For around forty years, ahead of every election, a host of opinion polls—sometimes contradictory—have told us what the result would be if we voted that very day. Other surveys are conducted too, particularly for commercial purposes. But how reliable are they?
-
A little history
-----------------
George Gallup is credited with pioneering opinion polling. This American from Iowa set out to predict the winner of the 1936 US presidential election. Working from a sample of 50,000 people, he forecast that the incumbent president, Franklin Roosevelt, would win with 56% of the vote, contradicting the Literary Digest, which predicted victory for the Republican candidate Landon on the strength of a questionnaire that had received 2.4 million responses. The Democratic candidate's landslide victory with 61% of the vote showed for the first time that a well-chosen sample was more reliable than one forty times larger but affected by an obvious bias—in this case, non-response bias.
The French sociologist Jean Stoetzel met Gallup in 1937 and recognized the value of such surveys for better understanding individual behavior. On his return, he founded the Institut français d'opinion publique (IFOP) and immediately polled the French public with the question Do you approve of the Munich Agreement? Only 57% of respondents answered yes, although the agreement secured an 87% majority in the Chamber of Deputies. The IFOP would rise to prominence during the 1965 presidential election by predicting that General de Gaulle would be forced into a runoff against François Mitterrand.
