Princeton Election Consortium

A first draft of electoral history. Since 2004

In State Races, As Much Suspense As 2006 and 2010 Combined

October 26th, 2014, 11:21pm by Sam Wang

Journalists and pundits have lavished considerable attention on the question of who will control the Senate in 2015. But a broader phenomenon has escaped notice: the sheer number of close state-level races, both in the Senate and in statehouses. At risk are many incumbents who were elected in previous wave years: in 2010 for Republican governors and in 2008 for Democratic senators.

State-level elections in 2014 have the largest number of close races in the last 10 years. This chart shows the number of gubernatorial and Senate races that have been won by three percentage points or less in midterm and Presidential-year elections since 2004. For 2014, the bars indicate races where polling margins are as small. There are 11 close gubernatorial races and 7 close Senate races, totaling 18 in all. This is equal to the last two midterm years, 2006 and 2010, combined.

A common theme in these races is the risk to Democratic and Republican incumbents alike. Six of the governors are Republicans were elected in 2010, a very good year for their party. Four of the Senators are incumbents elected in 2008, a very good year for Democrats. In this respect, the political pendulum is swinging back – in both directions at once.

The closest races are listed in the ActBlue (Democrats) page at left, and are also funded by the NRSC (Republicans). For readers on either side of the aisle, the selected races are highly effective targets for directing resources.

Nine of the eleven governors in close races are incumbents. Usually, they would be safe bets for re-election: from 2010 to 2013, the re-election rate of incumbent governors has been 88%. With 28 governors running for reelection this year, we might have expected three or four to be defeated. This year, two are already as good as headed out the door. Not included in this chart is Democrat Neil Abercrombie of Hawaii, who was defeated in the primary election. Also not shown is Republican Tom Corbett of Pennsylvania, who currently lags by nine percentage points in his race and is nearly certain to lose. So the rate of involuntary retirements is close to the historical record, over a week before Election Day and before even a single general election has been completed.

Since senators began to be elected by popular vote in 1914, their re-election rate has been 89%. Twenty-eight sitting Senators will be on the ballot next week. Potential losers include six close contests that feature an incumbent in the race: Alaska, Colorado, Kansas, Kentucky, New Hampshire, and North Carolina. Four of these incumbents are Democrats. In addition, two Democratic Senators lag by four points or more, Mark Pryor of Arkansas and Mary Landrieu of Louisiana. Because the number of potential Democratic losses is so large, a Republican takeover of the Senate looks more probable than not, though by a slim margin.

Tags: 2014 Election · governors · Senate

29 Comments so far ↓

  • Canadian fan

    The volume of ballots returned between Democrats and Republicans in Iowa does not reveal a key data point – how many are ” marginal voters ” – that is, those who don’t usually vote in midterms as opposed to those that do. The key to a Braley success will be to the degree to which marginal voters have been located and encouraged to vote. As the Democrats ground game in Iowa ( and other battleground states ) has specifically targeted voters than didn’t vote in 2010 – this will be key to not only Iowa but all close senate races. There have been a stream of articles during the month, however, that suggest the Democrats have been successful in this push. What Olav Grinde states is true. If Braley leads in early votes by six percent, and particularly if he leads independents by 3.4 %, we have the makings of a Braley win.

  • Edward G. Talbot

    I have been following the site off and on for months as there is too much noise on 538 ever since the move to ESPN, and I have come to appreciate the direct and simple (but not too simple) approach Dr. Wang takes, as well as comments largely devoid of hysteria. I have a question that is pretty simple and I apologize if it has been answered. Can someone direct me to an explanation of the “margin” column in the “Power of Your Vote” graphic is calculated? I know the data comes from HuffPo and I know it is median-based. I seem to recall seeing that it uses polls back to June and uses the most recent result for each pollster (LV poll if both RV and LV are available). But several of the values there now – Begich +3 and Hagan tied for example – don’t match what I just outlined and wouldn’t have matched that data hours or a day or two ago either (in case the graphic is just a little behind HuffPo data).

    So I figure I must be missing something. Any help is welcome.

    • JayBoy2k

      Under “Power of your vote”, there is Why do I care about this. Clip on that to get to FAQs. One answers the question of what polls are used and how.
      It no longer uses polls back to June, only the last week.

    • Edward G. Talbot

      Thanks, Jay! That all lines up now.

  • Olav Grinde

    Interesting news from Iowa in a Bloomberg article (referred to by Prof. Michael McDonald):

    “Among those who have already cast their ballots during Iowa’s early voting period, Braley had a six-point edge. Braley also has a small advantage among independent voters, 43.8 percent to 40.4 percent.”

    We know Iowa is close. These numbers seem significant, since independent voters account for 18.6% of those who have cast an early vote.

    Over 322,000 Iowans have already voted – a number that equals 28.5% of total ballots cast in the 2010 election.

    • Insidious Pall

      According to the Iowa SoS office, 41% of the returned ballots are Dems and 40% are Repubs. This is as of yesterday. The GOP is ahead of where they were in 2010 and even in 2012. But what it actually portends for the final tally is anyone’s guess.

  • Johnny Beat

    I’m probably in for a nasty Election night but I have this beautiful feeling that the Democrats are going to win all seven of these (along with Alaska & Kentucky.)

  • Keith

    Another Alaska poll showing Begich with a lead 50-42 based on the tightest likely voter screen.

  • Canadian fan

    In 2012, 19.96 % ( about 145,000 ) of the population of Alaska was found to be American Indian and/or Alaskan Native – by far the largest proportion of native population of any state in the entire county. As this race will likely be decided by less than 30,000 votes, this huge, largely untapped sector of the voting public could not only turn a race, but do so in a commanding way. One thing is certain. If Begich succeeds, it will act as a template for all future elections – local, state-side and national – that are held in Alaska.

    • Olav Grinde

      Interesting that the last four Alaska polls show:

      - Begich with sizable lead in two (6%, 10%)
      - One tied poll
      - Sullivan leading in just one (4%)

      (Previously Sullivan was leading in a long row of polls.)

      I realize that Alaska is considered notoriously difficult to pull, but could this be an indication that Begich may have turned the race around in his favor?

  • JayBoy2k

    ” But a broader phenomenon has escaped notice: the sheer number of close state-level races, both in the Senate and in statehouses….” Thanks Sam for an excellent article based on election data and a potential key insight.
    The real discussion is WHY — I do not know why, but would like to think that a politician in a “swing state” promised something — a fiscal policy, an action, moderation in voting, and did the opposite.. and the people of that state will make the politician pay!! What a concept. —
    It will be interesting to see how independents in each State vote compared to 2008 & 2010.

    Hard to think this is a wave response.. What about McConnell & Roberts, and Pryor and Landrieu? Each was 1st elected decades ago.

    Anti-Incumbent compared to recent election history – YES! — I think independents must be feeling abused by both parties.

    Would love to see more input on why or why not this phenomenon is happening. It may be a good thing!!!

    • Jeremy lehman

      This looks like a “throw the bums out!” election. Since the bums in the Senate were last elected in a democratic wave and the bums in governors mansions were last elected in a republican wave, there’s more of them from those parties (respectively) to throw out. But even still, there’s still opposite party bums (Roberts/McConnell in the Senate, Quinn & Hickenlooper in the governor’s mansions) to get rid of.

  • Matt McIrvin

    I think it’s not so much an anti-incumbent election per se, as one in which a lot of incumbents currently running were elected at very favorable times in marginal places. Regression to the mean, in other words.

    • Joseph

      One might call that an “anti-wave” election, I suppose. I like the idea that it’s a mixture of both anti-incumbent and anti-wave.

  • Joseph

    So what I hear you saying is that this is an anti-incumbent election. Doesn’t take much imagination to figure out why. The Reps have created the mother of all do-nothing Congresses. The problem for Democrats is that the Reps are innoculated against this negative feedback loop in the House by gerrymandering. It will take dynamite to explode them out of that power structure. A classic example is in my neck of the woods. California’s 6th district race is exceedingly close, with the Republican challenger accusing the Democrat incumbent of doing nothing in Congress. Utterly ironic! Of course, this was always a divided district, and thus a close election was probably inevitable. Still, the anti-incumbent sentiment is almost certainly having an impact, irony notwithstanding.

  • Amitabh Lath

    Frequency of polls is another indicator of a close race. Look at the difference in the number of polls conducted in Hawaii vs. Iowa.

  • Amitabh Lath

    How did you pick 3% as the window? Historically, did half the races that were within 3% in late October flip one way and half the other?

    • axt113

      Isn’t 3% around the typical margin of Error for most polls?

    • Kenny Johnson

      From some of Sam’s previous articles — in previous elections, the polls were off by as much as 3.5%. I believe that’s about how far off it was (on average) in 2010.

    • Amitabh Lath

      If you calculate statistical uncertainty as 1/sqrt(N) then the polls that have N>1111 would have 3% or lower. But most of the polls listed on Huffpost pollster have datasets in the 600 – 700 LV size

    • Sam Wang

      All the calculations – Katz’s, mine – depend on averages or medians of multiple polls. So the effective sample size is larger.

    • Sam Wang

      Josh Katz’s analysis at the NYT indicates that margins <4% are the only ones that err at this stage. Maybe I should have set the threshold at 3.9%.

  • Zathras

    Just to stretch the record a little further back:

    In 2002, there were 5 close Senate races:

    and 4 close Governor’s race:

  • Canadian fan

    An extremely revealing article, Sam – as always – replete with a very easy to read graph. As you say, this is something that has been given short shrift in the media, but is – as you show – of historical proportions. There is also the historical gravitational pull of the senatorial elections. As Republicans have never defeated more than two Democratic sitting senators in any one cycle since the Reagan wave of 1980 ( including those in their own waves of 1994 and 2010 ), in order to establish a majority, Republicans would almost certainly have to break this precedent. We could be headed towards the ultimate cliff-hanger – whereby the Senate majority is determined by a January Georgia run-off. If up to that point, the Republicans have 50 seats and Democrats have 48 seats ( including two independents ), Orman would be the 49th senator. If the Democrats won Georgia, Orman would become the 5oth senator. If he subsequently decided to caucus with the Democrats, that would decide the majority.

  • Jack Rems

    Of these 18 elections, how many might need a run-off? It seems to me run-offs generally favor Rs, because of low turnout. Any stats on this?

    • Froggy

      I’m pretty sure that it’s just Georgia and Louisiana that could end up in runoffs.