Blanche Lincoln has released an private poll which shows her down only 9 points to John Boozman in Arkansas. This contradicts public polls of the matchup which have shown her down by 31 points (Magellan), 29 points (Rasmussen), 25 points (Zata3), 23 points (PPP), 19 points (a Reuters/Ipsos poll that was just released) and 17 points (Mason-Dixon). It also contradicts my regression analysis, which says she should be down by about 19 points, given her (very low) approval ratings, Arkansas’ partisan identification, and the fact that she is being challenged by an experienced opponent. It doesn’t take a genius to determine which figure is the outlier there.
The usual rule of thumb is that publicly-released internal polls have a lean of about 5 points, so that if her campaign released a poll showing Lincoln down by 9 points, that means she’s really down by 14. But sometimes the bias in an internal poll is considerably more than 5 points, depending on the pollster, the candidate, and the circumstances of the race. In this case, the Lincoln campaign seems desperate to fend off the narrative that her campaign is dead in the water and doesn’t deserve fundraising or activist attention, and so the incentive to put out a favorable poll might be especially strong.
They’re desperate to fight off this narrative because it’s absolutely true. In a cycle where we have so many authentically competitive Senate races, it would be absurd for national Democrats to spend more than a pittance on her. (Likewise, it would be absurd for Republicans to give Boozman too much help — he shouldn’t need it.) Suppose that Lincoln is in fact only down by 9 points. That would translate to about a 15 percent chance of her winning, according to our model. That might justify spending something on her, but it would nevertheless place her behind the Democrats in Colorado, Delaware, Pennsylvania, Nevada, Illinois, Washington, Ohio, Kentucky, Missouri, California, Wisconsin, New Hampshire and North Carolina in the pecking order for resource allocation, all of whom are locked in races in which the odds are closer than 85/15 in one direction or the other.
But we should probably not take an internal poll quite that literally, especially since there’s plenty of other evidence to contradict it. If Lincoln is really down by 14 points — assuming her campaign poll has a 5-point house effect but is otherwise honest — her odds of coming back are only about 5 percent. If she’s down by 19 points, as my regression model suggests she “should” be, her odds are around 1 percent. And if she’s down by 25 points — as the average of the public polls shows — her chance of winning is only 0.2 percent, making her a 500:1 longshot, according to the model.
Candidates, particularly incumbent candidates, just don’t come back from deficits like this very often. Harry Reid is considered hugely lucky to have moved back into a dead heat in Nevada — but he was down by only about 10 points in his worst moments, not 20 or 25 points, like Lincoln is. Lincoln would probably need at least two major strokes of luck to come back against Boozman — two macaca moments. And there’s no particular reason to think that she’s liable to get them: this is a deeply unpopular incumbent (her approval ratings are in the 30s) in a deeply anti-incumbent climate, in a deeply red state in a deeply red cycle, and she struggled to win the majority of votes in her own party primary. A penny saved on Lincoln is a penny that could spare another Democrat.
EDIT: For those wondering, our model assigns Bill Halter only a 5 percent chance of beating Boozman in the hypothetical event that he magically replaced Lincoln on the ticket.