Polling shows a range of results on ballot measures in the run-up to the election. Often, they don’t accurately predict the outcome.

Polling is mathematically sound. It’s been studied enough that in many ways, it’s a science. Why doesn’t polling work?

How polling works.

Pollsters ask a thousand or a few thousand people, pollees, some questions.  The pollees are chosen at random.  Sometimes they’re completely at random, sometimes their demographics are known so that the demographics of the set of pollees together roughly resemble the demographics of the general population.  The mathematical theories behind polling then assert the answers are representative of the whole population.

If the pollees are similar to the population in general, or are chosen truly randomly, statisticians can calculate the odds that the answers of the pollees differ from the answers of the general population.  This gives them the “margin of error”, which is typically under 5% for a national poll where the sample is a thousand people or more.  The math is sound.  So why doesn’t polling work?

Answering a survey is different from voting.

A big part of the problem is: Answering a survey is different from voting.  Many surveys are done far ahead of an election. And answering a survey is part of a conversation. Voting is a committed action, a decision.

I was polled over the phone a few times before the last election.  Each time, the pollster posed an issue and then gave me some choices for answers.  I listened carefully, trying to quickly consider the issue and then weighed each answer.  By “weigh”, I meant I sensed my feelings about each answer, letting thoughts about it percolate up into my mind.  Then I picked an answer I felt was best and we moved on.

On the other hand, when I went to decide my votes, I sat down at my computer with the sample ballot and my voter guide.  I read each measure, scanned the long explanation and the pros and cons and looked carefully at who endorsed and opposed each measure. For the few I was uncertain about, I checked with my wife first, and on a couple of issues I talked to a friend who is more politically involved.  A few times I checked online with some organizations I respect.  I went through a similar process when I decided upon some of the candidates.

Sometimes when I vote I don’t vote that carefully. Almost always I check my answers against my wife’s, and we talk over any differences.

The huge difference in the process leads to a significant difference in the results.

But this is only the beginning of the problems with polling. We’ll look at more in part 2.