Every day, newspapers bombard us with opinion polls on "what the French think." They are regularly criticized for their methodology and interpretation, especially during election campaigns. Yet surveys can be conducted with much tighter controls. This is what Insee and many other official-statistics bodies do.
Why conduct a survey?
---------------------------
A survey is a statistical operation that provides information about a given population by questioning only a sample of that population, generally selected at random. Probability theory can then be used to draw conclusions about what is likely to hold in the population, given what was observed in the sample. Surveys—or survey sampling—belong to inferential statistics, so called because a result about the population as a whole is inferred from knowledge of only part of it.
A survey is conducted when questioning the entire population is impossible: it would take too much time or cost too much, or precise knowledge of the information sought is unnecessary and a good approximation obtained at lower cost will suffice. Insee's Labour Force Survey, for example, measures the number of people in employment and unemployment, but also establishes a breakdown of types of activity, average working hours, and so on. There is no need to question every month all 41 million people aged 15 to 64 who live in ordinary housing in France to obtain a good estimate of the employment and unemployment rates. Besides, there simply would not be enough time! So "only" around 90,000 people are surveyed each quarter, already providing a very high degree of precision.
