The obligatory post about the size of the GOP wave
[Posted by Karl]
After the Weekly Standard’s Jay Cost upped his prediction to a GOP gain of 61 House seats last week, Ace posed a series of questions to him on the Twitter:
T or F, @jaycost : When predicting, you (and other analysts) feel much more comfortable getting it wrong on the low side than high side.
Real Q @jaycost If you have a choice between guessing low or guessing high, by same deviation, it feels safer to be on low side, right?
I ask, @jaycost , because ppl like Charlie Cook keep making projections before immediately saying “but it’ll be much higher.”
I don’t think Jay answered. (Indeed, I think Jay has canceled his Twitter account.) Although I cannot speak for Jay, I thought they were good questions, especially because they are a good springboard for looking at the bar of expectations people are setting for Election Day.
In reverse order, Charlie Cook is rather notoriously cautious in his projections, primarily based on the perceived power of incumbency. As the Cook Political Report put it in its October 8, 2010 update:
Longtime readers will observe that while we rarely rate unindicted incumbents worse than a Toss Up to win reelection, today we are moving 13 incumbents, 12 Democrats and one Republican, into the opposite party’s column to reflect their underdog status. It’s not that these endangered members’ prospects have suddenly taken a turn for the worse, or even deteriorated gradually over the last several months. Most of these members have trailed all year, and it’s simply exceedingly rare to see a candidate in their position in October come back to win reelection, especially now that early voting will be underway in many states very soon.
We have historically treated the Toss Up column for incumbents as the political equivalent of the intensive care unit, and haven’t commonly used gradations of danger worse than Toss Up. But we have seen some incumbents that have fallen far enough behind and remain behind for so long that we could no longer justify referring to their races as Toss Ups. Overall, 51 House Democrats have trailed their GOP challengers in at least one public or private poll taken this year, a higher number than we have ever seen before. At this point in 2006, fewer than 20 House Republicans trailed their Democratic opponents…
Cook underestimated the 1994 GOP wave, but his overall record is pretty good, which is probably why he usually stays cautious, and why the ratings he has already made frighten the Dems.
Beyond Cook, the conventional wisdom on the wave may not be as groupthinky as we may think. For example, Jay Cost’s October 15th roundup of prognostications ranged from Stuart Rothenberg’s 37-45 seats (“with considerably larger gains in excess of 45 seats quite possible”) to Alan Abramowitz’s model predicting roughly 68 seats and Steve Lombardo (President and CEO of the Lombardo Consulting Group) predicting a GOP gain of 60-70 House seats. Since then, most of the lower estimates have been revised upwards, as time runs out for any serious changes in the dynamics of the elections. In fact, Jay Cost revised his own estimate from 57 seats to the aforementioned 61 (where he remains today). (Note: Patrick Ishmael slogs through a lot of micro-data and reaches 62-65 new GOP seats.)
Jay was fairly transparent is explaining how he arrived at last week’s estimate. Essentially, he assumed Dem and GOP turnout akin to 2004, but allocated the Independent voters in the proportions we have seen overall in recent public opinion polling. Those assumptions would give the Dems roughly 46% of the two-party vote. What Jay did not say is that Gallup’s generic ballot model suggests that 46% of the two-party vote would project to somewhere between 186-208 seats, with a “best” prediction of 197 seats. That “best” prediction would give the GOP 238 seats, unsurprisingly close to Jay’s estimate of 240 GOP seats. Jay’s estimate is also in the ballpark with the aforemetnioned Abramowitz model, which relies in part on the Gallup likely voter generic ballot question.
Jay’s Oct. 22nd projection is small-c conservative only to the degree that he is assuming that Dems fix most — but not all — of their enthusiasm gap. It is the equivalent of the GOP being +10 on the final pre-election Gallup generic ballot question, when it could be as high as +14 now. But even under the Gallup model and the Abramowitz model, a GOP +14 generic ballot would suggest ~253 GOP seats, only a baker’s dozen more than Jay’s estimate (and almost as many seats as the Dems held in 2008 after two very good cycles for them). Unless Democrats’ enthusiasm collapses, that seems like the most realistic best-case scenario.
If there is some small-c conservatism in not projecting the best-case scenario, I would note — as Jay did — that if the GOP wins 54% of the two-party vote, it will likely exceed its 1994 wave (230 seats) and mark the party’s best showing since 1946. As much as I may hope for the best-case scenario, it would be not only unfair, but also counter-productive to set the bar of expectations there.
—
Update: The Cook Political Report’s pre-election House outlook is now a Democratic net loss of 48 to 60 seats, with higher losses possible.
–Karl
Karl
over at the campaign spot they have posted two discussions with an anonymous operative only named “obi wan.” He has been a favorite there for years. I was honestly going to post on it, but i would gently suggest putting it in as an update to this post. its really interesting stuff.
That being said, let me get out a bullhorn and say this.
DON’T GET COCKY!!!
Aaron Worthing (e7d72e) — 10/26/2010 @ 10:46 amI’m getting pretty cocky. I know I should not, but a lot of the projections and my expectations appear to have been lower than the few results we’re seeing from early voting.
I guess I should read a few voter fraud stories and get less cocky. As Ace says, ‘be the wave’. Anybody motivated enough to discuss politics online ought to be motivated enough to phonebank and physically go to a tough election contest and GOTV.
Dustin (b54cdc) — 10/26/2010 @ 10:52 amThe GOP ends up with 230 seats in the House; 49 in the Senate. A good election night for them, but not quite the blowout some are saying.
I say this for two reasons:
(1) I think the GOP GOTV efforts isn’t as strong or organized as the Dem GOTV. I think the GOP is taking it for granted that the “angry” Tea Party people will come out in droves — and they certainly will, but not as much as I think many expect.
(2) I think a lot of the polls we see now have a 2-3% cellphone bias in favor of the GOP.
Kman (d25c82) — 10/26/2010 @ 11:32 amKarl – what is a cell phone bias?
JD (ae2c41) — 10/26/2010 @ 11:38 amJD: It is alleged that since polling org’s don’t call cell phones, that their results are skewed away from under 30’s who have, for the most part, cut the land-line link to society, so that the results are therefore leaning GOP since the under-30’s are not sufficiently counted.
AD-RtR/OS! (3cdbf9) — 10/26/2010 @ 11:46 amLOL LOL LOL LOL
That’s a huge blowout.
If Kman is really taking 3% away from the GOP’s polling, it’s not clear where he’s coming to his actual conclusions (which are right in line with polling).
His projection is conspicuously identical to the latest New York Times projection.
At any rate, this is typical lefty spin. David Plouffe said the same crap. If we don’t win everything instantly, what a stunning disappointment for the right!
I could get used to that level of failure. Those results are a massive proof the Tea Party is the most powerful factor in American politics, and that the failed Obama/Pelosi policies are rejected.
I’m sure we’ll here lots of this. Even if the only race we don’t win is Delaware, that will be pretend-proof that the Tea Party is weak.
Dustin (b54cdc) — 10/26/2010 @ 11:46 amKman clearly doesn’t understand how polling or political journalism works, at all. It’s as though he’s just repeating memes he’s read.
Pollsters already know they don’t get an accurate sample of lefties and righties, and overtly tip the scales because of various factors, from time of day, to refusal to participate, leading to the discrepancy.
This is something most people learn in high school, right?
Polling is not the direct ‘call and count’ that Kman seems to think it is.
Dustin (b54cdc) — 10/26/2010 @ 11:50 amDustin
don’t contradict kman. let the dems think that it won’t be so bad. then more of them will stay home.
Aaron Worthing (e7d72e) — 10/26/2010 @ 11:51 amEven if we were to accept that arse pull as a fact, how could that lead to a 3% overall jump?
JD (ae2c41) — 10/26/2010 @ 11:55 am.230
-178
____
0052
That’s 1994 level blow-out. And that’s a leftward skew of the polling data, to be sure. RCP shows 223 GOP seats with 33 toss-ups. And the graph from July til now is showing the GOP picking up steam. Surely, more than 7 of the 33 toss-ups will go GOP. And surely, some of the leaning Dems will go GOP.
Sounds like Kman doesn’t know much about the TEA Party. I give you the rally turn-outs of the TEA Party/912/10-2-10 compared to Coffee Party/ anti-10-2-10 (all three of them).
John Hitchcock (9e8ad9) — 10/26/2010 @ 11:56 amI think expectations are elevated to the point that less than a 60 seat gain in the House might be viewed almost like a victory for the Dems.
Gerald A (0843ed) — 10/26/2010 @ 11:58 amRight, but the cellphone phenomenon is a relatively new one, and most pollsters haven’t factored it in yet.
Kman (d25c82) — 10/26/2010 @ 12:02 pmFTFY
John Hitchcock (9e8ad9) — 10/26/2010 @ 12:05 pmJeez, do you guys ever get tired of putting words in other peoples’ mouths?
I didn’t say it was going to be a “failure”. In fact, I said it was going to be a good night for the GOP, but not as big a blowout as some think. Like, for example, John Hitchcock (#10)
Kman (d25c82) — 10/26/2010 @ 12:06 pmYep, like I thought, you have absolutely no idea how polls or political journalism works.
They don’t factor it in, in the way you’re pretending they do, simply subtracting X percent. They weigh these polls by asking questions to determine which faction a voter belongs to, weighed based on turnout predictions.
You’re absolutely right… the Democrats are going to lose just as badly as predicted (this is what you said, even though you also pretend you took out 3% that you have no basis for taking).
Of course you have no idea that people lacking landlines aren’t similarly supporting Republicans (they probably are). 3%, eh?
Dustin (b54cdc) — 10/26/2010 @ 12:07 pmFiveThirtyEight – which is now run out of the NYT but which is still run by baseball statistician Nate Silver, who has the best prediction record in 2008 – projects 51 seats.
http://fivethirtyeight.blogs.nytimes.com/2010/10/25/house-forecast-8-days-to-go/
http://elections.nytimes.com/2010/forecasts/house
aphrael (e0cdc9) — 10/26/2010 @ 12:07 pmre: cellphone bias
It’s a topic of discussion among pollsters and pol scientists in recent years. At this point, it’s not statistically significant in most cases. And some pollsters make an effort to include cellphone respondents.
The bigger problem with Kman’s claim is the unstated assumption that cellphone-only households will turn out in a midterm election at the same rate as mixed or landline households. Given that cellphone-only homes likely trend younger, that’s a dubious assumption.
To take a recent off-year example, the polls in VA and NJ in 2009 actually skewed 2-3% more Democratic than the final result. And the likely voter screens this year likely have the same skew.
Karl (fa7301) — 10/26/2010 @ 12:14 pm“Karl – what is a cell phone bias?”
http://www.google.com/search?q=cell+phone+bias
“Right, but the cellphone TEA Party phenomenon is a relatively new one, and most pollsters haven’t factored it in yet.”
How would a pollster do that?
imdw (47a9bf) — 10/26/2010 @ 12:19 pmI think the main jist of ‘don’t get cocky’ should apply mainly to GOP leaders who think this is support of a return to 2006 era spending. It’s not. The Democrats are praying the GOP blows it by going the Ted Stevens / Lisa Murkowski path, spending tons of money with a little graft for the members.
The Democrats have ruined our economy, inheriting a boom with low unemployment, and spending us to instability. We cannot afford dem-lite. The Tea Party is even more vicious towards big spending Republicans than it is towards big spending Democrats.
There’s another point that fraud occurs, last minute surprises will be trumpeted in the MSM again, and polls aren’t perfect, but that’s all typical static reasonable people already expected. The democrats are being repudiated to some degree approximately huge.
Dustin (b54cdc) — 10/26/2010 @ 12:20 pmWow. Where did I say they factor it in by “simply subtracting X percent”?
3% is my estimate, because I know some polls do take into account cellphone users. But yes, there is evidence of a right-leaning bias (bigger than 3%) if you poll only landline respondents:
Kman (d25c82) — 10/26/2010 @ 12:23 pmLink
Thanks, Karl. Go frig yourself, imdw. It looks like kmart stalkerish thingie just likes asserting things.
JD (ae2c41) — 10/26/2010 @ 12:26 pmI understand that sentence to some degree approximately not at all.
Kman (d25c82) — 10/26/2010 @ 12:26 pmBTW, the very latest from Nate: “Bunch of really awful polls for Democrats in House races today.”
Karl (fa7301) — 10/26/2010 @ 12:27 pmThe link I posted does link to the Pew report Kman linked — which (again) found the bias insignificant in most cases. But I’ll take the opportunity to note that studies of cellphone bias from 2008 — when Obama had an energized youth vote — impress me a lot less in the current environment.
Karl (fa7301) — 10/26/2010 @ 12:30 pmAnother way of looking at it: Even in 2008, the RCP average overestimated Obama’s spread by 0.3, so where’s the systematic cellphone bias?
Karl (fa7301) — 10/26/2010 @ 12:34 pmDemocrats have been talking about cellphone bias for years. In fact, those people who have no landline are also far less likely to vote. The 2008 election may have been an exception but it was the only one.
Mike K (568408) — 10/26/2010 @ 12:37 pmNot surprising. I’m sure you disagree anyway, since I said it.
the democrats are being repudiated to a degree that is huge. Even if they lose a little less than projected, or a little more, it’s a projection of huge repudiation. It’s unreasonable for someone to say the democrats aren’t being rejected in a huge degree by quibbling the way you are about a mere 49 seat GOP minority in the Senate.
That’s huge, too. The issues you bring up are static that everyone already knew all about, since polls are never perfect and a few surprises always occur, and the loser (democrats this year) always pretend this means they weren’t really rejected.
I see value in being magnanimous about this. We should all try to learn the lesson here, and see that politics go much deeper than mere elections and polls. Our culture will not abide GOP or Democrat deficit spending much more.
Dustin (b54cdc) — 10/26/2010 @ 12:40 pmI had a landline in 2008, as did most of my friends and colleagues. Now, probably half of them still have one. That’s why I think it will have an impact this time ’round.
I mean, is that scientific? No. But unless you’re Nate Silver (and I’m not), a lot of this is tea leaf reading……
Kman (d25c82) — 10/26/2010 @ 12:42 pmNow he’s bringing up a personal anecdote?
Oh man. anything to ignore the basic reality that the democrats are losing this election hugely.
Dustin (b54cdc) — 10/26/2010 @ 12:43 pmYou would think so, right?
Kman (d25c82) — 10/26/2010 @ 12:46 pmDustin:
Seriously, is the English language your first language? I’ve acknowledged (several times now) that the Dems are going to do very poorly on Election Day.
Kman (d25c82) — 10/26/2010 @ 12:49 pmBe ready, folks, for plenty of what Kman is parroting.
He’s just noting this isn’t the blowout some expected, guys. The Tea Party failed.
Dustin (b54cdc) — 10/26/2010 @ 12:54 pmI think that’s a separate question. The Tea Party isn’t necessarily the GOP, and vice versa.
Kman (d25c82) — 10/26/2010 @ 1:09 pmKman,
None of Pauline Kael’s friends voted for Nixon in ’72, either.
That’s a lightly snarky way of alerting you to the issue of sample size.
Cellphone bias will be a serious issue for pollsters at some point, but not quite yet. Again, if anything, the 2009 polls skewed Democrat, not GOP — so if there’s a pro-GOP cellphone bias, there’s still a bigger pro-Dem bias elsewhere.
Karl (fa7301) — 10/26/2010 @ 1:17 pm“That’s why I think it will have an impact this time ’round.”
Kman – Are you saying that the “smart set”, which obviously includes you but excludes Republicans, has dropped landlines disproportionately enough since 2008 to have an impact on polling results?
daleyrocks (940075) — 10/26/2010 @ 1:22 pmGood Allah, that kmart thing is mendoucheous.
JD (4d450e) — 10/26/2010 @ 1:30 pmIt’s easy to get Kman to refute himself, since he responds directly to reverse psychology. I tell him what he wants to say, and he denies that argument, saving everyone time.
I’ve done it about 20 times now and I’m surprised no one has noticed. It’s registering a solid 3 on my amuso-meter.
He admits he can’t say the Tea Party failed. He admits the Democrats are huge losers this year. He admits he’s got nothing substantial to say about polling.
Dustin (b54cdc) — 10/26/2010 @ 1:37 pmI was kind of hoping for 1894
Kevin M (73dcc9) — 10/26/2010 @ 4:03 pmThis election will be an accurate inventory of idiots reported nationally dy district.
dunce (b89258) — 10/26/2010 @ 6:40 pmYep, those dolts who vote really cause a lot of problems for the smart folk who really know what is best.
However, good user name. It’s a keeper. Tell your Mom. She’s upstairs.
Ag80 (743fd1) — 10/26/2010 @ 6:53 pmA few problems with your proposal:
With a secret ballot, it’s very hard to buy votes, because the buyer gets no proof that the seller actually voted as agreed. But with your proposal the buyer can demand to see the receipt and verify that the seller voted as instructed.
This problem also can enable vote intimidation: “Vote how we say, and prove it, or you will be very sorry.”
pst314 (48ad7b) — 10/26/2010 @ 7:26 pmIt might be 47 to 50 GOP win for the House seat. It is hard to tell. In 1994, the GOP took control of the House about 52 when former Pres. Bill Clinton was in office.
ml (fc559b) — 10/27/2010 @ 4:33 am