Public opinion consists of the views held by the population of a state that influence those in power. In a democratic state, politicians must listen to public opinion if they wish to keep their jobs. Dissatisfied constituents can vote out those who ignore their views. But regimes with other types of governments also need to pay attention to public opinion. If the public overwhelmingly opposes the government, the regime could be in serious danger of revolution or collapse.
We learn about public opinion through polling, which asks people their views and then compiles the results. Politicians and pundits in many countries rely on public opinion polls, and the media frequently reports on polls. Sampling a subset of the population allows pollsters, or the people who create and take the polls, to get a sense of overarching concerns and interests within a large population. Rather than polling every citizen, an expensive and time-consuming process, polls use samples. Pollsters hope that the opinions of the sample accurately reflect the population as a whole. Just as one does not need to taste every bite of stew to know that it needs more salt, one need not poll every person to learn public opinion.
To make sure that their poll results are accurate, pollsters seek good samples. The most obvious way to get a good sample is to include lots of people. But including more people does not guarantee that the poll will be accurate. Instead, a sample must be representative—that is, the sample must have the same basic characteristics as the population. If the population has a 15 percent poverty rate, for example, the sample should have a roughly equal portion of poor people. Pollsters have a number of techniques to ensure a representative sample, and they rely on statistical methods to measure the probability that a poll is accurate.
Pollsters rely heavily on probability and randomness to increase the chance of getting a good sample. In a probability sample, each person in the population has a known chance of being chosen as part of the sample. When pollsters assign each person an equal chance of being selected, they are using random selection.
Sampling error results from bad samples. A poll that falls prey to sampling error will inaccurately measure public opinion. A common source of sampling error is a skewed sample, one that does not match the population. Some popular types of polling—asking people as they walk down the street, for example, or online polls—produce very skewed samples and are therefore unreliable.