News & Events
You might also like
- Five Tips on Drainage Law
- 2014 Ohio Farm Bureau Presidents Trip to D.C.
- How OFBF members are working to change a law affecting road access
- Animals make our lives better
- A non-partisan look at the implications of the Affordable Care Act
The art and science behind polls
Knowing what’s on the minds of voters and consumers can be a valuable tool. Getting that knowledge can be as simple as doing a poll. But doing the right type of poll and doing it correctly is both an art and science, says a long-time pollster.
The key to find out whether there is an agenda to a poll is to find out who sponsored it, said Martin Saperstein, president of public-opinion research firm Saperstein Associates.
“Polls are used in a myriad of ways. You need to find out who paid for the poll. If an interested organization is paying for the poll, frequently they have an agenda and want to promote it,” said Saperstein, whose clients include Bob Evans Farms, Nationwide Insurance, Columbia Gas of Ohio and Ohio State University.
Saperstein talked about different types of polls and how they are used on a recent radio episode of Town Hall Ohio. He said that while Internet surveys are becoming increasingly popular, many agencies still rely on phone polls.
Does Sample Size Matter?
People often have the misconception that sample sizes need to be large to get an accurate picture of an issue, he said. Sample size has nothing to do with the size of the population, he said.
“Say you are in the kitchen and want to make soup. You stir the ingredients and randomize the ingredients. Then you want to see if it is salty enough. You put the spoon in and taste it. Do you use a bigger spoon if you’re cooking in a 55-gallon drum or 1-quart saucepan? No, you use the same size spoon,” he said.
Saperstein pointed out that sample size is determined by how much error a client can live with or how much money they have to spend on a survey. A larger sample size will result in a smaller error, but only to a point. For example, he said for 100 interviews, the error is plus or minus 10 percentage points. For 1,000 interviews, it is plus or minus 1 ½ percentage points. For 20,000 it is plus or minus .5 percentage points.
“The error decreases at a slower rate as the sample becomes larger and larger,” he said.
Problems and Influence
Many people think sampling error can cause the biggest problems with polls but Saperstein believes it is measurement.
“Polls are often driven by the sponsor who is paying for it and they shape the questions that predetermine the results,” he said, pointing out that the questions need to be valid and reliable.
Polls can have a lot of influence, in particular in the political arena.
“In politics, polling sometimes changes the agenda and sometimes it reinforces the agenda,” he said. “If you think back to the two Bush elections and think about how close the election was … if the impact is even tiny, it could have a huge impact on the elections, given the split on the nation.”