| I haven't been up with SSP as much as I would have liked to be over the past few days, but the brewing brouhaha in Alaska was too good for me to pass up.
First, let's take a quick look at the geography of the vote, by state house district (blue for Murkowski, red for Miller):
Murkowski did well in the southeastern part of the state, as well as the Arctic North, and the Aleutians. In the Anchorage-area inset below, you can also see that she did well in the Dem-leaning (and less GOP-vote heavy) Anchorage-city districts, but got absolutely crushed in the GOP-vote-rich Mat-Su Valley north of the city (no surprise, given Palin's involvement).
For the analysis and maps presented here, I use Alaska House of Representative Districts, numbering 40 in total. I also rely extensively on 2008 absentee and provisional ballot rates, which I think is a reliable assumption; Election Day only turnout between the two contests is remarkably similar:
I'd also hoped to use Young v. Parnell in 2008 as a proxy for Murkowski and Miller respectively, but the correlation there isn't as strong:
Between the 40 House Districts, there was a huge variation in the percentage of GOP ballots that were cast absentee in 2008, ranging from 2.96% in HD-39 (Nome), to 20.47% in the Anchorage-based HD-18.
Using this analysis, the outcome already begins to look grim for Lisa Murkowski. Here is a comparison of her Election Day performance and the percentage of votes cast absentee in 2008:
From this, I think we can expect fewer absentee ballots from her stronger areas, and more absentee ballots from areas more favorable to Miller.
Things look even worse for her when we start to factor in the fact that there is variation in the number of votes cast per HD, and therefore variation in the number of absentee ballots returned per HD. In 2008, this varied from a tiny 28 absentees in HD-39, to 741 in HD-34 (Rural Kenai). Assuming that each HD returns the same percentage of total absentees as they did in 2008, this is what we get:
That negative correlation we saw earlier becomes even stronger, as Miller strongholds are expected to return more of the absentee ballots outstanding. Adding the additional assumption of Murkowski's performance among absentees to be the same as her Election Day performance (tenuous, yes), she can be expected to receive about 48.51% of the absentee vote.
Repeating this analysis for provisional ballots, things are no better for Lisa.
While the relationship between her Election Day performance and provisional ballot rate seems to be better, adding into consideration the percentage of statewide provisionals per HD shows that familiar negative correlation for her. Making the same assumptions as above, Murkowski can be expected to get 48.82% of the provisional vote.
Extrapolating further, using the oft-cited 7,500 absentees and assuming 5,000 provisionals (5,200 were cast in 2008), we can expect Miller's margin to increase by 223 votes among absentees and 118 votes among provisionals. For Murkowski to pull this out assuming those figures for outstanding ballots, Murkowski would have to perform 8.04% better than she did on Election Day among absentees and provisionals...which is quite the tall order! More outstanding ballots means the swing she needs will be less, but it also means more ballots through which she could lose more ground. Given this, there isn't too much reason to be optimistic for Lisa, and we might just have to put the 'W' in the column for the Grizzly Momma.
You can also check out my dataset below the flip.