Cross posted from my comment on OpenLeft.com:
So I just did a quick spreadsheet on the delegate situation. If the undecideds and unviable candidates are taken out, we have a 59/41 allocation of delegates before rounding.
There are 241 district level delegates. The 53 CDs have the follow number of delegates, followed by the delegate allocation assuming every CD exactly reflects the statewide polling (ha!):
2 CDs: 3 delegates (that's 2 Clinton, 1 Obama per CD)
26 CDs: 4 delegates (2 Clinton, 2 Obama)
19 CDs: 5 delegates (3 Clinton, 2 Obama)
6 CDs: 6 delegates (4 Clinton, 2 Obama)
That gives us 137 Clinton, 104 Obama for the district delegates.
Now we do the statewide part. My read of the rules is that all statewide delegates (at-large and PLEO) are apportioned proportionally among the viable candidates. If I'm wrong about that, please correct me.
81 at-large delegates: 48 Clinton, 33 Obama
48 pledged PLEOs: 28 Clinton, 20 Obama
Total based on statewide vote: 76 Clinton, 53 Obama
Grand total for delegates allocated on Feb. 5:
Difference: 56 delegates
Now even if the poll today reflects exactly the statewide vote and the undecideds broke like the decides, that's not to say this would be the result. I have no idea how variations is geography would affect things. In general, if Obama does better in over-represented CDs (as happened in Nevada, for example), the total could be closer.
Or he could already be benefiting from favorable rounding enough that this could be close to a ceiling for this statewide breakdown. He's already evenly splitting the delegates in the 4-delegate CDs. He has to break 50% to get the fifth delegate in the 5-delegate and 3-delegate CDs. So, until his total goes to 50%, he 'only' picks up statewide delegates and, at some point, takes the 6th delegate in the 8 6-delegate CDs from Clinton. The next tipping point to get a lot of delegates is the magic 50%, at which point 21 delegates magically flip.
Of course, the actual results don't work like that, since the vote breakdown will certainly not be uniform statewide. But it's useful to see where the magic numbers are in each CD.
commissar added the following (which is completely correct):
I think there is an advantage for Clinton, with her strong Hispanic support, vs. Obama's strong black support, both of which is geographically concentrated.
Just to illustrate, I'll use the four-del. CDs:
26 CDs: 4 delegates (2 Clinton, 2 Obama) = 52 + 52 each
Let's assume that nine of them are heavily ethnic districts: six Hispanic and three black. Then assume in all nine, that support skews the results up to a 3-1 split. This gives us
17 split = 34 + 34
6 Hisp. = 18 + 6
3 Black = 3 + 9
26 TOTAL = 55 + 49
Again, my math is arbitrary, but I think the geo. concentration of ethnic supporters works to Hillary's advantage, because the variation in results will not break evenly.
I'll state this another way: To the extent we see CDs with big majorities that disrupt your model (and we will), those will tend to favor Clinton by approx 2:1, because of ethnic pockets.
So, yes -- I wasn't even attempting to estimate what regional variations would do to the results. And they are more likely to skew the delegate count than to balance out to no net effect.
For one thing, if two candidates have roughly equal support, but A has it spread out throughout a state and B has concentrated support in a region (or, say, the cities), A will get more delegates. In Iowa, Clinton and Edwards were probably helped by the system a bit. In NV and NH, Obama was.
In a further note, I used the same model for New York and Illinois based on the latest poll on pollster.com. The results:
HRC 51 / BHO 25 / JRE 11 / Oth 2 / Und 11
Renormalized after viability:
HRC 67 / BHO 33
150 HRC delegates / 85 BHO delegates = HRC +65
HRC 25 / BHO 50 / JRE 11 / Oth 2 / Und 11
HRC 33 / BHO 67
61 HRC delegates / 120 BHO delegates = BHO +59