Something smells fishy in Alaska, and it's not the salmon. Sen. Ted Stevens appears to be headed to reelection (pending 40,000 uncounted absentee ballots), despite his conviction on seven counts of violating federal ethics laws.

Stevens was widely thought to have little chance of winning. FiveThirtyEight.com's Nate Silver had rated the race as "safe Democratic," and projected a victory of 11.3 points for challenger Mark Begich.

Stevens is a titan of Alaskan politics, the state's iconic public figure. The Anchorage airport is named after him, and it's not uncommon to hear him referred to as "Uncle Ted." People have a soft spot for the guy. So, a candidate who voters might be embarrassed to admit they're supporting for emotional reasons turns in an unexpectedly strong performance. That seems like a familiar story.

If the national results have put to rest discussion of the Bradley effect, we may have found a replacement to argue about: the Stevens effect. Are people reluctant to admit to pollsters that they're voting for a felon? Alaska's only congressman, Don Young, is also headed for an unexpected reelection. The same theory could well apply to him. Young, though not indicted, was in electoral trouble in the first place for reasons similar to Stevens.

If those absentee ballots don't bring the results back in line with predictions, the Stevens effect seems at least a plausible theory. Or maybe nobody really knows how to poll Alaska.

(Full disclosure: I worked briefly as an organizer for the Alaska Democratic Party earlier this year.)