polling

Polling is the process of systematically collecting information about the opinions, attributes, or behaviors of a defined population by surveying a sample. It is used to gauge public opinion, market trends, and customer satisfaction, among other purposes. The central goal is to infer characteristics of the population from the sample data.

Polls rely on probabilistic sampling where every member of the population has a known, nonzero chance of selection. Common approaches include simple random sampling, stratified sampling, and multi-stage or cluster sampling. Non-probability samples, such as online opt-in panels, are also used but carry greater risk of bias. Data collection can be conducted by telephone, online surveys, mail, or in-person interviews.

Key concepts include the margin of error, which expresses the precision of survey estimates. A typical national poll aims for a 95% confidence level with about ±3 percentage points as a common benchmark, though actual margins depend on sample size and population variability. Increasing sample size reduces the margin of error according to the square root law, while weighting adjusts for demographics to reflect the population.

Process wise, polls involve questionnaire design, fieldwork, data cleaning, weighting, and analysis. They may employ tracking polls to monitor changes over time or exit polls to estimate voting outcomes on Election Day. Question wording, order, mode effects, and nonresponse can influence results.

Limitations include sampling error and biases such as nonresponse bias, social desirability bias, and coverage error. Results should be interpreted as snapshots with stated margins of error and caveats about population definition and timing. Modern polling developed in the 20th century with organizations like Gallup and Pew Research, which refined scientific methods for measuring public opinion.