| || || ||
| ||ONLINE EXTRAS/What Americans Think|
Predicting S.C. Primary Winner Virtually Impossible
By Richard Morin
Washington Post Polling Director
Monday, February 14, 2000
Voters in South Carolina primaries like to say they correct the mistakes of voters in New Hampshire primaries. Pollsters, smarting from a generally dismal showing in New Hampshire, hope South Carolina will make things right for them as well.
Good luck. Polling in New Hampshire is tough. Polling in South Carolina -- particularly this year -- is virtually impossible. Here's why. Any registered voter can vote in the Republican primary, regardless of party affiliation. That spells potential disaster for pollsters, who live or die by their ability to define through their questions the population of likely voters and then measure their preferences.
In a typical year, that means interviewing Republicans who regularly vote in primary elections and who intend to vote in this year's primary. Independents and Democrats largely can be ignored. Dave Woodard, a Clemson University political scientist who directs the Palmetto Poll, says independents traditionally have not been a factor in South Carolina primaries. Not so, this year, he says.
Recent surveys suggest that three factors may conspire to rewrite political history in South Carolina, and perhaps send some of America's boldest pollsters into retirement.
The first is the compelling personal biography of Arizona Sen. John McCain, who already is in the record books by virtue of his 19-point shellacking of Gov. George W. Bush in New Hampshire. His story induces the kind of whopper turnout among independents that made those New Hampshire pre-election tracking polls so wrong. Currently in South Carolina, McCain leads Bush by a whopping 57 percent to 33 percent margin among political independents, who constitute about one in four likely GOP primary voters, according to the latest Post-ABC News poll in South Carolina.
McCain, a former prisoner of war, also puts in play military veterans, who represent about one in eight adults in South Carolina. He currently leads Bush 52 percent to 41 percent among veterans. Among non-vets, he's tied with Bush in the Post-ABC poll.
These data also suggest that McCain is luring a small but potentially significant number of vets who are Democrats to vote in the GOP primary. Overall, about 8 percent of those who say they plan to vote in the South Carolina Republican primary are Democrats. Half of these party-jumpers are veterans or live in households that includes a military vet, and a lopsided majority of these vet Dems are voting for McCain.
The second factor that spells trouble for Bush and for pollsters is that South Carolina Democrats have canceled their primary, opting instead for caucuses. Woodard suggests this may free up thousands of independents and weak Democrats for the Republican primary. That's why he's telling reporters that the GOP primary is too close to call, even though his Palmetto Poll found Bush with a 20-point advantage barely a month ago.
The third factor complicating the pre-election polls in South Carolina is that few voters turn out. In the past three GOP presidential primaries, turnout as a percentage of the voting age population (pollsters call it VAP) has been six, eight and 10 percent.
Those numbers mean that, on average, there are only about eight real voters in every 100 South Carolina adults. Thus polling a sample of true likely voters is extremely costly and time-consuming. On average, only one of every 12 people you contact on the phone will end up voting in the primary. (That's very different from general presidential elections, where about half of all adults end up casting ballots.)
The scarcity of primary voters poses a real problem for pre-election tracking polls, which require separate, large samples of likely voters each night to measure changes in voter preferences. Do the math. First assume that turnout will be about 10 percent. To capture a sample of 200 truly likely voters in a single night, the typical sample size in pre-election tracking polls, requires that pollsters screen nearly 2,000 adults. (These "screens" are a series of three to five questions that ask intention to vote, interest in the campaign and past voting behavior. The answers are then combined and only those most likely to turn out are included in the sample of likely voters.)
Say it takes only two minutes for interviewers to introduce the survey, randomly select a respondent and then ask the screening questions. That's 4,000 minutes just to identify 200 likely voters, or more than 60 hours of interviews, tough to do in a single three-hour night of interviewing, even with a small army of interviewers.
So pollsters cut corners to save time and trim costs. They don't ask so many screener questions, which means they overestimate turnout by including too many people in their sample of likely voters. Instead of 8 or 10 percent, their sample may include be 20 or 30 percent of all adults questioned. One recent South Carolina poll had a turnout estimate of 43 percent. That means these pollsters were betting that more than four in 10 South Carolina adults were projected to vote in the Republican primary, or eight times the likely proportion. Ain't gonna happen.
And turnout matters, particularly in South Carolina. Recent surveys suggest that the higher the turnout, the better McCain does. In the recent Post-ABC News poll, McCain wins big if the election were held today and turnout tops 20 percent (or twice the recent historical high). But Bush would win by four to six points if only about 10 percent of the voting age population actually voted.
That's why primary election tracking polls seem so, well, all-over-the-place. Depending which corners are cut, the results can vary widely from poll to poll. Overestimate turnout and you overestimate McCain's share of the vote. Underestimate turnout and you overstate Bush's support.
Archive: What Americans Think