Although Ted Stevens holds a small lead in Alaska and is the favorite
to retain his seat, the outcome is not as inevitable as it might appear
to be. Stevens currently holds a lead of 3,353 votes, or about 1.5
percent of the votes tallied so far. But, there are quite a large
number of ballots yet to count. According to Roll Call,
these include "at least 40,000 absentee ballot, 9,000 early voting
ballots, and an undetermined number of questionable ballots".
Indeed, it seems possible that the number of "questionable" ballots could be quite high. So far, about 220 thousand votes have been processed in Alaska. This compares with 313 thousand votes cast in 2004. After adding back in the roughly 50,000 absentee and early ballots that Roll Call accounts for, that would get us to 270 thousand ballots, or about a 14 percent drop from 2004. It seems unlikely that turnout would drop by 14 percent in Alaska given the presence of both a high-profile senate race and Sarah Palin at the top of the ticket.
But even if Begich were to make up ground and win a narrow victory, this would seem to represent a catastrophic failure of polling, as three polls conducted following the guilty verdict in Stevens' corruption trial had Begich leading by margins of 7, 8 and 22 points, respectively.
The emerging conventional wisdom is that there was some sort of a Bradley Effect in this contest -- voters told pollsters that they weren't about to vote for that rascal Ted Stevens, when in fact they were perfectly happy to. Convicted felons are the new black, it would seem.
The problem with this theory is that the polling failures in Alaska weren't unique to Stevens. They also applied to the presidential race, as well as Alaska's at-large House seat. In each case, the Republican outperformed his pre-election polling by margins ranging from 12 to 14 points:
Contest Projection Result Delta
AL-ALL Berkowitz +6.4 (i) Young +7.7 GOP +14.1
AL-Sen Begich +12.9 (ii) Stevens +1.5 GOP +14.4
AL-Pres McCain +13.9 (iii) McCain +25.3 GOP +12.4
(i) Pollster.com Trend Estimate
(ii) FiveThirtyEight Polling Average
(iii) FiveThirtyEight Trend-Adjusted Estimate
are three plausible explanations I can think of to explain this
discrepancy. The first and most likely is that the Democratic vote
became complacent and did not bother to turn out. The outcome of the
presidential contest was not going to be close in Alaska, and Barack
Obama's victory in the Electoral College was apparent as of about 4 PM
local time. Begich supporters, moreover, may have looked at the polls
and concluded that their candidate was far enough ahead that they
didn't have to bother to vote. Meanwhile, the Republican base was going
to turn out no matter what because of their enthusiasm for Sarah Palin.
There seems to be a sort of danger zone at about 10 points wherein a
candidate is far enough ahead that many of his supporters assume the
race is in the bag, but not so far ahead that he is immune to poor
turnout (a similar dynamic affected then-Governor Jim Blanchard of
Michigan in his 1990 race against John Engler).
The second possibility is that a substantial percentage of the Democratic vote is tied up in the early and absentee ballots that have yet to be counted. We know that Barack Obama overperformed among early voters in many states, and Alaska may be no exception. (Although, I would guess that the absentee vote is predominately rural, whereas Begich's base is in Anchorage).
The third possibility is that a lot of those "questionable" ballots are Democratic ones, and that there have been irregularities in the voting tally. Although this is the least likely possibility, Alaska is a provincial state with some history of corruption, and Democrats ought to be making sure that too many of their ballots haven't been disqualified.