Quote:
Originally Posted by Fred314X
Aren't you fascinated by a "national" poll that questioned precisely 1,767 registered voters? Good golly, Miss Molly--they must be 1,767 of the most important people ever, to be telling the whole country of more than 275 million what we believe!
|
When I retired I left my statistical sampling book at work but the number sampled as representative of the population is based on randomness, confidence level and margin of error. Here it is made simple:
Margin of Error and Confidence Levels Made Simple
According to the Census Bureau for the 2008 Presidential election there were 225,449,000 people living in the US of which 206,072,000 were citizens and of which 146, 311,000 were registered to vote. For their results to be representative of registered voters, the sample size has to be based on 146,311,000. However, most pollsters say "likely voters" so they must know what percentage of registered voters are likely to vote in a non-presidential election year and apply that percent to 146,311,000 to come up with a total population number on which to calculate their sample size based on confidence level and margin of error.
I'm not sure what confidence level (99%, 95%, 90%, 85%, 80%) they use but in some polls I see, it appears the pollsters use plus/minus 3% for their margin of error because they seem to call anything within 3 percentage points (as in Jones 46%, Smith 48%) a statistical dead heat (tie). Obviously, a poll with a larger confidence level and low margin of error (as in "I am 99% confident, plus or minus 1%") is more desirable than "I am 80% confident, plus or minus 5%" and the sample size will reflect that.
You should pay attention to likely voters vs registered voters vs adults when comparing polls.
You should pay attention to how they ask the question. Four pollsters trying to determine Nancy Pelosi's approval rating could ask the question like this:
Do you approve of how Nancy Pelosi is doing her job?
Do you approve of how Democrat Nancy Pelosi is doing her job?
Do you approve of how Speaker of the House Nancy Pelosi is doing her job?
Do you approve of how Speaker of the House Nancy Pelosi (D-California) is doing her job?
In Number One, people may not know who Nancy Pelosi is and you might get a high number of I Don't Know responses. In Number 2 they may not know who Nancy Pelosi is but they have good or bad feelings about Democrats and they respond accordingly. In Number 3, they might not know anything Nancy Pelosi has done or who she is but the know the Speaker of the House sits behind the President at his State of the Union address and bangs a gavel. How badly could she be doing in that job? And, they respond accordingly. All kinds of for/against bias could impact the responses to the 4th question. All four pollsters will report their results as "Nancy Pelosi's approval rating."
You should pay attention to whether they are weighting their sample (i.e., polling more Democrats than Republicans or vice versa)
And you should look at the confidence level and margin of error.