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

Races I’ll Be Watching On Election Night

October 31st, 2014, 12:05am by Sam Wang

Senate polls in individual states have moved around…but the Meta-Margin and the average seat count have stayed stable. Nonetheless, the crystal ball is cloudy. Why is that? The Midterm Polling Curse. Spoooooky!

As I wrote last week, everyone’s calculations are, to an extent, built on sand. Historically, in any given year midterm polls have been off in the same direction by a median of 2 or 3 percentage points. Depending on the year, either Democrats or Republicans end up outperforming polls. In current poll medians, six races are within less than 2 percentage points: Alaska, Colorado, Georgia, Iowa, Kansas, and North Carolina. Therefore all six of these races could be won by Republicans…or all six could be won by Democrats.

The other races total 48 Republicans and 46 Democrats/Independents. Republicans are slightly favored to take control, since an even split of the six close races would give them the 51 seats they need*. However, the likely possibilities range anywhere from a Republican majority of 54-46 to a Democratic majority of 52-48. As of today, cranking through the math and the uncertainties gives a probability of 55% for a Republican takeover.

I am thinking about how to guess on Election Night where in this range the race will land. I’ve decided that key races to watch are…Kentucky and New Hampshire. Wait, what?? Here is why.

The polls will be off, on average, by some amount in one direction or the other. Let’s call that average amount Delta. On Election Night, I will be watching returns carefully for clues about how large Delta is. In particular, I’ll be watching Kentucky and New Hampshire. Even though both races have a clear favorite, they have the advantage that voting ends fairly early in the evening. If either party outperforms polls in these states, that might indicate a broader trend nationwide.

First, here is a map of poll closing times, courtesy of Daily Kos Elections.

I should start with a major caveat. Initial returns will often be hard to interpret because within each state, sparsely-populated counties are likely to report earlier than densely-populated areas. Therefore, care is necessary in estimating the statewide vote accurately before all the votes are counted. The Huffington Post has a method for doing this. I’ll be doing something similar at Reuters headquarters on Election Night in New York City. Within permissible limits, I’ll share that information as the evening unfolds.

Here is the order in which I will be watching races. All times are Eastern.

6:00-7:00pm: Kentucky polls close. Mitch McConnell (R) has this race wrapped up, with a 6.0 ± 1.1% (median ± SEM) lead in opinion polls. If his actual lead is 8% or more, he is outperforming polls. If he wins by 4% or less, he is underperforming. The amount of over/underperformance is our first estimate of Delta.

7:00pm: New Hampshire, Florida, and Georgia. New Hampshire closes in most polling stations (some will stay open until 8:00pm). This small state counts fast. Currently, Jeanne Shaheen (D) leads Scott Brown (R) by 3.0 ± 0.7% and will probably win. If she wins by 1% or less, Republicans are likely to outperform polls in other states as well. Of course, if she loses, that presages an evening of wins for Republicans. On the other hand, if she wins by 5% or more, that’s a good sign for Democrats in other states.

Also closing (except for the panhandle) is Florida. In the governor’s race, Charlie Crist (D) currently leads Governor Rick Scott (R) by 1.0 ± 0.9% in opinion surveys. If Scott pulls out a win, that may indicate a good night for Republicans. However, we may not find out the answer until late in the evening, because Florida has late-reporting urban counties such as Miami-Dade.

Finally, the Georgia Senate race is currently tied (0.0 ± 0.9%). This one’s tough to use to calculate Delta, since it may be trending toward David Perdue (R). An alternative is to watch the governor’s race, where incumbent Nathan Deal (R) leads Jason Carter (D) by 3.0 ± 0.8%.

7:30pm: North Carolina polls close. Accurate estimation of the vote may be hard because the state has rural and urban areas. However, the race is basically tied, with Senator Kay Hagan (D) leading Thom Tillis (R) by only 0.5 ± 0.9%. So it’s an important one to watch.

8:00pm: Kansas polls close.

9:00pm: Colorado polls close.

10:00pm: Iowa polls close.

Midnight-1:00am: Alaska polls close.


*In addition to simple polling math, two additional issues may arise. First, if the Georgia Senate goes to a January runoff, this will have an unknown effect on voting. Normally, a runoff would favor Republicans because of lower turnout. But if Senate control hinges on such a runoff, turnout might be higher. Second, if Independent Greg Orman wins and the Senate is 50R, 49D/I, he has the option of caucusing with Democrats, who have aided him; or Republicans, who have attacked him at every turn. I am assuming he would choose the Democrats.

Tags: 2014 Election · Senate

42 Comments so far ↓

  • Davey

    Should be exciting – our whole group here in CA look forward to reading your analysis as the results come in. We are hoping for some early chaos – Kentucky can go against expectations, or Michigan, or both. It would send the pundits scurrying like a kicked anthill…and honestly, one of the great joys of Californian life is watching East Coasters on Election Day slowly go insane from stress and fatigue. By nine pm here Chris Matthews and Sean Hannity both look like they were hit by a train. Our system is sound enough to withstand a little confusion, so I’ll be the first person in history to say it: Grimes/Land 2014!

    • Insidious Pall

      Michigan, I am nearly certain, is a done deal, ship sailed, train left station, bird flown, people have spoken. Peters has doubled his lead in MI just in the last week. Land’s TV presence is so indelibly awkward that each ad that runs results in another point drop in the polls. The only remaining question in MI is will it affect Snyder’s re-election? There will likely be substantial ticket-splitting; Snyder should prevail, albeit less certain than the senatorial race.

    • Olav Grinde

      Michigan appears to be missing from Dr Wang’s “Power of Your Vote” table. I seem to recall seeing it there previously…

    • Davey

      You guys are too serious…I was aware I was describing a scenario that would come along something like once in a hundred thousand simulations. But it is the instances in life where those rare occurrences pop up that reminds us that we are not truly in control of our destinies, and that our fates cannot be calculated by machines. Should Grimes and Land simultaneously win, the entire nation would take notice, and we would be forced to question everything we thought about money and news and skew in politics, and in a way there would be something healthy about that.

    • Sam Wang

      If that combination happened, I know my head would feel like exploding….

  • wendy fleet

    I wonder if clikking pec every 23 +/- 5 minutes all day and most of the night changes anything — some Epsilon Factor? I breathlessly check the meta-margin — like scrying tea leaves. Or is all this clikking just wearing out my mouse?

  • Canadian fan

    By far the best election night guide I’ve seen. Superb work, Sam. I’ve always watched election night returns with excitement, but you provide us with a method for spotting clues of trends in the early results. I seem to recall that for Kentucky – and other states that straddle two time-zones – they wait until they are all closed before giving out results. But perhaps not. The meta-margin is indeed stable – even if some isolated polls are not – and that gives reassurance. The point both you and Nate make regarding election night waves – that can occur to either party – helps a great deal in sifting through the early results and looking for clues. What makes this all so unique is that these are the closest collection of senate races in over a decade. That was not supposed to happen with a map like this. But it has. That’s a clue in and of itself.

    • Olav Grinde

      I am considering a prudent course of action: booking a one-way, December flight to Europe for me and my wife – cancelling if only if the Democratic Party prevails on the 4th of November. ;)

    • Amitabh Lath

      Olav, I think Europe has its own problems with the right wing. I remember a trip to CERN in 2012 and there were Marine Le Pen posters all over the little French town where I was staying.

  • Infidel753

    It’s going to be a nail-biter. Comparing polls to actual results will basically reveal how accurate the polls’ likely-voter screens were. That in turn will tell us how relatively effective the Democrats’ emphasis on GOTV, and the Republicans’ efforts to make voting more difficult for groups that don’t usually vote for them, have been.

    From what I’m seeing on Republican sites, they’re vastly confident. If it looks like they’re actually not going to do well, I’ll be spending a lot of the night reading them. If you really want to see people acting like they’ve been hit by a train, that will be the place.

  • Insidious Pall

    This is an excellent aid for flipping through the results. Invariably, I am flipping, clicking, switching my way through the sources trying to keep up with poll closings. This will be different – thanks, Sam.

  • Just curious

    Does anyone know why American news organizations don’t publish exit polls the moment the polls close? In other countries they literally count down the seconds and show the poll results.

    Just because it may be more difficult in the US in some regards doesn’t mean it’s impossible. And if they don’t know who has won – what’s the problem, just say it’s not the final result and give viewers the numbers. First exit polls, then projections every few minutes.

    • Insidious Pall

      The fear is that exit poll results can affect voters waiting in line when the polls close (alleged in FL 2000) or, in the case of presidential elections, that people in western states may not bother if the race has already been called. In the latter case, news organizations at times will release demographic and turnout data immediately, but not actual count. That said, it seems to me there has been a trend back to early release of actual results in recent years.

    • Matt McIrvin

      Another problem is that exit-poll results are usually adjusted using statistics collected on the polled sample over the course of the evening, and raw results released early can give a very inaccurate picture.

      On top of that, they have an unwarranted reputation for being extremely accurate, and when they don’t match the results, it usually starts people yelling about the election being rigged. But I don’t suppose media outlets would have major scruples about that.

    • Sam Wang

      Actually, exit polls are reweighted based on measures such as PEC-style poll medians, or actual election results. They are not an independent measure.

  • mediaglyphic

    Given that turnout is likely to be one of the biggest determinants of the election, i wonder if anyone has redone the analysis doen in 2012 using google search data to predict who will vote.

    This isn’t polling data, but its quantitative. I’ve emailed the author of this piece to ask how he performed this analysis, but received no response yet. Does anyone here know how to perform this analysis?

  • Zathras

    With all the analysis on polling averages, has anyone done an analysis of polling spreads? I have looked at the 10 closest races in 2010, and the 10 closest races in 2014, looking at polls for the last 2 weeks of the campaign, and the average standard deviation of the polls in 2014 is about 50% higher than the average standard deviation of the polls in 2010. Anyone see an analysis like this?

  • Dan Ferrisi

    Would you be more surprised to see Republicans gain a net +9 seats, or to see Democrats only lose a net -4 seats?

  • kevin cripe

    I have seen that polls over the last couple cycles have been off by about 3.5% in favor of republicans. (especially noticeable in places like Colorado). I have a theory about it. If a poll was taken in Colorado and 10% of respondents were African-American that poll would skew democratic. However, that group along with gender differences and other racial differences are screened. To my knowledge there is one group which is not normally screened that is the evangelical vote. It is much different than say the catholic vote because it slants way republican(maybe 80%) If you don’t poll the right % of Evangelicals the polls will skew dramatically one way or the other. Those would be easy statistics to compute. Is that a possibility?
    Kevin Cripe

  • bks

    I can’t blame anyone for staying up late and watching the returns, but I’m a morning person so I will turn in early and get a good night’s sleep before attempting to absorb the results.

    I just drove from Chicago to Fort Myers and back, and the strangest political ads on the radio were the ones for the Medical Marijuana iniative in Florida. I didn’t even know it was on the ballot there, but I will be looking forward to the results. Outside of those, pretty much ever ad I heard was an attack ad. No wonder the simian-on-the-street has such a low opinion of the government. –bks

  • Michael Clark

    Why do Americans such early poll-closing times? In Ireland, polls close at 22:00, allowing people to vote after finishing work. Indeed, in many Dublin polling stations, a majority of votes are cast after the time that polls would close in the US. Kentucky is particularly ridiculous with its 18:00 closing time…are Kentuckians afraid to vote in the dark?

    • securecare

      Our agricultural history which we still haven’t gotten beyond yet.

    • Kenny

      We do have laws that allow employees to take time off to vote.

    • Matt McIrvin

      On the whole, Americans don’t particularly want everyone to be able to vote.

    • WoodyinNYC

      Indiana and Kentucky use the early voting hours to depress turnout. It’s no accident of history.

      Years ago when unions began to grow, the Repubs feared that the unwashed masses could be led by union activists from the factory door to the polling place.

      In those days, before many places had union contracts, factory workers were still on the job at 6 p.m. and well after. So the legislatures arranged to have the polls closed BEFORE working people get off work.

      A very successful strategy, keeping Indiana deep red and Kentucky mostly red as well. Who is going to change it? And how?

  • tfitz


    Are the polls winding down such that the meta=margin will likely be static?

    • Froggy

      The number of polls being released is increasing, but the last few days they’ve not resulted in any great movement. When there are a lot of recent polls already out there (e.g., Colorado, with about 15 polls in the last two weeks), it takes multiple new polls all pulling in the same direction to make a sizable shift in the median.

    • Sam Wang

      I wonder if in the home stretch I should tighten the window to 7 days.

    • Froggy

      Tightening the window to 7 days would mean relying on very few polls in states that aren’t polled often, Alaska for instance. A better approach might be to keep the window at 14 days, but limit the number of polls used if there are a large number of polls in that time period. Only considering the last seven polls would allow for more volatility.

    • Sam Wang

      Agree re Alaska. Good suggestion.

      Search this site for “deviance minimization” or “variance minimization” to see how I’ve dealt with last-minute movement in the past.

  • CRM

    C’mon, New Hampshire…baby needs new shoes! ;)

  • Savanna

    Even though I don’t really want to know the answer to this question, is Charley Baker still ahead by a lot in Massachusetts?
    Why do the dems of Massachusetts keep nominating this woman?

  • JayBoy2k

    You once provided a pointer to Upshot. Right now, Upshot indicates that PEC is calling 52 R and 48 D states using a list of every state probabilities; It used to be 51-49 before Alaska switched from D to R; plus separately a 59% probability of Rs taking the Senate.
    This seems to match the Power of the Vote Column. If you make an adjustment, it would be good to understand how the adjustment (versus a poll) changed the Power of the Vote Column.

  • Diedealan

    I think it is safe to say, at the weekend before, that the most likely result is for a 50/50 election, with the Senate not decided until after the runoffs and re-counts.
    If this is the case, then please watch out for how politics will take over. Party switching, especially by vulnerable Blue State 2016 Republican senators is possible. And, the most intriguing possibility is Ormand wins Kansas to become the greatest freshman Senator king-maker of all time. He could be the deciding factor, but not before he gets a lot of things he wants from the party desperate to hold the Senate.

  • G-Horn

    This is just a hunch, but I think Mitch underperforms. I live in Louisville, and have a pretty good feel for the political mood here. McConnell is not a popular politician by any stretch. I know he’ll win because of the rural-red vote and because he painted Grimes as an Obama lackey. But I think people are much more motivated to turn out for her and against him. The rest of the state? I’m always disappointed, but I think she’ll keep it closer than 6%. If she’s had any success in Paducah and Western KY, she might keep it within 4%.

    • Donny

      Or Alison could win. I’m a political junkie, and I’ve never seen a politician hated as much as McConnell is. If the Anti-Mitch turnout is extremely high, he could get beat. Let’s not forget that greater than 50% of Registered Voters in KY are Democrats, too. I personally know people who registered to vote for the first time ever JUST to vote against Mitch. That’s just pure hatred of the man… I agree that if Mitch does win, the margin is likely to be small, probably 4-5% or less.

  • Donny

    And what does it mean if McConnell loses? Don’t be too surprised if Alison wins Heidi Heitkamp style. I live in KY, and these polls are extremely strange based on what’s going on here on the ground. KY (like Alaska) is an extremely hard state to poll since the different parts of the state are SO different and numerous other factors. Namely, Dems have a huge registration advantage (greater than 50%) and control most statewide offices. The votes are there if they show up to vote. If Mitch wins, according to my number crunching based on polling plus past data, it will be less than 4% I believe.

  • Michael Allen

    The polls in Kentucky might contain a large error. A lot of poor white people in Kentucky finally have Health Insurance due to the Obamacare Medicaid expansion. And many of them know the Democrats brought them this relief. However, they may be too embarrassed to admit they are going to vote Democrat. To their friends, neighbors, or pollsters. Don’t believe me, just google “Bradley Effect.” Voters who feel a sense of stigma about their choice of candidates, don’t always tell pollsters the truth. And it could be in Kentucky that some conservative voters will not admit to ANYONE that they plan to vote for a Democrat.

    • Donny

      Yes, not only that, but the poorest people tend to vote for the Democrat. At least according to exit polls from the last time Mitch ran in 2008. Most of these people don’t have a landline, cell phone, or internet. They certainly aren’t going to be counted in polls. They may not like everything Democrat, but they sure know the Republicans (especially Mitch) aren’t for them. The East Part of the state is the poorest, and it tends to be kind to Democrats despite being a MUCH different Dem voter than say urban Louisville. Lastly, many of these polls have oddities in the numbers. For example, the last one that came out had 49% of voters as conservative. The last time Mitch ran, it was 39%. Maybe they are more conservative now, or maybe they oversampled Mitch-friendly voters by like.. a lot. The old exit polls are on still.

Leave a Comment