Republicans have a 59 percent chance of taking back the Senate, according to FiveThirtyEight’s latest forecast. We’ve got a new poll in Alaska, a state that was in desperate need of new data.
The GOP’s path to a Senate majority becomes a lot easier if it picks up Alaska. Republican challenger Dan Sullivan has a 64 percent chance of defeating Democratic Sen. Mark Begich in the Last Frontier. Sullivan’s chances are up from Monday, thanks to a Public Policy Polling (PPP) survey. PPP found Sullivan ahead 43 percent to 41 percent (with third-party candidates included in the matchup). That’s a big change from PPP’s last survey. Two months ago, PPP showed Begich ahead by 6 percentage points.
PPP is the latest nonpartisan survey to move in Sullivan’s direction. Sullivan has led all three nonpartisan polls released since his primary victory last month by an average of 3.3 percentage points. In the two months before the primary, two nonpartisan polls taken in Alaska by PPP and YouGov had Begich ahead by an average of 9 points.
On the other hand, polls in Alaska sponsored by liberal-leaning groups, such as the AFL-CIO and Senate Majority PAC, paint a different picture. These three surveys (two conducted by Harstad Strategies and one by Hays Research) all have Begich up — by an average of 3.7 percentage points.
Should we pay any attention to the partisan polls?
In past years, surveys sponsored by candidates and party groups (e.g. the Democratic Senatorial Campaign Committee) have had a bias in favor of the group’s preferred candidate of 4.25 percentage points. Moving the average of the sponsored polls in Alaska by 4.25 points toward Sullivan would go a long way toward bringing the sponsored and non-sponsored surveys into alignment.
But it’s not an open-and-shut case. Sponsored polls that show their preferred candidate ahead have had less of a bias (about 1 percentage point). That pattern may not apply to polls from PACs, which haven’t been releasing surveys for long.
Finally, the polls released by liberal groups have been “higher quality,” conducted by live interviewers and have called cellphones. The non-sponsored surveys have been conducted with robo-calls (PPP and Rasmussen Reports) or online with a non-probability sample (YouGov). Although YouGov has historically done well in the United States, it’s clear that it is having trouble maintaining its panel (a key component to non-probability sampling) in Alaska. That could lead to a higher error rate than normal.
Of course, we built a model so we don’t have to weigh all these factors in an ad hoc way. FiveThirtyEight’s forecast currently leans more toward the nonpartisan polls. Sullivan is projected to win by 2.5 percentage points.