Polls don't always reflect election reality

By: Matt Stevens

Posted: 11/3/08

Anyone who has ever followed a political campaign, regardless of whether it was at the national, state, or local level, is familiar with polls. Every election season, people hear about which way polls show the election going. Over the course of the last several weeks, voters have heard that Sen. Barack Obama, according to polls, is on track to win the presidency by a comfortable margin. Given how many people are expressing their disapproval - in both conversations and in signs and bumper stickers - of the current Bush administration, and the fact that professional pollsters have gone out of their way to perfect their tactics, these poll results may very well turn out to be accurate. However, ever since the birth of scientific polling, pre-election polls and Election Day exit polls have, on occasion, wrongly predicted which candidate would win the election. This should come as no surprise, as there are so many different factors that pollsters must take into account when conducting a poll. If pollsters do not account for just one factor, voters are likely to be in for a big surprise on Election Day.

To begin, when pollsters ask voters questions, they must make sure that they give people choices if they want to get a truly accurate picture. For example, when a USA Today/Gallup Poll asked respondents if they thought that Bill Clinton was a good or bad president, 71 percent answered that he was a good president. Yet, when USA Today and Gallup asked people if they thought that Clinton was an outstanding, above average, average, below average, or poor president, only 49 percent ranked Clinton outstanding or even above average.

Pollsters have to make sure that they do not ask questions that present a certain candidate or point of view in a more favorable light than the other. For example, a pollster's asking voters if they see Barack Obama as "a strong leader capable of leading the U.S. to new heights," while also asking them if they see John McCain as a "strong leader," would almost certainly get an overwhelmingly pro-Obama response. Pollsters must make sure that their samples (people that they are interviewing) are representative of the entire voting population in order to get an accurate result.

A pollster whose interviewees were practically all White upper-class suburbanites would almost certainly be predicting an overwhelming McCain victory. Another common polling mistake is too much overreliance on telephones to reach people. For starters, only people with regular telephones are reached for interviews. People who only use cell phones are not going to be included in the survey. And even if a household is randomly called, chances are that the sample will wind up being biased in favor of Democrats because, generally speaking, women tend to do more telephone answering than men do, and women tend to be more liberal and inclined toward Democrats. Lastly, if pollsters conducting the infamous Election Day exit polls want to make a truly accurate prediction, they must make sure that they sample diverse polling places that will be representative of the voting population. If a hypothetical pollster only went to polling places in rural areas, he or she would almost certainly be predicting a McCain victory, for example.

George Gallup is considered the father of the modern political poll. Gallup was the first person to poll using scientific methods. Gallup came to prominence in 1936, when his methods correctly predicted a landslide victory for President Franklin Roosevelt. Gallup's use of scientific methods in polling all but discredited the old methods used for polling, like the one used by the magazine Literary Digest that same year, which consisted of sending out questionnaires to automobile owners. This poll projected that Kansas Gov. Alf Landon would easily defeat the president! Its flaw? By sending the questionnaire only to automobile owners in the middle of the Great Depression, wealthy Americans were overrepresented in the sample. And since wealthy Americans tend to be Republicans, it is no wonder the Literary Digest poll got the result that it did. Despite Gallup's revolutionary polling techniques, mistakes are still made by pollsters, which lead to inaccurate predictions about elections.

In 1948, many voters were unhappy with the administration of President Harry Truman. Since the Republicans had gained control of Congress two years before and polls were showing wide support for the Republican presidential nominee, New York Gov. Thomas Dewey, most people were expecting Truman to lose the election. Truman went on to win, of course. So why were those polls so wrong? The pollsters had relied too much on the telephone to reach people for interviews. In 1948, telephones were still something of a luxury that were mostly possessed by wealthy people, which led to such overwhelming support for Dewey in the polls.

In the early evening of Election Night 2004, when a close election had been predicted, exit polls projected that Sen. John Kerry would comfortably defeat President George W. Bush. Yet, by the next morning, the actual returns showed that Bush would comfortably defeat Kerry. Why were the exit polls so wrong? According to Warren Mitofsky, a veteran pollster, "interviewers assigned to talk to voters as they left the polls appeared to be slightly more inclined to seek out Kerry voters than Bush voters."

So what is the point of mentioning all the ways in which polls may be incorrect? As promising as the election looks for Sen. Obama right now, Americans must remember that polls have been wrong before and there is always the chance that they will be wrong again. With this in mind, voters should decide for whom to vote based on who they think will make the better president rather than who looks like the winner - because history just may repeat itself.