Poll Analysis: Romney wins some, loses some

99.5% probability of winning 0.5% probability of winning
Mean of 323 electoral votesMean of 215 electoral votes

There were lots of new polls released in the two days since my previous analysis:

WISt Norbert College17-May22-May4065.04943O+6

The good news for Romney is that Arizona gives him a modest +7% lead over Obama. And Romney will need it, now that Secretary of State Ken Bennett (R-AZ) has concluded that Obama really was born in Hawaii. Although there is an occasional poll showing Obama ahead, the larger trend shows the state in Romney’s column:

Other good news for Romney is that North Carolina has him up by +2% over Obama, and he takes three of the four May polls. The larger trend has favored Obama since late 2010, but things are clearly changing in the state:

ObamaRomney25Apr12-25May12North Carolina

Florida is up to its old tricks of bouncing between candidates. After being down -6% in the previous poll, Obama scores +5% over Romeny. The overall trend in FL looks more favorable to Obama, but that can clearly change:


In good news for Obama, we finally get the first poll out of Maryland. Obama has a dominating +23% lead over Romney.

Obama takes another Ohio poll with a +5% lead over Romney. Obama has now led in the last ten Ohio polls, dating back to late February.

Virginia gives Obama +4% over Romney. The recent trend still favors Obama in the state, but there is at least a hint that the state is swinging toward Romney:


Both Wisconsin polls give Obama the lead over Romney, and solidify his lead in the state.

Today, after 100,000 simulated elections, Obama wins 99,494 times and Romney wins 506 times (including the 51 ties). Obama receives (on average) 323 to Romney’s 215 electoral votes. In an election held now, Obama would have a 99.5% probability of winning and Romney, a 0.5% probability of winning.

The probability is a slight improvement for Romney, up from a 0.2%. On the other hand, Romney’s expected electoral vote total drops from 220 to 215, as Obama’s increases from 318 to 323.

Electoral College Map

AlabamaAlaskaArizonaArkansasCaliforniaColoradoConnecticutDelawareFloridaGeorgiaHawaiiIdahoIllinoisIndianaIowaKansasKentuckyLousianaMaineMarylandMassachusettesMichiganMinnesotaMississippiMissouriMontanaNebraskaNevadaNew HampshireNew JerseyNew MexicoNew YorkNorth CarolinaNorth DakotaOhioOklahomaOregonPennsylvaniaRhode IslandSouth CarolinaSouth DakotaTennesseeTexasUtahVermontVirginiaD.C.WashingtonWest VirginiaWisconsinWyoming

Electoral College Map

GeorgiaAlaskaArizonaArkansasCaliforniaColoradoDelawareConnecticutFloridaMississippiAlabamaHawaiiIdahoIllinoisIndianaIowaKansasKentuckyLouisianaMaineMarylandMassachusettsMichiganMinnesotaMissouriMontanaNebraskaNevadaNew HampshireNew JerseyNew MexicoNew YorkNorth CarolinaNorth DakotaOhioOklahomaOregonPennsylvaniaRhode IslandSouth CarolinaSouth DakotaTennesseeTexasUtahVermontVirginiaD.C.WashingtonWest VirginiaWisconsinWyoming

Here is the distribution of electoral votes [FAQ] from the simulations:

Ten most probable electoral vote outcomes for Obama:

  • 328 electoral votes with a 3.77% probability
  • 318 electoral votes with a 3.42% probability
  • 312 electoral votes with a 3.24% probability
  • 319 electoral votes with a 3.19% probability
  • 313 electoral votes with a 3.04% probability
  • 327 electoral votes with a 2.88% probability
  • 303 electoral votes with a 2.86% probability
  • 322 electoral votes with a 2.53% probability
  • 321 electoral votes with a 2.22% probability
  • 304 electoral votes with a 2.13% probability

After 100,000 simulations:

  • Obama wins 99.5%, Romney wins 0.5%.
  • Average (SE) EC votes for Obama: 323.3 (19.6)
  • Average (SE) EC votes for Romney: 214.7 (19.6)
  • Median (95% CI) EC votes for Obama: 322 (287, 364)
  • Median (95% CI) EC votes for Romney: 216 (174, 251)

Each column of this table shows the electoral vote total aggregated by different criteria for the probability of winning a state (Safe=100%, Strong=90%+, Leans=60%+, Weak=50%+):

ThresholdSafe+ Strong+ Leans+ Weak
Safe Obama105
Strong Obama159264
Leans Obama3939303
Weak Obama151515318
Weak Romney101010220
Leans Romney4040210
Strong Romney132170
Safe Romney38

This table summarizes results by state. Click on the poll count to see the individual polls included for the state.

84VotespollsVotesObamaRomney% wins% wins
AL91*754 37.8 62.2 0.0100.0
AZ1121329 45.6 54.4 1.2 98.8
AR61*679 36.8 63.2 0.0100.0
CA551796 56.2 43.8 99.3 0.7
CO91*564 50.0 50.0 49.4 50.6
CT71*1460 58.9 41.1100.0 0.0
FL2954448 49.3 50.7 26.3 73.7
GA162952 43.7 56.3 0.3 99.7
HI41*517 64.8 35.2100.0 0.0
IL201*546 61.5 38.5100.0 0.0
IN111*447 45.0 55.0 6.8 93.2
IA611086 55.4 44.6 99.4 0.6
KS61*442 45.0 55.0 6.8 93.2
KY81*528 45.5 54.5 7.3 92.7
LA81*542 41.1 58.9 0.1 99.9
ME21552 54.3 45.7 92.5 7.5
ME111*488 64.8 35.2100.0 0.0
ME211*421 53.7 46.3 85.7 14.3
MD101792 62.4 37.6100.0 0.0
MA1121015 62.7 37.3100.0 0.0
MI161508 53.3 46.7 85.6 14.4
MN101467 58.0 42.0 99.3 0.7
MS61*717 40.0 60.0 0.0100.0
MO101*465 48.4 51.6 31.7 68.3
MT321278 46.9 53.1 6.2 93.8
NE21460 42.4 57.6 0.8 99.2
NE111*389 45.5 54.5 10.0 90.0
NE211*252 49.6 50.4 45.2 54.8
NE311*284 35.9 64.1 0.0100.0
NV61480 54.2 45.8 90.0 10.0
NH411093 56.4 43.6 99.8 0.2
NJ1422077 56.5 43.5100.0 0.0
NM51*494 57.5 42.5 99.1 0.9
NY291720 60.7 39.3100.0 0.0
NC1553611 50.3 49.7 58.1 41.9
ND31*480 41.3 58.8 0.3 99.7
OH1843674 52.0 48.0 96.3 3.7
OK71448 30.4 69.6 0.0100.0
OR711327 52.0 48.0 84.2 15.8
PA2032063 54.3 45.7 99.7 0.3
RI41*495 59.4 40.6 99.8 0.2
SC91*1833 51.7 48.3 84.9 15.1
SD31*442 44.3 55.7 4.9 95.1
TN111654 46.0 54.0 7.3 92.7
TX381460 38.9 61.1 0.0100.0
UT61*688 33.0 67.0 0.0100.0
VT31528 67.8 32.2100.0 0.0
VA1332502 53.3 46.7 98.9 1.1
WA121481 57.8 42.2 99.1 0.9
WV51373 40.8 59.2 0.7 99.3
WI1063349 52.6 47.4 98.4 1.6

* An older poll was used (i.e. no recent polls exist).

Details of the methods are given in the FAQ.

The most recent analysis in this match-up can be found from this page.


  1. 1

    rhp6033 spews:

    THIS is the Romney “bump” from becoming the presumptive candidate? He’s got one more chance at a “bump” as a result of the convention, that’s a lot of ground to make up without much to support it. Romeny’s electoral vote strategy seems to be non-existent – he’s relying upon the same sates McCain carried, plus a couple more – the “southern & mid-west red states” – to pull off a win, which is a failing strategy.

  2. 2

    rhp6033 spews:

    Romney’s one big advantage is that there is NO WAY he’s going to lose Utah. And the Confederate states are also solidly in his pocket, no need for him to waste money competing there. Everything else is either solidly Democrate or in play.

  3. 4

    YellowPup spews:

    Darryl: Always appreciate the regular poll analyses. What about the gov/recall race in your old home state of Wisconsin? Granted, not a local or national race, but one with national significance, and where it’s been running pretty close.

    @1: Glad to see that Romney’s presumptive bump can be measured in .1%s.

  4. 5

    GetFactsFirst spews:

    Are you weighing all these polls equally in determining your projected State by State results?
    I’d have to say any poll that shows Obama or Romney winning by 10+ points ought to be discounted or eliminated.

  5. 7

    wharfrat spews:

    As a veteran of numerous prez campaigns starting with LBJ in 64 I’m well aware that things will get a lot tighter. I think that Obama will pull it off but keeping the Senate out of Republican hands is even more important. I’m personally looking forward to getting out of Las Vegas and back to Eastern WA to work and vote to dump McMorris-Rogers.

  6. 9


    GetFactsFirst @ 6

    “RealClearPolitics poll of Polls shows Obama @ +1.6% points today.”

    That RCP average is based on national polls and then simply averages the polls—pretty simplistic stuff. First, using national polls is an approximation of the popular vote. Yet we elect presidents through the electoral college, not the popular vote.

    So, my analyses differ in that I only use state head-to-head polls, and then I use the rules of the electoral college.

    Additionally, I am particularly interested in predicting the probability that each candidate would win, based on current polls—something that averaging cannot address. Therefore, I use a statistical method called Monte Carlo analysis. Essentially, I simulate 100,000 elections using samples corresponding to all recent state head-to-head polls. The analysis yields unbiased estimates of the probability that each candidate would win, based entirely on the evidence contained in the polls.

    “Have you done an analysis using the RealClear polls results??”

    Well…RCP doesn’t have any “poll results”, per se. They simply collect polls and create averages. They do collect state head-to-head polls as well, but they don’t aggregate the results according to the rules of the electoral college.

    I don’t do the Monte Carlo analyses on the national polls, simply because the winner of the popular vote isn’t always the winner of the election (as we learned in 2000).

  7. 10


    Getfactsfirst @ 5,

    “I’d have to say any poll that shows Obama or Romney winning by 10+ points ought to be discounted or eliminated.”

    WTF? There are many good criteria for “discounting” polls, a double digit spread isn’t one of them.

    Consider this: In 2004 there were 34 states* in which the state was won by a double digit lead. In 2008, there were 39 states* in which the state was won by a double digit lead.

    For obvious reasons, states that go for the D or R candidate by double digits are usually not polled as often as “purple” states. Still, there are plenty of them, and we should expect numerous accurate polls with double digit leads for one candidate or the other.

    * An actual state or D.C.

  8. 11


    Yellowpup @ 4

    “What about the gov/recall race in your old home state of Wisconsin? Granted, not a local or national race, but one with national significance, and where it’s been running pretty close.”

    Thanks for the suggestion. There have been a bunch of polls taken for that race this week. If I get a chance over the weekend, I’ll do an analysis of them….

  9. 12

    proud leftist spews:

    The more I see of Mitt Romney, the less I like him. I think that will be the country’s reaction to him as well. The guy just doesn’t connect, on any level, with the average person. The national polls indicate a close race, but I think Darryl’s numbers show what the electoral vote will be. Not that close.

  10. 13

    GetFactsFirst spews:


    “So, my analyses differ in that I only use state head-to-head polls, and then I use the rules of the electoral college.”

    I understand and agree..as it is the EC that really matters. The question is really about what polls & data you include in your State by State analysis. Seems like outlier polls that use different criteria about who they poll (say % of Known D’s and R’s) can really skew the results. What I meant to say was why even include a poll that shows Obama up by 25% or Romney up by 11% Nationally…their State by State data is likely to be less credible.

    If Walker wins big in Wisconsin recall (say by 4 points or more), how do you think that will translate into the Nov. Presidential Vote??
    Statistics revolves around credibility of underlying data and assumptions. You can eventually get whatever result you want by the data and assumptions you choose. I appreciate that you are explaining that Darryl. It would be interesting in how things would change if you yanked the “outliers” on both sides.

  11. 14


    GetFactsFirst @ 13,

    Essentially…I use every valid poll that is released to the public. (As it happens, my collection of poll is far more complete than RCP.)

    But in order to be included, a few “quality control” criteria must be met. The FAQ gives a strict set of criterion for inclusion:

    To be considered acceptable, each poll must come from a reputable pollster and must included the following information:

    1. The name of the poll or polling firm
    2. The inclusive dates on which the poll was taken
    3. The state in which the poll was taken
    4. The number of individuals polled
    5. The counts of or percentage of individuals supporting each candidate

    We usually do not include internet-based polls or polls from discredited pollsters (say…Research 2000 or Strategic Vision). We also ignore polls released by party organizations or candidates. The problem is that such polls are released strategically. Including them would bias the results.

    If a poll meets these criteria, I will include it regardless of whether or not it looks like an “outlier.”

    The idea of picking and choosing “outliers” is far more troublesome than including all polls. As the election season moves on, there will usually be multiple “recent” polls per state included in the analysis, so “outliers” are at least watered down.

  12. 15

    GetFactsFirst spews:

    Here is the RealClearPolitics poll of polls on the Wisconsin Recall–

    Walker up by 4 in the Democrat Poll and 5-6 in several others and 12 in yet another.
    I know some Wisconsin lifelong State Employee Union members nearing retirement who will vote for Walker. Others are choosing Walker because of the Recall precedent. They fear that if Walker barely loses, the R;s will launch a Recall after 1 year. Too costly just because you are sore losers. Plus retirees, even government workers, love Walker’s efforts to keep down government costs, bring in new private enterprise etc.