Slaying the gerrymander
(Welcome, New York Times readers!) Thanks to commenters on this topic. Your feedback has shaped my thinking on this subject. I recall bein...
House: prediction update and GOTV advice
Today, Democrats lead the generic Congressional ballot, as they have for most of the campaign. However, taking control of the House does not come with a popular vote win, because redistricting and incumbency give a +2.5% advantage to Republicans. Therefore the probability of a change in control is only 18-33%. This is in a zone I define as knife-edge, meaning that substantial uncertainty remains. I give lists and maps of swing Congressional districts to guide Get-Out-The-Vote (GOTV) efforts.
If you are reading this in the Midwest, this is your lucky day. I have travel advice for you. However, it will involve hard work – and, if you are a Democrat, some uphill climbs.
Here is an updated House prediction – and recommended places for GOTV – all over the US.
This is a season of inversions. Yesterday I highlighted the possibility (about 1 in 4) of an electoral vote “mismatch,” i.e. the chance that President Obama would be re-elected but Mitt Romney would win the popular vote. However, there is a second undemocratic risk, one that will persist for years: the House of Representatives.
Several weeks ago, I pointed out (“Bush v. Gore times five,” October 16) that partisan redistricting of an unusual intensity has created a situation in which one side, the GOP, has an advantage of 2.5% before any votes are counted.
Things are close, and a Democratic takeover is not ruled out. For those on either side who want to fight!, I give maps of swing districts to show where your Get Out The Vote efforts can be most effective.
>>>
House prediction update
Today I give two new updates, based on (a) my prior analysis and (b) on Pollster.com data. They are not all that different.
The Pollster.com generic Congressional polls give a post-debate-1 median of D+1.0+/-0.5% (n=11, Oct. 4-22). However, 10% are still undecided. I estimate the outcome as D+1.0+/-3.0%. Republican retention of the House is likely but not certain. What is certain is that the House will be more closely divided.
Pollster.com also gives a breakdown of individual races. Individual district polls, where available, give a similar picture. They have 17 tossup races, 19 leaning D, and 18 leaning R – a total of 54 races in play. The remainder are relatively safe, 171 D and 210 R.
From these numbers come the following predictions:
- Generic ballot-based: 213 +/- 9 D, 222 +/- 9 R.
- District-by-district: 199 +/- 12 D, 236 +/- 12 R.
- Democratic takeover probability: 18-33%.
- Combined Bayesian prediction: 208 +/- 4 D, 227 +/- 4 R.
In summary: conditions are unfavorable for a Democratic takeover of the House – but it’s not out of the question. One thing is for certain: the 113th Congress will be more closely divided.
>>>
Swing Congressional districts near you
Are you interested in doing something to influence the outcome? One strategy is to give to the DCCC [ActBlue] (or Republicans, Crossroads GPS).
Another is to Get Out the Vote (GOTV) in a swing Congressional district near you. I don’t have a search app yet (see “Crowdsourcing request,” October 23). But I do have the following maps to help you plan your GOTV activity. Get out the walking shoes!
Midwest (see map at the top of this post): IA-03 and 04; MN-08; WI-07; MI-01 and 11; IN-08; IL-10, 11, 12, 13, and 17; OH-06 and 16; and KY-06.
South: TX-23, NC-07, GA-12, FL-10, FL-18, FL-22, FL-26.
Northeast (NY/PA/NJ/CT): PA-08 and 12; NY-01, 11, 18, 21, 24, and 27; NJ-03; and CT-05.
New England: RI-01; NH-01 and NH-02 (also CT-05, above).
Southwest: UT-04; AZ-01 and 09; and CO-03 and 06.
and finally,
West: WA-01; NV-03 and 04; and CA-07, 09, 10, 24, 26, 36, 41, and 52.
Many thanks to Lyndon Estes for the maps.
Topics:
Interesting analysis. It’s been my belief all along that the house would not flip until possibly the 2014 midterms. But, close is better than a greater majority for the obstructionist party (OP).
Agreed that close is better. If Boehner (or, in a coup, Cantor) is left with a tiny majority, picking off defectors will get easier and the ability of the GOP to keep their obstructive caucus together gets more difficult.
Beautiful maps. Thank you.
So much for Trump’s “big” announcement. What a joke: http://news.yahoo.com/blogs/ticket/trump-announcement-obama-161111445–election.html
LOL LOL LOL LOL LOL LOL Thanks!! Nathan
How about Mitt floating The Donald as his Secretary of the Treasury? That ought to win him some votes. 😉
Thanks very much, Dr. Wang!
Now that early voting results are being reported, does it make sense to adjust the state-level probabilities accordingly?
For example, if 25% of the voters in North Carolina have already voted and we know the D/R registration breakdown, and can compare those results to similar days in 2008, shouldn’t that count for even more than polling results?
At a minimum, shouldn’t the forward cone of uncertainty narrow since 1/4 of the result is locked in with current data?
I don’t believe voting results are being reported — only the party registration of those who have voted early. That is a very different thing, although there is likely to be a strong correlation.
I think it’s a matter of “If it ain’t broke……..
Should the model go awry then the good doctor will probably look cure what ails it, but until then he won’t be tinkering much.
I agree that only party registration is being reported, but other polling data gives a good notion of how O/R/? will break down given D/R/I registration.
The “ain’t broke” argument is reasonable, but the hard data approach Prof. Wang uses would seem to argue for making the best use of the hard data available, maybe not this cycle but in future ones.
As mentioned, actual votes are not being reported, just totals of votes returned by party, which means we don’t have actual voter preference information. Polls do give us actual voter preference data.
Also, we don’t know what the final turnout will be, so we don’t know what fraction of the total voters have voted.
Related, we don’t know if turnout to date, compared to the same period last time around, reflects changes in voter sentiment, increased publicity around voting mechanics (voter ID etc), better GOTV organization by one campaign or the other, or what. Meaning, there’s nothing we can presume about how the proportion of early voters from each party reflects the eventual proportion of voters from each party (which, as noted, is not the same as which candidate they vote for).
So the polls still give us much better predictive information. The early vote information is interesting, and is useful to the campaigns in assessing their effectiveness at turning out their likely partisans, but doesn’t tell us who will be President.
I beg to differ.
If you get to the point where 100% of the voters registered Democratic have voted, and 0% of the voters registered Republican have voted, with 2 days to go, there is a hell of a signal there.
So somewhere between no information about early voting and that kind of monster signal, there is still useful information to be gleaned.
Donald Trump is such a brilliant idea man! [Sarcasm alert]. Should we start a campaign to raise money for Romney’s favorite charity other than the LDS Church if he will release his tax returns? I’m particularly interested to know if he took advantage of the tax amnesty program for those who held dodgy off-shore tax avoidance accounts.
What software was used for the maps?
R, specifically with the packages rgdal and sp
Gallup edge closing, down 1% today from yesterday, R-50%, O-47%….
Yes, although Gallup’s methodology has been called into question of late with some ‘of the planet’ numbers coming through, they are clearly picking up a favorable trend moving back towards Obama.
Cherry Hill is no longer in NJ03, it is part of NJ01.
This makes it a tougher get. Otherwise we could drive down there, knock on doors, get some White Castle…
Great resource, Dr. W!
May I revisit the whole intrade issue here (as opposed to yesterday’s update)?
Specifically, I’ve been working on a theory of intrade possibly having a GOP bias, independent of the recent transparent manipulation attempts. Obviously painting with a pretty broad brush here, but wouldn’t you guess a larger number of GOP stalwarts would have the disposable income to more easily afford to wager on the presidential outcome?
Speaking personally, I would certainly be fairly confidant of my chances of weighing in on this, but I don’t have lots of extra cash laying around–not to mention the queasy ethical implications I’m not very comfortable with (also something that high rollers in the GOP might not care about ;).
…or am I off base?
I’m not saying that you are off base, but given their business plan you would have to invest a lot to get a decent return this late in the race.
Even if you bought at 5/share the max return is the difference between purchase and max return per share of $10.
It sounds paranoid, but I believe that Romney is manipulating everything that gives the illusion of Mitt-mentum.
James Fallows said it best.
I’m not precisely sure how intrade works, but one thing I know about sportsbetting is that most participants are idiots and the lines reflect how they lay their bets. So with a perception of Romney tying up with Obama in the media, you’d see a bunch of those idiots think he’s undervalued on intrade. Successful gambling is really just line shopping based on how the lines move away from where they truly should be
Intrade is much more visible in the media than the other betting sites. So it makes sense for some R moneybags to drop a few k$ (lunch money for these guys) so the red media can run few “Intrade showing momentum for Romney!” articles.
If indeed it all comes down to enthusiasm, then this is a cheap way to buy some.
@wufwugy – I think you’re suggesting something about sports betting that would be remarkable if true. By your logic, if I’m following correctly, there should be winning and losing bets that are apparent to an “informed” bettor, and would enable him to consistently do better than the odds.
But if this were true, given the size and liquidity of the sports betting market, this would be a license to print money for this “informed” bettor. Therefore we’d expect the entrance of professional, “informed” bettors, whose bets would even the odds across the market.
In other words, in theory this sort of arbitrage against “idiot” bettors should be impossible. Is this not the case?
As to InTrade, I think the value of political events is poorly valued as a function of participants over-estimating the importance of prominent but transient events; i.e. betting with the “conventional wisdom” rather than dispassionately evaluating the underlying dynamics.
@James, that assumes knowing how to shop for lines correctly is easy.
Walking around for Obama yesterday, I was supposed to talk up Jared Polis (Co-2) as well as Obama. Well, personally I voted for Susan Hall. Maybe I should nip down the road and walk around some CO-6 neighborhoods instead. The TV has just been brutal to Miklosi though. Even the Denver Post, despite complaining about Coffman frequently, endorsed him in 2012! Sadly, there was is little bravery at that paper, though they did endorse Obama for Prez.
Methodology question, and I apologize because I know Dr. Wang has answered this before, I just can’t find it offhand…
As the election comes closer, the prediction ‘strike zone’ and confidence bands shrink, correct, because the predictive power of the MM is greater up close?
I’m confused by the D-R popular vote margin vs. seat margin graph: why would it be linear?
It makes sense to me that the Republicans have a 2.5% re-districting advantage when the vote margin is close to 0.
Intuitively the redistricting advantage comes from gerry-mandering “safe” districts where your voters are less concentrated than your opponents’ “safe” districts.
But at some D-R margin — I’m not sure if it’s 10% or 20% or whatever — those less concentrated “safe” districts become the battleground districts and the re-districting advantage ought to turn into a disadvantage. Am I missing something?
Loos like someone wasted a lot of money on Intrade attempting to give a false impression of Romney momentum. Obama today up to 59.9 an increase of $0.48
I would say Colorado, Arizona, Texas, upstate New York, and Minnesota, Wisconsin, Nevada and Michigan are out of reach without a number list and a phone bank. At this point the prioritizing rule of thumb would be smaller is better.
I hope you succeed in taking this further. Having something like this ready to go in August 2014 would really sit up and work.
Also, keep an eye on Ohio Issue 2, a constitutional amendment to take redistricting out of the hands of the legislature. If it passes, it could be the prototype of many more around the country.
Wisconsin 7 can easily be manipulated. Eau Claire, Wausau and Superior are all small TV markets with slow news. If you can stage a GOTV rally in any of the cities of the district and let those markets know, you have news coverage. I know, I worked for WAOW for five years.
I had a quick math question. Gallup is now at 3 points (50/47) and this is from a 7 day average. Yesterday, they were 5 points apart (51/46). Would a 2 point swing require a huge single day swing? I’m thinking you have to have a really amazing day added, lose a big outlier, or both for this kind of shift. Is it easy to figure out what the exact poll for today is from the change in the average? Or is that contingent on the day that got removed? Does a 2 point swing from one new day of polls in a 7 day average imply a large bounce from debate 3?
I saw the crosstab for Gallup several days ago and it had 22% of African Americans voting for Romney. No one thinks that’s going to happen. This is why averaging polls is important. Gallup by itself (most by themselves but Gallup especially) aren’t that reliable.
It’s impossible to know the magnitude of the change in the spread when Gallup reports just integers. Rounded to tenths of a percent, it could have been 50.6/46.4 moving to 50.4/46.6, for a change of only 0.4% (4.2 to 3.8).
Recall also that there are *two* changes that occur to the rolling sample every day: the calls from seven days ago roll out, and the calls from last night roll in. So if you had some favorable effects for one candidate in both of those changes it will magnify today’s apparent jump, even though part of the change is from a week ago.
Hey Sam is it possible we can get an expected demographic breakdown of the swing states? From what I’ve found from various sources suggests that due to turnout and demographic expectations, Obama will be able to win even if he only wins white men by as low as 30-35% in swing states.
Also because my prediction is that he will win white men by over 40% in swing states, the polls are a bit off and Obama is going to crush Romney in EV
Two questions:
1) What are we called upon to do here? Doorknocking?
2) Who’s up for putting together a crew to make a difference and eat donuts in the central valley between San Jose and Sacramento?
I mean, hey, look, it’s all YELLOW.
http://www.youtube.com/watch?v=Phyg_uIPQII
or I suppose conversely it could have been 51.4/44.5 to 49.5/47.4 for a change of 4.8% (6.9 to 2.1). 0.4% to 4.8% is quite a range. Are the internals ever made public?
thanks!
We are now down where polls don’t matter. Getting out the vote, which side does that best in swing states, will win.
First time poster, long time lurker, including the last month of the 08 election. Question: is it just me or is this race shaping up much like the last incumbent election in 04? Comparing the EV movement, there seem to be some real similarities: both incumbents hit EV totals in the 320+ range before plunging sharply after the first debate before more or less plateauing for the balance of the race. The difference, of course, is that Kerry had a big surge after the DNC and was way out in front before being Swiftboated, while Romney has never at any point of the polling led the EV race.
My question is whether there is any evidence to suggest that this is a normal pattern for an incumbent election, in that the only real undecideds are “baked in” after the first debate and the rest of the election is all about firing up your base. One prior election does not a trend make (obviously), but wondered if there was any evidence to be gleaned from even earlier elections (84, 92 and 96).
Dr. Wang…sorry for using space here to ask this. But looking at the map has inspired me.
I spent ’09 and ’10 covering (photojournalism side of things) crisis elections (Haiti, Afghanistan, Juarez, the malay/thai border).
I’m back in the States now, not covering our elections…but thought maybe since I am here, I’d spend Nov. 6th somewhere interesting…put myself in a really compelling county…and just capture the mood that day.
Maybe PEC can use its collective imagination. What might be a really good (challenging? symbolic? uplifting? dangerous? unique?) place where everyone else isn’t going to be? I’ve done a little brainstorming, but so far, my ideas are a little on the cliche side (poorest county in the country, native american reservation…).
Food for thought!
Thanks,
Jeff
Jeff try and contact the international team, see what they suggest
http://www.huffingtonpost.com/2012/10/24/connie-mack-un-united-nations_n_2008694.html?ref=topbar
I wish this were a joke, but it isn’t. Check out some of the right wing militias. I’m sure they will be getting ready for “the revolution” if the president gets re-elected. Years ago I had one such wacko sitting behind me in a restaurant explaining to someone how Bill Clinton was going to use the military to keep from giving up power when his term was up. I recently overheard two people talking about the coming revolution if Obama gets re-elected. It may be more dangerous to your mental than physical health though. 😉
Parma, Ohio, Jeff. It’s critical and there will not be one single member of the national press within 25 miles.
–bks
How about eastern Nebraska, specifically Douglas county, which includes Omaha? I heard a fascinating piece on ATC today about how this one county could theoretically come into play since last time it went for Obama, and NE awards their electoral votes by district.
How about a fly-on-the-wall report of a ballot machine computer hacker? I’m sure Karl Rove can put you in touch with several. 😉
For Democrats who want to do GOTV in Ohio. Here is a list of districts and the OFA offices that you should contact.
16th:
Parma 44129- 81,601 pop.
Strongsville 44136- 44,750
Brunswick 44212- 34,255
WestLake 44145- 32,729
Medina 44256- 26,678
Wooster 44691- 26,119
Wadsworth 44281- 21,567
Berea 44017- 19,093
6th:
New Philadelphia 44663- 17,288 pop.
Minerva – 3,715
Carrollton 44615- 3,242
Spending 11/6 with right wing militia…not a bad idea actually.
Unless it turns out to be like a Heaven’s Gate cult.
Interesting idea though. :}
Jeff, I think from your perspective, you might want to pick a Florida district with significant minority and elderly populations, but close prediction of outcome. What you would be looking for from the journalistic/photojournalistic angle is what visible signs there are (and Florida is where to expect them) of vote suppression tactics, voter intimidation (mental more than physical), etc.
I envision a really good shot of an elderly black lady being turned away at the polling place because she didn’t have enough i.d. with her, or was incorrectly “purged” from the rolls, so forth.
The suggestion by another commenter above to contact the international observer team might lead to some cross-correlation with my suggestion.
As with Afghanistan, etc.: Come home safely!
For the purposes of the indelible image, it’s too bad the International Observer Teams aren’t wearing body armor and blue helmets.
Or at least a blue beret.
I asked a while ago whether any pollster with a website can be included in the meta-margin. Well, there seems to be a new pollster with a website. Pharos Research Group has released a bunch of swing state and Senate polls, with some good numbers for Dems, including O+5 in Ohio.
David Nir at Daily Kos raised some questions about them and got responses from the pollster, who is a former employee in Gallup. At any rate I don’t see why they should be viewed as less credible than e.g. Gravis. Why are they not included in the aggregates?
The MM calculator automatically feeds off the pollster.com RSS feed. You need to talk to Mark Blumenthal at the Huffington Post to have him include the Pharos polls over there. Once they are added to pollster.com, they will update here as well.
Yeah, I realized as I was writing this that PEC probably uses some other database. I’m just still kind of bothered by the idea anyone can just feed numbers into the polling aggregates. The response that so far there hasn’t been enough garbage to mess up the aggregates is not totally reassuring to me as that could change at any time with the proliferation of pollsters and the increasing emphasis on building narrative in the horserace coverage.
I do communicate with Mark Blumenthal from time to time. He is a former professional pollster, and is very engaged in pollsters’ methods. He is on the lookout for problems. I respect and trust him.
Paul, that is very interesting!
Today, Electoral-Vote.com does list the Pharos Research polls:
Florida: Obama 47, Romney 47 (tied)
Ohio: Obama +5%
Pennsylvania: Obama +3%
Florida Senate: Nelson 52, McGillicuddy 44
Indiana Sene: Donnelly 46, Mourdock 46
Montana Senate: Tester 48, Rehberg 46
North Dakota Senate: Heitkamp 49, Berg 48
Ohio Senate: Brown 52, Mandel 31
Pennsylvania Senate: Casey 52, Smith 42
http://www.electoral-vote.com/
I apologize for posting this note on a thread dedicated to another topic… but it happens to be the active one!
We all know that Rasmussen’s polls err on the Republican side. But is it due only to the fact that they conduct automated polls and leave behind cell-phone only voters? i.e. they have a bias, but an “honest” and constant one? Or do they “manage” that bias more actively?
I want to share with all of you a small analysis I just did:
1. Take the margins between the two candidates as computed by 538, Pollster and RCP for the 10 assumed most competitive states (Co, Fl, Ia, Nv, NH, NC, Oh, Pa, Va, Wi) and average them.
2. Calculate the difference between each of these state averages and Rasmussen’s most recent margin for that state.
3. Plot this difference against the state’s “competitivity” expressed as the absolute value of the average margin calculated under point 1 above.
4. Do a linear regression analysis of this data series.
… and you get the message that the more competitive the state is, the more Rasmussen’s margin is biased towards Romney. R2 is maybe not sufficient to take this data to court, but nevertheless quite good.
Example: if the state is little competitive like Pa, Rasmussen’s bias is 0.0%. If the state is VERY competitive like Co, Rasmussen’s bias is 4.3%. The other 8 states fall neatly on the line in between!
It may be something to study over a longer duration than my snapshot analysis…
François Badoux: “…the more competitive the state is, the more Rasmussen’s margin is biased towards Romney.”
The pattern you’re unveiling is hugely interesting, imho. So far I’ve only read anecdotal evidence or commentary on precisely how Rasmussen is biased.
Interestingly, today Electoral-Vote.com has a lengthy analysis of the Rasmussen bias, with hard numbers. These can be downloaded for more exhaustive analysis.
http://www.electoral-vote.com/evp2012/Pres/Maps/Oct25.html#item-8
Nate Silver obliquely mentioned this site in his last blog post, saying that some people accuse him of having a Romney slant,but other prediction sites have Obama at 80 to 95% chance of winning (probably remembering the Talk of The Nation discussion)
Link to RAND which is holding steady for any who are interested:
https://mmicdata.rand.org/alp/index.php?page=election
New PPP number from Virginia: O+5
That should move VA from O+1 to O+1.5 at the next MM update. There will be 6 rather than 5 polls to average: O+5, O+3, O+1, R+2, O+2, R+3.
Brian, yeah. Good news for my state. I was concerned about the reaction from the Tidewater area regarding taking what the President said about the Navy. Romney left out the fact that Obama approved contracts to build new ships already. But even the military has said they want a leaner, more high tech one, so this makes defense contractors happy here in NoVa.
Interesting. 16 of these are in areas where both POTUS candidates are competing (NOT including PA, MI, AZ) re: GOTV out of roughly 40-45. Is that an unusually large percentage?
Did the MM just tick down?
Reason: No. Yesterday at this time it was 1.46. I’m looking at yesterday morning’s map now. It went down very slightly later in the day but today at 1.52 it’s higher. The EV is even more stable.
Jay, thanks. I thought it was 1.54 last night. Did the new PPP for Va get factored in?
Reason: Just click on the VA row in the Power of Your Vote table near the top right of this page and it will takesyou to the latest VA polls on Pollster.com. Once the new PPP is listed there, the next MM update will reflect it.
That I wouldnt know, but I notice not much changes it much. While we’ve been talking the same figures reloaded.
Chris, thanks.
Is CO-06 incorrectly highlighted on the map? The shape looks nothing like the district 6 maps from other sites. Thanks!
Redistricting. The new district is shown.
Not that it matters, but Gallup is unchanged from yesterday. Except O approval lost 2 and Disapprove gained 2. Any ideas how it could be exactly the same when there was a huge trend towards O compared to the day before? Just curious.