## Friday, May 23, 2008 at 6:33 pm by Darryl

## Poll Analysis: Clinton Still Beats McCain

Clinton |
McCain |

99.9% probability of winning | 0.1% probability of winning |

Mean of 313 electoral votes | Mean of 225 electoral votes |

From yesterday’s analysis, Sen. Hillary Clinton had a 99.9% probability of defeating Sen. John McCain in a general election. There were three new polls today that contribute to this match-up. Not much has changed.

After 10,000 simulated elections, Clinton wins 9,994 times (plus the House would give her the one tie), and McCain wins 5 times. Clinton still has a 99.9% probability of winning and McCain has a 0.1% probability of winning a general election held today.

The long term trends in this race can be seen from a series of elections simulated every 7 days using polls from 23 Sep 2007 to 23 May 2008, and including polls from the preceding 1 month (FAQ).

For the current time period, this is the distribution of electoral votes [FAQ]:

- 10000 simulations: Clinton wins 99.9%, McCain wins 0.1%.
- Average ( SE) EC votes for Clinton: 313.2 ( 13.4)
- Average (SE) EC votes for McCain: 224.8 ( 13.4)
- Median (95% CI) EC votes for Clinton: 314 (286, 338)
- Median (95% CI) EC votes for McCain: 224 (200, 252)

State | EC Votes | # polls | Total Votes | % Clinton | % McCain | Clinton %wins | McCain %wins |
---|---|---|---|---|---|---|---|

Alabama | 9 | 1^{*} |
811 | 39.6 | 60.4 | 0.0 | 100.0 |

Alaska | 3 | 2 | 997 | 40.3 | 59.7 | 0.0 | 100.0 |

Arizona | 10 | 1 | 519 | 41.0 | 59.0 | 0.0 | 100.0 |

Arkansas | 6 | 1 | 460 | 57.6 | 42.4 | 99.7 | 0.3 |

California | 55 | 2 | 1412 | 57.6 | 42.4 | 100.0 | 0.0 |

Colorado | 9 | 1 | 455 | 48.4 | 51.6 | 22.4 | 77.6 |

Connecticut | 7 | 1^{*} |
1477 | 51.7 | 48.3 | 91.7 | 8.3 |

Delaware | 3 | 1^{*} |
532 | 52.8 | 47.2 | 90.9 | 9.1 |

D.C. | 3 | 0 | (100) | (0) | |||

Florida | 27 | 3 | 2973 | 54.1 | 45.9 | 100.0 | 0.0 |

Georgia | 15 | 1 | 425 | 43.5 | 56.5 | 0.8 | 99.2 |

Hawaii | 4 | 1^{*} |
487 | 52.4 | 47.6 | 86.0 | 14.0 |

Idaho | 4 | 1^{*} |
548 | 29.9 | 70.1 | 0.0 | 100.0 |

Illinois | 21 | 1^{*} |
512 | 56.4 | 43.6 | 99.6 | 0.4 |

Indiana | 11 | 2 | 1748 | 49.2 | 50.8 | 20.8 | 79.2 |

Iowa | 7 | 1 | 435 | 48.3 | 51.7 | 22.4 | 77.6 |

Kansas | 6 | 1 | 460 | 42.4 | 57.6 | 0.2 | 99.8 |

Kentucky | 8 | 1 | 564 | 43.6 | 56.4 | 0.3 | 99.7 |

Louisiana | 9 | 1^{*} |
470 | 38.3 | 61.7 | 0.0 | 100.0 |

Maine | 4 | 1 | 445 | 57.3 | 42.7 | 99.6 | 0.4 |

Maryland | 10 | 1^{*} |
552 | 55.1 | 44.9 | 98.4 | 1.6 |

Massachusetts | 12 | 1 | 455 | 60.4 | 39.6 | 100.0 | 0.0 |

Michigan | 17 | 1 | 440 | 50.0 | 50.0 | 50.0 | 50.0 |

Minnesota | 10 | 1 | 994 | 55.0 | 45.0 | 100.0 | 0.0 |

Mississippi | 6 | 1 | 546 | 39.6 | 60.4 | 0.0 | 100.0 |

Missouri | 11 | 2 | 1872 | 50.5 | 49.5 | 72.6 | 27.4 |

Montana | 3 | 1^{*} |
450 | 40.0 | 60.0 | 0.0 | 100.0 |

Nebraska | 5 | 1 | 455 | 37.4 | 62.6 | 0.0 | 100.0 |

Nevada | 5 | 1 | 435 | 52.9 | 47.1 | 88.9 | 11.1 |

New Hampshire | 4 | 3 | 1315 | 51.0 | 49.0 | 79.3 | 20.7 |

New Jersey | 15 | 1 | 723 | 57.8 | 42.2 | 100.0 | 0.0 |

New Mexico | 5 | 1 | 440 | 53.4 | 46.6 | 92.3 | 7.7 |

New York | 31 | 2 | 1015 | 60.8 | 39.2 | 100.0 | 0.0 |

North Carolina | 15 | 4 | 2128 | 47.8 | 52.2 | 3.3 | 96.7 |

North Dakota | 3 | 1^{*} |
511 | 39.3 | 60.7 | 0.0 | 100.0 |

Ohio | 20 | 3 | 2541 | 54.6 | 45.4 | 100.0 | 0.0 |

Oklahoma | 7 | 1^{*} |
552 | 45.7 | 54.3 | 2.8 | 97.2 |

Oregon | 7 | 1 | 430 | 53.5 | 46.5 | 92.8 | 7.3 |

Pennsylvania | 21 | 5 | 4349 | 56.8 | 43.2 | 100.0 | 0.0 |

Rhode Island | 4 | 1^{*} |
571 | 59.4 | 40.6 | 100.0 | 0.0 |

South Carolina | 8 | 1^{*} |
536 | 46.6 | 53.4 | 6.0 | 94.0 |

South Dakota | 3 | 1^{*} |
440 | 43.2 | 56.8 | 0.5 | 99.5 |

Tennessee | 11 | 1^{*} |
450 | 42.2 | 57.8 | 0.1 | 99.9 |

Texas | 34 | 2 | 1006 | 44.0 | 56.0 | 0.0 | 100.0 |

Utah | 5 | 1 | 514 | 23.5 | 76.5 | 0.0 | 100.0 |

Vermont | 3 | 1^{*} |
551 | 55.7 | 44.3 | 99.2 | 0.9 |

Virginia | 13 | 2 | 1122 | 45.6 | 54.4 | 0.1 | 99.9 |

Washington | 11 | 2 | 1065 | 52.4 | 47.6 | 93.2 | 6.8 |

West Virginia | 5 | 1^{*} |
549 | 52.8 | 47.2 | 91.2 | 8.8 |

Wisconsin | 10 | 1 | 450 | 47.8 | 52.2 | 16.3 | 83.7 |

Wyoming | 3 | 1^{*} |
508 | 31.5 | 68.5 | 0.0 | 100.0 |

^{*} denotes that an older poll was used

Details of the methods are given in the FAQ.

The most recent analysis in this and other match-ups can be found from this page.

Sunday, May 25th, 2008 at 6:10 am

This is some of the craziest stuff I’ve ever seen! Pick a swing state, Ohio, say. You simulate this as a guaranteed win on the basis that every voter randomly picks a candidate with 54.6% chance that it’s Clinton. Of course this gives you Clinton every time. The truth is that 54.6% is just a best guess at the moment. There is a margin of error. The correct way to do this is pick a number from the normal distribution centred on 54.6 and with the appropriate variance. That’s the odds of Clinton winning the state.

Surely you can see that Clinton is not 99.9% certain to win the nomination. Your forecast basically adds up to just taking the poll results and assuming they’re perfect.

Sunday, May 25th, 2008 at 6:23 am

You don’t know what you’re doing.

For an example of someone who does, go here: http://www.fivethirtyeight.com

Sunday, May 25th, 2008 at 6:59 am

Wow, talk about making completely hopeless analysis look very nice. Might even call it a polished turd.

Sunday, May 25th, 2008 at 7:41 am

Okay, really. Did you *really* think your math was good when you showed Clinton with a 100% chance of winning Florida? With a 75% chance of winning Missouri, even though she only had a 1 point advantage in the polls? Really? Did you really not notice that a completely absurd number of your predictions had probabilities of 90+? Really? Come on.

Sunday, May 25th, 2008 at 7:50 am

I’m sorry, you can’t ever suggest or outright say that Hillary Clinton would do better than Our Lord Jesus Obama. That is wrong, incorrect, and completely false as it is an absolute fact that Obama will bring us all together, make us love one another, and cause us to end war, hunger, and suffering. Please stop spreading these terrible lies that Hillary Clinton actually does better than Obama AND changes the map for Democrats. She’s evil, horrible, terrible, and mean. She is not allowed to win. These are just the facts.

Sunday, May 25th, 2008 at 9:20 am

Even if a good model was used, to use current polling data to predict what will happen in November is silly.

The notion that she or any candidate has a 99.9% chance of winning the White House is absurd.

A lot will happen between now and the election which will change the polls (both McCain and Obama will make mistakes. Events will happen that nobody can predict). And the actual results may be different from the polls.

And while I don’t think Clinton is evil, there are independents and Republicans who do and won’t vote for her.

I guess I shouldn’t be surprised that pro-Clinton blogs are picking this up without the least bit of skepticism.

Sunday, May 25th, 2008 at 3:24 pm

Steve,

“Okay, really. Did you *really* think your math was good when you showed Clinton with a 100% chance of winning Florida?”There isn’t really any “math” behind the methods, but I’ll try to address your specific comments. In Florida, there are three current polls, a Rasmussen with 500 individuals giving Clinton a 47% to 41% edge, and two Quinnipiac polls, one with 1,419 individuasl giving Clinton a 48% to 41% edge, and another with 1,411 individuals giving Clinton a 49% to 41% edge. The statistical evidence is overwhelming…when the polls are polled, a McCain win is well outside the margin of error (if the election were held today).

“With a 75% chance of winning Missouri, even though she only had a 1 point advantage in the polls?”In Missouri there are two current polls, a huge SurveyUSA poll (1,523 individuals) that gives Clinton a 48% to 46% lead, and a small (500 person) Rasmussen poll that gives McCain a 45% to 43% edge. When the polls are combined, the larger SurveyUSA poll dominates the results. Clinton wins about 75% of the simulated elections.

“Really?”Yep!

“Did you really not notice that a completely absurd number of your predictions had probabilities of 90+? Really?”Why is that absurd? I mean, look at it this way…when pollsters say a poll gives a result that is “within the margin of error” or “statistically tied” all they mean is that the probability that one candidate wins over the other is under 95%. When a poll is “outside of the margin of error” or “statistically significant” that just means that the probability the “winning” candidate being truly is ahead of the trailing candidate is greater than 95%.

This is all elementary statistical sampling theory. I am not sure where your disbelief arises.

Sunday, May 25th, 2008 at 3:32 pm

the real facts,

“Even if a good model was used, to use current polling data to predict what will happen in November is silly.”Indeed! That is why I am not making such a prediction. All of my posts point out that the results only apply to a general election held today. The FAQ mentions that as well. Are you reading impaired or something?

“The notion that she or any candidate has a 99.9% chance of winning the White House is absurd.”But, the numbers coming out of the polls don’t support your claim of absurdity.

A lot will happen between now and the election which will change the polls (both McCain and Obama will make mistakes. Events will happen that nobody can predict). And the actual results may be different from the polls.”First…the results only apply to a hypothetical general election held right now. Secondly, my analyses give a distribution of outcomes, not a single outcome. The actual results of an election held right now would likely fall within the distribution of results generated from the polling data.

And while I donâ€™t think Clinton is evil, there are independents and Republicans who do and wonâ€™t vote for her.”But…of course, the pollsters try to capture this effect. So your point is irrelevant.

“I guess I shouldnâ€™t be surprised that pro-Clinton blogs are picking this up without the least bit of skepticism.”I wouldn’t be surprised either. When McCain was beating Clinton, the pro-McCain folks certainly picked it up “without the least bit of skepticism.”

Sunday, May 25th, 2008 at 3:45 pm

John H,

“This is some of the craziest stuff Iâ€™ve ever seen!”Hmmm…7.5 years of a NeoCon-inspired Wingnut shit-fer-brains running the country and THIS is the craziest stuff you have ever seen??????

“Pick a swing state, Ohio, say. You simulate this as a guaranteed win on the basis that every voter randomly picks a candidate with 54.6% chance that itâ€™s Clinton. Of course this gives you Clinton every time.”Ummm…I’m not sure what you are getting at here. If I flip a coin that has a 54.6% probability of heads, then I will more often get heads than tails for each flip, but not every time. However, if I have 1,000 such coins and flip ‘em all, the probability of getting more tails than heads is something like 0.04%

The truth is that 54.6% is just a best guess at the moment.”The 54.6% is a summary statistic. The read data are the number of people who said they would vote for McCain and the number of people who said they would vote for Clinton or Obama in the poll.

“There is a margin of error.”Indeed. The distribution of electoral votes you see in my posts quantifies that error. That is why there is a distribution over a range of electoral votes. The probability just comes from adding up the area to the left or right of the winning number (269 for the Democrat and 270 for McCain).

“The correct way to do this is pick a number from the normal distribution centred on 54.6 and with the appropriate variance.”My method is using a binomial distribution to find the variance of outcomes. What you have suggested is using the Normal approximation to the binomial distribution. My method is the correct way—your’s is an approximation that 20th century statisticians used before computers were readily available.

“Surely you can see that Clinton is not 99.9% certain to win the nomination.”This work (1) has nothing whatsoever to do with “the nomination.” I am simulating hypothetical general election match-ups. (2) My analyses only pertain to right now, not November. I make that clear in my posts.

Your forecast basically adds up to just taking the poll results and assuming theyâ€™re perfect.Um…no. My analyses were specifically designed to avoid simple pooling of polls and specifically designed to generate a distribution of election outcomes.