Primary season is all but over, but we still hear much about electability. There are ubiquitous arguments on who won swing or key states (defined by what suits your point). We see pretty maps that are at the whim of the latest poll. Better are poll-averaging algorithms, like Pollster.com, which reduce outlier impact but aren't too predictive so far out.
Then, like manna from heaven, there's Poblano's analysis at 538. Not to usurp his amazing work, but to satiate my own obsessive compulsive curiosity, I compiled 538's data on Obama/Clinton match-ups vs. McCain, added Pollster.com data for comparison, and sought to identify "swing" states based on blue/purple/red and Cook Report categories.
The goal: To play with 538's data in looking at Obama & Clinton's relative "electability" and to see how the who-won-which-state stuff bears out in current trends. Results, analysis below the fold.
I. Analysis of Win Probabilities
538 is special because, not only does it weigh polls by pollster reliability and recentness, but it also weighs them against a demographic-based regression formula. Even more fun, it runs simulations to provide a Win probability for each candidate against McCain.
As indicated above, I pulled together some charts ranking states by Obama's, and then Clinton's, 538 Win probability versus McCain, and organized the data as follows:
Base: 80-100% probability of Dem winning
Likely/Lean: 60-79% probability of Dem winning
Toss-up: 41%-59% probability of Dem winning
In play/Stretch: 21-40% probability of Dem winning
Out of play: 0-20% of Dem winning
Sure, the lines are somewhat arbitrary, but as Wittgentstein might note, Isn't it enough to say that the data stands roughly here? (Aphorism 88)
Anyway, Obama's chart is here: ObamaWinAnalysis-May19_1.pdf
Clinton's chart is here: ClintonWinAnalysis-May19_1.pdf
[Footnotes: (1) The data is current through Monday afternoon, May 19, including the latest OH polling. (2) 538 Win %s drift slightly from day to day, even without additional polling, so depending on when this diary is actually posted and read, some numbers may be off by a point or so.]
To summarize the high points for Obama:
Base -- 207 electoral votes -- 5 purple states (WA, OR, NJ, MN, IA)
Likely/Lean -- 41 EVs -- 4 purple states (NM, NV, WI, PA)
Toss-Up -- 46 EVs -- 3 purple states (CO, MI, OH)
In play/stretch -- 58 EVs -- 2 purple states (NH, FL)
Out of play -- 183 EVs -- 1 purple state (MO)
High points for Clinton:
Base -- 175 EVs -- 2 purple (NJ, PA), 1 red (AR)
Likely/Lean -- 97 EVs -- 4 purple (OH, MN, FL, WA), 1 red (WV)
Toss-Up -- 29 EVs -- 3 purple (OR, NM, MI)
In play/stretch -- 48 EVs -- 5 purple (MO, WI, NH, IA, NV)
Out of play -- 189 EVs -- 1 purple (CO)
In sum, Clinton has 284 EVs from states with 50% or higher probability to Obama's 274, and she has more EVs in the Base and Likely/Lean range. She also has a greater number of EVs in her Base+Likely/Lean categories.
But Obama is still breaking 270 EVs, and look at his performance among purple states: 5 of the 15 purple states now sit in his "base" range, 9 of 15 are likely/lean Obama or better, and 12 of 15 are toss-up or better.
It's also worth looking at where one candidate is doing materially, or "categorically" better than the other -- for example, where a state is a likely/lean state for one candidate but a toss-up or worse for the other.
Reviewing the data from above we find:
Obama is performing "categorically" better than Clinton in 10 states: CO, CT, HI, IA, NM, NV, OR, WA, WI, VA.
Clinton is performing "categorically" better than Obama in 6 states -- AR, FL, MO, OH, PA, and WV.
As noted below and often on this site, these figures support a general contrast, where Clinton is strong in the 2000 and 2004 battlegrounds of FL and OH (but potentially weaker in having to defend more blue turf); meanwhile, Obama's strength lies in the breadth of support across the Democratic base and mid-sized purple states, but he'll need to cobble together more smaller states if he can't win Ohio. As we will see, this trend holds up when we look at Obama's and Clinton's relative advantages across the categories assigned by the Cook Report.
II. Comparisons by Cook Categories
The Cook Report's most recent electoral vote outlook isn't based on any assumptions about who the candidates are and thus gives a baseline that is neutral yet more informative than the generic blue/purple/red divisions.
So I've compiled a chart that sorts states by Cook Category and then provides data for (1) Obama's advantage over McCain, (2) Clinton's advantage over McCain, and (3) the difference in Win % between Obama and Clinton (so a negative number means Clinton is relatively stronger). The chart focuses on 538's Win %s, but also provides the Pollster.com and 538.com polling averages for comparison. (You will see that Pollster.com trends are fairly consistent with 538's numbers).
States where Obama has a higher Win % than Clinton are in blue; states where Clinton has the advantage are in pink. (Wife over my shoulder note: First, you're an insane dork. Second, what's up with coloring Clinton pink, asshole?! Auth: Didn't even think about the coloring thing. Electoral Vote.com assigned Clinton pink in its similar comparison, so I just kinda went with that.)
So, here's the full chart: CookAnalysis-May19_1.pdf (highlights discussed below).
Starting with Cook's "Toss Up" states Obama has the advantage over Clinton in all five:
State Obama Win% - Clinton Win %
MN 19
CO 49
IA 52
NV 50
NM 15
Among the "Lean Dem" states, a similar advantage for Obama:
State Obama Win% - Clinton Win %
PA -18
MI 14
WI 31
OR 34
NH 2 (essentially a tie given natural fluctuation in the simulator)
Among "Likely Dem" states (WA, ME, DE), it's Obama 3-for-3, though WA is the only state where the difference is significant (numerically and in the realistic potential of different outcomes).
The differences for Solid D and Solid R states isn't generally material since they're generally "solid" in any event. So the main point for now is that, among the Toss-Up and Dem states considered in play, Obama has the clear advantage.
But when we get to the Republican states in play, it's Clinton's turn. She dominates the Lean Republican category, consisting of FL, OH, AR, and then VA (the one Obama-favoring state). She also dominates the Likely Republican category, which includes MO, WV, and then AZ (which is more than just "likely" red once McCain's nomination is taken into account).
III. Thoughts and Conclusions
I really compiled the data just to see how it would look and not to prove any point. After all, the nomination battle is over; Obama supporters really don't have to prove anything. But since we've come so far, here's some closing thoughts:
1. Neither Candidate Has A Compelling Electability Argument Over The Other
Of course, any set of polling-related data carries the usual caveats: we are far off from the election, and the data doesn't take into account internals (e.g., favorability, experience, trustworthiness), how the campaigns will be run, or GOTV efforts, among thousands of other variable. Even in the week I've been compiling and updating the info, there's been significant shifts -- for example, polls giving both candidates boosts in WA and NM. And there are some states -- e.g., NC, AK, MO -- where Obama will compete but are counted as "out of play" given the current Win %.
But even ignoring all that, the data still doesn't clearly favor one candidate over another. Both have clear, but very different, paths to 270+, and the differences are essentially what we've seen since the 50-state SUSA results in early March: Clinton's strategy centers on Florida and Ohio to run up EVs while Obama enjoys stronger and broader support among the mid-sized blue-leaning and toss-up states.
From there, any argument about electability necessarily collapses into "subjective" forward-looking factors about how well each candidate will actually run his or her campaign against McCain. But at that point, the whole point of trying to argue based on "objective" data goes out the window, and you just end up with Clinton supporters arguing for her and Obama supporters arguing for him.
2. What About That Who Won Which State Stuff?
As people have been pointing out for a long time, there isn't much to the argument about which states each candidate has won.
Of the 50 states+DC, I've assigned a winner to 47 (I'm counting OR and KY as foregone conclusions, but have left SD and MT blank. I'm giving Clinton FL, but not MI because Obama wasn't on the ballot. Also, consistent with Chuck Todd's tally at First Read, no winner for TX because of the split Clinton-won-the-primary-but-Obama-won-the-caucaus-and-total-delegates thing).
Of those 47 states, as reflected above, there are only 13 states where the primary winner is doing materially, or "categorically," better than the primary loser. There are 3 states (NV, NM for Obama; MO for Clinton) where the primary loser is actually doing categorically better than the primary winner.
And, yeah, because I'm obsessive compulsive, I also did a chart organizing the data by who won the primary to see as clearly as possible how the primary "winner" is currently fairing compared to the other candidate.
That chart is here: Obama-Clintonstates-May19.pdf
If you take a look, you'll see that Obama maintains a Win % advantage in 22 out of 28 states he won; he and Clinton are tied in 3, and Clinton maintains an advantage in 3 (two of which are de minimis).
On the flip side, however, Obama currently maintains a Win % advantage in 8 of 19 Clinton states -- including the "big" states of CA and NJ, as well as IN where he was supposedly rejected by white people. Clinton maintains an advantage in 10 of the 19 she won, and there's 1 tie.
So where does that leave us? I think it means:
(a) Who won which state has very little connection to who will win or who is even currently favored.
(b) To the extent a correlation exists, it's this: If one candidate currently has a "categorical" advantage over another, you can fairly safely bet that the candidate won the primary. But that's just intuitively obvious and has no predictive value for going forward.
(c) Obama overall has been "improving" in states won by Clinton, as seen by the fact that he now has a relative advantage in 8 of the 19 Clinton states.
Well, that about wraps it up for now. Thanks for listening and I hope you have as fun mulling over the figures as I did. And of course, a million kudos to Poblano and 538.com for giving me something to obsess about.