“Polling is an art, but it’s largely a scientific endeavour.”¹
In this election season there are more “scientific” opinion polls than you can shake a stick at, should you believe them?
Maybe, most sample about 1000 – 1500 individual. How they select them is the first factor. There are two dominate methods:
- Random digit dialing sample of landlines – used by organizations such as Pew Research Center
- Registration based sampling uses voter registrations or sometime list of voters in a prior election.
The first polls people who may not be eligible to vote, while the second, if it uses up to date voter registrations, at least polls those who are eligible to vote.
Disregard any nationwide poll, we don’t elect based on a nationwide vote, it comes down to the electoral college. If you don’t see a map and a count of electoral vote, ignore it.
Disregard any two candidate polls, there are 4 or more candidates on the ballot.
Understand margin of error, if it is given as 2% each individual’s results can change by as much as 2% thus 48 to 52 with a 2% margin of error could be 46 – 50 to 50 – 54, a tie or a 8% blow out.
Real Clear Politics does an average of poll, thus no margin of error is given. Give their site a visit, have a look around.
- Electoral College – RCP Electoral Map
- Your state if it is a Battleground state.
¹ Michael Link, president and chief executive of Abt SRBI Polling, in Nature Magazine, October 19, 2016