Clinton campaign manager out: this is a sign of trouble but should not be viewed as a move of desperation. A strategic move at this point, followed by wins before March 4th (like WI) will give the perception that the shake-up worked.

It also looks like Obama may be headed to victory in the Maine caucuses. After you read the rest of this, you’ll see why that has to be considered an upset, exceeding the CT one Obama previously landed.

Now on to my original post…

With exit polling, numerous potential predictive factors have been advanced as the reasons Obama or Clinton have performed as they have. Generally speaking, a high percentage of Black voters is considered favorable to Obama. A high percentage of Latino voters is considered favorable to Clinton. Older women favor Clinton. Younger voters favor Obama. And so on.

However, exceptions have occurred in some states for some of these. So I went exploring to determine which factors have proven the most accurate predictors of each candidate’s success. I’ve found the top two predictors are 1) the percentage of Black citizens in a state and 2) the red state/Clinton unpopularity factor. The definition of the latter is any red state that Bill Clinton lost in both the 92 and 96 elections.

Bill Clinton was extraordinarily popular among Black voters, but Obama has negated that. It’s uncertain where the exact dividing line is but it’s between 17% and 21%. As an absolute, every state that has Black citizens making up at least 20.9% of its populace has been won by Obama. He’s won 5 and 4 remain. As an absolute, Clinton has won many where the Black population falls below 17%. Between 17% and 20.9%, the only state that’s voted so far is NY (17.4%) but Clinton won that as her home state. Virginia’s the only other state in that gray area, at 19.9%, and polls today indicate Obama is likely to win that one easily.

Of the remaining states, however, there’s several that Clinton may not be able to touch despite other favorable predictors. Consider the list of the 21 states -plus DC - where Black citizens make up 10% or more of the population. (I colored Clinton’s wins orange and Obama’s green, leaving the primaries ahead in black).

DC 56.5
MS 37.1
LA 31.7
GA 29.9
MD 29.5
SC 29.0
AL 26.3
NC 21.7
DE 20.9
VA 19.9
NY 17.4
TN 16.9
FL 15.8
AR 15.7
IL 15.0
NJ 14.5
MI 14.3
OH 12.0
TX 11.9
MO 11.5
PA 10.7
CT 10.2*

This predictor suggests Obama should easily win DC, MS and MD. Since the polls indicate he’ll win VA easily, too, the only one left in that top section where Hillary might try to compete is NC. Anyone notice lately her suggestion that there’ll be a place in her administration for John Edwards? Based on this predictor, I doubt that she’ll win, but keeping it close can help in the delegate count.

The second predictor could prove even more troublesome for Clinton. Among the red states where her husband lost twice, she has yet to win a single one. So let’s look at that partial list, too. I also included notes that show the average margin of loss in 92 and 96, and if there was a variance of 3% or more in the vote margins of those two elections, I noted that, too.

UT -19.89
NE -17.94
ID -16.07 (5 pts worse in 96)
AK -13.35 (8 pts worse in 96)
KS -11.68 (13 pts worse in 96 due to favorite son Dole)
ND –9.43 (5 pts better in 96)
WY -9.29 (7 pts worse in 96)
OK -8.22
SC -7.09
MS -7.02 (4 pts better in 96)
AL -6.87
IN -5.85
TX -4.21
SD -3.49
VA -3.17
NC -2.74 (4 pts worse in 96)

Coupled with the first predictor, MS should go big for Obama. And on this factor alone, WY should go for Obama. But since the declining margin of loss below AL is not clearly tested, we can’t say for sure how IN, TX, SD, VA and NC will vote, can we? Combining both predictors, it’s fair to say that VA and NC are likely to go for Obama. But that leaves IN, TX, and SD to guess at.

However, just below that, where states voted once for Clinton in 92 or 96, that list continues:

GA -0.24
MT -0.19
AZ +0.14 (4 points better in 96)
TN +1.12 (7 pts worse in 96)
CO +1.45 (6 pts worse in 96)
NV +1.83 (Clinton won this one twice)
FL +1.91 (8 pts better in 96) (The next two Clinton won twice)
KY +2.09
OH +4.10 (4 pts better in 96)

Using just these two predictors, MT, KY and OH should be competitive, but it’s still hard to say whether IN, TX and SD will. And, as we all know, TX and OH are two of the three biggest remaining that Clinton has to count on. So are there other predictors we can use to make a better guess?

Let’s try the other oft-cited racial demographic. Does a high percentage of Latino citizens in a state favor Clinton? Here’s the list of the highest in that group:

NM 44.0
CA 35.9
AZ 29.2
TX 35.7
NV 24.4
FL 20.2
CO 19.7
NY 16.3
NJ 15.6
IL 14.7
CT, UT 11.2
RI 11.0
OR 10.2

With this predictor, it suggests the dividing line is somewhere between CO’s 19.7% and FL’s 20.2%. Looking back to the Black citizen percentages, both could have a dividing line between 19% and 20%. So the remaining primaries of RI and OR would not be impacted by any of these three predictors, but the Latino citizen percentage suggests TX should go for Hillary, even though the Clinton Unpopularity factor suggests a closer race.

But it’s really not that clear. Despite what this list looks like, let’s remember that the most Latino state of all, NM at 44%, was only won by 1% by Clinton. Out of 136,000 voters, she leads by 1,100 votes, as they near a final result with 99% of the votes counted. At this point, it looks like she’ll get 13 delegates there, to Obama’s 12. So even if we predict a TX win for Clinton, the margin could be close enough to negate any real advantage there. In conclusion, the Latino citizen percentage has not proven as reliable as the other two indicators.

If Clinton falls behind in the delegate count before March 4th, she may be dependent on OH to catch up, and maybe the other big state, PA, that votes on 4/22 (plus whatever smaller states she can shine in.) So are there any other predictive factors I’ve found that can aid this guesswork? Yes.

There’s two. There’s the inverse of the Clinton Unpopularity factor, measured two ways.

The first one: every state that Clinton widened his margin in 96 by at least 8% has been won by Hillary… except one: CT.
Here’s that list:

FL +1.91 (8 pts better in 96)
NH +5.59 (almost 9 pts better in 96)
NJ +10.12 (15 pts better in 96)
CT +12.29 (12 pts better in 96)
OR +14.02 (8 pts better in 96)
ME +14.60 (13 pts better in 96)
HI +18.34 (14 pts better in 96)
NY +22.36 (13 pts better in 96)
RI +25.46 (15 pts better in 96)
MA +25.95 (15 pts better in 96)

So other than the Obama upset in CT, and the probability that his former home state of HI could go his way like AR went for Hillary, this factor indicates Clinton’s the favorite in ME and RI.

That leaves the other measure of Clinton Popularity, the highest average margins of victory in the 92 and 96 elections. It’s a whole lot harder defining this dividing line, though. Here’s the list of states where he won by an average of 10% or more:

NJ +10.12 (15 pts better in 96)
MI +10.30 (6 pts better in 96) Remember, Obama’s name was not on this ballot, though.
DE +11.72 (7 pts better in 96)
WA +11.99
CT +12.29 (12 pts better in 96)
CA +13.15
MN +13.89 (5 pts better in 96)
WV +13.89
OR +14.02 (8 pts better in 96)
ME +14.60 (13 pts better in 96)
MD +15.08
IL +15.88 (3 pts better in 96) An exception because it’s Obama’s home state.
AR +17.34
HI +18.34 (14 pts better in 96) Possible exception as Obama’s former home state.
VT +18.88 (6 pts better in 96)
NY +22.36 (13 pts better in 96)
RI +25.46 (15 pts better in 96)
MA +25.95 (15 pts better in 96)
DC +75.70

Using our previous predictive indicators, we already ascertained that CT was an Obama upset (or, if you prefer, an aberrant from the most accurate predictors). So this list suggests the dividing line may be around 12% to 13% average margin of Bill Clinton’s victories. All the states above a 16% margin have gone solid for Hillary, so maybe the dividing line’s 16%.

Working from the bottom of the list, in the primaries to come, the best analysis is this:

DC: The more reliable Black citizen percentage factor trumps all other factors, so this should be Obama’s.

RI: It shows up in both Clinton Popularity lists, so this should be an easy Clinton win.

VT: Only in one Clinton Popularity list but in the solid Clinton area. Clinton should win it.

HI: It shows up in both Clinton Popularity lists, so this should normally be an easy Clinton win. Since it’s also Obama’s former home state, that may trump everything, but it could prove close and has the potential to swing either way.

MD: it might be competitive but it’s also trumped by the Black citizen percentage factor, so it should be an Obama win.

ME: It shows up in both Clinton Popularity lists, so this should be an easy Clinton win. It’s also one of the three whitest states, with less than 2.9% Blacks and less than 2.9% Latinos in its citizenry.

OR: It shows up in both Clinton Popularity lists, but is near both dividing lines. It favors Clinton but may not impact the overall delegate count by much, no matter who wins it.

WV: Only on one Clinton Popularity list, near the divider, and also a very white state, with 3.3% Black citizens. Clinton should be favored but margin can’t be guessed at.

That latter group contains 8 of the remaining primaries, with only HI as a puzzle. From our previous factors, though, we also predicted:

MS for Obama.

NC probable Obama.

VA probable Obama.

WY probable Obama.

IN leans Obama, could be close.

SD leans Obama, could be close.

TX contradicting factors slightly favoring Clinton, may be close.

MT too close to call.

KY slightly favors Clinton.

OH slightly favors Clinton.

That’s 18 of the remaining primaries. Using these predictors, HI, OR, IN, SD, TX and MT are the hardest to guess. That leaves only two others that none of these factors lends a clue to: PA and WI. Both lean slightly towards Clinton but they belong with the other six hardest to guess.

Conclusion: The perception of momentum, plus media bias, could impact all 8 in the hard-to-guess category, as well as KY and OH. From all the data available, these are the strongest predictors I can find and they indicate Obama has good odds of winning 8, with Clinton winning RI, VT, ME and WV, and 8 too close to call.

Any way you slice it, OH and TX on March 4th and PA on April 22nd remain the three biggest slates of delegates. All three lean towards Clinton but none of them should be taken for granted. For all of these, I recommend reading the polls and compare them to these predictive factors. Let’s see if there’s any that upset my predictions. And on all the too-close ones I’ve identified, if the polls show margins within 10 pts, compare them to tracking polls to see if momentum is narrowing or widening that margin.

In the big three, Clinton will need a solid margin of victory in one or two to overcome the lead Obama’s likely to have before March 4th. Venturing the best guess I can make from these predictors, I suspect TX will be close, so it will be OH and PA where she’s gotta shine.

3:30 Update, on Maine: Should Obama win Maine today, as now seems quite possible (he leads by 17% with 70% counted), the only way this would NOT qualify as an upset is if I set the dividing line too low in the two Clinton Popularity lists. However, if we adjust those lines upward to accomodate Maine, that would move both OR and WV out of the too-close-to-call group into Leans Obama.

I’ll be looking for an ugly campaign (on both sides) for Wisconsin as that remains the only place she can break her fall before March 4th.

One final note: I’m not advancing these to promote demographic division. I’m just reading what appears to be there as useful and accurate predictors. As my first commenter noted, plenty of white support exists for Obama, just as it did in the first Iowa primary. If anything’s changed since then, I suspect that Democrats have moved past the concerns of 2004 and 2006 about ‘electability’, considering both of the remaining candidates very electable over McCain. Now electability has become a Republican issue. They’re thinking defensively while Dems are thinking offensively and the latter is a clear sign of strength no matter which candidate prevails.