Weighting changes how much each completed response contributes to an estimate. Researchers may adjust for unequal selection probabilities or bring known characteristics closer to population benchmarks. Useful weights can reduce imbalance, but extreme weights make an estimate depend heavily on a few cases. Weighting improves observed data; it does not create answers from people who were never covered or never responded.

A cross-sectional survey observes a population at one time. Repeated cross-sections draw new samples over time to track population-level change. A panel returns to the same people, which can show individual change but also faces attrition as participants leave. When interpreting a trend, first ask whether the design follows the population, the same individuals, or a rolling window of recent interviews.

Video lesson: Science Behind the News: Opinion Polls & Random Sampling