Scientific polls accurately reflect and describe public opinion. The key factor that dictates whether a poll is scientific or not is whether the group who is surveyed is selected randomly. In a randomly selected group, it is the surveyors, not the surveyed, who select potential respondents. Polls in which people choose to participate, such as Internet polls, are not scientific.

Randomness

In a scientific poll, the people who are polled are selected at random. Mark Blumenthal, writing at Pollster.com, notes that random essentially means that each member of the population has an equal chance of being selected. If the group selected is truly random, a small sample of 1,000 adults can reflect the attitudes of millions.

Margin of Error

Even if a poll is truly random, another poll conducted with the same method but different respondents can get a different result. According to Sheldon R. Gawiser, Ph.D. and G. Evans Witt, writing for The National Council on Public Polls, neither survey may be wrong. Every poll has a margin of error, which reflects the possible range of responses in randomly selected group. Margin of error decreases as the number of people who respond to a poll increases.

Other Factors

The way in which a question is worded can influence a response, though Mark Blumenthal notes that pollsters may disagree on what constitutes neutral questioning. The order in which questions are asked also influences responses as well, and so can the tone of the questioner. A national poll may be skewed if its respondents hail disproportionately from one region of the country.

Recent Issues

Cell phone users may not appear in a poll.
Cell phone users may not appear in a poll.

In the 1990s, pollsters had an easy time phoning households, Blumenthal notes, because 93 percent of homes had land-based telephones. Today, pollsters relying on databases of landline telephones report getting responses from a disproportionate number of older, white voters. Younger people tend to rely on cell phones. The lack of response from younger respondents may influence the accuracy of polls.

What Makes a Poll Unscientific?

An unscientific poll is one in which the respondents aren't selected randomly, or the questions are designed to elicit a particular result. For example, an Internet poll on a newspaper website, in which readers decide whether to participate, can be said only to reflect the responses of readers who care enough to vote, and not the general population. The answers may be skewed because only people who respond may be people who truly care about that issue. In some cases, pressure groups or special interests use unscientific polls and poorly worded poll questions to create the impression that their view is more popular than it is.