A primer on polls, believe or not

AS FALL approaches (starts Sunday), y'all probably are wondering who'll be leading in the polls for governor this time next year.

You're not?

Why, because you think you know? Because you don't like or trust polls?

Well, if it's the latter, you're not alone.

The international research and data firm Kantar just released a poll on polls that says 75 percent of Americans don't trust polls, thinking that they're biased.

(Which, of course, means 75 percent don't trust that poll.)

But, hey, there are polls and there are polls: left-leaning, right-leaning, candidates' polls, media polls, good polls, bad polls.

Since you'll soon see lots of 'em - on Gov. Corbett, state issues, the big-field Democratic primary - you should know there are polls you can trust.

Kantar's survey found the most-trusted polls (with 64 percent of respondents claiming "some trust") are those conducted by academics.

With good reason.

In Pennsylvania, we've seen consistently accurate polling from Muhlenberg College, Franklin & Marshall College and Quinnipiac University.

All three did better than the Washington Post/ABC News poll and Time magazine, according to a New York Times rating of 50 polls for accuracy in 2012.

In fact, Muhlenberg and F&M ranked ninth among the 50, tied with the Times.

(Disclosure: F&M pollster Terry Madonna has done the Daily News' polls for decades.)

I chatted with Madonna, Muhlenberg polling director Chris Borick and Q-Poll boss Doug Schwartz to offer you, dear voter, a short primer on polls.

"I think the biggest misconception is that all polls are the same," said Schwartz, "the average voter isn't aware of all the methodology that goes into it."

For example, Muhlenberg, F&M and Quinnipiac do "live calls" to landlines and cellphones; many other polls do not.

Some polls are automated. Some are Internet-only. Some don't call cellphones. Some "screen" for registered voters or "likely voters," still others for just adults.

Residential phone numbers are bulk-purchased from companies such as Marketing System Group in Horsham and Survey Sampling International in Connecticut.

Cell numbers are bought, too; certain numerical prefixes or exchanges are dedicated to cellphones.

Polling calls usually are placed from 6 to 9 on weeknights, 10 a.m. to 2 p.m. Saturdays and 5 p.m. to 9 p.m. Sundays. Pollsters try to keep calls under 15 minutes and hope to get up to 30 questions answered.

In recent years, according to Borick, polling entered "a tricky environment."

Many landlines have caller ID, leading to blocked calls. Many cellphone owners are underage or on the go; the former unable to pass screens for registered voters, the latter unwilling to be surveyed while, say, driving or shopping.

But pollsters are doing better with cellphones.

"In 2004, exit polls told us only 7 percent of voters were cellphone only," says Schwartz. "Now it's 25 percent to 30 percent and going up each election. Not making cellphone calls risk not representing up to a third of the electorate."

Also, polls are "weighted," or adjusted, to reflect demographics if samples do not.

Using census data, voting patterns, registration figures and more, pollsters give mathematically added "weight" to some answers so variables such as gender, age and income closely match actual demographic data.

And do weighted demographics for registered voters offer a glimpse at who candidates might seek to target?

"I think as a general proposition you can say that," Madonna said.

Borick and Schwartz agreed.

So here's a fun nugget.

Data from three recent Pennsylvania polls suggest candidates might want to appeal to this voter: a married Democratic woman age 45 to 64, who's Protestant, doesn't have a college degree and lives in the Philly suburbs where she is not a Penn State football fan.

All this, of course, comes with a margin of error - I mean, if you can believe it.