Electoral Vote Predictor

Nov 19, Projected U.S. Senate: 51 Democrats 49 Republicans

Sun, 19 Nov 2006 06:00:00 EDT

Incumbent congressman Dave Reichert (R) barely held onto his seat, beating newbie Darcy Burner (D), 51% to 49% in WA-08 . Similarly, in Wyoming WY-AL, incumbent Barbara Cubin (R) beat off a totally unknown challenger, Gary Taruner (D), by only 1012 votes.

Nov 18, Projected U.S. Senate: 51 Democrats 49 Republicans

Sat, 18 Nov 2006 06:00:00 EDT

CQ Politics has a nice set of biographies of the new members of Congress.

Nov 16, Projected U.S. Senate: 51 Democrats 49 Republicans

Thu, 16 Nov 2006 06:00:00 EDT

The Wall Street Journal has an article on which pollsters did best. The phone polls of Rasmussen (automated) and Mason-Dixon (human) were the best, off by slightly less than 4% on average. Next best were Zogby (human) and SurveyUSA (automated), off by slightly more than 4%. Zogby's interactive (Internet) polls were off by an average of almost 9%, although they predicted 18 of the 19 Senate winners correctly. SurveyUSA posted a detailed comparison of its polls with those of its competitors.

Nov 15, Projected U.S. Senate: 51 Democrats 49 Republicans

Wed, 15 Nov 2006 06:00:00 EDT

We are getting close to the end of the line now. I will keep posting sporadically, at least until all the House races are finalized. After that, I might post from time to time, but I am not going to consider all polls pitting the dozen Republicans running for President against the dozen Democrats running for President. Check back once in a while, then click on "Previous report" to go backward in time until you have seen everything you missed.

I am likely to do this again in 2008, but not seriously until we know who the presidential candidates are.

For a constant stream of political news, check Political Wire every day.

One race has just been called. According to the Hartford Courant, Joe Courtney (D) beat incumbent Rob Simmons (R) by 91 votes.

Nov 13, Projected U.S. Senate: 51 Democrats 49 Republicans

Mon, 13 Nov 2006 06:00:00 EDT

Yesterday we looked at how well the Senate pollsters did. Today we will look at the House pollsters. Below is a table that gives the final vote in column 2, followed by the results of 5 pollsters. Each cell consists of the Democratic percentage minus the Repubican percentage. For example, in AR-02, the Democrat - Republican value is 22, meaning the Democrat won by 22%. The last poll from SurveyyUSA predicted a Democratic win by 21%. The data is available in .csv format.

CD	Vote	M-D	Research 2000	RT Strategies	SurveyUSA	Zogby
AR-02	22				21
AZ-01	-9			-3
AZ-05	5				2
AZ-08	12					12
CA-04	-5			0	-7
CA-11	4			2
CA-26	-32				-19
CA-48	-31				-19
CA-50	-17				-14
CO-03	27				19
CO-04	-3	-10		3
CO-05	-20	-7			-9
CO-07	13	6		5	16	14
CT-02	0		1	6		-5
CT-04	-3		4	-9		7
CT-05	12		3	8
FL-05	-20				-14
FL-09	-12				-12
FL-13	0			2
FL-16	1		7	3
FL-22	4		-5	2		9
FL-24	-14					-2
IA-01	12					7
IA-02	2			-2
IA-03	6		12
ID-01	-5	-2
IL-06	-4			1		14
IL-08	5			5
IL-10	-6			2
IN-02	8		3	3		13
IN-03	-10		-12
IN-07	8		5
IN-08	22		7	10
IN-09	3		2	8	-2	2
KY-03	3			6	5
KY-04	-9			4	-6	3
MN-01	6			-3
MN-02	-18				-8
MN-05	35				25
MN-06	-8			-1	-7	-10
MT-AL	-10	-24
NC-08	0			4
NC-11	8			9		5
NH-01	2		-8
NH-02	8		-1	3
NJ-07	-1			-3
NM-01	0				2	9
NV-02	-4	-8	-8
NV-03	-1	-7		-7
NY-03	-12			-7
NY-19	2			2
NY-20	6			11
NY-25	-2			9
NY-26	-4				-4	-1
NY-29	-2			11
OH-01	-6			2
OH-02	-2			-5	-8
OH-06	22				19
OH-12	-6			-5
OH-18	24					20
OK-05	-24				-22
PA-04	5			-4
PA-06	-2			5		5
PA-08	1			3
SC-05	14				19
SD-AL	41	40
TX-22	10					8
VA-02	-2	-2		5		-8
VA-05	-19				-26
VT-AL	8		10
WA-02	32				29
WA-05	-10			-5
WA-08	0			2	0
WI-08	2			6
WY-AL	0	-7

To see how well a pollster did, we can subtract the actual vote from the prediction. For example, in the table below, if we subtract from SurveyyUSA's prediction for AR-02 (21) the actual vote total (22) we get -1. In other words, Mason-Dixon predicted that the Democrat would do 1% worse than he actually did. A perfect pollster would have a column consisting entirely of 0s. Since the margin of error in predicting each candidate's score is about 4%, the margin of error in predicting the difference is about 8%. Thus to a first approximation, any number less than or equal to 8 should be considered a correct prediction. Anything more than 8 is wrong. A negative number means the Democrat id better than expected. A positive one means the Deocrat did worse than expected.

CD	M-D	Research 2000	RT Strategies	SurveyUSA	Zogby
AR-02				-1
AZ-01			6
AZ-05				-3
AZ-08					0
CA-04			5	-2
CA-11			-2
CA-26				13
CA-48				12
CA-50				3
CO-03				-8
CO-04	-7		6
CO-05	13			11
CO-07	-7		-8	3	1
CT-02		1	6		-5
CT-04		7	-6		10
CT-05		-9	-4
FL-05				6
FL-09				0
FL-13			2
FL-16		6	2
FL-22		-9	-2		5
FL-24					12
IA-01					-5
IA-02			-4
IA-03		6
ID-01	3
IL-06			5		18
IL-08			0
IL-10			8
IN-02		-5	-5		5
IN-03		-2
IN-07		-3
IN-08		-15	-12
IN-09		-1	5	-5	-1
KY-03			3	2
KY-04			13	3	12
MN-01			-9
MN-02				10
MN-05				-10
MN-06			7	1	-2
MT-AL	-14
NC-08			4
NC-11			1		-3
NH-01		-10
NH-02		-9	-5
NJ-07			-2
NM-01				2	9
NV-02	-4	-4
NV-03	-6		-6
NY-03			5
NY-19			0
NY-20			5
NY-25			11
NY-26				0	3
NY-29			13
OH-01			8
OH-02			-3	-6
OH-06				-3
OH-12			1
OH-18					-4
OK-05				2
PA-04			-9
PA-06			7		7
PA-08			2
SC-05				5
SD-AL	-1
TX-22					-2
VA-02	0		7		-6
VA-05				-7
VT-AL		2
WA-02				-3
WA-05			5
WA-08			2	0
WI-08			4
WY-AL	-7

Nov 12, Projected U.S. Senate: 51 Democrats 49 Republicans

Sun, 12 Nov 2006 06:00:00 EDT

Let's take a look at how the pollsters did. Below is a table given the final vote in column 2, followed by the results of 8 pollsters. Each cell consists of the Democratic percentage minus the Repubican percentage. For example, in Arizona, the Democrat - Republican value is -9, meaning the Republican won by 9%. In California, the last poll from Mason-Dixon predicted a Democratic win by 30% (it was actually only 24%). The data is available in .csv format.

State	Vote	Gallup	M-D	Polimetrix	Quinn.	Ras.	Res. 2000	SurveyUSA	Zogby
Arizona	-9		-8	-4		-9		-13	-8
California	24		30	32		23		29
Connecticut	-10			-4	-12	-8	-12	-11	-12
Delaware	41
Florida	22		24	26	35	17	18	23	29
Hawaii	24
Indiana	-87
Maine	-53					-46
Maryland	10	15	3	6		5		3	5
Massachusetts	38			40					30
Michigan	16		16	15		16	11	10	7
Minnesota	20	10	10	18		14		16	8
Mississippi	-29					-37
Missouri	3	4	1	0		-1	0	6	3
Montana	1	9	0			2			1
Nebraska	28					20
Nevada	-14		-17	-11		-12	-14		-21
New Jersey	8	10	7	6	5	5	6		12
New Mexico	42					24			20
New York	36				35	30			24
North Dakota	40					22
Ohio	12	2	6	18	12	12		12	7
Pennsylvania	18	18	13	16	10	13			8
Rhode Island	6	3	-1			8			14
Tennessee	-3	-3	-12	-5		-4		-5	-10
Texas	-26			-36		-26			-19
Utah	-31		-30	-32
Vermont	33					32	21
Virginia	0.2	-3	1	0		0		8	1
Washington	17		16	14		11		8	4
West Virginia	30			34		33
Wisconsin	37			45		27			9
Wyoming	-40		-41			-27

To see how well a pollster did, we can subtract the actual difference between the candidates from the predicted on. For example, in the table below, if we subtract Mason-Dixon's prediction for Arizona (-8) the actual vote total (-9) we get 1. In other words, Mason-Dixon predicted that the Democrat would do 1% better than he actually did. A perfect pollster would have a column consisting entirely of 0s. Since the margin of error in predicting each candidate's score is about 4%, the margin of error in predicting the difference is about 8%. Thus to a first approximation, any number less than or equal to 8 should be considered a correct prediction. Anything more than 8 is wrong.

State	Gallup	M-D	Polimetrix	Quinn.	Ras.	Res. 2000	SurveyUSA	Zogby
Arizona		1	5		0		-4	1
California		6	8		-1		5
Connecticut			6	-2	2	10	-1	-2
Delaware
Florida		2	4	13	-5	-22	1	7
Hawaii
Indiana
Maine					7
Maryland	5	-7	-4		-5		-7	-5
Massachusetts			2					-8
Michigan		0	-1		0	-16	-6	-9
Minnesota	-10	-10	-2		-6		-4	-12
Mississippi					-8
Missouri	1	-2	-3		-4	1	3	0
Montana	8	-1			1			0
Nebraska					-8
Nevada		-3	3		2	14		-7
New Jersey	2	-1	-2	-3	-3	2		4
New Mexico					-18			-22
New York				-1	-6			-12
North Dakota					-18
Ohio	-10	-6	6	0	0		0	-5
Pennsylvania	0	-5	-2	-8	-5			-10
Rhode Island	-3	-7			2			8
Tennessee	0	-9	-2		-1		-2	-7
Texas			-10		0			7
Utah		1	-1
Vermont					-1	-33
Virginia	-3	0	0		0		7	0
Washington		-1	-3		-6		-9	-13
West Virginia			4		3
Wisconsin			8		-10			-28
Wyoming		-1			13

I may not be able to get to the computer for a couple of days, but I will try to do this analysis for the House sooner or later.