UPDATE: As commenters have pointed out, polls and predictions are a fun fix for political junkies. What matters is GOTV! Make O-mentum a reality! [EVENTS / VOLUNTEER / PHONE BANK]
UPDATED:
I revised the title to reflect the reality that these projections really mean that Obama and Clinton are essentially within 50-75 delegates (3%-4%) of each other. My earlier title was meant to get people pumped, but I think it is safer that I temper expectations.
Obama has momentum and he's closing. But that's all I'd read into the polls and these projections. For the record, Obama trailing Clinton by anything less than 100 delegates is a big victory in my books. GOTV, volunteer, phonebank, vote!
Note: I have maintained a consistent methodology for the last week and have always had Clinton leading the projecting delegate count. (see here and here) After adding in today's polls, I have Obama up for the first time. Obama quite clearly has the Big Mo'!
.
Primer: Delegate Process Explained
.
A while back, I posted a diary outlining my method for projecting delegate counts based on the popular vote percentage. This model was based on data from the 2004 Democratic primaries, which also featured a three candidate dynamic (Kerry, Edwards, Dean/Clark). However, with John Edwards out of the race, I reconfigured my model because many polls include him as a candidate. Moreover, there are many Democrats who will still vote for Edwards on Tuesday.
.
Current 'Super Tuesday' Projection**: Obama 783, Clinton 777, Edwards 7
** - Does not include delegates from states where no recent poll (2 weeks) is available.
Delegates from States with No Recent Poll: 111 (AR, KS, ID, AK, ND)
.
BACKGROUND:
.
#1: Adjusting for John Edwards' Departure (adj. JE)
In 2004, when the field had winnowed to two candidates (Kerry, Edwards), the worst combined performance of the remaining candidates (Dean, Clark, Lieberman, Kucinich, Sharpton) was around 4% of the vote. Therefore, the column 'adj. JE' starts out with a base of 4%, regardless of whether not John Edwards was included in the original sample. In order to account for Edwards die-hards (those who will vote for him no matter what) I add to this 4%, a specified fraction (0.35) of the Edwards vote above 4%. Let me clear this up with a quick example:
Let's say a poll has John Edwards at 14%. I move 4% out this 14% to the 'adj. JE' column to start out. I then add in 0.35 x (Edwards' vote share above 4%), i.e 14% minus the base 4%. Mathematically...
adj. JE = 4% + [0.35 x (14% - 4%)] = 4% + 3.5% = 7.5%
Where do I get this 0.35 from? In exit polls thus far, 35% of Edwards supporters indicated that they had no second choice. I expect, at the very least, that this contingent of Edwards supporters will remain loyal to their candidate.
.
#2: Weighting Polls (factoring in momentum)
If you have been following the polls recently, you'll no doubt notice a strong movement towards Barack Obama. To account for this movement, and working with the assumption that the most recent poll is the most relevant one (accuracy aside), I weight polls by the day they were published. For example, a poll that came out today, carries a weight of 10. A poll from 2 days ago, carries a weight of (10-2) = 8. To prevent any negative weighting, I assign all polls published more than nine days ago a weight = 1.
It is my hope that when tabulating poll averages this greater emphasis more recent polls adds to the accuracy of my projections.
.
#3: Undecideds
Nothing special here, but I thought I'd be thorough. With the newly factored in support for Edwards (adj. JE), I then find out what the new number of undecided voters is. This, quite simply is....
Undecided = 100% - Clinton Support - Obama Support - 'adj. JE'
.
#4: 'Pushing' Undeiceds
Most polls include a sizeable chunk of undecided voters. How do we 'push' undecideds in these polls? Let us first assume that undecideds are choosing between Clinton and Obama. I think this is justified because the percentage of late deciders choosing between Clinton and Obama has gone up with each race. With John Edwards now officially out of the race, I expect almost none of the late deciders to choose Edwards or Gravel.
Iowa: Clinton 27% - Obama 32%
New Hampshire: Clinton 37% - Edwards 36%
Nevada: Clinton 41% - Obama 42%
South Carolina: Clinton 21% - Obama 51%
Average 'Break' of Clinton-Obama Late Deciders: Clinton 43%, Obama 57%
I 'push' undecided voters into the Clinton and Obama columns in the above percentages.
.
#5: From Polling Percentages to Delegates
For detailed information on how I go from vote percentages to delegate shares, look here. Truth be told, much of the analysis in that diary has been rendered useless by John Edwards' departure. Most of the procedures become relevant only in the event that John Edwards gets more than 6% or 7% of the vote in a given state. There are indications that he might just secure such percentages in some states, bot not in a manner that will make him a player in the delegate count.
.
STATE-BY-STATE RESULTS:
Note: As I explained at the top, I have not made any projections for the states without any relevant polling data. Most of these are small states, whose 'moods' could be very different from that of the nation at large. I am not comfortable 'projecting' them until I have atleast one recent poll. I leave it up to you decide on how these states will 'break'.
.
.
SPREADSHEET_DATA (cut and paste into your own spreadsheet app)