Perspective

Pigs, pols and polls

By STEVE KIRCHER, Times Research Manager
Published December 9, 2007

The Iowa caucuses, coming up in just a few weeks, have been at the center stage of presidential politics for at least a generation.

But that farm state, which has nearly six times as many hogs as people, made a more significant contribution to American politics 40 years before the Iowa caucuses became famous: scientific political polling.

Two native Iowans, who were lifelong friends, made that contribution and became famous well beyond the state borders.

Chances are you haven't heard of Mike Cowles (pronounced Coles). He died in 1985 one week after his family-owned newspaper, the Des Moines Register, was sold to Gannett, the United States' largest newspaper company. He's best known as the founder of Look magazine, and his family owned many newspapers, including the Gainesville Sun and the Lakeland Ledger.

I directed the Des Moines Register's Iowa Poll from 1983-87, but I never met Mike Cowles. He had retired from the paper and lived on Long Island. I did meet the other, better-known native Iowan, George Gallup, founder of the Gallup Poll, briefly before a political event in Des Moines.

In the late 1920s, Cowles teamed up with George Gallup while Gallup was working on his doctorate degree at the University of Iowa. Cowles asked Gallup to conduct scientific surveys of newspaper readers.

Gallup also conducted surveys for his mother-in-law, Ola Babcock Miller, when she ran for Iowa's secretary of state in 1932. She won, becoming Iowa's first woman secretary of state.

At the time, scientific polls were uncommon. What makes a poll scientific is how people are chosen for the survey. In his academic studies, Gallup learned about the statistical law that says if people are randomly selected from the population, you don't need to interview very many of them to get an accurate read on the opinions of the whole population.

Gallup proved it in the 1936 U.S. presidential election. By that time, he had moved to Chicago where he was applying his research techniques for an ad agency, Young and Rubicam. His methods were so successful that the ad agency grew steadily during the Great Depression of the 1930s.

In 1936, Gallup conducted a national poll of 3,000 randomly selected adults. It said Democrat Franklin Roosevelt would beat the Republican governor of Kansas, Alf Landon, by a comfortable margin. The leading poll at the time said otherwise.

The Literary Digest was a popular general-interest magazine and was known for its presidential polls. For the 1936 election, it mailed questionnaires to 10-million people. Two-million responded and they strongly preferred Republican Landon. Those 2-million accounted for nearly 2 percent of the U.S. population while the Gallup survey included only 0.002 percent of the country's population.

Landon won only two states. Roosevelt won 46. The actual vote was 61 percent Roosevelt, 39 percent Landon. A landslide. The era of scientific public opinion polls was born. And two years later, the Literary Digest went out of existence, discredited by is unscientific poll. How could the 3,000 people in the Gallup poll more accurately reflect American voters than the 2-million Literary Digest respondents? Two words: random selection.

Say, God forbid, you have a swimming pool full of black and white marbles, evenly distributed. A statistical principle, which is formally known as the central limits theorem, says that if you take just a quart of marbles from the pool, that sample of a couple hundred marbles will accurately represent that whole pool full of marbles. If the quart is 75 percent white marbles, then chances are very good that the whole pool is 75 percent white marbles.

If you find that hard to believe, you should know that it's been proven by repeated statistical tests. Also, the Gallup Organization is thriving today, but the Literary Digest is not.

Even though the Literary Digest heard from 2-million Americans, the magazine's poll had a major flaw that prevented it from relying upon statistical laws. In random selection, everyone has an equal chance of being included in the poll. The Literary Digest mailed only to people who had a phone or a motor vehicle. In 1936, those who had a phone or car tended to be affluent, and during the Great Depression, they voted Republican.

One of the challenges facing pollsters in Iowa today is identifying who actually will caucus on Jan. 3. The Iowa Poll relies upon two tools. First, it randomly selects people from the registered voter list. Second, it only interviews people who say they will definitely or probably attend a caucus.

The poll also asks other questions that can be used to further test caucus turn-out:

- "Have you attended caucuses in the past or will this be your first caucus?"

- "The caucuses will be held on Jan. 3, just a couple days after New Year's - a time when some people take holiday vacations. Have you changed travel or vacation plans in order to be able to participate on caucus night?"

- "Jan. 3 is also the same night as the Orange Bowl. How tempting do you think it will be to stay home that night to see the football game?"

Surprisingly, a random sample begins to accurately represent the whole population with as few as 30 interviews. Pollsters conduct more interviews to increase the precision of their polls, increasing the chance the survey will closely mirror the larger population.

When I directed the Iowa Poll, one issue we didn't have to deal with was people who have only cell phones and no land lines. Today around 15 percent of households have only cell phones, and younger people are even more likely to have only a cell phone. Fortunately, a majority still has land lines and those who don't are similar to those who do.

Other growing obstacles to getting a representative sample are increased use of answering machines and call blocking, plus people's busy lives keeping them away from home when the pollster is calling. A couple tactics help reduce these impediments. Phone numbers are tried four to six times on different days at different times. Using the name of a well-known entity, like the Iowa Poll or the Gallup Poll, helps reduce the impact of call screening and respondent refusal.

Mike Cowles (remember him?) started the Iowa Poll in 1943 after he left the U.S. Office of War Information, the propaganda arm of the U.S. government during World War II. Cowles reportedly used polling to keep track of the public mood during the war.

The Iowa Poll is hardly just about politics. Back in 1989, the Iowa Poll determined that 11 percent of Iowans name their cars, 86 percent do not, 2 percent have no cars and 1 percent were not sure whether they named them or not. (The most common car names are Betsy and Betty.) In 1977, the Iowa Poll determined that only 5 percent of Iowans expected to go to hell, but 31 percent knew someone else who would go there. In 1978, when asked which of life's pleasures - food, drink, books, friends and sex - they would give up, sex was chosen most often.

Cowles started the Iowa Poll because he believed it would distinguish his newspaper and help boost advertising and circulation. He already had used Gallup's readership polls to increase the Register's circulation. One survey found that Iowans preferred to look at pictures rather than read stories, so he increased the number of pictures in the paper, and circulation went up. Another survey said that readers preferred comics over front- page news, and advertisers started to use the comics to appeal to consumers.

There is art as well as science in polling. Back in 1984, I entrusted my reputation and that of the newspaper on a sample of 64 randomly chosen Iowans in predicting the caucus results.

In 1984, the Iowa caucuses were still ascending in prominence and our polling procedures were ramping up too. The 1972 caucuses were the first to influence presidential politics when George McGovern performed well and beat expectations. Jimmy Carter noted McGovern's success and campaigned extensively in Iowa before the 1976 caucuses.

In 1984, former Vice President Walter Mondale was the Democratic front-runner, with former astronaut John Glenn striving to be ahead of the pack. Interest was minimal in the Republican caucuses since the popular incumbent, Ronald Reagan, was running for re-election.

At the time, few public opinion polls were conducted in Iowa besides the Iowa Poll. So far this year, 50 polls have been published on the Iowa caucuses. Among the polling organizations: CBS-New York Times, ABC-Washington Post, American Research Group, Zogby, the University of Iowa and the Los Angeles Times-Bloomberg.

In 1984, we at the Iowa Poll had been increasing the frequency of our polls and moving the timing of the last poll closer to caucus day. We also were asking more questions to identify likely caucus attendees.

We conducted several hundred telephone interviews with Iowa voters before the caucuses, but in analyzing the data, I felt a group of 64 survey participants were most likely to attend. They said John Glenn was losing ground.

Glenn was highly critical of that survey. How could 64 interviews represent Iowa caucusgoers, he asked. Heck, that's not even one voter from each of Iowa's 99 counties, he pointed out.

On caucus day, Mondale won, Colorado Sen. Gary Hart was second and Glenn was in fifth place, just ahead of former Florida Gov. Reubin Askew. Statistical principles, thank goodness, had won again.

A PRIMER ON POLLS

With the Iowa caucuses closing in, here are a few tips in how to read the polls.

- Location, location, location. Forget the national polls right now. Pay more attention to the polls where the electorate will speak first, Iowa and New Hampshire. The outcome in those two states will most likely change the national polls anyway.

- For signs of change, see how the candidates are doing in non-horse race questions. Who is most likable? Who is most trustworthy?

- CHECK THE ORDER: Does the "who would you vote for" question come at the start or the end of the survey? It's best to have it at the beginning because the response is not influenced by other questions in the poll, which may change from survey to survey.

- TRENDS: Who is rising or falling consistently in several polls?

- WHO'S BEING POLLED: A recent USA Today/Gallup poll had former New York Mayor Rudy Giuliani at 25 percent and Mike Huckabee at 16 percent; the Iowa Poll taken around the same time had the opposite (Giuliani 13 percent, Huckabee 29 percent). The USA Today/Gallup poll is national, the Iowa Poll includes only Iowans.

- POLL SPONSORS: Give more weight to polls done by organizations without a political interest to exploit, like news organizations and universities.

[Last modified December 10, 2007, 07:51:00]