Diving Deep For Dem Delegates (State Poll Analysis) By: Mark W Adams

Be warned, this it more wonkish than anyone who isn't getting paid should ever write. I have no idea what possessed me except the chance to predict the future. Follow me as I work through the democratic primaries state-by-state.

According to my calculations, after Tsunami Tuesday once all the elected and Superdelegates are counted, Clinton will have AT LEAST 928(54%) on the morning of February 6th. Barack Obama can count on a minimum of 569 delegates (33%), and Edwards will have no less than 220 delegates (13%) going into Louisiana and the rest of the country. Again, this is a MINIMUM total with several states ignored for lack of data.

Adjusted Chart of Delegates Through Tsunami Tuesday

Distributing uncommitted voters via same percentage of those expressing preferences based on state polling data. (Probable delegates awarded in parenthesis):

Table Of Tsunami Tuesday Delegates

Date	State	Del	S.D.	Clinton Poll(Del)		Obama Poll(Del)		Edwards Poll(Del)		Unknown Not Awarded
1/3	Iowa	45	11		14		16		15
1/8	N.H.	22	8		9		9		4
1/15	Mich	128	29	0%	-	0%	-	-	-	100%	(157)
1/19	Nev	25	8	36%	(9)	34%	(9)	30%	(7)	-	-
1/26	S.Car	45	9	35%	(16)	49%	(22)	16%	(7)	-	-
1/29	Florida	185	25	0%	-	0%	-	0%	-	100%	(210)
2/5	Alab	52	8	58%	(34)	31%	(18)	11%	-	29%	(17)
2/5	Alaska	13	5	-	-	-	-	-	-	100%	(18)
2/5	Ariz	56	11	41%	(27)	23%	(15)	-	-	36%	(24)
2/5	Ark	35	12	65%	(22)	19%	(7)	16%	(6)	-	-
2/5	Calif	370	70	53%	(195)	32%	(119)	15%	56	-	-
2/5	Col	55	16	48%	(26)	27%	(15)	25%	(14)	-	-
2/5	Del	15	8	41%	(9)	17%	(4)	-	-	42%	(10)
2/5	Geor	87	17	39%	(33)	41%	(36)	20%	(18)	-	-
2/5	Idaho	18	5	39%	(7)	42%	(8)	19%	(3)	-	-
2/5	Illinois	153	32	25%	(46)	50%	(93)	-	-	25%	(46)
2/5	MO	72	16	47%	(32)	27%	(20)	26%	(20)	-	-
2/5	N.Jers	107	20	51%	(65)	17%	(22)	-	-	32%	(41)
2/5	N.Mex	26	12	-	-	-	-	-	-	100	(26)
2/5	N.York	232	48	55%	(154)	17%	(48)	-	-	28%	(78)
2/5	N.Car	91	19	38%	(35)	30%	(27)	32%	(29)	19%	(21)
2/5	N.Dak	13	8	-	-	-	-	-	-	100%	(21)
2/5	Okla	38	9	46%	(17)	20%	(8)	34%	(13)	-	-
2/5	Tenn	68	17	-	-	-	-	-	-	100%	(68)
2/5	Utah	23	6	-	-	-	-	-	-	100%	(23)
Tot	Del	-----	-		750		496		192	--
Pct	Del	-		45%		30%		22%

Note, the percentages are not the polling percentages, but the effective distribution of support when undecided votes follow the known trends.

So, where'd I come up with this voodoo?

Keep Reading to see my sausage recipe.


In this episode I go beyond the simple arithmetic of multiplying national polling data with available delegates to see where we are and where we might end up. I've worked up through Tsunami Tuesday (Feb. 5th) so far, and then later will get to later contests. So far the data, especially on Edwards, fits nicely with national trends I'm using to make sure I'm not all wet.

The good folks at Wikipedia are keeping track of the state-by-state primary polls for us, and now that I've spent the weekend updating both my computer software and my brain's hardware, I think I can try and make some use of all that data, and more importantly present it to you in a way we both can understand without consulting a statistician.

Real Clear Politics is also doing journeyman work nationally and pretty much logs every poll out there and keeps a running average of the national polls. Another source that cuts through some of the white noise is USA Elections Polls who have compiled the Feb. 5th Tsunami Tuesday races in one place and keeps track of the big states with all the juicy delegates -- and just about everything else too.

One glance tells us that while all this data is useful, there are serious problems reaching any conclusions without really diving into the numbers. Using the national polls as a benchmark, we see John Edwards holding in the mid to low teens, slightly lower than he was before Iowa and New Hampshire. Now we all know that when dealing with individual contests nation-wide it's foolish to base any conclusions just by that data, especially when we have large undecided/unknown numbers in the mix -- let alone those margins of error.

It's the undecideds that will turn this thing one way or another. What did Olbermann call it, the Keith rule? The margin of error multiplied by the percent of undecideds equal a poll that doesn't tell you anything reliable? Something like that.

Now, the trick is to predict not just who will win, but how. With three viable candidates, the chances that nobody gets to the magic number, 2025 delegates, increases the chances that there's a spoiler/kingmaker role to be played. With Edwards at the bottom of the totem pole, it's his role I'm focusing on to either "prove" or debunk my idea that if he were to quit running, Hillary could seal up her win by just receiving a small fraction of his supporters. This gets us half-way there. (I know, this is a long post, but it's a long season with lots of states to go.)

NATIONAL POLLING DATA

CNN's Poll (pdf) asking who would you "support" if it were a two person race pushes Hillary over 50% by 3 points, exactly the polls margin of error. The poll is striking for it's certainty. 3 points is a low margin of error, but registering only 2% with no opinion it unheard of. The poll may indeed be accurate, and reflective of what will eventually happen, but it's not asking who you will "vote for" but who you "most likely will support." I don't know if this makes any difference and how hard people were "pushed" to answer, but every other national poll has from 10% to 20% unknown. Maybe this is CNN's attempt to crush MSNBC and the Olbermann Rule in one fell swoop.

Even with CNN's poll included, RCP's average remains at 11.8% undecided, Edwards pulls in 13.2% in the average of the national polls, which fits with where CNN has him (12%), with the only "outlier" being USA Today/Gallup at 20%. This is not out of line with my state by state analysis. I have Edwards at 13% of the delegates awarded by Feb 5. At that rate, if he holds steady, Edwards will command about 526 votes at the convention. That makes him a player.

For the sake of trying to be reasonable, I'm not simply going with RCP's average numbers, but throwing out the outliers for our base line numbers, what we'll use to compare the other data to make sure it's in the ballpark. That means that Edwards national numbers stand at 11.5% (5 points down from the data I was using Friday) when I throw out Gallup, and the undecided are
14% when I get rid of CNN's preposterously accurate number of people who have made up their minds. Again, this checks out with the state by state data.

If you add up Edwards support plus the undecided vote, that leaves 25.5% up in the air if John quit today. That's about 895 of the 3512 non-superdelegates. Only Gallup (go figure) has Clinton and Obama's support even at 33%-33%. I believe that their track record of over sampling in favor of Obama (actually undersampling Hillary's support) is the stuff of legend since they had him beating Hillary in New Hampshire by 13 points (so did Zogby). All the other polls have Hillary up between 4% and 15% (RCP average 7.4%)

A 50-50 split is the most optimist result Obama could get of the undecided/Edwards vote. Since he trails in the Superdelegate endorsements 177 to 72, even if all of Edwards 28 superdelegates went to Obama, he loses. Obama needs ALL of Edwards elected delegates, the ones he has now and the ones he's going to get to win, at least under this broad brush look at the national polling data.

EARLY STATE DELEGATE ANALYSIS

MICHGAN 1/15 (0)

A beauty contest. You wolverines really blew it.

Vote Mitt! Confound the Opposition!

NEVADA

1/19 (25) [New Poll out today! Keep reading]

This caucus is almost a black hole. The most recent polling data is six weeks old and while ARG still had Richardson at 2% and Dodd and Biden at 4%, Hillary was an amazing +27% to win (Cl-45%, Ob-18%). Edwards at 14% was at the threshold the DNC rules requires to award a candidate any state's delegates -- 15%, the same place he was in the same poll last October (with Clinton at 51% and Obama only at 11%).

This state is truly in flux with Dodd, Biden and Richardson out and the Cullinary Union endorsing Obama. (The grape vine tells me there are a lot of rank and file SEUI members still endeared to Edwards for marching with them during their strike.)

Despite the volatility, since Edwards has remained steady without campaigning there, and the anti-Hillary/Obama votes represented by the combined 10% that was going to other candidates added to the 10% unsure mean I can predict with confidence that again the very least Edwards will get is the DNC required 15% of the Nevada vote, or at least 3 delegates.

Working with the tightest poll numbers so far (Cl-45% Ob-18% Ed-14%) that still leaves about 23% undecided/unknown. The candidates respective portion of the voters with

known

preferences works out to and effective split of CL-58.44% Ob-23.38% Ed-18.18%.

Distributing ALL the delegates according to the tightest/latest poll leaves a delegate count of Cl-14, Ob-7, Ed-4.

I'll be using this method throughout the rest of the states where Edwards at least is polling in the double digits.

[BUT WAIT! THERE'S MORE! The race just got a whole lot tighter -- tied in fact, at least statistically. According to the very, very latest poll, it's Clinton 32%, Obama 30%, and Edwards 27% -- in a caucus where anything can happen. Effective percentage after distribution of unknowns: 35.96% 33.71% 30.34% with a delegate division of 9, 9, 7, or 9, 8, 8,
(HT - Bowers).]

SOUTH CAROLINA

1/26 (45)

It's certain that Obama will win, and win big in certain states, and this could be one of them. He's still looking good in South Carolina at 40%-42%, hitting 50% in Survey USA(pdf), beating Hillary by 7%-20%. But Edwards is campaigning hard. The table of the Wiki data awards Edwards zero out of S.C.'s 45 elected delegates, putting his support at 14%, just below the 15% threshold cusp. But the poll they use (Rasmussen) still has Richardson in the mix at 2%. All the other January polls have Edwards in the 15%-16% range with other/unsure between 2% and 11.7%.

Let's give John the benefit of the doubt in the State were he was born and was in 4th place at this stage but still won in 2004. I'm putting him at 15% and thus winning 6 of the S.C. delegates next week.

Applying the methodology from before, their effective percentage becomes Cl-34.88% Ob-48.84% Ed-16.28%. (even better than I thought.)

TSUNAMI TUESDAY

UNDER-POLLED STATES

According to USA Election Polls, of the 17 states where we have polling data, Edwards is safely above the DNC's 15% threshold in 5, within the margin of error in 3 more, plus New Mexico(26) is completely up for grabs since Richardson was dominating his home state and only Clinton was polling (barely) above 15% as of the last poll there in September. The data the Wiki is using to awards delegates is 10 months old, and even then had Richardson polling at 33% right after he officially announced. That info might as well have been from the last century.

Utah(23)

Supposedly there is an ARG Poll for Utah cited by both Wiki and USA Election Polls, but good luck finding it, and note the poll is almost a year old -- completely worthless. Best to declare it a non-contest or use the national polls as the baseline.

Tennessee(68)

Obama and Edwards are tied in the last poll -- which is from last March and is also virtually useless.

Likewise the Arizona(56) data is too old to be much use, and while it consistently had Clinton winning, Edwards has been anywhere from 2nd at 18% to fourth behind Al Gore. I'm leaving these states alone as far as handicapping the results let alone John's viability.

Idaho(18)

No polls since July. Calculating that poll divides up the unsure Clinton 39.24%(7), Obama 41.77%(8,) Edwards 18.99%(3). This state is a Republican stronghold, and when Gore was put in the same poll Clinton and Obama were still tight but switched places. The difference with the national polls in this small state should not affect the outcome too much in the likely event that things have changed since summer there.

STATES RECENTLY POLLED

Colorado(55)

Nobody has done a poll in the home of the August convention since September, but that beats the data for the spring and summer polls of the states listed above. With data this old, the latest national polls would likely be as accurate as ARG's Poll(pdf). Applying preferences to the unsure in the survey, the effective support is Clinton 4%(26), Obama 27%(15), Edwards 25%(14) -- I awarded Obama and Clinton an extra delegate each to account for national trends.

Georgia(87)

The are fresh polls in Georgia, and Obama, Clinton, Edwards have been steady at 1,2,3 respectively since summer. While it looks like the folks doing the Wiki awarded some remainder delegates to Edwards as a son of the South, I distributed them to the other two in line with the national sentiment. Clinton 38.75%(33), Obama 41.25%(36), Edwards 20.00%(18)

Arkansas(35)

was polled(pdf) between Christmas and New Years that had Clinton at 57%, Obama at 17%, and Edwards at 14% with 12% other/unsure -- which didn't make it into the Wiki. Were it not for the fact that Edwards is barely viable there, I would have put Arkansas in Hillary's home turf column and wrote it off. At it stands, their effective percentages after uncommitted distribution are Hillary-65% Barack-19% John-16%, getting 22, 7 and 5 delegates respectively.

Missouri(72)

Depending on where the Biden/Dodd/Richardson vote goes, Edwards could squeak out one of his few second place finishes in this bell-weather state. The last poll(pdf) had Obama and Edwards just one point apart, behind Hillary by five points. Clinton 49%(32), Obama 27%(20), Edwards 26%(20).

Oklahoma(38)

another state where Edwards beats Obama according to the December Sooner Poll(pdf). Adjusting for uncommitted voters, the state breaks Clinton 45.95%(17), Obama 20.27%(8) 33.78%(13)

HOME BASE STATES:

ILLINOIS (153), NEW YORK(232), NEW JERSEY(107), CONNECTICUT(48), N.CAR.(91)

Both Hillary and Barack will see good numbers from their regional support base on Feb. 5. It's doubtful that Edwards will meet the 15% threshold either in ObamOprah's Illinois or New York, New Jersey or Connecticut, the tri-state area any popular New Yorker can usually count on. Edwards is only polling in the single digits in all these states. Wiki's numbers will suffice and Edwards gets no delegates there.

North Carolina(91)

, sadly, is not the exclusive turf of John Edwards. In fact, the last poll (pdf) from December) has him in second place behind Clinton, but not as far back as he was in November(pdf). Clinton 38.27%(35) Obama 29.63%(27) Edwards 32.10%(29), but things could change if the kids in Chapel Hill start hitting the pavement for John.

CALIFORNIA (370)

The big prize Tsunami Tuesday of course is California, and I'm surprised that there still isn't much out there in the way of polls. Maybe it's just too expensive to sample the Golden State. Note that while Barack Obama, John Edwards and John Kerry descended on South Carolina, Hillary Clinton got on a plane for the West Coast where she's been dominating all the polls. Edwards placed at 13% and 14% in two pre-Christmas polls, but Survey USA's (pdf) early December poll had him at 16% in a contest just between the "Big Three," which is why I initially guessed Edwards receives 59 of California's 370 elected delegates. Hillary takes away half of them, leaving Obama with about 34%.

Using the latest poll (Survey USA Mid Dec.-pdf), and distributing the uncommitted as before, we end up with an effective distribution of Clinton-53%(195) Obama-32%(119) Edwards-15.05%(56). He's just barely viable in this poll where his rating is 14% before looking at undecided and gaining only one point by "doing the math," but with Wiki using the The Field Poll(pdf), and it's whopping 42%unsure/other, I've got to be closer to reality than their figures where Edwards only comes in at 13% and eliminated by them.

For what it's worth, when I use Wiki's numbers -- because of the huge pool of unsure voters -- Edwards effective percentage actually goes up to 18%. I'll stick with the more conservative and poll -- but there's no way Edwards gets shut out of California.

Adjusting undecided votes and the 15% threshold.

By the end of Tsunami Tuesday, Wikipedia only has Edwards with 147 (regular) delegates based on their 15% threshold and just taking the latest state poll numbers without assigning uncommitted voters. Just a straight data drop. I've looked at each state more carefully and believe this is a more accurate representation of what will be the state of play by Feb. 6th. Even with 5 states completely incalculable, Edwards should get significantly more delegates that you would think at first blush.

Now I hate to pick a fight with Wikipedia, since it represents the collective wisdom of the online community, but cutting off the delegate award for him when he's just a hair under 15% in old polls or ones with significant percentages of undecided voters is folly. Think about it. Since an election does not have a box marked "unsure", for Edwards to be at 15% or below, Obama and Clinton would have to share over 85% of the actual vote (splitting 43%/42% or one getting even closer to 50%) to keep Edwards nonviable in any particular state.

I know us social scientists leave "doing the math" to the engineers and astrophysicists, but if you want to keep calling it Political "Science" and not Political "Theater", you got to do the homework.