Day

to

Day

Politics

                            

 

1)  What does a poll tell us?

2)  How is this Poll Average calculated?

3)  Why is this Poll Average better? 

4)  What is 'margin of error'? 

5)  What determines if the state is a toss up?

6)  What determines if the state supports the candidate?

7) How are the battleground state probabilities calculated?

 

 

 

 

 

 

 

What does a poll tell us?

 A poll is conducted on a sample of individuals out of the population.  The results are an educated guess about the population.  It is an educated guess because we know the range in our uncertainty about the true population.  The margin of error tells us that we have 95% confidence that a candidate’s estimated percentage of supporters falls within plus or minus 1 margin of error.  For example:

     Candidate A has an estimated percentage of support of 49%.  If the margin of error is 3%, then    we have 95% confidence that the candidate’s percentage of support in the population is between 46% and 52%. 




How is this Poll Average calculated?

The Day to Day Politics Poll Average uses the statistically correct method for conducting a poll average.  It utilizes the number of people conducted in each poll to accurately reflect the information of each poll instead of treating each poll with the same weight.  A poll conducted on 100 people should not have the same weight as a poll conducted on 3000 people as other poll averages do.  Thus, the Day to Day Politics Poll Average has several advantages over other poll averages.  First, it properly gives more weight to larger polls, since they more accurately reflect the population average.  Second, the Day to Day Politics Poll Average statistically has a margin of error, which reveals the measure of uncertainty in the Poll Average, just as the margin of error in a single poll measures the uncertainty of that poll. 

Finally, the Poll Average is conducted on the last week to 10 days, but any more than that would not accurately reflect the changing views of many Americans. 




Why is this Poll Average better? 

This poll average correctly combines different polls based on their different samples sizes and gives more weight to polls that have larger sample sizes, since those polls have smaller ranges of uncertainty.  Also, this Poll Average has a margin of error, which gives us the range in uncertainty around the estimated percentages of support.    The Poll Average usually has a margin of error of less than 1%, which means we can more accurately estimate the percentage of support within the population than any single poll or other types of poll averages. 




What is 'margin of error'? 

The margin of error tells us that we have 95% confidence that a candidate's estimated percentage of supporters falls within plus or minus 1 margin of error.  For example:

 

Candidate A has an estimated percentage of support of 49%.  If the margin of error is 3%, then we have 95% confidence that his percentage of support in the population is between 46% and 52%. 




What determines if a state is a toss up?

A state is considered a toss up if after we analyze the polls for that state, there is less than a 95% probability that any candidate will win the state. 




What determines if a state supports the candidate?

A state is considered supporting the candidate if after we analyze the polls for that state, there is more than a 95% probability that the candidate will win that state. 




How are the battleground state probabilities calculated?

The polls for a state are collected over the previous 1-2 weeks and the election within that state is simulated 50,000 times from those polls.  Then the number of times the candidate wins the election in those simulations is divided by 50,000 and the probability is reported. 

      (Technical note:  What is taking place here is we are treating the polls as multinomial distributions and using WinBugs to sample from the posterior distribution.  Then we estimate the  probability that candidate A’s support is actually greater than candidate B’s support)