On a sample of 300 to 500 people, this margin is logically wider. A difference of one or two points from one poll to another then loses its relevance, because it remains within the uncertainty interval.
Six months before the presidential election, the polls already published can no longer be counted. Nicolas Hubé, a professor of information and communication sciences, urges readers to approach them with distance rather than take them as verdicts.
This proliferation of figures and curves, as the vote approaches, complicates interpretation. A few reflexes make it possible to step back from this flood of results.
This margin always appears in the polling institutes' original studies. It appears far less often in the press articles that relay this work.
The sample size, for its part, must be specified in the institutes' official documents. In principle, it appears in the articles.
The ranges can thus overlap. A candidate placed"between 12 and 15 points"and a rival"between 10 and 14":"it is much less simple to explain,"says Nicolas Hubé. The margin of error is the first of them. It remains"fairly large,"on the order of 3 to 4 points depending on the number of people surveyed, Nicolas Hubé notes.