Raleigh

Barack Obama is up five points. No, he's up 20.

Beverly Perdue's lead is 26 points. Wait a minute! She's statistically tied with Richard Moore.

Anyone paying attention to polls this election season probably has a case of whiplash by now. Different polling outfits have come out with some markedly different numbers, and even tracking polls performed by the same companies have reported dramatic shifts.

In early March, a Democratic polling firm in Raleigh, Public Policy Polling, showed Lt. Gov. Beverly Perdue leading State Treasurer Richard Moore by well over 20 percentage points in the race for the Democratic nomination for governor. A poll a few weeks later by the same firm showed the two in a virtual dead heat.

Another poll from Survey USA earlier this month showed the two Democrats tied at 40 percent each.

In the Republican gubernatorial primary, the same polling companies have alternately put Fred Smith and Pat McCrory in virtual ties or shown McCrory way ahead.

And a survey commissioned by McClatchy Company newspapers showed better than 50 percent of Democrats and Republicans undecided in the governor's race just a month out from the election.

So, what's going on here? Do the polls reflect real swings in voter attitudes and preferences?

For the most part, they probably don't.

What's really happening this election season, particularly on the Democratic side, is that pollsters aren't real sure what election turnout will look like on May 6. So grabbing a true, realistic sample of likely voters for their surveys is more difficult than ever.

In other states, the Barack Obama phenomenon has led to a dramatic increase in voter turnout. With the Democratic presidential primary still in play, elections officials expect the same kind of turnout here. But how much of an increase will we see? Will it be double, triple or something far less?

The McClatchy survey was criticized because it reflected what appeared to be unlikely increases in voter turnout. The company that performed the poll didn't do the normal tricks -- asking or actually determining whether respondents had voted in previous primaries -- to try to cull their sample to actual likely voters.

Some of the criticism came from Public Policy Polling. But that company shifted its survey sample in late March. It started polling equal numbers of Democrats who had voted in the two previous primaries and those who had voted in neither but had voted in the past general election. The intent was to take into account the Obama factor.

Maybe that's the way to do it. But firms like Public Policy Polling and Survey USA, though they've accurately predicted races in the past, have always been a bit suspect because their polling is electronic, with no live person on the other end of the telephone line. Does that skew the demographics of who does and doesn't participate?

Well, the primary is 18 days away. We'll see soon enough whether any of the numbers mean much.

Scott Mooneyham writes for Capitol Press Association. Contact him at smooneyh@ncinsider.com.