« Blue Rumble
Rossi Still Campaigning on ‘We Wuz Robbed!’ »

Sunday, March 9, 2008 at 10:49 pm by Darryl

On Those 50 SurveyUSA Polls

A few days ago SurveyUSA published head-to-head presidential match-up polls for all 50 states. The polls were taken between the 26th and 28th of February (a few took until the 29th), and they typically surveyed about 600 people. SurveyUSA published a summary page in which they tallied the electoral votes for each candidate in a general election.

For the Sen. Barack Obama–Sen. John McCain match-up the results suggested that, if the election were held in late February, Obama would take 280 electoral votes and McCain would take 258 electoral votes. The Sen. Hillary Clinton–McCain match-up gave Clinton a victory with 276 electoral votes to McCain’s 262.

The day after the polls were release, I posted new analyses for these match-ups that included all poll taken within the past month. The Obama–McCain match-up suggested that Obama would win with a 100.0% probability and an average of 309 electoral votes to McCain’s 229 electoral votes. The Clinton–McCain match-up suggested that McCain woud win. He had an 89.1% chance of beating Clinton with a mean of 291 electoral votes to Clinton’s 247 electoral votes.

A received a couple of emails from people asking about this discrepancy, and my Drinking Liberally buddy, Nicholas Beaudrot suggested I run my simulation analyses with only the SurveyUSA polls included. Great suggestion. First the results of my SurveyUSA-only polls.

The results of 10,000 simulated elections using the SurveyUSA polls gives Obama 8,832 wins (plus 54 ties) and McCain 1,114 wins. Obama has an 88.9% probability of winning with an average of 302 electoral votes and McCain an 11.1% probability of winning with an average of 236 electoral votes. Here is the map with links to the individual state results:

Electoral College Map

Washington Oregon Idaho Montana Wyoming North Dakota South Dakota Minnesota Wisconsin Iowa Illinios Indiana Colorado Utah New Mexico Texas Nebraska Oklahoma Kansas Nevada California Arizona Missouri Arkansas Louisiana Mississippi Alabama Tennessee Georgia Kentucky Ohio Michigan Pennsylvania West Virginia Virginia Hawaii Alaska North Carolina South Carolina Florida New York Vermont New Hampshire Maine Massachusetts Rhode Island Connecticut New Jersey Delaware Maryland District of Columbia

For the Clinton–McCain match-up, Clinton wins 6,797 times (plus 98 ties) and McCain wins 3,105 times. This gives Clinton a 69% probability of winning with an average of 277 electoral votes and McCain a 31% probability of winning with an average of 261 electoral votes.

Update: Alex emailed me and pointed out that the Tennessee results were incorrect for Clinton. I misread my own writing and entered 46% McCain to 40% Clinton. The correct poll numbers are 46% to 46%. When that correction is made,
Clinton has a 75.2% (plus 0.9% for ties) probability of winning, and McCain has a 23.9% probability of winning. Clinton’s mean electoral votes is 282 to McCain mean electoral votes of 256. Many thanks for Alex for pointing out the error!

Electoral College Map

Washington Oregon Idaho Montana Wyoming North Dakota South Dakota Minnesota Wisconsin Iowa Illinios Indiana Colorado Utah New Mexico Texas Nebraska Oklahoma Kansas Nevada California Arizona Missouri Arkansas Louisiana Mississippi Alabama Tennessee Georgia Kentucky Ohio Michigan Pennsylvania West Virginia Virginia Hawaii Alaska North Carolina South Carolina Florida New York Vermont New Hampshire Maine Massachusetts Rhode Island Connecticut New Jersey Delaware Maryland District of Columbia

The Clinton–McCain single-poll simulation analysis matches the SurveyUSA expectation for electoral votes almost perfectly. But what the SurveyUSA methods cannot do is predict the probability of winning. The simulation analysis is helpful for that, and shows about a two-thirds chance for Clinton to win.

For Obama–McCain single-poll simulation analysis Obama does better than the SurveyUSA expectation for electoral votes (302 versus 280). And the analysis suggests about a 90% probability that Obama beats McCain.

But the biggest discrepancy is between my analyses using all polls in the last month versus just the 50-state SurveyUSA polls. There are several likely reasons for the discrepancies. First, the dynamics of the race have been changing substantially in the last month, particularly as the Democratic forerunners have their streaks of primary victory. Clinton has improved substantially over the month. Obama seems to have has lost a little of his shine. So the differences probably reflect genuine secular change in the public perceptions.

But another reason has to do with pollsters and methodology. Every pollster has some bias that reflects sample selection, delivery mechanism (robo-poll versus live human questioner), wording of polls, rates of re-contact, etc. It seems like SurveyUSA produces polls that are slightly more favorable to Democratic candidates than does Rasmussen or Strategic Vision. But there is another factor as well. There were a plethora of polls conducted in mid-February by smaller, less experienced political pollsters. I’ve not specifically examined “the little guys” versus “the big guys,” but it may be a factor.

I chose a one month window precisely because I wanted multiple pollsters represented, and I though public opinion would be relatively stable over a one month window early in the campaign. As the general election approaches, and more and more polls come out, I’ll switch to smaller windows–first 2-week and then one-week windows.

What these data suggest to me is that Clinton should gradually do better as older polls drop out and new polls come in. Obama will slip a bit over the same period.

Obama–McCain EC Votes # polls Total Votes % Obama % McCain Obama %wins McCain %wins
Alabama 9 1* 567 42.5 57.5 0.0 100.0
Alaska 3 1* 561 47.2 52.8 10.6 89.5
Arizona 10 1* 552 43.3 56.7 0.0 100.0
Arkansas 6 1* 504 38.3 61.7 0.0 100.0
California 55 1* 539 56.0 44.0 99.8 0.2
Colorado 9 1* 573 55.0 45.0 99.2 0.8
Connecticut 7 1* 570 61.8 38.2 100.0 0.0
Delaware 3 1* 553 55.0 45.0 98.8 1.2
D.C. 3 0 (100) (0)
Florida 27 1* 581 48.9 51.1 29.4 70.6
Georgia 15 1* 563 43.2 56.8 0.0 100.0
Hawaii 4 1* 546 66.3 33.7 100.0 0.0
Idaho 4 1* 553 42.9 57.1 0.0 100.0
Illinois 21 1* 546 65.9 34.1 100.0 0.0
Indiana 11 1* 524 45.0 55.0 0.8 99.2
Iowa 7 1* 564 55.0 45.0 99.1 0.9
Kansas 6 1* 544 45.0 55.0 1.0 99.0
Kentucky 8 1* 527 38.0 62.0 0.0 100.0
Louisiana 9 1* 650 42.0 58.0 0.0 100.0
Maine 4 1* 588 57.7 42.3 100.0 0.0
Maryland 10 1* 577 57.0 43.0 100.0 0.0
Massachusetts 12 1* 579 53.9 46.1 97.7 2.3
Michigan 17 1* 585 50.6 49.4 61.6 38.4
Minnesota 10 1* 553 53.9 46.1 97.9 2.1
Mississippi 6 1* 591 43.1 56.9 0.1 99.9
Missouri 11 1* 568 46.7 53.3 5.3 94.7
Montana 3 1* 551 45.4 54.6 1.7 98.3
Nebraska 5 1* 542 48.3 51.7 19.2 80.8
Nevada 5 1* 532 52.8 47.2 91.3 8.7
New Hampshire 4 1* 573 51.1 48.9 72.1 27.9
New Jersey 15 1* 540 50.0 50.0 49.5 50.5
New Mexico 5 1* 558 53.8 46.2 97.8 2.2
New York 31 1* 533 57.8 42.2 100.0 0.0
North Carolina 15 1* 487 48.9 51.1 33.0 67.0
North Dakota 3 1* 503 52.3 47.7 83.7 16.3
Ohio 20 1* 566 55.5 44.5 99.6 0.4
Oklahoma 7 1* 569 40.1 59.9 0.0 100.0
Oregon 7 1* 565 54.5 45.5 98.8 1.2
Pennsylvania 21 1* 541 47.1 52.9 10.4 89.6
Rhode Island 4 1* 572 58.2 41.8 100.0 0.0
South Carolina 8 1* 554 48.4 51.6 22.6 77.4
South Dakota 3 1* 569 47.8 52.2 15.2 84.8
Tennessee 11 1* 569 41.3 58.7 0.0 100.0
Texas 34 1* 558 49.5 50.5 40.4 59.6
Utah 5 1* 541 43.8 56.2 0.1 99.9
Vermont 3 1* 576 68.4 31.6 100.0 0.0
Virginia 13 1* 592 50.0 50.0 48.8 51.2
Washington 11 1* 571 57.8 42.2 100.0 0.0
West Virginia 5 1* 540 39.8 60.2 0.0 100.0
Wisconsin 10 1* 543 56.0 44.0 99.9 0.0
Wyoming 3 1* 508 39.4 60.6 0.0 100.0

* Only the SurveyUSA 50 state polls are included.

Clinton–McCain EC Votes # polls Total Votes % Clinton % McCain Clinton %wins McCain %wins
Alabama 9 1* 556 44.6 55.4 0.9 99.1
Alaska 3 1* 556 37.8 62.2 0.0 100.0
Arizona 10 1* 575 41.9 58.1 0.0 100.0
Arkansas 6 1* 533 56.1 43.9 99.4 0.6
California 55 1* 533 55.5 44.5 99.1 0.9
Colorado 9 1* 568 46.7 53.3 5.5 94.5
Connecticut 7 1* 576 55.6 44.4 99.4 0.6
Delaware 3 1* 532 52.8 47.2 91.5 8.6
D.C. 3 0 (100) (0)
Florida 27 1* 589 54.8 45.2 98.7 1.3
Georgia 15 1* 539 38.4 61.6 0.0 100.0
Hawaii 4 1* 487 52.4 47.6 86.9 13.1
Idaho 4 1* 548 29.9 70.1 0.0 100.0
Illinois 21 1* 512 56.4 43.6 99.7 0.3
Indiana 11 1* 515 40.4 59.6 0.0 100.0
Iowa 7 1* 539 47.1 52.9 8.4 91.6
Kansas 6 1* 558 45.2 54.8 1.4 98.6
Kentucky 8 1* 551 45.0 55.0 1.3 98.7
Louisiana 9 1* 557 44.5 55.5 0.9 99.2
Maine 4 1* 575 53.4 46.6 94.9 5.1
Maryland 10 1* 552 55.1 44.9 98.8 1.2
Massachusetts 12 1* 587 59.8 40.2 100.0 0.0
Michigan 17 1* 566 50.0 50.0 49.8 50.2
Minnesota 10 1* 548 52.2 47.8 86.6 13.4
Mississippi 6 1* 581 45.1 54.9 1.2 98.8
Missouri 11 1* 583 47.9 52.1 13.6 86.4
Montana 3 1* 550 38.4 61.6 0.0 100.0
Nebraska 5 1* 542 34.5 65.5 0.0 100.0
Nevada 5 1* 551 45.6 54.4 2.2 97.8
New Hampshire 4 1* 573 45.5 54.5 2.1 97.9
New Jersey 15 1* 568 52.8 47.2 91.9 8.1
New Mexico 5 1* 566 50.0 50.0 50.2 49.8
New York 31 1* 527 62.4 37.6 100.0 0.0
North Carolina 15 1* 478 45.6 54.4 3.1 96.9
North Dakota 3 1* 511 39.3 60.7 0.0 100.0
Ohio 20 1* 566 55.5 44.5 99.3 0.7
Oklahoma 7 1* 552 45.7 54.3 2.5 97.5
Oregon 7 1* 571 47.3 52.7 8.9 91.1
Pennsylvania 21 1* 566 50.5 49.5 61.9 38.1
Rhode Island 4 1* 571 59.4 40.6 100.0 0.0
South Carolina 8 1* 536 46.6 53.4 5.9 94.1
South Dakota 3 1* 582 43.5 56.5 0.2 99.8
Tennessee 11 1* 533 46.5 53.5 5.6 94.4
Texas 34 1* 546 46.2 53.8 3.9 96.1
Utah 5 1* 559 29.3 70.7 0.0 100.0
Vermont 3 1* 551 55.7 44.3 99.4 0.6
Virginia 13 1* 569 44.5 55.5 0.8 99.2
Washington 11 1* 571 48.9 51.1 26.6 73.3
West Virginia 5 1* 549 52.8 47.2 91.7 8.3
Wisconsin 10 1* 550 52.2 47.8 86.5 13.5
Wyoming 3 1* 508 31.5 68.5 0.0 100.0

* Only the SurveyUSA 50 state polls are included.

Details of the simulation methods are given in the FAQ. The most recent match-ups can be found from this page.

Posted in Poll analysis, politics

  • Both comments and trackbacks are currently closed.
  • Trackback URI: http://hominidviews.com/wp-trackback.php?p=1370
  • Comments RSS 2.0

One Response to “On Those 50 SurveyUSA Polls”

  1. JPF Says:

    Monday, March 10th, 2008 at 1:09 am

    Perhaps the surveys should be weighted by recent-ness. Like giving each poll a bounded half-life of one or two weeks: a poll from yesterday would get full weight, and a poll from a week ago might get half weight, 2 weeks = 25%, etc., and then anything over a month ago is excluded.

AWSOM Powered