Electoral Prospects
What is going to happen on 2 November?
No one really knows, to be sure. But there is a convergence of opinion among those who follow electoral trends. Nearly everyone supposes that 2010 will resemble 1994, when the Republicans picked up fifty-four seats in the House of Representatives.
For what it is worth, I think that they all have it wrong.
Scott Rasmussen recently predicted that the Republicans will gain fifty-five seats in the House. In August, Larry Sabato at the University of Virginia estimated that there would be a pick-up of forty-seven seats. Next week, he will tweak those numbers. Last week, he reported that, if forced to do so right then, he would increase his estimate of Republican gains by single digits. In short, there is hardly any difference between his prediction, that of Rasmussen, and the RealClearPolitics average suggesting that the Republicans will gain fifty-six seats.
Of course, some prognosticators come in a bit lower. Charlie Cook puts the pick-up at fifty-one seats, and Nate Silver of The New York Times has it at forty-nine. Others – Jay Cost at The Weekly Standard comes to mind – put it a bit higher: in his case, at sixty-one.
What they are all doing is understandable. They are interpreting the present in light of the recent past. Most of the time such a procedure makes excellent sense. But it is good to remember that every once in a while there is a sharp change. Barack Obama, Rahm Emanuel, Nancy Pelosi, and Harry Reid set out to change things, and perhaps they have succeeded in ways that they did not anticipate.
One must keep in mind that no Sovietologist predicted the collapse of the USSR and its dismemberment. Decades of comparative stability had lulled them to sleep. Only in the aftermath did the experts become aware that there had been ample warning signs.
There are plenty of warning signs now. Over the last twenty-two months, we have been confronted with one damned thing after another. First, in February, 2009, came the emergence of the Tea-Party Movement; then, in August, 2009, we witnessed the confrontations at the town-hall meetings. In November of that year, everyone was surprised by the verdict handed down in the gubernatorial elections in Virginia and New Jersey. In January, 2010 came the biggest surprise of all: the election of Scott Brown to Ted Kennedy’s senatorial seat in Massachusetts. And in subsequent months we were faced with the purge of a number of putative RINOs in the course of the Republican selection process and with the nomination by the Republican Party of a host of Tea-Party candidates. “As isolated incidents,” I wrote in a recent post, “each of these events might be dismissed. Taken together, they portend an electoral upheaval without recent precedent.”
What about the polling data? As Jay Cost recently remarked, the quite considerable disparity in that data turns on a crucial question as yet unanswered: “the partisan composition of the electorate remains the critical unresolved issue of this cycle. Every pollster is making a guess as to what the electorate will look like, and these guesses are at least as important as their final numbers.” “In fact” I added in the post linked above, “the guesses determine their final numbers.”
Moreover, as Sean Trende intimated this morning in a post on RealClearPolitics, the guessing that is going on makes no sense. When one examines the cross-tabs, as he has done, one discovers that what many of the polls are reporting is premised on the notion that the percentage of Democrats in this year’s electorate will more nearly resemble the percentage in 2008, when that party made major gains in the House, than the percentage in 2004 – which is not at all what we saw in Virginia, New Jersey, and Massachusetts in November, 2009 and January, 2010. Put simply, they ignore the enthusiasm gap.
In the post linked above, I drew attention to something else that is unprecedented. The Gallup Organization has been collecting generic ballot data for almost sixty years. For the most part, they focus on registered voters and not on likely voters. In recent elections, however, when October comes around, they apply a screen to their sample for the purpose of focusing in on those most likely to vote. Here it is worth keeping in mind that in no midterm election since 1974 – when eighteen-year-olds were given the right to vote – has the turn-out of registered voters exceeded 39.6%.
This year, Gallup’s generic ballot data suggests that, if there were a 40% turn-out on 2 November, the Republicans would have a 17% advantage. Shocked by this, the Gallup Organization ran the numbers a second time on the assumption that there would be a 55% turn-out (which they have not seen fit to do in the past). In this case, they concluded, the Republicans would outpoll the Democrats by 11%. The latter number is close to what Pew, Rasmussen, Fox News, and CNN/Opinion Research report. At this stage in a midterm election season, the Republicans have never – since Gallup began collecting generic ballot data – enjoyed an 11%, much less a 17%, advantage.
What does this mean? First, 2010 is not 1994. This time, the tide favoring the Republicans is much, much stronger. My own instinct – expressed in the post linked above – is that the rule of thumb Lou Cannon used to apply to the polls when Ronald Reagan was a candidate applies this year to the Republican Party as a whole. One should take the most reliable of the polls – those of Scott Rasmussen – and add 5% to the expected results for every Republican candidate.
For a time I was alone in thinking along these lines. Now I have company. After examining the crosstabs and considering the degree to which the pollsters are basing their estimates on the highly dubious presumption that, in its partisan makeup, the voting public in 2010 will more nearly resemble that of 2008 than 2004, Sean Trende suggested this morning that one would be well-advised to add 3% or 4% to the results for each of the Republican candidates.
If one does that, then my prediction that the Republicans will gain seventy to one hundred seats in the House and take control of the Senate makes perfect sense.
- Comment (12)
- · Quote
- · UnfollowFollow (0)
Comments :
Jun '10
Re: Electoral Prospects
Laying in the Moët & Chandon - if we hit 70/10 I'll send some to you
May '10
Re: Electoral Prospects
My prediction is 91 seats, and a takeover of the Senate. My reasoning: the polling that's been done this year seems way off. How is Tom Tancredo within the MOE in supposedly purple CO? How did the Obamacare referendum get close to 3/4s of the vote in MO? My intuition is great, and 91 feels like the number.
May '10
Re: Electoral Prospects
As one who spent a career analyzing data, I love the way that you handle the numbers and think about them.
I believe you are correct.
I hope you are proven to be a pessimist. ;-)
Oct '10
Re: Electoral Prospects
I'm sticking with my 85/10 prediction, but I'll be thrilled if that ends up being too low!
Re: Electoral Prospects
Paul, boy do you know how to pick up a guy's spirits. I just got off a plane from Tokyo, stumbling home feeling jet-lagged to news that Proposition 23, our effort to sideline cap-and-trade here in the Golden State, is sinking in the polls. My fingers are crossed that this turn out model is similarly skewed--although as Peter Robinson has noted on the podcasts, California seems to exist in its own weird electorally disconnected bubble.
Sep '10
Re: Electoral Prospects
It seems to me that Republicans poll low during most elections. I remember seeing John Edwards on CNN the night before the '04 election, and he seemed pretty smug and sure of himself that Kerry/Edwards would win based on late polling data. Thank goodness he was wrong along with those polls.
I have been intuiting that Republicans would fare about 5% better than current polls show. Your article validates my feelings. I hope we are both right.
Oct '10
Re: Electoral Prospects
Paul A. Rahe:
If one does that, then my prediction that the Republicans will gain seventy to one hundred seats in the House and take control of the Senate makes perfect sense.
I remember the conventional wisdom in 1994 which was that the Republicans would do well enough to take over the House, but the Senate was unreachable. While I don't think this election is a straight repeat of 1994, it does show that these sort of political uprisings are very hard to poll.
I just have this odd feeling that the pollster's models won't be able to account for a significantly increased number of voters who would normally be hard to detect because they don't traditionally vote in these mid-term elections.
I think the conventional wisdom will end up being the worst case scenario for the Republicans. I will not be surprised, therefore, if the Senate goes Republican and they pick up a much larger number of seats in the House. I do expect many surprises both good and bad on election night.
This feels like it's going to be a much bigger storm hitting the political class than the polls can detect.
Re: Electoral Prospects
I will drink it.
Jul '10
Re: Electoral Prospects
I'm calling the GOP at 242 in the House, 47 in the Senate.
Aug '10
Re: Electoral Prospects
Are any of you even a little superstitious?
May '10
Re: Electoral Prospects
It seems there are still a number of House races that no one has quality polling on? I wonder if the # of seats in play is established by starting at 0, and adding races where the Republican is ahead (or MOE); or if you start from the total, 435, and subtract for every race where the Democrat is ahead...?
Jun '10
Re: Electoral Prospects
I had a similar conversation with a friend of mine a day or two ago. He was worried about what appeared to be a late Democratic surge. I told him that the current polling models were missing a variable this time around, a multiplier of some kind. I agree with Dr. Rahe, last year's polling models may not be sufficient to explain what's happening now. I reckon when the election is over, the pollsters will have to tinker with their models to discover the discrepancy between their predictions and the final outcome. I'm convinced Republicans will take most races that fall within the 3% margin of error. The knew variable might be labeled the "intensity factor." We'll know in about ten days. I'm predicting 90 House seats and 10 in the Senate.