This evening KY3/MSU released a poll of 785 people taken statewide. The online report did not indicate the date(s) of the interviews, the margin of error for the full sample, or the voter screen. The results were weighted. The results do indicated the margin of error for the three congressional race subsamples.
…If the 2010 election for U.S. Senate from Missouri were held today, would you vote for Robin Carnahan, the Democrat, or Roy Blunt, the Republican?
Robin Carnahan – 48.4%
Roy Blunt – 48.8%
Other – 2.8%
The formula to calculate the margin of error (95% confidence level) is 1.96 x the square root of the multiplied percentages of each candidate which has been divided by the total sample.
1.96 x square root ((Carnahan x Blunt)/785)
For our purposes here we’ll call ’em even at 50% each. That’s a 3.5% margin of error (3.497%, if you don’t want to round to one decimal place).
A commenter responding to questions about the poll added the following:
….Margin of error is +/-3.5 and although there were more female respondents, the data was weighted appropriately to compensate.
Since the poll was conducted with Random Digit Dialing. Since most younger people use only cell phones (cell numbers are not included in this type of poll) there are more older respondents. Again, though, the data is weighted to compensate. The poll results gained from this sample can be generalized to the larger state population with a 95% confidence level….
It’d be nice to know the interview screen – adult, registered voter, or likely voter – and the period of time in which the interviews took place.