2006 CANDIDATES
Ben Cardin
Michael S. Steele
Kevin Zeese

DATE 1					MOE 2	S 3
NOV 5		47.6	50.9	1.5	+/- 1.7	3207
NOV 3		47.2	51.2	1.6	+/- 1.7	3344
OCT 27		45.7	52.8	1.5	+/- 1.7	3351
OCT 17		45.2	53.3	1.5	+/- 1.7	3501
OCT 10		44.2	54.7	1.1	+/- 1.6	3618
OCT 3		44.6	54.0	1.4	+/- 1.6	3640
SEP 26		45.3	53.2	1.5	+/- 1.7	3462

DATE 1	POLLSTER				MOE 2	S 3
NOV 5	SURVEY USA	46.9	50.0	3.1	+/- 3.8	663
NOV 3	MASON-DIXON	48.4	51.6	0	+/- 3.9	625
NOV 2	SURVEY USA	49.0	49.0	2.0	+/- 3.7	694
OCT 31	RASMUSSEN	47.4	52.6	0	+/- 4.4	500
OCT 31	REUTERS/ZOGBY	46.3	51.6	2.1	+/- 4.0	600
OCT 30	BALTIMORE SUN	45.3	51.6	3.2	+/- 3.5	800
OCT 27	ZOGBY/WSJ	44.7	54.3	1.0	+/- 3.8	650
OCT 26	RASMUSSEN	46.9	52.1	1.0	+/- 4.4	500
OCT 26	WASHINGTON POST	43.9	55.1	1.0	+/- 3.1	1003
OCT 17	SURVEY USA	48.4	48.4	3.2	+/- 3.7	698
OCT 10	RASMUSSEN	45.7	53.2	1.1	+/- 4.4	500
OCT 3	REUTERS/ZOGBY	44.6	54.2	1.2	+/- 3.1	1000
OCT 1	USATODAY/GALLUP	41.5	57.4	1.1	+/- 3.8	678
SEP 26	MASON-DIXON	46.1	52.8	1.1	+/- 3.9	625
SEP 25	BALTIMORE SUN/POTOMAC	43.4	55.4	1.1	+/- 3.4	815
SEP 19	SURVEY USA	48.9	48.0	3.1	+/- 4.3	522
SEP 13	RASMUSSEN	45.7	53.2	1.1	+/- 4.4	500
SEP 11	ZOGBY/WSJ	44.4	54.3	1.1	+/- 3.1	1000

NOTES
¹ Date is the last day that the poll was taken. Often polls are embargoed for several days until their release date.
² Margin of Error. MOE is related to sample size, 0.98 divided by the square root of the sample size at 95% confidence (19 times out of 20). For example, a sample of 1000 people has a MOE of +/- 3.1% and a sample of 2000 people has a MOE of +/- 2.2%. For rolling averages, the samples are assumed to be independent, thus sample sizes are aggregated to produce an overall MOE. In practice, some polling firms use lists with repeat calls; in these case, it is more precise to reduct these repeat calls from the overall sample size, thus the actual margin of error might be slightly higher. Example: 5 surveys of 1000 people would normally produce a MOE of +/- 1.4%, if samples were independent. If 25% of calls were repeat calls, however, a more accurate MOE would be based on 4000, not 5000, people, producing a MOE of +/- 1.5%.
³ Sample size.
⁴ Poll numbers exclude undecided voters (and in some cases include estimated for Others), thus are pro-rated to 100%. For example, if the raw poll data shows the Republican candidate with 47%, the Democratic candidate with 41%, Others with 2% and there were 10% undecided, the results excluding undecideds are: Republican 47/90 = 52.2, Democrat 41/90 = 45.6, Others 2/90 = 2.2.

JOIN THE CONVERSATION BELOW