There’s been some excitement this week about Pennsylvania, where two internal polls commissioned by Democrats show a tight Senate race. One of the polls actually puts their candidate, Joe Sestak, ahead by 3 points; the other has him 1 point behind Pat Toomey. Further contributing to the buzz is the disclosure that Republicans are upping their ad buys in the state. On the other hand, Rasmussen Reports provides for a buzzkill of sorts, having released a poll this morning showing Mr. Toomey 10 points ahead — the largest advantage that Rasmussen has given him all year.
Our forecast yesterday evening, which did not yet include the Rasmussen poll (nor the internal polls; we do not use them in our Senate predictions), showed a projected 7-point victory for Mr. Toomey and gave him a 92 percent chance of emerging as the victor. None of the new information would lead me to deviate greatly from that forecast.
The internal polls. I’m not sure why people take polls released by campaigns at face value. This does not mean that campaigns don’t have very good pollsters working for them. But the subset of polls which they release to the general public is another matter, and are almost always designed to drive media narrative. For an instructive example, Google the term “internal polls”: the first result is a blog post, circa late October 2008, entitled “McCain’s Internal Polls Looking VERY Good.”
What we’ve found is that is that polls commissioned by campaigns and released to the public show, on average, a result that is about 6 points points more favorable to their candidate’s standing than nonpartisan polls released at the same time. (Other analysts have found similar results.) So, just as a first cut, you might take a Democratic internal poll that shows a tied race and “translate” it into nonpartisan terms by adding 6 points to the Republican’s margin. That would mean Toomey +6, a figure that is well within the range of the nonpartisan surveys of the race.
The 6-point rule of thumb is problematic, however, because internal polls tend to exaggerate relatively more or less under different circumstances — from not at all (once in a great while, there may even be an incentive for a candidate to lowball his own numbers), to a double-digit fib under other conditions.
One circumstance where I tend to be particularly suspicious of internal polls is when a candidate is on what I’d call the “threshold of viability.” That is, a poll could conceivably change the perception of whether he has a realistic chance to win his race or not — and therefore, could impact the allocation of scarce resources like activist energy and national party funds.
Mr. Sestak is arguably near to such a threshold. If the polls — which had shown about a 7-point lead for Mr. Toomey, although there have been few public polls released in the past two weeks – were to show a somewhat tighter race instead (say, 3 or 4 points), the race would clearly be winnable for him and would be among the more important battlegrounds in the country. But a 3-point swing in the other direction — putting Mr. Toomey about 10 points ahead — would mean that Mr. Sestak was a longshot. This is the threshold of viability.
The NRSC ad buy. Arguably this is the more compelling indicator: perhaps it suggests that Republican internal polls show the race tightening too. But a couple of things to keep in mind:
First, while candidate and campaign committees have access to some information that the public lacks, they also suffer from a disadvantage, which is that it is hard for them to detach themselves from the race. Campaigns can and do behave emotionally, and depending on their mood, can either read their data through rose-colored glasses, or with their tinfoil hats on. Even the smartest campaigns can have trouble viewing the evidence objectively. So attempting to discern their motives can be a dangerous game — particularly when campaigns will occasionally engage in feints of various kinds to deceive gullible reporters. A few people, like David Wasserman of the Cook Political Report, know how to separate out the signal from the noise — but most gossip and speculation is simply gossip and speculation.
The second reason is a little through-the-looking-glass. Suppose you accept our model’s estimate that, with his roughly 7 point lead, Mr. Toomey has a 92 percent chance of winning. That estimate is formed from an evaluation of polling averages in past races. It is likely, however, that the candidates who held those 7-point leads in the past were continuing to put money into their races (in part, because people tend to perceive races as being closer than they really are, and most people think of a 7-point race as being quite close).
In other words, that 92 percent win percentage is conditional upon Mr. Toomey continuing to run an aggressive campaign. Were he to take the next three weeks off and vacation in Bermuda, his position would be considerably less secure than that. The NRSC’s ad buy might therefore be a reasonable precaution, even if Mr. Toomey’s polls show no tightening in the race.
The Rasmussen poll. Perhaps, as Rasmussen suggests, Mr. Toomey’s lead is expanding instead? That could easily be the case.
Rasmussen Reports polls have a modest Republican lean, however, which seems to be larger in some circumstances than others. For instance, there was a period of about 15 days when no one but Rasmussen was polling the Senate race in Washington, and a series of Rasmussen showed the Republican, Dino Rossi, gaining ground. Three other polls of the state have finally come out in the last 24 hours, however, and they all showed Democrat Patty Murray either holding steady or improving her position. Likewise, a Rasmussen poll commissioned by Fox News earlier this week suggested tightening in the Connecticut Senate race, which hadn’t been surveyed in a week or so — but then a Quinnipiac poll out this morning had the race moving in the other direction. Our statistical model continues to take Rasmussen polls at face value, which is as it should be. But in general, their polls can march somewhat to their own drummer, and I have grown somewhat skeptical of buying into trends implied by Rasmussen polls until they are confirmed by other pollsters.
But didn’t Mr. Sestak make a big, late comeback in the primary? Yes, he did — overcoming what had been a 30-point polling deficit to win his race against Arlen Specter by 8 points. And he ran a smart campaign, conserving his resources until late in the race when voters were just tuning in.
Primaries, however, are much different beasts than general elections — and many of the principles which are critical to understanding the former will mislead you in the latter. The polling in general elections is much more accurate. There are fewer swing voters. Turnout is much higher, which reduces volatility. Voters formulate their preferences earlier, and hold onto them more stubbornly.
All of these things mean that last-minute comebacks are rare, and a campaign strategy contingent upon one is probably dubious. One of Mr. Sestak’s problems, instead, is that Mr. Toomey — who had an extremely conservative voting record when he represented Pennsylvania in the U.S. House — quickly worked to establish himself as the moderate in the race (for instance, by endorsing Sonia Sotomayor) while Mr. Sestak and Mr. Specter were fighting for the Democratic nomination. It is probably too late for Mr. Sestak to alter the fundamental dynamics of the contest.
Still, he has a fighting chance: the percentage of undecided voters in the race remains at about 15 percent across an average of polls, which is higher than all but a couple of other competitive Senate races. And polling suggests that there is an especially large enthusiasm gap in Pennsylvania, which Mr. Sestak could potentially close.
But these things can be discerned from the public polling, rather than trying to read tea leaves from internal polling and advertising spending.