Caucus fever gives pollsters a headache

Future Tense for Dec. 17, 2007

By Steve Welker

By Steve Welker

I'm five years and 1,000 miles away from Iowa. Nonetheless, I've caught Hawkeye caucus fever. Neither time nor distance provides prophylactic protection against this quadriennial infection.

I hit RealClearPolitics.com every morning and check the latest polls. I already read political news in the North Carolina newspapers, Google News and the New York Times. Now I've added the DesMoinesRegister.com, IowaCaucus.com (from the Cedar Rapids Gazette) and the Quad-City Times' Elections 2008.

I conducted my own straw poll in Mount Airy and Dobson. "How much time do you spend each day reading about the Iowa caucuses?" The average can be stated in nanoseconds; that is, almost no time at all.

Like the man said, maybe you just had to be there. Personally, I was there in Iowa when the caucuses started and I covered or attended precinct caucuses every four years when presidential candidates were running plus plenty of "off-year" caucuses, too.

A lot of Iowans never catch a case of caucus fever as bad as mine. Even among the infectees, many don't pay much attention to the polls. Iowans who understand how the caucuses work also understand that polls aren't a very good predictor of results. The pollsters themselves will admit -- though usually not on the record -- that they're doing well if they can predict the rank order of the top candidates.

The people who supposedly take the pulse of America also admit -- usually after a few beers or in late-night forums on the Internet -- to feeling increasingly sick about their own future.

The pollsters blame their problems on technology, but I believe they put themselves on the road to extinction by overpromising what they could do. How long would you listen to a racetrack tout who couldn’t pick a winner? The fact is, pollsters have miscalled presidential races since Truman-Dewey in 1948 through Bush-Kerry in 2004 when last-minute polls gave Kerry a 1.5-million-vote edge. More recently, the pollsters misgauged last year's backlash against moderate Republicans when Jim Leach lost his congressional seat in Iowa. Closer to home, the USA Today/Gallup poll miscalled the Jim Webb-George Allen race on the eve of the Virginia election.

Political polling has a long, often-checkered (and often black) history. The Harrisburg Pennsylvanian published the first U.S. presidential straw poll in 1824, according to Terry Madonna and Mike Young at Franklin & Marshall College’s Center for Politics and Public Affairs.

(The phrases “straw vote” and “straw poll” have been around even longer. In the 1600s, John Seldon wrote, “Take a straw and throw it up into the Air — you may see by that which way the Wind is.” Perhaps anticipating the poll-confounding influences of Internet bloggers and rock-star rallies, Seldon added, “More solid things do not show the Complexion of the times so well, as Ballads and Libels.”)

The Harrisburg newspaper surveyed groups of citizens in Wilmington during July 1824, asking about their presidential favorites. Military hero Andrew Jackson had a commanding 70 percent lead (335 votes) over Secretary of State John Quincy Adams (23 percent). “Old Hickory” won the popular vote in the fall election, beating Adams by 77,000 votes, but neither candidate had a majority of electoral votes. In the U.S. House of Representatives, Adams triumphed.

The Pennsylvanian’s poll is what the media today describe as an “unscientific” survey and the pollsters themselves label “a non-probability sample.” For the next 110 years, some polls were surprisingly accurate, but most failed because they didn’t question a representative sample of voters. One of the most notorious examples involved the Literary Digest poll. It gained a wide following during presidential elections in the early 20th century, because it used huge samples. In 1936, the Literary Digest mailed out more than 10 million questionnaires and 2 million people responded. The Literary Digest confidently predicted Republican Alf Landon would overwhelm Democrat Franklin Roosevelt. Instead, Roosevelt won 46 of 48 states and 63 percent of the vote. Totally discredited, the Literary Digest soon went out of business.

The Digest poll failed because of bias in its “sampling frame.” The mailings went to people who had 1936 auto registrations and who were listed in telephone books. In 1936, millions of Roosevelt supporters were too poor to have phones or cars, so the sample missed scores of likely voters.

That historic goof gave rise to scientific polling, starting with George Gallup and the much-respected, Iowa-based Gallup Poll.

Even the “scientific” pollsters made huge mistakes.

In 1948, Gallup and two competing pollsters, Crossley and Roper, all predicted that Thomas Dewey would defeat Harry Truman. Several “straw” polls that year went the other way. A Kansas City feed company printed donkeys and elephants on its burlap sacks, letting farmers register their preferences by which ones they bought. Some 20,000 farmers were “polled” and 54 percent favored Truman. A similar “popcorn poll” at movie theaters produced the same improbable results. In the end, Truman beat Dewey by more than 2 million votes and carried 28 states (303 electoral votes) to Dewey’s 16 (189 electoral votes).

Gallup and Roper conducted their polls by telephoning a random sampling of likely voters. Many of Truman’s supporters either didn’t have telephones or were working in the fields when the pollsters called. Then the key difference was between people who checked their ballots with well-manicured fingers and those with nails rimed by dirt and grease.

Year after year, in one election after another, the pollsters learned from their mistakes. In particular, the predictions based on exit polls have become highly accurate, which is how the networks can “call” many races just moments after the polls close. However, pre-election polling remains as much craft as science.

Now technology has laid new landmines in the pollsters’ path.

One technological change is the rise of cell phones. One in every eight U.S. households has junked “land-line” telephones in favor of cell phones. Many of those cell numbers don't appear in any directory, so it's hard for pollsters to find them.

Caller ID frustrates pollsters, too. Many people don’t like paying for the minutes when pollsters call and many won’t waste time on surveys. If a call comes from an unknown source, many people won't even pick up the phone.

Squeezed by the two technologies, pollsters find fewer people to call.

A third technological challenge is the Internet itself. News and opinion have never changed faster. Public sentiment swings at hurricane velocities. Polls sometimes become outdated before they’re published, especially when a new disclosure or rumor swings the public’s mood like a weathervane in an Iowa twister. Some pollsters now conduct “rolling” polls, averaging results from three or four days’ calls and then discarding the oldest day’s results as each new day’s arrive. Several sites on the Internet, such as RealClearPolitics.com, now conglomerate the leading polls, averaging the results to smooth out variations.

I expect more pollsters will use the Internet for surveys, because it’s fast, cheaper than snail mail or in-person interviews and well-suited to handling large numbers of people. However, relying on the Internet poses its own problems for pollsters, but that's a subject for another day.

As for Jan. 3’s outcome in Iowa, I predict the pollsters will be right — and wrong — just as they’ve been in the past and will be again ... in the future.

——

Steve Welker is the editor of SurryBusiness.com. His e-mail address is steve@stevewelker.net.

This article, as well as reader comments can be found at stevewelker.net/article/58. Steve Welker is an editor and columnist living in Mount Airy, North Carolina. More of his writings can be found online at stevewelker.net