Polimetrics

2008-01-18T07:31:00.000-08:00

Primary season - a lot has been said about the NH primary polling errors. I am sure it will be a case study for future stat texts.

Just updating the links now and I will be building the presidential model.

2007-06-30T14:27:00.000-07:00

What I'm working on:

My congressional model was dead-on accurate. I am going to try to use some of the techniques to correct the "tail" bias on the presidential model for 2008.

2006-11-04T07:24:00.000-08:00

4 days before the election I show 78 GOP House seats competitive or potentially competitive compared to only 3 for the democrats.

The dems are almost certain to win the house - my point estimate is currently a net change of 35 seats, an increase of 12 seats from my last post.

Realistic estimates show the democrats pciking up between 20 and 52 seats. My probable interval is between 30 and 41 seats.

The Senate will be the interesting battle. As has been written elsewhere, I go with this ranking:

PA, OH, RI Likely Dem
MT Lean Dem
MO, VA Tossup Leaning Dem
TN Lean GOP
AZ Likely GOP

The Dem seats at risk are NJ and MD - I don't see the GOP taking either seat. Today MD is their best bet, but its still a longshot.

Good outcome for GOP: lose 5 senate seats, lose less than 30 house seats
Good Outcome for the Dems: win 7 senate seats, win more than 40 house seats
Probable outcome: Dems take 35 House seats, 6 Senate Seats

Here are the results to date:

154 (154 GOP 0 Dem) Safe GOP (Lead >15%)
15 ( 15 GOP 0 Dem) Likely GOP (10-15% Lead)
11 ( 11 GOP 0 Dem) Lean GOP (5-10% Lead)
11 ( 11 GOP 0 Dem) Barely GOP (2-5% Lead)
6 ( 6 GOP 0 Dem) Tossup GOP (0-2% Lead)
5 ( 5 GOP 0 Dem) Tossup Dem (0-2% Lead)
10 ( 10 GOP 0 Dem) Barely Dem (2-5% Lead)
14 ( 13 GOP 1 Dem) Lean Dem (5-10% Lead)
7 ( 5 GOP 2 Dem) Likely Dem (10-15% Lead)
202 ( 2 GOP 200 Dem) Solid Dem (Lead >15%)

ST DI Cur Dem Lead Status Result
AZ 1 R -3.0% Barely R
AZ 5 R 0.7% Tossup Switch to D
AZ 8 R 13.5% Likely D Switch to D
CA 4 R -6.5% Lean R
CA 11 R 2.4% Barely D Switch to D
CA 50 R -4.3% Barely R
CO 4 R 1.6% Tossup Switch to D
CO 5 R -6.4% Lean R
CO 6 R -14.1% Likely R
CO 7 R 13.5% Likely D Switch to D
CT 2 R 1.2% Tossup Switch to D
CT 4 R 2.5% Barely D Switch to D
CT 5 R 6.7% Lean D Switch to D
FL 5 R -13.5% Likely R
FL 8 R -9.4% Lean R
FL 13 R 7.3% Lean D Switch to D
FL 16 R 10.1% Likely D Switch to D
FL 22 R 0.0% Tossup
FL 24 R -14.5% Likely R
GA 12 D 7.2% Lean D
IA 1 R 3.8% Barely D Switch to D
IA 2 R -3.0% Barely R
ID 1 R -1.0% Tossup
IL 6 R 5.0% Barely D Switch to D
IL 8 D 11.5% Likely D
IL 10 R -3.1% Barely R
IL 14 R -10.7% Likely R
IL 19 R -14.7% Likely R
IN 2 R 7.9% Lean D Switch to D
IN 3 R -12.8% Likely R
IN 7 D 12.1% Likely D
IN 8 R 15.3% Solid D Switch to D
IN 9 R 3.0% Barely D Switch to D
KS 2 R -2.1% Barely R
KY 2 R -13.2% Likely R
KY 3 R 4.8% Barely D Switch to D
KY 4 R 2.5% Barely D Switch to D
MI 7 R -9.8% Lean R
MI 8 R -12.8% Likely R
MI 9 R -14.1% Likely R
MN 1 R -1.2% Tossup
MN 2 R -11.1% Likely R
MN 6 R -4.1% Barely R
NC 8 R -2.2% Barely R
NC 11 R 8.1% Lean D Switch to D
NE 1 R -10.9% Likely R
NE 3 R 0.0% Tossup Switch to D
NH 1 R -12.1% Likely R
NH 2 R 3.4% Barely D Switch to D
NJ 5 R -13.8% Likely R
NJ 7 R -3.1% Barely R
NM 1 R 7.3% Lean D Switch to D
NV 2 R -5.5% Lean R
NV 3 R -7.3% Lean R
NY 3 R -5.2% Lean R
NY 13 R -12.8% Likely R
NY 19 R 3.4% Barely D Switch to D
NY 20 R 1.5% Tossup Switch to D
NY 24 R 12.2% Likely D Switch to D
NY 25 R 5.4% Lean D Switch to D
NY 26 R 5.2% Lean D Switch to D
NY 29 R 6.6% Lean D Switch to D
OH 1 R 5.0% Lean D Switch to D
OH 2 R -1.5% Tossup
OH 12 R -6.7% Lean R
OH 15 R 7.8% Lean D Switch to D
OH 18 R 15.1% Solid D Switch to D
PA 4 R -2.4% Barely R
PA 6 R 5.5% Lean D Switch to D
PA 7 R 9.0% Lean D Switch to D
PA 8 R -2.3% Barely R
PA 10 R 12.2% Likely D Switch to D
TX 22 R 8.2% Lean D Switch to D
TX 23 R -6.8% Lean R
VA 2 R -0.8% Tossup
VA 10 R -5.1% Lean R
WA 5 R -6.3% Lean R
WA 8 R -0.5% Tossup
WI 8 R 4.1% Barely D Switch to D
WV 2 R -11.5% Likely R
WY 1 R -4.2% Barely R

2006-10-22T16:21:00.000-07:00

2006 Election Status - House Model

To come up with a quanitifiable model for the House is a challenge: different districts are often in the same city and area code, voters often cannot even name their member of congress are the just the beginning of difficulties pollsters encounter. Another method is to look at the GOP/Dem vote in past elections which is not always avialable for districts that cross county lines. Still, I gave it a shot and here are the results.

The model incorporates the following:

micro statistics:
the ratings experts: Sabato, Cook, CQ Politics, Rothenberg and Bowers
polls: all polls adjusted for time and partisan bias
macro statistics:
generic congressional ballot
direction of country

Here are the results to date:

166 (166 GOP 0 Dem) Safe GOP (Lead >15%)
10 ( 10 GOP 0 Dem) Likely GOP (10-15% Lead)
16 ( 16 GOP 0 Dem) Lean GOP (5-10% Lead)
4 ( 4 GOP 0 Dem) Barely GOP (2-5% Lead)
13 ( 13 GOP 0 Dem) Tossup GOP (0-2% Lead)
1 ( 1 GOP 0 Dem) Tossup Dem (0-2% Lead)
7 ( 7 GOP 0 Dem) Barely Dem (2-5% Lead)
16 ( 13 GOP 3 Dem) Lean Dem (5-10% Lead)
5 ( 1 GOP 4 Dem) Likely Dem (10-15% Lead)
197 ( 1 GOP 196 Dem) Solid Dem (Lead >15%)

If all races hold exactly as predicted, the Dems would win 23 seats and would control the House 226 to 209.

Even if the Dems lost all races they led by 5% or less, they would still win 15 seats and barely win the House 218-217.

However, the GOP has much more to lose. If the GOP loses all seats they have a 5% or less lead in, they would lose a staggering 40 seats and the new House would be 243-192 Dem, a dramatic change in power. In fact, there are 13 seats where the GOP holds a bare lead of under 2% while there is only 1 seat where the Dems have a bare lead.

The GOP currently have 66 seats at risk (competitive) while the Dems have only 7. In fact, every Dem seat has a projected lead for the Dems of over 5%. It is highly possible that the Dems will not lose a single seat in the House.

I know many democrats are afraid of saying they are going to win the house due to the results of 2020 and 2004 where the results and polls were not in alignment. I have written about the 2004 race in particular where I showed the states that were the most off from the model were NY, NJ and CT - the three states most affected by the terrorist attacks of 9/11. I also conjectured that these states which undersampled GOP voters the most were the ones where the GOP message of fear was most effective - and that scared voters are also most likely to avoid strangers - in particular the exit polsters.

Is the scared voter still causing bias to the polls? There is no way to measure that accurately people who choose not be polled or if the message of fear will be effective in this election. I will limit myself to the statistics which currenlt show an overwhleming lead for the Dems, much larger then the lead they apparently had in 2004.

ST DI Cur Dem Lead Status Result
AZ 01 R -5.5% Lean R
AZ 05 R -7.3% Lean R
AZ 08 R 15.5% Solid D Switch to D
CA 04 R -8.3% Lean R
CA 11 R -0.5% Tossup
CA 50 R -9.0% Lean R
CO 04 R -6.7% Lean R
CO 05 R -8.7% Lean R
CO 07 R 8.2% Lean D Switch to D
CT 02 R -0.6% Tossup
CT 04 R 2.1% Barely D Switch to D
CT 05 R -1.3% Tossup
FL 08 R -12.7% Likely R
FL 13 R 6.8% Lean D Switch to D
FL 16 R 8.3% Lean D Switch to D
FL 22 R -0.9% Tossup
GA 08 D 9.7% Lean D
GA 12 D 6.3% Lean D
IA 01 R 4.7% Barely D Switch to D
IA 02 R -6.5% Lean R
IA 03 D 13.4% Likely D
ID 01 R -8.6% Lean R
IL 06 R 0.5% Tossup Switch to D
IL 08 D 8.7% Lean D Switch to D
IL 10 R -10.2% Likely R
IL 14 R -14.1% Likely R
IN 02 R 8.2% Lean D
IN 03 R -14.0% Likely R
IN 07 D 12.0% Likely D
IN 08 R 15.0% Likely D Switch to D
IN 09 R 7.3% Lean D Switch to D
KS 02 R -11.8% Likely R
KY 02 R -13.6% Likely R
KY 03 R -1.9% Tossup
KY 04 R -2.3% Barely R
MI 07 R -14.7% Likely R
MI 08 R -14.7% Likely R
MN 01 R -3.4% Barely R
MN 06 R 3.6% Barely D Switch to D
NC 08 R -1.1% Tossup
NC 11 R 7.3% Lean D Switch to D
NE 01 R -12.6% Likely R
NH 02 R -8.0% Lean R
NJ 07 R -5.3% Lean R
NM 01 R 8.5% Lean D Switch to D
NV 02 R -4.5% Barely R
NV 03 R -7.3% Lean R
NY 03 R -6.9% Lean R
NY 19 R -1.7% Tossup
NY 20 R -1.1% Tossup
NY 24 R 7.4% Lean D Switch to D
NY 25 R -1.8% Tossup
NY 26 R 9.2% Lean D Switch to D
NY 29 R -0.7% Tossup
OH 01 R -0.2% Tossup
OH 02 R -3.3% Barely R
OH 12 R -13.7% Likely R
OH 15 R 4.6% Barely D Switch to D
OH 18 R 8.3% Lean D Switch to D
PA 04 R -8.9% Lean R
PA 06 R 5.1% Lean D Switch to D
PA 07 R 4.3% Barely D Switch to D
PA 08 R -1.8% Tossup
PA 10 R 8.8% Lean D Switch to D
TX 22 R 8.6% Lean D Switch to D
TX 23 R -7.4% Lean R
VA 02 R 3.0% Barely D Switch to D
VA 10 R -8.5% Lean R
VT 01 D 10.8% Likely D
WA 05 R -10.0% Lean R
WA 08 R -1.2% Tossup
WI 08 R 3.2% Barely D Switch to D
WV 01 D 14.5% Likely D
WY 01 R -8.7% Lean R

2006-08-02T21:16:00.000-07:00

Starting to modify this neglected site for the 2006 elections.
More to follow.

2005-11-15T17:34:00.000-08:00

A year later - still chugging along.

I'm currently working on revising my presidential election model. The 2004 model worked well on the battleground states but not so well on the deep blue and red states.

More later.

2004-11-08T09:43:00.000-08:00

What's coming up at Polimetrics?

I am currently working on compiling a county list of all votes in 2004 to compare them to the 2000 election. Of particular interest to me are:

1. Why were the exit polls so wrong on this election?

Earlier I posted on a possible explanation as to the reason (Voters who are more afraid of terrorism were more likely to vote for Bush and also more likely to avoid exit polls) but I admit there are other possibilities:

a. Bush voters are less likely to talk to pollsters than Bush voters in 2000. This would be a major change from 2000 where the margins of error were random.

b. Kerry voters were more motivated to talk to exit pollsters than Gore voters were. Again, this would be a first.

c. The sample precincts used by exit pollsters to cluster sample have changed radically from 2000. Based on some counties I have already looked at, this seems unlikely - but results need to be finalized.

d. The exit polls were correct, but there was an error in the vote count. This seems the less likely occurrence of all (unlike Florida 2000 where many Gore votes were either made incorrectly or counted incorrectly) but this is the first election that incorporated major use of the paperless ballot. Fortunately, this is easily tested after the 2004 votes are completed by seeing if the change from 2000 for paperless electronic voting counties is significantly different than the other counties.

2. What is the battleground for 2008?

My preliminary analysis shows that the extreme states (NY, HI, ID, etc) became less extreme, contrary to my model. My model preformed well in ranking the states but could not predict the dramatic movement by the extreme states. Most of the battleground moved to the left, the middle state was Ohio in this election. This may be good news for the Democrats in the future as it now appears they can win the election without winning the popular vote (assuming they can be competitive), a major switch from 2000. Again, there are many absentee and provisional votes to be counted which may affect the final results.

2004-11-03T20:57:00.000-08:00

Why were the exit polls so wrong?

After reading the 6PM exit polls, it seemed clear that my projection and Zogby's were dead on - that we were heading to a close Kerry victory - instead we get not only a close Bush electoral victory but about a 3 million popular vote victory - a popular vote much different than 2000 - different in that if Ohio had shifted 150,000 votes, the popular vote would have shifted only 2 million - Kerry would have been the one who won the electoral vote and lost the popular vote. What happened?

I will research this thouroughly once I have all the final numbers. Here are my preliminary thoughts.

First, the exit polls and the actual vote:

6PM exits Vote Result
Kerry Bush Kerry Bush Diff
PA 53 46 51 49 -2.5
FL 51 49 47 52 -3.5
NC 48 52 43 56 -4.5
OH 51 49 49 51 -2
MO 46 54 46 54 0
AR 47 53 45 54 -1.5
MI 51 47 51 48 -0.5
NM 50 49 49 50 -1
LA 43 56 42 57 -1
CO 48 51 46 53 -2
AZ 45 55 44 55 -0.5
MN 54 44 51 48 -3.5
WI 52 47 50 49 -2
IA 49 49 49 50 -0.5

Every exit poll in every battleground state erred in favor of Kerry. The odds of this occuring randomly is about 1 in 16000. Clearly, the exit polls had a systemic problem.

At Mystery Pollster there is a discussion as to possible reasons as to why the polling was wrong. Blumenthal suggests that the Incumbent Rule may not have held up in this election because of the war. That may be true, but that doesn't explain the bad exit polls where there are no undecideds. There is somethign else at work here.

One of the commenters (a New Yorker) said he was a closet Bush supporter but didn't want to tell anyone becuase of the anti Bush sentiment. Perhaps that played a role.

During the election night, I was comparing 2004 to 2000 county returns of the key states. I noticed that the increase in turnout in the Bush counties was higher than the increase in turnout in the Gore counties. Additionally in Florida, although the blue counties voted for Kerry at approximately the same percentage as 2000 for Gore, the red counties voted more heavily for Bush than they did in 2000. This combination of greater increase in turnout and higher percentage for Bush in the Red counties explains explains Bush's 370,000 vote win in Florida.

How did the exit polls miss this? It's not early voting as exit polls account for that in independent sampling and includes them in their projections. I have a theory that I will test later when I have more data:

Fear won this election for Bush.

1. Conservative Christians voted in much larger numbers than they ever have driven to the polls by two gut issues: fear of terrorism and fear of gay marriage. Also, this group is more distrustful of pollsters and was therefore undersampled.

2. A quick analysis of the returns also showed the reason that the popular voted was so skewed. Some blue states - particularly NY - did much better for Bush than the expected Partisan Index for a 3% Bush victory. I will do a county by county analysis later. NY of course was deeply affected by 9/11 and may have been more swayed by Bush's campaign of fear.

3. The gender gap decreased significantly from 2000 - scared moms?

Fear is a powerful tool - it built the Pyramids. I believe the data will show Bush's brilliant campaign of fear successfully brought out of his silent base - the secret army that the pollsters missed.

2004-11-02T22:37:00.000-08:00

My projection shows Kerry will not catch up in Ohio.

The popular vote will be close - the elctoral vote is over.

My model had the states ranked correctly, the exit polls were again all wrong.

2004-11-02T15:42:00.000-08:00

LATE EXIT POLLS

Kerry Bush

PA 53 46
FL 51 49
NC 48 52
OH 51 49
MO 46 54
AR 47 53
MI 51 47
NM 50 49
LA 43 56
CO 48 51
AZ 45 55
MN 54 44
WI 52 47
IA 49 49

SENATE AK: Knowles 50 - Murk 47 CO Salazar 51, Coors 47 FL Castor 51, Martinez 48 KY Bunning 52, Mongiardo 48 LA Vitter may get the 50 percent that would elect him, 51 percent NC Burr 50 Bowles 48 OK Coburn 51 Carson 46 SC DeMint 50.2 - Tenenbaum 49.8SD Thune 50 Daschle 50

2004-11-02T14:31:00.000-08:00

ZOGBY PREDICTION MATCHES POLIMETRICS!!!

The top pollster in the country has called the election for Kerry 311 to 213 with NV and CO too close to call. Yesterday, we made exactly the same call on the 48 states he called, plus we didn't wimp out on two states. Our call gave NV to Kerry and CO to Bush making our prediction 316-222.

Thank you John for seeing the light.

Update: More evidence of a Kerry victory. Both the tradesports and Iowa Future Markets have dramatically moved towards Kerry. Gamblers know.

2004-11-02T14:22:00.000-08:00

Afternoon Exit Polls

FL: 50/49 - KERRY
OH: 52/47 - KERRY
MI: 51/48 - KERRY
PA: 58/42 - KERRY
IA: 50/48 - KERRY
WI: 53/47 - KERRY
MN: 57/42 - KERRY
NH: 58/41 - KERRY
ME: 55/44 - KERRY
NM: 49/49 - TIE
NV: 48/49 - BUSH
CO: 49/50 - BUSH
AR: 45/54 - BUSH
NC: 47/53 - BUSH

I will not vouch for these from mydd.com but they look good for Kerry.

Another indicator that change is the air is the Florida Latino Vote:

Hispanics 53-46 Bush (2000 was 65-35 Bush)
Cubans 68-32 Bush (2000 was 82-17 Bush)

2004-11-02T13:05:00.001-08:00

early exit poll numbers

For what they're worth:

State
Kerry
Bush

AZ CO LA PA OH FL MI NM MN WI IA NH
45 48 42 60 52 51 51 50 58 52 49 57
55 51 57 40 48 48 47 48 40 43 49 41

With all the early voting, I have no idea how to interpret these numbers. 4 years ago, they were wrong, but I would rather be Kerry than Bush with these numbers.

2004-11-01T19:37:00.000-08:00

Senate and House Projections

from easiest to hardest:

Things have imprioved for the GOP, mainly because the contested seats are in red states.

Illinois: Democrat pick-up. Sure thing

Georgia: Republican "pick-up" Nearly sure thing

South Carolina: Republican pick-up - state is just too GOP for even a moron like DeMint to throw away.

Louisiana: Runoff in December. No one will get 50%

Kentucky: Republican Hold - Bunning will resign sometime after the election and be replaced with a Republican who is not ill.

Alaska: Democrat pick-up - Knowles is popular.

Colorado: Democratic pick-up - Salazar looks good here

Florida: Democratic Hold - Kerry's win will give Castor the edge.

Oklahoma: Republican hold - Another GOP stronghold

South Dakota: Republican pick-up - The Incumbant rule works against the minority leader.

North Carolina: Democratic Hold - this is just a guess because I won't pick a tie.

That makes it 51-47-1 with LA to be resolved. The GOP holds as states have recently been breaking their way. I believe the GOP will hold the House, but not gain seats so Kerry will have a GOP congress. With Delay in charge, the House will be extremely unfriendly to Kerry, but the new President will have a much better relationship with his stomping grounds. He is well respected by several GOP senators: Hagel, McCain, Luger, Spector, Chafee, Snowe to name a few. He knows how the Senate works. Expect new Kerry legeslation to be negotiated with his buddy GOP Senators.

2004-11-01T17:50:00.000-08:00

Final Projection 11/1 6:00 PM
Kerry 60,502,000 50.0% 291 EV
Bush 58,524,000 48.4% 247 EV
Other 1,936,000 1.6%

click here for electoral map

Probability Bush wins each state:
NM 24.9%
IA 29.7%
FL 29.8%
OH 54.0%

Overall Winning Probability:
Kerry 83.8% Bush 16.1%

Bush -/+ Today (Yesterday)

-75.8 (-75.8) District of Columbia
-32.3 (-32.2) Massachusetts
-28.1 (-28.0) Rhode Island
-27.8 (-27.6) New York
-20.9 (-20.8) Connecticut
-19.1 (-18.7) New Jersey
-18.0 (-18.9) Maryland
-17.8 (-17.7) Delaware
-17.5 (-17.4) Vermont
-14.1 (-14.0) Hawaii
-14.0 (-13.7) California
-13.6 (-14.4) Illinois
-10.6 (-12.2) Maine
-9.1 (-8.5) Washington
-6.4 (-6.4) Michigan
-6.1 (-6.3) Pennsylvania
-5.4 (-4.6) Minnesota
-5.4 (-5.4) Oregon
-4.7 (-3.4) New Hampshire
-3.7 (-3.6) Wisconsin
-1.7 (-1.6) New Mexico
-1.3 (-1.4) Iowa
-1.3 (-1.0) Florida
0.3 (1.0) Ohio
2.0 (0.6) Nevada
3.4 (3.4) Missouri
4.6 (4.7) Virginia
5.3 (4.8) Colorado
5.6 (7.9) West Virginia
7.1 (7.2) Arkansas
8.2 (4.4) Arizona
10.7 (10.8) Louisiana
12.3 (11.8) North Carolina
13.0 (12.6) Georgia
13.3 (10.8) Tennessee
14.1 (14.2) Mississippi
15.2 (17.8) Alabama
16.2 (17.4) South Carolina
18.0 (9.4) Indiana
19.3 (18.1) Kentucky
21.9 (22.0) Kansas
22.0 (22.1) Texas
22.4 (22.5) South Dakota
22.7 (22.8) Alaska
23.5 (23.6) Montana
23.9 (24.0) North Dakota
27.1 (27.2) Nebraska
28.3 (27.9) Oklahoma
35.9 (36.0) Idaho
37.0 (37.1) Wyoming
43.4 (43.5) Utah

Current Battleground Chart
Polimetrics Partisan Index
Polimetrics Partisan Index - Electoral Vote Analysis

This is my final run which again shows little change from yesterday as undecided voters continue to break towards Kerry as expected. Ohio is nearly tied now which makes Kerry a 5-1 Favorite tomorrow.

MY PREDICTION

Kerry will win, and my model is probably conservative:

If I go by only the polls, Kerry adds Ohio and increases his margin in Florida for a 311-227 win.

Every poll and early voting has shown turnout will be high. I rarely believe the turnout predictions, but empirical evidence supports that claim. For example:
1/3 of Florida voters have already voted.
There are 6 million more registered voters than my original projection based on US Census data. Heavy turnout is bad for Bush. As Sabato says, "most of those new voters will not be coming out to say, 'Good job Mr. President.'.

Conservatives have Kerry ahead or tied. Sabato predicts a 269-269 tie. Fox poll has Kerry ahead in the nation and Florida using conservative likely voter screens. Gallup has the race a tie with another sample that is plus 5 GOP.

My final prediction: Kerry wins 51.5%-47.0%-1.5%, 121 million total votes. He will add Nevada for a 316-222 win. There will be no lawsuits.

2004-10-31T21:44:00.000-08:00

Projection 10/31 9:00 PM
Kerry 60,604,000 50.1% 291 EV
Bush 58,422,000 48.3% 247 EV
Other 1,936,000 1.6%

click here for electoral map

Probability Bush wins each state:
NM 26.5%
IA 29.3%
FL 35.1%
NV 59.4%
OH 65.8%

Overall Winning Probability:
Kerry 76.9% Bush 23.3%

Bush -/+ Today (Yesterday)

-75.8 (-75.8) District of Columbia
-32.2 (-31.6) Massachusetts
-28.0 (-27.5) Rhode Island
-27.6 (-27.0) New York
-20.8 (-22.0) Connecticut
-18.9 (-20.1) Maryland
-18.7 (-17.7) New Jersey
-17.7 (-17.1) Delaware
-17.4 (-16.8) Vermont
-14.4 (-14.5) Illinois
-14.0 (-13.4) Hawaii
-13.7 (-13.1) California
-12.2 (-11.6) Maine
-8.5 (-7.9) Washington
-6.4 (-4.6) Michigan
-6.3 (-5.7) Pennsylvania
-5.4 (-4.4) Oregon
-4.6 (-4.2) Minnesota
-3.6 (-2.1) Wisconsin
-3.4 (-5.0) New Hampshire
-1.6 (0.0) New Mexico
-1.4 (-0.1) Iowa
-1.0 (-0.5) Florida
0.6 (3.7) Nevada
1.0 (0.8) Ohio
3.4 (4.1) Missouri
4.4 (5.0) Arizona
4.7 (5.2) Virginia
4.8 (5.1) Colorado
7.2 (7.0) Arkansas
7.9 (6.3) West Virginia
9.4 (17.2) Indiana
10.8 (11.4) Tennessee
10.8 (11.4) Louisiana
11.8 (12.4) North Carolina
12.6 (13.2) Georgia
14.2 (14.8) Mississippi
17.4 (18.0) South Carolina
17.8 (18.4) Alabama
18.1 (18.7) Kentucky
22.0 (22.6) Kansas
22.1 (22.7) Texas
22.5 (23.1) South Dakota
22.8 (23.3) Alaska
23.6 (24.2) Montana
24.0 (24.6) North Dakota
27.2 (27.8) Nebraska
27.9 (28.5) Oklahoma
36.0 (36.6) Idaho
37.1 (37.7) Wyoming
43.5 (43.8) Utah

Current Battleground Chart
Polimetrics Partisan Index
Polimetrics Partisan Index - Electoral Vote Analysis

The critical state of PA has moved steadily towards Kerry leaving Bush a ghost of chance there. The weird polling data has dropped out of the Zogby 4 day average of NM sending that state back to Kerry.

Spooky day with the National tracking polls - the one that most favored Kerry was Fox. Go figure.

The polls are now coming out at a rate of 60 per day. They don't show much change. If anything, the bin Laden tape actually hurt Bush's numbers, but I believe we are seeing more of the undecideds deciding for Kerry as expected.

I haven't written much about the senate, but appears the democrats would need a lot of luck to win. They need a net gain of one - and as I see it now, they will swap IL and GA, likely gains in AK and CO, likely loss in SC, and fight in OK, NC, SD. LA will probably go to a December runoff if the GOP doesn't win outright on 11/2.

The tossup senate races are in solid Bush states and three are current Dems. Even if they split 2-2, the GOP holds the senate.

2004-10-30T17:28:00.000-07:00

Projection 10/30 5:00 PM
Kerry 60,192,000 49.8% 286 EV
Bush 58,836,000 48.6% 252 EV
Other 1,936,000 1.6%

click here for electoral map

Probability Bush wins each state:
FL 42.0%
IA 48.3%
NM 50.6%
OH 62.4%

Overall Winning Probability:
Kerry 73.8% Bush 26.2%

New Mexico is essentially a tie right now, Bush leads by about 1000 votes - right about where it was in 2000. If Gore had won NH in 2000, NM would have been litigated as well as Florida. Now, as it was then, NM is irrelevent unless WI comes back into play.

Colorado seems out of play as well - I never thought it was in play unless Kerry wins comfortably. Keep in mind in my projections, I only concern myself with scenarios that could actually change the outcome - last time only FL and NH were close enough to change the outcome. This time, it is FL or OH. Cheney must be working on his tan in Hawaii.

A year ago experts broke the battleground into three parts - the Southwest (NM, NV, CO, AZ) the Great Lakes (MI PA MN WI IA OH) and Florida (FL). I remember many telling me that the Southwest strategy was the best chance for the democrats. I simply replied with battleground probabilities showing the most important states in a close election. The top state is Florida - by far. 80% of the simulations in a tied election had Florida going with the winner. The six Great lake States came next. Kerry has intelligently focused most of his time doing the Midwest - Florida shuffle.

Three days to go and there is little movement in the polls. Yesterday, all the lazy media talked about was the effect the Osama Bin Laden video would have on the electorate. Does this qualify as an "October surprise?" I don't think so. Killing Bin Laden would have some effect on the election, but knowing that he is alive and hates America is not news. Frankly, all I care about the video is if it can be analyzed for any information to his whereabouts.

With all the early voting, the exit polls may be more suspect than the last election. In my county, 40% are voting absentee. Based on polling, the early voting has been about 50-50 between Kerry and Bush, and improvement for the Democrats from 2000 when Bush won the early voting by about 15%. Whether this is a portend for the election or just the democrats GOTV strategy of getting their voters to the polls early remains to be seen. I will say this: someone who has voted has a 100% chance of voting. Things happen on election day to prevent people from voting - work, life, patience get in the way. There is no down side in getting your supporters to vote early.

2004-10-29T14:39:00.000-07:00

Projection 10/29 3:00 PM
Kerry 60,274,000 49.8% 291 EV
Bush 58,752,000 48.6% 247 EV
Other 1,936,000 1.6%

click here for electoral map

Probability Bush wins each state:
IA 38.9%
FL 45.5%
NM 48.0%
OH 51.2%

Overall Winning Probability:
Kerry 76.7% Bush 23.3%
Current Battleground Chart
Polimetrics Partisan Index

Very slight movement in polls today - this run is almost identical with 2 days ago.

One new piece of information was that are now approximately 143,000,000 registered voters, a staggering 5,000,000 more than my projection. Therefore I now predict the turnout will be over 120,000,000. I calculate 84% to 86% of the registered voters will vote - I derive this number by starting with last election's 83% and increasing it slightly due to interest and early voting.

2004-10-28T17:01:00.000-07:00

Projection 10/28 5:00 PM
Kerry 58,718,000 50.1% 311 EV
Bush 56,561,000 48.3% 227 EV
Other 1,872,000 1.6%

click here for electoral map

Probability Bush wins each state:
NM 26.4%
IA 27.9%
FL 37.4%
OH 44.8%

Current Battleground Chart
Polimetrics Partisan Index

Some significant changes in the polls today - always subject to change.
Colorado has now passed Nevada and may be in play.
Wisconsin and Nevada both leave the battleground. It was a close call not taking WI out yesterday.
The four states in the battleground all now favor Kerry, albeit the two most critical are by razor thin margins.
Kerry's winning percentage has increased to over 80%.

Now one might ask how Kerry's winning percentage could be so high when both Ohio and Florida are so close. This is because Kerry only needs one while Bush needs them both. If these states were both 50-50, Kerry would have a 75% chance of getting at least one. Since they are both slightly Kerry, his winning percentage increases.

Also, removing WI from the battleground makes IA and NM unimportant - combined they can not replace the loss of OH for Bush. As of today, OH and FL are the only states that matter.

John Zogby said today that Kerry will win. I will not be so bold, but it is clear today why Zogby would say that.

2004-10-27T20:04:00.000-07:00

Projection 10/27 8:30 PM
Kerry 58,623,000 50.0% 291 EV
Bush 56,653,000 48.4% 247 EV
Other 1,872,000 1.6%

click here for electoral map

Probability Bush wins each state:
WI 20.3%
IA 33.1%
NM 34.6%
FL 37.3%
OH 56.8%
NV 75.7%

Overall Winning Probability:
Kerry 76.7% Bush 23.3%

Today showed continued improvement for Kerry in the battleground while national numbers remained flat. PA, one of the original Big 3, now appears to be out of reach. Iowa has swung to Kerry. Additionally, WI has moved to more than +2 Kerry for the first time in weeks. NV now projects to a closer race than WI and is therefore moved to the current battleground along with WI, IA, NM, FL and OH. This leaves 64 possibilities where Kerry wins 43, Bush wins 20 and 3 ties which also count as a Bush win.

It also looks like the magic national popular number has returned to 0.6% that Kerry's votes must exceed Bush's to win Florida, the continuing middle state - the same percentage by which Gore won the populur vote in 2000.

A note about the incumbent rule whic says undecided usually break heavily to the challenger: This is almost always true when the challenger is less well known in the region being contested, which is clearly true of Kerry. An example where it may not hold is if the challenger was also a national figure. For example, Hillary Clinton would not benefit as much as Kerry from the incumbent rule. This rule also applies to state or local races - for example - I would not expect former Governor Knowles to benefit from the incumbent rule in the Alaskan senate race.

2004-10-26T15:26:00.000-07:00

Projection 10/26 3:30 PM
Kerry 58,669,000 50.1% 284 EV
Bush 56,608,000 48.3% 254 EV
Other 1,872,000 1.6%

click here for electoral map

Probability Bush wins each state:
WI 31.5%
NM 38.5%
FL 38.6%
IA 51.5%
OH 55.6%

Overall Winning Probability:
Kerry 72.6% Bush 27.4%

Latest run, check out the current battleground chart and compare it to the "Polimetrics Partisan Index I ran last in April. Only Ohio is a suprise among the critical states. I have conjectured that projected outcome can be explained more by the demographic and cultural make-up of a state than by early polls. These final polls are looking much more like the partisan index.

Kerry's numbers improved slightly, but all movement is within the margins of error for various polling agencies. It really comes down to get-out-the-vote.

2004-10-25T23:09:00.000-07:00

REVISED Projection 10/25 11:30 PM

Kerry 58,456,000 49.9% 284 EV
Bush 56,830,000 48.5% 254 EV
Other 1,872,000 1.6%

Probability of Bush Winning:

NM 33.6%
WI 36.4%
FL 41.3%
IA 58.2%
OH 67.5%

Overall Winning Probability:

Kerry 65.8% Bush 34.2%

Ok, the earlier results are modified due to a revision in the Gallup numbers and an addition of the LA Times poll. I also added individual probabilities for the 5 critical states and an overall probability of winning based on independence among the results from these 5 states. This may be an invalid assumption because there is almost certainly a correlation among the expected results for WI and IA and possibly OH.

These results are more in line with prior runs. One can see how even the smallest outlier can really swing the numbers. Anyone who says they can predict a winner today must be looking at numbers not reported to the public.

If you want to really get educated about polling check out the articles by The Mystery Pollster . This series of daily articles from an experienced pollster really gets behind the calculation of the numbers and discusses the recent controverisies.

2004-10-25T16:16:00.000-07:00

Projection 10/25 4:30 PM

Kerry 58,087,000 49.6% 257 EV
Bush 57,192,000 48.8% 281 EV
Other 1,872,000 1.6%

I might as well call this run of the model "Election 2000". Kerry wins the popular vote, Bush wins the Electoral vote and Florida ends up at 0.1, right in the middle.

A couple of things about today's run: it includes the latest Gallup poll which has been highly critcized due to its sampling of republicans in much higher percentages than the 2000 exit polls. I don't judge Gallup here - I include them as I would any other pollster. Without Gallup's poll, the results would be similar to yesterday's run.

Current discussion is that the five states that will decide the election are NM WI FL IA OH. Other states may be in play but for a candidate to win them means they have probably already won the election. For example, Kerry may win Nevada, or Bush may win Minnesota, but if that happens, the middle states have probably already gone to the winner.

There are 32 possible outcomes for these 5 states: Kerry wins 20 of them, Bush wins 10 and there are 2 ties which also means Bush wins. (Bush needs 42 Evotes, Kerry needs 28 Evotes from these 5 states).

Current Probabilities of victory based on a 2.5% Standard Deviation for the partisan Index:
Kerry 51.5% Bush 48.5%

Very Close Indeed

2004-10-24T15:04:00.000-07:00

Projection 10/24 3:30 PM

Kerry 58,600,000 50.0% 284 EV
Bush 56,677,000 48.4% 254 EV
Other 1,872,000 1.6%

Not much change from yesterday - the new polls look better for Kerry only because the undecideds have decreased and are moving heavily towards Kerry, as expected.

Hawaii has had a couple of strange polls that look more like a battleground poll. This state has always been rated solid Kerry - the partisan index rating is +23 Kerry, so I don't put much faith in these polls. However, Hawaii has now dropped behind California in my ranking, still safe Kerry.

This leads me to a discussion of variability in the partisan index. When testing this model on the 2000 election, there was one state that fell outside the margin of error on projection from the 1980-1996 election data and demographic factors: West Virginia moved eight percent away from the expected partisan index in 2000. Now you might say this discredits the model, but the reality is one would expect 1 in 20 states to fall outside a 95% confidence interval, about two standard deviations from the mean. West Virginia was not only the only state outside of the margin of error, but actually moved outside a 99% margin of error, nearly 3 standard deviations from the mean.

What will the surprise state be this year?

2004-10-23T21:23:00.000-07:00

Projection 10/23 10:00 PM

Kerry 58,366,000 49.8% 284 EV
Bush 56,913,000 48.6% 254 EV
Other 1,872,000 1.6%

Florida returns to the middle state, but all results are well within the margin of error. Today the battleground is reduced to FL, OH, WI and IA.

The news today was Bush focusing the remainder of his campaign on WI, IA, FL, and NM. He either must feel safe with OH or is writing it off. Seems strange based on these stats, but he may be viewing the strategy of winning WI, IA and FL which is a 271-267 win or even NM instead of IA for a 269-269 tie. The actual polls have Kerry in the lead in OH, it's only the elasticity of the partisan index that puts Ohio just below Florida in my model.

Unlike Florida, losing Ohio doesn't leave Bush hopeless. Still, I find it hard to believe Bush can lose Ohio without also losing Wisconsin.

Early voting statistics in Iowa are about 2-1 in favor of registered Democrats. Will the democratic strategy of getting people to early vote succeed? It will be interesting to measure this effect after the election.

2004-10-22T22:13:00.000-07:00

Projection 10/22 10:00 PM

Kerry 58,929,000 50.3% 304 EV
Bush 56,352,000 48.1% 234 EV
Other 1,872,000 1.6%

Most states are behaving consistant with the partisan index - most notably the continued reddening of Iowa. The one major surprise has been Ohio which is now ranked on the blue side of center. There have been many polls this week rangiong from +6 Kerry to +2 Bush, but no poll has Bush over 47%, a definite sign of trouble for an incumbent. The movement of Ohio now makes Wisconsin the key state - Bush would need to swing 1.7% to win WI and FL to pull off a 271-267 victory. Note the Popular vote difference in 2.2%, so Bush could again win the election and lose the vote.

2004-10-21T19:03:00.000-07:00

Projection 10/21 7:00 PM

Kerry 58,983,000 50.3% 304 EV
Bush 56,288,000 48.1% 234 EV
Other 1,872,000 1.6%

Main difference today is Iowa and Ohio switching. Good polling in OH and FL for Kerry today. Two good polls for Bush in IA.

Today Gallup had Kerry +6 in Ohio and -7 in Wisconsin! Does that make any sense? WI is 2-3% more Dem by the partisan index, so Gallup's results would be a net deviation of 16% from expectation!

2004-10-20T19:17:00.000-07:00

Projection 10/20 7:15 PM

Kerry 59,193,000 50.5% 291 EV
Bush 56,087,000 47.9% 247 EV
Other 1,872,000 1.6%

A small improvement for Kerry based on national polls and some new state polls. Ohio calculates to a difference of about 8,000 votes out of 5.1 million to be cast!

Although the national polls bounce around due to sampling error and methodology, internals of polls have been remarkably consistent: Republicans are about 90-9-1 for Bush, Democrats are about 89-10-1 for Kerry and Independents are about 50-45-5 for Kerry. Deviations in the polls are often due to likely voter screens and the demographics of the sampled group. It seems logical to weight by party ID which is so highly correlatated to the vote, but some pollsters argue that party ID is self-selected and changes with the mood of the voter. We will have to wait for the vote to measure the wisdom of these pollsters.

2004-10-19T15:41:00.000-07:00

Projection 10/19 4PM PDT.....

Kerry 59,001,000 50.4% 291 EV
Bush 56,470,000 48.2% 247 EV
Other 1,689,000 1.4%

After months of waiting, I have published the first battleground projection - better late than never! The poll shows a near dead heat (big suprise) - with Kerry up by 0.6% in Florida, the current decisive state.

What makes this projection different than the hundreds of projections already out there, many of which I have linked?

1. The projection uses a partisan index - a best estimate correlation among the states based on past elections and demographic projections by the US Census Bureau. For example, midwestern states such as Iowa continue to trend towards the right while states like Arizona have trended left due to internal migration. Using multiple regression on variables such as ethnicity trends, population growth, regional factors, etc, I determine a partisan index - what many call blueness and redness. Final results have been adjusted for the candidate's home region based on past elections since 1980.

2. The projection takes into account national polls, for adjusting the partisan index to current status and state polls to adjust the partisan index within its determimed 95% confidence band (about 3-5% based on past performance.) In other words, because the methodology for state polling is not consistant, I use a Baysian approach in adjusting all state results by the partisan index. For example, in this run the most extreme adjustment was South Dakota which has an adjusted partisan index of about 28% but has reported a state poll with a 10% difference. The actual result of +18.6 is a more reasonable estimate than the reported poll. Note, these adjustments are automatic based on model parameters - not based on any subjective analysis.

3. Undecided voters are assumed to break in accordance with past elections when one of the candidates is an incumbent. Third party voters are determined based on an expected national vote of about 1.5% adjusted by each state's third party relative voting trends since 1980.

4. Using US census data, I have also projected total votes for each state to determine a predicted national share.

Why I am I doing this?

I am a Math/Statistics professor who has always been interested in the study of political statistics. This is my latest attempt at making an empirical tool to predict and study presidential elections. I hope to do for politics what Bill James has done for baseball - to educate people, expose misconceptions and ignorance, cut through the rhetoric, and to examine politics using purely statistical methods. Most of all, I want to have fun with this - politics is the most entertaining game in town.

For the record, I am not unbiased - who is today? I have already cast my ballot for the challenger in this election. My rhetoric, (probably like yours), is posted all over the Internet at places like Daily Kos. I will keep this place rhetoric free - notice there are many good data links from "the other side" - like "Election Projection", one of my favorites.

When it comes creating political numeracy, I hope we are all on the same side.

Mo

2004-08-04T16:29:00.000-07:00

New Links have been added to those making electoral predictions. I like Election Projection the the best as it supports the position of high correlation among the states based on the national trend. State head-to head polls at this time are highly suspect.