Why do researchers study a sample of a thousand voters rather than every voter in a country? The obvious reason is practical: surveying an entire population, called a census, is often too slow, too costly, or simply impossible. There is also a deeper reason. A well-chosen sample, if carefully constructed, can tell us almost as much as a full census, at a fraction of the cost.
Consider a cook tasting a pot of soup. She need not drink the whole pot to judge whether it needs more salt; one spoonful, provided the soup has been stirred, reveals the character of the whole. This is the essence of a representative sample: a subset that mirrors the population it is drawn from. A sample drawn only from wealthy neighbourhoods, or only from the loudest opinions, has not stirred the pot; it has merely tasted what floats on top.