It’s a long
way from the banner headline “Dewey Defeats Truman” in the Chicago Tribune.
I’m not
talking about politics and the press, but about polling, which most Americans
are probably sick of about now. Back in 1948, public opinion polling
was in its infancy, contributing to that journalistic fiasco.
The just completed inauguration of President Obama concluded the 2012 election, where we were deluged with more polls than ever, and they
remarkably foretold the results, correctly predicting the electoral outcome in
every state.
How? Americans are justifiably suspicious of polling, but it is a fact of our lives, and not just in politics. This is a version of an article I wrote for the Oklahoma Press Association Publisher.
Forget the poo-pooers
who argued with what the polls showed because they disagreed or didn’t want to
believe, or thought they were biased.
Forget your political views of Huffington Post. But its poll aggregator
of hundreds of polls nailed the results.
Hence,
here’s
Prof’s Press Primer on Polling, Part One
Definition
of terms is first.
- Population—The
group to be surveyed, such as likely voters, residents of Hennessey, Thunder
season ticket holders.
- Random—Random
does not mean “haphazard.” It means that every person in the population has an
equal chance of being chosen. It’s easy in a classroom—you put every name in a
hat and have a few names pulled out. Bigger groups require phone numbers or
addresses, all more easily available than ever with computer data.
- Sample—The
portion of the population to be chosen randomly to ask the poll questions
- Valid—A poll
is valid if the results collected from the sample can be applied to the entire
population.
- Margin of
error—Expressed as a plus and minus percentage. Every poll has flaws and
variables that will affect the accuracy of the results, but the larger the
sample, the lower the margin of error (If you poll everyone in the population
there will be no margin of error, but that isn’t possible in most cases).
Now the key
question—how big a sample do you need to conduct an accurate poll?
You’re not
going to believe the answer. So first things first. Timing, wording of
questions, training of the pollsters, polling methods, and other factors also
affect a poll’s validity, not just the sample size. But sample size is not
dependent on population size.
That said,
to get a sense of how people in Stillwater might vote on any issue, or people
in Oklahoma on another issue, or people in the United States, for a five percent margin of error, you need
roughly only 400 registered, or likely, voters selected randomly. Yep, that’s
all.
Here’s how
the margin of error figures. Suppose the results come back showing Panhandle
residents favor seceding from the state by a 52-48 percent margin. The results
are within the margin of error so the election could go either way—it could be
52-43, or 47-52, or any combination. If on the other hand it was 75-25,
Oklahoma, you have a problem.
Most
national polls try to have a sample of about 1,200 people—that produces a
margin of error of about plus and minus three percent.
Also
important in polling is the timing. A poll or electability two days after
Romney winning the first debate is valid that day. But as fast as things change
in this digital news country, it wouldn’t be valid in five days.
'Other factors can affect outcome...
you could live in Florida'
Other
factors can affect outcome. People who say they will vote and don’t show up. Or
a Hurricane could shut the place down. Or you could live in Florida.
As with
everything in journalism, sources also matter in polls. Who conducted it?
But that’s a
separate subject—tomorrow, an American’s checklist for evaluating a poll (should you believe it?)
Hint: USA
Today once ran a story and headline at the top of the page about most American
women wouldn’t remarry the same man, based on a Women’s Day survey. What was
wrong with that? And why Romney was shocked that he didn't win.
