There are polls, and then there are polls

Richard Fulton
This week I am continuing with the discussion of political opinion polling on the Clinton-Trump presidential election.
As I began exploring last week, those polls are complex with multiple points at which error could creep into the final figures reported. Polls use small samples of large pools of people to come up with their results.
So the first question to be asked is what group of people are you targeting? For political polls at election time it is either Registered Voters or Probable Voters. Obviously the first is a larger and more diverse pool of people than the second.
The second question to be asked is about the sample; how was it determined and how large was it. There is a complex proprietary formula each company uses that is different in some regards from others: thus hard to evaluate.
Live phone calls use to be used by most national pollsters, but now they use online sources. Live phone calls are used by marketers today since they don’t care about sampling only selling. Fake political pollsters use live phone calls to gather information on you, not the election per se.
A national poll usually has a sample size of between 700 and 2000. Some, like NBC News, have used samples as high as 15,000. How you choose those to sample is also a secret kept by each polling operation.
The few who use live phone calls usually believe they need a sample of only 700-900 to get a low error. Remember, error factor means so many points plus or minus the reported numbers.
If we use some examples from polls from last week we can see how they can vary in results when asking who people will most likely vote for in the presidential race. The examples I use generally are those that are not connected with any political or ideological organization, thus with less bias in wording and more error.
There still are badly worded questions in many polls, however.
Most polls in the last week or so have had Clinton ahead, but they vary in a how much ahead. The range is from plus 3 percent to plus 9 percent. A few have Trump ahead by plus 1 percent to plus 4 percent.
Last week NBC News commissioned a survey that had a sample size of 17,000 people with an astounding 1 percent error factor. They found Clinton ahead by 6 percent points. In theory this should be the most accurate poll of the bunch given good polling practices. But many used fewer samples to get similar results.
I found two surveys sponsored by CNN that had significantly different results. One had Clinton ahead by plus 9 percent points while the other had Trump ahead by three percent points. The latter had a sample of only 882 using Registered Voters as their pool of participants.
Look closely at the polls. YouGov/Economist poll last week had Clinton ahead by plus 3 percent but had an error factor of 4.1 percent (bad) and used a sample of only 906 Registered Voters. Don’t pay much attention to this poll.
Only one poll last week had Clinton ahead with a majority of the votes, UPI. Most of the more reliable polls still have Clinton ahead but by varying margins. The national numbers remain flexible with so long to go before the election.
But remember, we do not elect our president by a national popular vote. We don’t elect them at all by a strictly popular vote. The president is the one who receives a majority (and it must be a majority) of the electoral college vote.
This vote is accumulated by the winner of the plurality of votes in each state which determines how each state casts it electoral college votes. Win one more vote in Missouri and you get all of Missouri’s electoral college votes.
If you are a political junky, you can follow the real election by searching the polls of each of the states. Think of the complexity of this given the possibilities of errors in each of those state polls.
The New York Times looks at these state polls for you, adds them up and gives the odds of which candidate will win the election. Currently they give Clinton a 77 percent chance of winning.
Remember, George W. Bush won the electoral college majority and lost the national popular vote.
