We have been kicking around various "what if" scenarios on a potential three-way Senate race in Florida this afternoon. It's an incomplete and highly speculative exercise, but also, well, fun...and we want to let our readers in on it via a spreadsheet embedded at the end of this post. I'll get to my hunches below. First, let me explain the context.
As implied in my post last week, there is great potential for change in the early "horse race" numbers in Florida. Whenever pollsters ask voters to choose between well known incumbent and a lesser known challenger, the margin between the two usually narrows as the campaign gets underway. Kendrick Meek's 26% name recognition (on the Quinnipiac poll) and $3.4 million cash on-hand qualifies him as just such a challenger.
Add the wild-card of an independent candidacy by Republican Governor Charlie Crist and you have a situation where the early trial-heat results are almost certain to change between now and November. But in what direction?
Unlike most independent and third party candidates, Crist begins with near universal name recognition. He is currently the best known of the three candidates. But can Crist maintain (and grow) his current support or will it fizzle? Historical precedent exists for either scenario. In one corner are candidates like Joe Lieberman, Bernie Sanders or Lowell Weicker -- well known public figures who held their early support and won. In the other are initially well known independent candidates like Ross Perot in 1992, Kinky Friedman in Texas in 2006 or Marshall Coleman in Virginia in 1994 -- well known independent candidates whose support fades as the election approaches. (Both Harry Enten, who contributed to this post, and Chris Cillizza offer reasons why Crist is no Lieberman).
With these cautions in mind, consider the cross-tabulations by party identification for three recent polls that tested a three-way trial-heat between Democrat Meek, likely Republican nominee Marco Rubio and Crist as an independent. (I have not yet included the recent Rasmussen survey, because their full party tabulation and party composition are hidden behind a subscription wall. I emailed Rasmussen to request their results and will update this post accordingly should they respond). (Update - 4/22: Rasmussen fielded a new survey yesterday and provided both the results by party and their party composition included in the chart below).
As a pollster, what intrigues me most about this table is the potential for change. Specifically, I have three questions:
What is Charlie Crist's floor among Republicans? How much of his support among Republicans will persist as the reality of an independent candidacy sinks in and Rubio wins more endorsements from prominent Republicans?
What is the ceiling on Kendrick Meek's potential support among Democrats? Meek's Democratic support will almost certainly rise as his recognition increases, but how high? (The related converse: What is Crist's floor among Democrats?)
What is Meek's potential among independents? As I have written previously, many that initially identify as independents are really closet partisans that "lean" to a party.
Now I have a few hunches about the answers to those questions -- and a spreadsheet that allows you to plug in your own answers and play "what if" -- but first some warnings.
The table above includes the party composition of each poll and, for reference, the results for partisan composition from both the 2008 and 2006 exit polls. You have probably already noticed some big differences in party composition for the polls, let me make this warning as loud and clear as I can: It is a MISTAKE to assume that differences in party ID are ONLY about sampling. Yes, different polls may sample different kinds of people; some may include more Democrats or more Republicans due to either their likely voter model or some hidden response bias. But polls also differ in the way they measure party identification. Some push harder than other for an answer, and the push is not just about the question text. It also involves the mode (live interviewer or automated) and how pollsters that use live interviewers train them to handle uncertain respondents.
Complicating all of this further is that the above tabulations differ in the way they define the independent category. Quinnipiac excludes those who identify with a third party, or who say they "don't know" what they consider themselves, Research 2000 appears to include them.
But back to my hunches: Meek's support is almost certain to rise to at least 70% among Democrats and 80% is not unreasonable (Crist won just 14% of Democrats in 2006). A goal of 30% for Meek among independents also seems reasonable (given that many independent identifiers lean and typically vote Democratic). So if Crist holds a third of Republicans, and we assume a party composition that's the average of the three polls, Meek leads by a comfortable margin.
On the other hand, flip party composition to match the 2006 exit poll, drop Meek to 75% among Democrats (bumping Christ up to 20%) and you get a Meek-Rubio dead heat. Bump the Republican party ID advantage up a point or two and Rubio wins narrowly. Drop Meek to 70% among Democrats and bump Rubio up to 75% among Republicans and Rubio wins comfortably.
One thing becomes clear in all of this: Crist has a hard time prevailing unless he grows his support beyond what he currently receives in a three-way race. And that won't be easy.
But I readily concede that my hunches are educated guesses, at best. What's your take?
PS: Thanks to the Quinnipiac University poll for sharing their party composition data.