# Princeton Election Consortium

### Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

## Senate prediction 2012 (Election Day final)

#### November 6th, 2012, 11:00am by Sam Wang

Briefly, my predictions are
President: (mode) Obama 332, Romney 206 EV, (median) Obama 309, Romney 229 EV. Two-candidate popular vote: Obama 51.1%, Romney 48.9%.
House: Democrats win 2-22 seats. D 205+/-10, R 230+/-10 seats, Republicans retain control.
Senate: Democrats win 1-3 seats. D/I 55 +/1, R 45 +/- 1 seats, Democrats retain control. More on the Senate here.
The poll-based median Senate outcome is 55 Democratic/Independent seats, 45 Republican seats. It is very focused: 85% of the probability is concentrated in the range of 54-56 D/I seats.

Win probabilities:
>90% Democratic. Connecticut (Murphy), Massachusetts (Warren), Missouri (McCaskill), North Dakota (Heitkamp), Virginia (Kaine).
81-89% Democratic. Indiana (Donnelly), Wisconsin (Baldwin).
71-80% Democratic. Montana (Tester).
75% Republican. Nevada (Heller).
>90% Republican: Arizona (Flake).

Update: Wow, I’m getting it good and hard in comments. I made a transcription error on North Dakota – the probability there is 75%. Sorry!

Also, I should not have expressed these primarily as probabilities. The probability function is a nonlinear function of margin and SEM. That means it (a) can offend intuition, and (b) is susceptible to user error.

In regard to the latter, it would be better for me to add in a possible bias of b=-2% to +2%, as defined before. That slop factor will give a more realistic assessment given poll accuracy. The “slopped-up” probabilities are below.

State (n) median+/-SEM D win%
Arizona (4) Flake +5.5+/-3.8% 12%
Connecticut (8) Murphy +6.0+/-1.3% 99.8%
Indiana (10) Donnelly +2.5+/-2.0% 84%
Massachusetts (8) Warren +4.5+/-2.0% 96%
Missouri (7) McCaskill +6.0+/-2.9% 96%
Montana (5) Tester +1.0+/-1.3% 69%
Nevada (7) Heller +2.0+/-2.9% 27%
North Dakota (3) Heitkamp +2.0+/-2.3% 75%
Virginia (8) Kaine +3.0+/-1.1% 96%
Wisconsin (9) Baldwin +1.0+/-0.9% 72%

Tags: 2012 Election · Senate

### 91 Comments so far ↓

• Grateful

Good morning Dr. Wang –

That looks about right to me.

Just voted in Northern VA, took about 2 hours total from start to finish.

Everyone have a great day!

• Rebecca

I have a masters in statistics and find these statistics very reassuring.

• Kerr

Thanks for all you do, Sam. I work in data warehousing and analysis, and I appreciate your scientific approach to poll aggregation. I hope, for reasons both scientific and political, that your EVpredictions are as close to the mark as they’ve been in the past.

Being from MO, I am particularly happy to see a confident prediction for McCaskill. Akin really hurt himself with his comments.

• Pat

Sam, both your mode and your median (332 and 309) do not correspond to the current estimators (303 and 312). Are you deviating a little from your own numbers (snapshot and histogram) to make your prediction? Or is it based on an earlier or future update to the topline numbers?

• SkepEng

I was thinking the same… How is the mode at 332? It doesn’t coincide with the one in the histogram.

• Frank Katz

Check out Intrade. People who put their money where their mouth is are buying Obama and dumping Romney at a furious rate. POTUS is now up 4.5% TODAY. Another confirmation of Dr. Wang’s predictions.

• Muhahahahaz

I just opened an account recently, and my money finally got there this morning. Oh how I wished I had opened it sooner!

I’m not gonna buy “Obama wins” for more than \$7 (it’s currently at \$7.37+), but there are plenty of other good bets to be had. :-)

• Paul G

I am glad about the current Senate outlook, but I have a (pessimistic) question about the long term trajectory of the Senate. Historically, good candidates have been able to win Senate seats all over the country. However, as the parties become more ideologically sorted and the public becomes more partisan, it seems to be getting more difficult to play on the other partys turf. (Recent exceptions due to Republicans nominating terrible candidates notwithsatnding.)

In the long run, this seems to be a terrible dynamic for the Democrats. According to the Cook PVI, there are 27 lean Republican states and only 22 lean Democrat states. But when you look at the very partisan states, it is much, much worse. 22 states have a PVI of R > +5, while only 10 have a PVI of D > +5.

Right now the Democrats hold a bunch of seats in deep red states but that could change soon, such as in 2014 when a bunch of them are up for re-election.

If the Senate becomes better sorted by party, and all the strongly partisan states elect their own type, we cold be looking at a situation where the Republicans start off with around 40 safe Senate seats, the Democrats with only 20, and they fight over the other 40 “middle state” swing seats.

I hope this is not where we are headed, but I dont see any reason why not…

• Rieux

Well… do you expect the Republicans to stop nominating idiots like Christine O’Donnell, Sharon Angle, Todd (actually William) Akin, or Richard Mourdock any time soon?

• E A Dawsoni

It’s certainly one of the strongest long-term stressors on current constitutional arrangements, and likely to become more so.

Medium “long-term”, to some extent even the reddest states will likely become bluer in statewide races as they urbanize – the correlation of population density with voting behavior is a lot stronger than many realize, for example, play with the sliders at:

http://elections.nytimes.com/2008/results/prensident/explorer.html

setting them to 1-1000 and then 2000 – top of scale, for starters.

• rich thaler

in addition to the prospects of Republicans coninuing to nominate hard right misogynistic candidates, demographics will also help reduce or eliminate the Republicans potential in many southern and southwestern states. Black and hispanic growth rates outpace those of white voters. Unless republicans adopt more immigrant and minority friendly policies, as well as more woman friendly policies, they are looking at an increasingly bleak future.

• John

Republicans should have a long term natural advantage in the Senate, Governors and state legislatures as they dominate the low population prairie states.

Considering that the national vote is basically split along party lines, but Republicans have advantages in the low population states (and therefore have more control over gerrymandering in total), they should have a minor advantage in the House too.

• Fred

Demographics will sort that out. Every year there are less and less angry old white people relative to everyone else.

• E A Dawsoni

For some reason my link above doe not work.

“county level map election 2008”

and select :

“Electoral Explorer – Election Results 2008 – The New York Times”.

• Steve16748

Good topic Paul, I’m seeing the almost everywhere things are getting bluer, as the old Montana ranch wife said; “progress is made one funeral at a time”. There are 22 to 24 reasonably blue states now, not counting Virginia, Florida, N. Carolina, Montana in that mix. Add in Arizona, Missouri, Indiana, and even Texas as someday blues and we stand a pretty decent chance of majorities in the Senate in the future. Lets have the guts to get rid of the filibuster and get on showing the public that we pass legislation that makes life in America better. It could be a virtuous cycle.

• Dr. Wang, can you please add an event for Hurricane Sandy on your chart(s).

I want to make sure that people see that the Ro-mentum had died long before Sandy, and that there was a steady, slow rise for Obama since about the middle of October.

That narrative is going to be played to save punditry over the next month if Romney loses

• Robin Colgrove

Just voted as a family unit in Needham Ma for Warren. I’ve been doing door-to-door canvassing for Warren from back when she was _trailing_ in the polls. This was my (college freshman) son’s first vote. He had not gotten around to mailing in his absentee ballot till too late so we drove out last night to Tufts to bring him home to vote. It took some doing but we did convince the (very nice) poll supervisor that since we showed her the unused absentee ballot, it was OK for my son to vote. It made me think that the situation can be much different for people who _don’t_ have friendly, sympathetic poll workers and savvy, determined family on their side!

• Khan

Very, very different. Good parenting :D

• NickP-LA

Does it really matter that the democrats are controlling the senate? Repubs are filibustering everything.

• BrianTH

The exact mix of Democrats elected to the Senate may end up determining whether or not there is filibuster reform (at the end of the day, the filibuster is just a matter of Senate rules).

• RocketDoctor

Reid has stated that he was wrong not to do filibuster reform at the start of the last Senate.

If one is to believe him, we’ll have some filibuster reform in the next Senate. My only fear is that it is faux reform without any teeth. We shall see.

• E L

FYI: The filibuster is a Senate rule, not a US law or part of the US Constitution. A majority vote of the Senate sets the Senate rules at the opening session of each two year term. Once the rules are set for the term it takes a 2/3 to change them. But 2/3rds is only the traditional rule for changing a rule and could be changed at the beginning of the two year term by majority vote.

• John

The time to do filibuster reform was in 2008.

If he does it now, which I strongly doubt, it will be just in time for the Senate to flip back to the Republicans in 2014 as many vulnerable Dems that won in 2008 come up for re-election in an off cycle year.

Also, what would be the point of doing it now given that the House is controlled by extreme right wingers? The only thing this would help with is maybe USSC nominations and treaties — in exchange you hand Republicans a club to beat Democrats down with on everything in 2014.

It would be an extremely dumb move if he did it now. But, of course, he’s a leader of the Democratic Party (my party) so that would be par for the course.

• Olav Grinde

@John, I’m very skeptical of filibuster reform for precisely those reasons. In a future Republican-dominated Senate, filibustering may well be the only way for Democrats to block heinous Republic-sponsored legislation.

• BrianTH

I think there is a very good argument to the effect that the filibuster over the long run serves the interest of those who typically prefer the status quo and, when necessary, legislative inaction. On net, that may work our poorly for progressive Democrats.

• Shawn Huckaby

I say just go back to the original Jimmy Stewart version. No more implied filibusters. Get your butt on the floor and start reading. As soon as you can’t keep going, it’s over. Maybe if it’s actual work we won’t see it abused as often.

While we’re at it let’s get rid of anonymous “holds” on nominees.

• John

@BrianTH: My biggest concern is that Republicans get enough suicidal legislators, especially if the filibuster is weakened, to gut SS, Medicare and/or Medicaid.

They’ll take a short term hit to get a long term weakening of those programs.

• Live-blogging at _The Guardian_ is underway:
http://www.guardian.co.uk/world/2012/nov/06/election-day-obama-romney-live

–bks

• Cook

Hi Dr. Wang! Are you sure Heitkamp has a >90% chance of winning in North Dakota? It seems like the reverse would be more likely. Thanks!

• Eric

I second the question about Heitkamp. This would be the exact opposite of 538 which predicts 92% that Berg wins.

• Thanks for all your work, Sam. I make the rounds on this, the Votemaster, 538, and a few others. Your frequent updates and modestly sized pieces of commentary make your site the most helpful.

• Carl

So when it comes to judge how well you did, we should look at… what? The mode forecast and the state-by-state predictions?

• RandyH

Thank you Sam for all of the work,Now lets hope All of the votes are counted right.

• DelMartian

Article on Sam and other quants, but featuring Sam, in today’s Chronicle of Higher Education “Percolator”:

http://chronicle.com/blogs/percolator/the-rise-of-the-poll-quants-or-the-demise-of-the-tv-pundits/31636?cid=at&utm_source=at&utm_medium=en

• Richard

I think Obama wins Ohio, Wisconsin, Iowa, New Hampshire, Virginia, Florida and obviously wins re-election. My prediction is based largely on the polls but also a recognition that Team Obama has a superior ground game in all the battleground states. Thus, in Electoral Votes I predict Obama gets 332 to Romney’s 206, in popular vote Obama gets 50.9 to Romney’s 48.3%. Thanks to Dr. Wang and Andrew and everyone else at PEC for a great website and great campaign year! Cheers!

• Greg Britten

I live in NH and can attest to the thorough work of the Obama folks. We got a call this morning to make sure we knew where our polling place was and that we were going to vote (in a couple of hours we will).

Many thanks to Sam for his thorough, and I am sure exhausting, work.

• I don’t know if this is real or not, but this voting machine video just got posted to YouTube and I saw it via George Takei:

Anyone know if this is legit? (the video, not the voting machine)

• Yes, the video is real, this is not the first machine to show this problem this year. The claim is that they were “miscalibrated” – but they only seem to be miscalibrated in GOP leaning ways.

• Khan

If one more person sends me an email talking about early voting in Ohio by county and decides to not include the fact that the GOP county they are touting with increased turnout has only 10,000 people versus the Dem county with 1.2 million I will scream.

• Muhahahahaz

Lol

Thanks for the great work Dr. Wang. I do hope your predictions turn out to be correct.

Have any other Obama supporters considered wagering a little bit on Romney just as a hedge?

So, how early does everyone think before we know who has won the presidential election? Late in the evening? Early morning hours? Next week?

Personally, I am predicting that we’ll know if it’s wrapped up fairly early on, as soon as the first significant results start rolling out of Ohio (polls close 6:30 CT). If Obama shows a strong lead early on in Ohio, it’s pretty much over. But, OTOH, if there’s no clear leader…it could be a long night.

(PS Like almost everyone else, I’m eying the 303/332 split prediction, which pretty much matches what I’ve been thinking for the past three months.)

• rich thaler

Sam Wang epitomizes the theme “Princeton in the nation’s service!”. sir you are a national treasure and a hero to math geeks, quant jocks and thoughtful people everywhere!!! as an alum, i am so proud that you are on our team!!

• bflobillw

How do I kill the next 7 hours?

Sam, thanks for all your good work. That is, of course, assuming you are correct (at least correct enough!). If you and Nate are incorrect…… oh god, I don’t even want to think about that.

• BrianTH

All the main aggregators, including Sam and Silver, are necessarily broadly right about what the polls are suggesting–they differ a bit in detail, but there is too much consensus in the overall polling picture for that not to be the case. In that sense, the aggregators themselves can’t really be introducing any highly-significant methodological errors at this point (although details may matter for future contests). Instead, a surprise outcome would have to be the result of the pollsters themselves making major, and systematically biased, methodological errors.

IMHO, despite all the talk of swing states I’ve always thought the entire election hinges on Ohio. But my question is, where do we look in Ohio to get a good idea how the rest of the state’s vote tally will turn out? Looking at the election results from 2008, the majority of (rural) counties were strongly Republican, counter-balanced by the much larger, urban Democratic-leaning counties. (No surprises there, that’s a common trend in many states.)

But there are a handful of counties that have relatively large population and also were pretty close to evenly Democratic/Republican in 2008. Notably: Clark (Springfield), Hamilton (Cincinnati), Lake (Cleveland suburbs), Montgomery (Dayton), Stark (Canton) and Wood (Bowling Green).

Should we be looking at the early results on Tuesday night from some of these counties to get an idea of where the rest of the state is heading? In particular I’m looking at Lake County (suburbs NE of Cleveland), which was about dead-even split between parties in 2008 and also one of the most populous counties in the state. Wikipedia even states, and I quote, “A 2008 analysis of Ohio presidential election results from 1960 to 2004 found that no other county more closely follows the statewide Ohio voting pattern. Lake County doesn’t always vote with the winner, but consistently is closer to the winner’s Ohio vote percentage than any other Ohio county.” More or less the definition of a bellwether county, methinks.

So, ideas? Or should we look elsewhere? Or nowhere at all?

• Lake County is a decent indicator, but among bellwether and swing counties I would watch Hamilton (Cincinnati) and Stark (Canton). Stark typically approximates the statewide vote in nearly all statewide elections (not just Presidential). In 2008, Obama won Stark County by one point less than statewide. In 2004, Kerry lost by one point more than statewide.

Hamilton is a traditionally Republican county that has trended Democratic. Kerry got 46% of the vote there, while Obama became the first Democrat to carry it since LBJ in 1964. Obama got 53% of the vote there. Cincinnati is heavily Democratic, as Obama carried 25 of 26 Wards in the city, and took 68% of the citywide vote. Cincinnati normally gets swamped by the Republican suburbs, but many of the inner-ring suburbs within Hamilton County have trended Democratic in recent years as many Republicans have moved into further outlying areas in surrounding counties Butler, Clermont and Warren. Those three counties have grown immensely in population in the past 30 years and are overwhelmingly Republican, joining small town and rural Western Ohio as the core of the Republican base in the state. Western Ohio, apart from Hamilton, Montgomery and Lucas Counties, IS the Republican base in Ohio. It’s the most conservative part of the state.

• DelMartian

Don’t we need more decimal places on the Bayesian prediction?

• Muhahahahaz

Never! He has a 100% chance now. ;-)

• 1blueinnola

Thanks for the great ride this time around Sam! The truth will always win!

• Khan

I think it’s important to note that Flake from Arizona might wind up in prison instead of office after his campaign was caught robocalling Democrats telling them to go to the wrong polling locations. FBI and DOJ opened up a federal investigation already.

• shma

I’m somewhat new to this site and I’m wondering if the same methodology used in the Presidential race is used to get the probabilities for senate races.

If so, I don’t find that the polls are showing a >90% win probability for Heitkamp :

http://elections.huffingtonpost.com/pollster/2012-north-dakota-senate-berg-vs-heitkamp

Someone can double-check my numbers , but using either the last 3 or 4 polls (4 polls were taken within a week of the most recent poll) I get that the z-score is 1/1.4826 and the probability of a win is 75%.

• Use median-based statistics. Then take the median absolute deviation to calculate an estimated SD and SEM.

• shma

Sorry, I was at lunch.

I actually was using median based statistics for my calculation.

The last four polls are [B+2,H+1,H+2,H+4] for a median of H+1.5.

The absolute deviations for the median are then [3.5, 0.5, 0.5, 2.5] giving a MAD of 1.5, and a z score of 1.5/(1.5*1.4826)=0.6745.

Then the probability of a win is
0.5*[ erf( .6745/sqrt(2) ) +1] = 0.75

In fact, the answer should be exactly 75%, since the scale factor for the MAD is inverse of the normal distribution’s CDF evaluated at 0.75.

Now it’s entirely possible that I have made a stupid mistake somewhere, but the calculation seems pretty straight-forward.

• shma

Aha, I see my stupid mistake.

• RFco

Prof. Wang, I love your work, but I hope this doesn’t happen again:

http://election.princeton.edu/2008/11/17/how-meta-analysis-did-2004-2008/#comment-2507

I was very disappointed that time, and I am glad someone had brought up the issue (I’m also glad you answered, even if I don’t completely agree with your perspective). Please be as objective as possible, as you would for your academic research.

Good luck!

• wheelers cat

AMG i love twitter so much.

• Khan

It angers me that half the country finds that sort of just outright, patent dishonesty acceptable.

• wheelers cat

not for long Khan.
im sensing a sea change.
either it is The Return of the Poll Jedi or the Rise of the Numerati.
i get them mixed up.
;)

• Grateful

Whoa…

• E L

From the Guardian:

Robocall tells Florida voters they can vote tomorrow

The Miami Herald reports that the elections office in Pinellas county, Florida mistakenly made robocalls to thousands of voters telling them they have until 7pm tomorrow to vote. “The calls went out between 8 a.m. and 8:30 a.m. because of a glitch with the SOE’s phone system,” the Herald reported

• Shawn Huckaby

Rats, something, something, ship?

• 538 Refugee

New lawsuit in Ohio. Evidently if you call your vote tabulating software ‘experimental’ you can sneak it through the mandated channels.
http://www.cbsnews.com/8301-505124_162-57545531/ohio-faces-controversy-over-voting-machines-update/

• 538 Refugee

In Ohio, a lawsuit alleges that Secretary of State Jon Husted and Election Systems & Software, an Omaha, Neb., company that makes electronic voting systems used in the state, improperly approved the use of untested, non-state certified software in voting machines to help tally results.

“Our position is we don’t know what’s in [the software], but it seems to violate state law and federal law,” Fitrakis said. “Under Ohio law, it’s supposed to be tested by the Ohio board of voting machine standards.”

The complaint further alleges that the state contract for the software was not open to bidding and did not undergo public review.

Fitrakis says that Ohio officials have described the software as “experimental. Although that might legally eliminate the need to test and certify the technology, the court challenge could hinge on whether software that has been deployed for actual use, rather than for testing and evaluation, can fairly be called experimental.

• Froggy

538, just this morning I was talking Mrs. Froggy off of the ledge concerning this very issue. Here’s an article that someone here linked to (thanks!) that hopefully will help calm fears concerning this:

• Muhahahahaz

Lol… such bs. If it’s being used in the election, that’s not “experimental” any more! I hope this nonsense gets taken care of…

• 538 Refugee

Mr. Froggy. Thanks. Since it was on CBSnews I thought it might be a legit story. I live in Ohio but I’m off to cast my meaningless ballot anyhow. Not sure why I bother sometimes. ;)

• Khan

Pinellas County Florida has nearly a million people and went +8 Democrat in 2008. County elections office “accidentally” robocalls hundreds–maybe thousands of people telling them they can vote TOMORROW too.

Accident my ********

• Don’t go to this alleged miami herald link . for me it connected to a VISA ad demanding my email that had no close button, and wouldnt let me return to PEC either.
[expletive deleted]

• Steve

Thanks for the great site Sam. I sure hope your predictions are right. I noticed in your senate predictions that you have Heitkamp (ND) >90% for the democrats while 538 et. al give him only and 8% chance. Is this an error? If so should we deduct one seat from all values on your histogram?

• Khan

Another scandal brewing.

GOP’s massive elections group–True the Vote has lost their observer license and all TTV watchers have been banned from in and around Columbus, OH polling stations.

Possible criminal charges coming for fraud and suppression.

• Khan

Police just showed up in Philly as GOP group True the Vote were allegedly sending a group of ex-Navy Seals to remove “black panthers” standing outside voting areas.

• Ben

What are your N. Dakota polls that have Heitkamp > 90% to win?

• Fred

Well Sam and Nate are within 1 EV on their snapshots. Nates still seem to be changing though. Last night around 2am he was showing 315 and now he is showing 313.

• LET THE DISTORTIONS BEGIN !
I wanted to link to this story on my FacebooT but wont because the headline reports the early NH returns of Obama 28 to Romney 15 as “tied in New Hampshire.”

Long rant short, GRRR!

http://www.ibtimes.com/us-elections-2012-midnight-voting-results-obama-romney-tied-new-hampshire-hamlet-860612

• Khan

I have to admit something.

I crank called True the Vote’s “voter integrity hotline” and told them a pack of wild Unicorns kept stealing my ballot.

• wheelers cat

were they BLACK wild unicorns?

• Hmm… Dr. Wang Sen. Heitkamp at >90% odds of reelection in North Dakota, while Nate Silver has about 90% odds that she will lose her seat. Can someone explain this?

• Everyone, I am using polling data at Pollster.com and using median-based statistics. I don’t think there’s a math error – please advise if otherwise. And of course I am prepared to be wrong.

• shma

Silver’s unadjusted polling average is Berg+3.9, while the median of the four polls that Wang uses is Heitkamp +1.5.

The difference is due to the fact that Wang only uses polls that fall in a seven day window. In this particular case, that window excludes two polls taken in mid-october that show Berg up by a large margin.

Silver also applies an adjustment based on state ‘fundamentals’ which greatly benefits the Republican candidate.

• Michael K

I believe Fivethirtyeight also discards polls sponsored by the campaigns (because those are often released selectively) which means the polls by the Mellman Group get tossed out.

• Sam, Andrew- a PECadillo- PEC’d get more hits if its link posted in FacebooT showed the current prediction. Doable methinks via thumbnail or as headline.

• Sdkasl

I think Sam is very tired! He says Heitkamp has a >90% chance of winning the Senate seat in North Dakota. Every other site has her opponent, Bert, in the mid 90s. Also, Sam has Tester’s chance of being reelected in Montana at >90%, but the other sites have him at about 33%. What gives, Sam?

• Kathy

Can I get a “money back guarantee” if these forecast are wrong? I’d be devastated….

• OK, OK, guys. You caught me in an error here regarding North Dakota…though Heitkamp is still favored. I’m very sorry!

See the updated entry. I’ve included polling medians, and also given probabilities based on the possibility of an overall polling bias of -2% to +2%. That should be enough information for you to fact-check me more directly.

You may now resume dogpiling on the rabbit.

• More thanks. I don’t quite see why you should have to call every state. It seems also unfair to me, though perhaps I misunderstand, that due to a lack of transparency, 538 is less accountable, yet can easily crib from PEC. Btw in recent days that dogpile has become part of me.

May the best rabbit win.

• To everyone who flayed me on the Heitkamp prediction: um, I got that one right! Of the 10 closest races, 10 out of 10, in fact.