential campaigns of both Bushes here in Louisiana and had been asked to work for Romney’s first run for President.  He has worked in the campaigns of Senator David Vitter (all campaigns Vitter ever ran starting with State Rep.), Congressman Scalise (also starting from his first State Rep. race), Congressman Flemming, Congressman Parker, Congressman Matt Cartwright, and several congressional candidates including Helena Moreno, to name a few of the federal campaigns.

Bush-Romney-Vitter pollster, and a registered-voter survey released two weeks before Election Day?

Does not compute.

Unless Grimm is just doing what his client wants, and what his client wants is a registered-voter poll.

So why is Raycom asking for a poll of registered voters?

There are reasons for this, one imagines. Not screening for likely voters makes for a cheaper poll, because if you screen out people who aren’t certain they’ll vote you’re going to have to make more calls to generate a decent sample (this one has a samle of 605 respondents), plus a likely voter screen question adds phone time and makes each response cost more.

But the poll did ask the favorability question on Cassidy and Landrieu; Cassidy was at 26 percent Extremely or Very Favorable, 21 percent Somewhat Favorable, and 35 percent Not Very Favorable, while Landrieu was 34 percent Extremely or Very Favorable, 13 percent Somewhat Favorable and 46 percent Not Very Favorable. So that’s phone time being expended on getting some qualifications on the sample.

Is Raycom trying to play up the undecideds in advance of the debate that will air on their stations next week? A poll that doesn’t screen for likely voters who know the race and who they’ll be voting for would be a way to do that. And this way they were able to run what most other polls show as a 10-15 percent undecided number all the way up to 23 percent.

The other explanation would be that Raycom is carrying water for Landrieu by doing a registered-voter poll 15 days out from the election. That one doesn’t really seem likely, if for no other reason than that if you’re going to run such a poll you’re going to hire a Democrat pollster to do it rather than Grimm.

Hiring a Democrat pollster would likely get you some better numbers than what Grimm shows. Landrieu at 36 and 40 is Landrieu getting killed on Election Day in December.