It’s silly season for election watchers, that time of year where every uptick or downtick in the polls is assumed to have great meaning: the first tip-off to a last-minute comeback, or the first sign that a candidate who seemed to have his race locked up won’t be able to close the deal.
Late comebacks in Senate races do sometimes occur — it was at about this point in the 2006 election, for instance, that Jim Webb. a Democrat, emerged with a lead over George Allen in the Virginia Senate race, before going on to narrowly defeat him. In many or most cases, however, rumors of a candidate’s demise — or comeback — may be greatly exaggerated, given how little time there remains to make up even fairly small polling deficits.
Take the Senate race in California, for instance. There have been six independent polls released there in the last week or so. Five show a lead for Barbara Boxer, the Democrat, but by margins ranging from 1 to 5 points. The sixth poll, from Wilson Research Strategies (ordinarily a Republican polling firm, but this particular survey meets our standards for being nonpartisan) shows Carly Fiorina with a 3-point lead.
If you look at just those polls, you’d get the impression that the race is — at most — leaning toward Ms. Boxer, and may even be too close to call. But is Ms. Fiorina — who had seemed to be down by 4 or 5 points in the polls before — in fact gaining ground?
Actually, it’s not so clear. Five of the six polls were also in the field in late September or very early October (Wilson Research Strategies is again the exception; this was the first time they’ve polled the race.) Four of the five indeed show Ms. Fiorina having improved her standing. But none have it improving by more than 3 points, and the fifth poll shows a small improvement by Ms. Boxer instead. On average, these polls show Ms. Fiorina having gained only about 1 point.
Ms. Fiorina will take that point, to be sure. She also had a strong fund-raising quarter, and looks strong in comparison to Meg Whitman, the Republican nominee for governor, whose standing has eroded. She’s plenty capable of winning.
Nevertheless, she remains the underdog. Any tightening in the race has been fairly subtle, and could easily just reflect statistical noise. The perception that she has improved in polls by more than a percentage point or so stems in part from the fact that the mix of pollsters who have surveyed the state most recently happen to be somewhat Republican-friendly. If polls that had been more Democrat-friendly before — like CNN or Public Policy Polling, which had shown Ms. Boxer with a lead in the high single-digits — were to show the race closing to within a point or so, that could be more meaningful.
Meanwhile, time is ticking off the clock, and many Californians are already sending in their absentee ballots. A candidate who is trailing in the polls has to make up some ground each week to have a strong chance of coming back; gaining just a point or so in a two-week stretch may not be much better than treading water.
But I don’t mean to pick on Ms. Fiorina — the same story exists in several other states.
In Wisconsin, for instance, a new poll from St. Norbert College shows Russ Feingold just 2 points behind Ron Johnson — his best figure in a nonpartisan survey in some time. But this is the first time that St. Norbert has gone into the field: we have no idea whether they’re in fact capturing tightening (could be), or their poll just happens to be a bit goofy (could also be). None of the several pollsters who had Mr. Johnson ahead by about 7 points before have yet weighed in with news of a materially tighter race.
Likewise, in Washington, the apparent movement toward Dino Rossi has mostly to do with the particular mix of pollsters who have been active in the state in the last week or so — side-by-side comparisons of polls conducted by the same firm do not show him gaining any particular ground.
There is one state where a candidate in a fairly tight race does seem to have made up ground: Pennsylvania, where three polls now show a roughly tied race between the Democrat Joe Sestak and the Republican Pat Toomey, who had held a roughly 7-point lead before. In particular, Mr. Sestak’s standing in the Quinnipiac poll — where he’s gone from 7 points behind to just 2, is impressive, since that poll has generally had good news for Republican candidates this cycle.
Still, for reasons that I’ve articulated before, our model is deliberately designed to be quite conservative about the way it incorporates new surveys. And so it’s not ready to completely forget about the polls from a couple of weeks ago that showed Mr. Toomey with a clearer advantage, particularly since most of these pollsters have yet to re-survey the race.
There’s also the matter of the Rasmussen Reports poll that shows Mr. Toomey still leading by 10 points, and which is only about a week old. Rasmussen polls are generally Republican-leaning, but you can apply an adjustment for this (which we do; it amounts to 1 or 2 points) and it still shows a fairly clear lead for him.
In addition, there is the fact that the fundamentals of the race might not seem to favor Mr. Sestak. Pennsylvania is just a couple of points more Democratic than the country as a whole, but the Republican advantage on the generic ballot is more like 6 or 7 points. Given equally strong candidates, Republicans would probably be favored in an open-seat race there in a cycle like this one. And Democrats have a fairly weak nominee for governor, which could harm Mr. Sestak at the margins.
With all that said, our model’s forecast for Mr. Sestak — it has him with just a 16 percent chance of winning, although this is up from 6 percent before — certainly does seem a little stingy. Although I’m not sure that this qualifies as much more than superstition, he has made a couple of late comebacks before. His fund-raising has held up well and is on par with Mr. Toomey’s, and he’s generating a lot of excitement on liberal blogs, which can translate into more people making phone calls and knocking on doors.
But I’m not sure that the model is being a lot stingy with Mr. Sestak. Taking a simple average of the four polls that were very recently in the field — Quinnipiac, Public Policy Polling, Morning Call, and Rasmussen — would show Mr. Toomey with a lead of about 3 points. We instead show him ahead by 4 points. A lead of either 3 or 4 points — which roughly matches the one that Ms. Boxer and Ms. Murray seem to have at the moment, for instance — is quite meaningful at this stage of the race.
Of the dozen or so Senate and governor candidates to whom the model attributes just a 10 or 20 percent chance of winning, it’s likely that at least one or two of them will come back. If you want to make the case that Mr. Sestak or Ms. Fiorina are especially likely to be one of them, I have no real problem with that. (For that matter, I have my own “hunches” about the races: for instance, I like the odds the model is offering for these two candidates — Ms. Fiorina and Mr. Sestak — but I might not say the same for, say, Mr. Rossi, who seems to be the wrong type of candidate for this cycle.)
In general, however, over the past decade or so, the polls have been quite reliable late in the race, and such comebacks have been fairly unusual.
Their overall effect of all these changes, meanwhile, is quite neutral. Republicans have an 18 percent chance of winning the Senate, according to our forecast — little changed from the 17 percent chance they had earlier this week.