Las Vegas Sun

March 29, 2024

Don’t get duped by an inaccurate poll

The closer we get to Election Day, the more you can expect to be bombarded with polls declaring one candidate the leader, or surveys about how a demographic will vote.

​But not all polls should be trusted. Many statisticians cringe at the way some are conducted.

Here are tips on how to make sense of the numbers:

Context is everything. Knowing the results of a survey is never enough. You also have to pay attention to why the survey was taken, who conducted it, who paid for it, what questions were asked, who was targeted and who answered.

“In absence of that, it’s really just a bunch of numbers,” said Sallie Keller, a Virginia Tech University statistics professor.

A small sample size is better. It may seem counterintuitive, but a poll surveying as few as 1,000 Americans could be indicative of the country’s entire population if the poll is well conceived, the respondents are well selected and the poll is well executed.

If all that holds true, “you would be better off to get 90 percent of the 100 surveyed, rather than 20 percent of the 1 million,” Keller said.

It matters who conducts the poll and reports the results. Many political campaigns and advocacy groups hire unbiased polling companies and report results accurately. But sometimes, some smooth over undesirable results.

About the margin of error. A smaller margin of error means pollsters are more confident about their findings. It doesn’t, however, account for biases conducted during the poll.

What trends tell you. Not much. Combining the results of 20 polls is like trying to pay for a meal using currency from 20 countries. Polls all are calculated differently. Still, statisticians take comfort when other polls corroborate their own.

Beware of telephone surveys. In the age of caller ID and voicemail, telephone surveys are more imprecise than ever.

Moreover, surveys conducted by landline will target a specific demographic — one that likely isn’t reflective of the general population.

Beware of Internet surveys. Experts call these surveys of convenience. With one click, anyone can blast a survey to the world and report the results of whatever email addresses respond.

More often than not, experts say, the polls ignore centuries of social science about how to ask questions and whom to ask.

The bottom line. Trust your intuition, and remember the words of Ron Wasserstein, director of the American Statistical Association: “Everything you see is an estimate.”

Join the Discussion:

Check this out for a full explanation of our conversion to the LiveFyre commenting system and instructions on how to sign up for an account.

Full comments policy