Professional Documents
Culture Documents
Political Polls
Political Polls
Political Polls
It is only an estimate
Reasons :
• Asking about voting intentions gives only prediction
• Margin of error is given as 1.91%. Total survey error
need to assessed and evaluated.
• Obviously, we do not know exactly how many
people in Canada will vote liberal.
The case says that the survey is accurate within
1.9% points.
• 1.9 % is the margin of error
• The poll survey reports 38% vote for Liberals
and 26% vote for Conservatives.
• But the true values would be as high as 39.9%,
or as low as 36.1% for liberals.
• 38%‡1.9%
• Would fall within the range of 24.1% to 27.9%
for conservatives.
• 26%‡1.9%
The case says that the • 2,638 is the total number of individuals who completed the
sample consisted of
survey.
• If we estimate the number of samples by taking confidence as
2,638 people.
95% and margin error as 1.9%, then
n=((z-( α/2))^2p^*(1-p^*))/E^2
Here, taking p^* as 0.5
n=2660
• So, necessary sample size for given confidence and margin
error is 2,660.
The case also says
the results are
accurate 19 times
out of 20.
• This means that if we were to conduct the same survey 20 times, we would find that the vote for liberals would fall between
36.1% and 39.9% and the vote for conservatives would fall between 24.1% to 27.9% in 19 out of the 20 times
• In only one out of 20 surveys, we can expect the voting percentage to be higher or lower than these percentages.
• Confidence interval is 95%.
• 95% of all samples taken from the same population will have the same margin of error.
The Ontario poll has 940 people, where as the
national poll had 2,638 people.
In Ontario Poll,
In National Poll,
Confidence=95%
Confidence=95%
Error margin=3.2%
Error margin=3.2%
Taking p^* as 0.5
Taking p^* as 0.5
n=((z-( α/2))^2p^*(1-p^*))/E^2
n=((z-( α/2))^2p^*(1-p^*))/E^2
=937.89
=2660
=938
There might be population difference in both the areas due to which there is the
difference in the sample size
How does the smaller sample size affect the results? How can
the pollsters still be confident that the poll is right 19 times
out of 20?
• Because margin of error is lower with larger
sample sizes, the results of the poll are less
accurate when they are smaller.
n=((z-( α/2))^2p^*(1-p^*))/E^2
Here, E∝1/n
• Now, if n decreases and E increases, (z-( α/2)
will remain constant according to the
formula above.
• Hence the 19/20 confidence level remains
the same.
• We don't agree with the journalist's
conclusion.
• For the following reasons:
• Sampling bias might affect survey results -
Does the sample have the proper
distribution of the target population's age
groups? For the purpose of achieving the
proper demographic proportions, is the data
weighted by factors such as age, gender,
education, race, etc.?
The journalist said that • The case provides a margin of error (3.2).
Therefore, the drop is not too great if the
support drops true value turns out to be 37.2%.