(From comments crossposted at Swing State Project)
I decided to analyze the individual precincts, and compare their performance to that against the Jefferson-Carter race in 2006 (Jefferson won that by about 12 points.) This would seem to give us a far more accurate result than county-to-county comparison, as partial results will skew the data - certain precincts, for example, are giving 100% of their vote to Jefferson.
Overall, Cao is doing far better in Jefferson Parish (ironically) than Carter did, but somewhat worse in Orleans Parish.
I analyzed 59 individual precincts in Jefferson (these were just what were reporting when I did this analysis). Absentees/etc. are ignored here...
Jefferson won these in 2006, 5997-2734. 68.6% of the 2-way vote.
Cao is winning these now, 6752-3985. 62.9% of the 2-way vote.
Overall, Cao is doing ~30% better than Jefferson did, and overall turnout is up by 22%.
Rather than add up 192 precinct data in Orleans, I took out the first 25 that were reporting.
Cao is winning these, 1068-874. 55% of the two-way vote
Carter won them in 2006, 1141-758. 60% of the two-way vote
Turnout between the two races is up by 2-3%. But Cao is performing 5% worse than Carter in Orleans.
I'm predicting the final results (minus absentees/provisional!) to be:
Total 2-party votes in Jefferson (ignoring absentees/provisionals/etc): 20,785 Cao 60.9% (12659), Jefferson 39.1% (8127)
Total 2-party votes in Orleans (same caveats as above): 46287
Cao-43.6% (20,181)
Jefferson - 56.4% (26,106)
Final results:
Cao - 32,840 - 49%
Jefferson - 34233 - 51%
A VERY close race! This is effectively a tossup once we take into account the large error, and lack of accounting for early/absentee/provisional votes (absentees should favor Cao, provisionals favor Jefferson... they should cancel out, but that adds up even more error.)
EDIT: Analyzed more precincts from Orleans, total of 48 now...
Cao is winning these with 61.2% of the vote; turnout is 3,931
Carter won these with 64.3% of the vote; turnout was 4,285.
Overall, Cao is doing 3% worse than Carter, and turnout is down 8%.
This looks VERY bad for Jefferson... I'm now predicting:
Total 2-party votes in Jefferson (ignoring absentees/provisionals/etc): 20,785 Cao 60.9% (12659), Jefferson 39.1% (8127)
Total 2-party votes in Orleans (same caveats as above): 41,505
Cao-45.6% (18,926)
Jefferson - 54.4% (22,579)
Final results:
Cao - 31585 - 50.7%
Jefferson - 30706 - 49.3%
EDITED
Final results: Cao wins with 51.4% of the two-way vote. Orleans had higher turnout than I expected, but the stronger Cao performance there more than balanced it out.
My predictions seem to have been especially accurate in Jefferson parish.
More thoughts, copied from my comment below:
I think its pretty amazing how accurate I was able to be with a sample of only 20% of the overall precincts. Maybe I should do this more often!
But thinking about it, this was a pretty easy race to predict, because of how racially polarized it was. There would essentially be only one real variable in determining vote count - how much of the African-American vote Jefferson was able to hold. So as long as I got a decent precinct sample, I would be able to determine that quite accurately. This would probably have a MUCH larger error for more complex CDs.
Thanks for reading this! Though I'm kind of annoyed that most people seem to be ignoring my work.
Final comments:
This was not a Democrat vs. Republican race. This was a Jefferson vs. Not-Jefferson race. As my analysis shows, there was an extremely correlation between how Karen Carter performed in a precinct in 2006, and how Cao did in 2008. Cao's victory came because he was able to hold Jefferson to about the same margin as Carter did in Orleans, and improve dramatically on Carter in Jefferson parish.