A math teacher's primer to the polls
You will drive yourselves crazy if you live and die by day to day polls. Polls are snap shots and blurry ones at that. Polls have MOE's and they are not trivial. Most MOE's are at least 3.5 giving a range of 7. Obama won by 7.2% in 2008. Bush won by 7.8% in 1988. Neither of those are outside of all MOE's for even well done polls. There is a 0% chance Hillary will win by that amount this time. Obama won by 3.9% in 2012. I think Hillary will at best split the difference. A win by 5.5% is about the best we can expect. That is well within the MOE of any poll. Of course some polls will show her losing. Other will show her up by double digits. That is what an MOE means. Say she is really up by 5 points right now. Then any result ranging from her being down by 2 to up by 12 is within the MOE. There is around a 5% chance of getting a result the same as or more extreme than those on any one poll. Given that there are dozens of polls being produced the odds of one of them appearing becomes fairly strong. For example if you have 15 polls the probability of one of them being this extreme is 36.5%.
So what should you look for? One, look for poll averages. The MOE for a poll average is vastly lower. Instead of the 3.5% that the best poll has a poll average is likely about 0.5. If a candidate is behind in the poll average by 2 points then that candidate is quite likely to be behind in real life. Conversely a candidate is behind in a single poll by 2 points then he is statistically tied. Two, look for trend. If a candidate is losing support over time in the same poll consistently then that candidate is likely losing ground, conversely if a candidate is gaining support over time in the same poll consistently then that candidate is likely gaining ground.
So relax, polls will go up, polls will go down. Averages and trends matter. They are what stay in focus.
= new reply since forum marked as read
