ORIGINAL FRENCH ARTICLE: En finir avec le mésusage des sondages !
by Jean Paul Moatti*
Translated Monday 26 March 2007, by
The French presidential campaign has been marked by a “serenade of election surveys” to such a degree that one can legitimately wonder whether such procedures do not threaten our democratic ideals. What is going on?
The French presidential campaign has been marked by a “serenade of election surveys” to such a degree that one can legitimately wonder whether such actions do not threaten our democratic ideals. What is going on?
Surveys are effectively statistical devices which poll a limited sample—generally a thousand registered voters for pre-election surveys—in order to extrapolate from their responses the opinions of the general population. Under a certain number of rigorous technical conditions (the sample must be selected “randomly,” the characteristics of those who refused to participate should not differ too much from those who agreed, etc.) it is indeed possible to make reliable estimates from surveys.
The only snag is that the media presents the results of political surveys as if they provided a definitive outcome; while in reality, they are always estimates that include a substantial margin of error. This margin of error becomes all the more significant when the sample size is small and when the percentage of, for example, those who have the intention to vote for a certain candidate approaches 50%.
When the latest survey reports that Ségolène Royal will outdo Sarkozy in the first round of elections, readers or television viewers take that as a “true” statement, since the 29% credited for the first round is higher than the 28% allotted in the second round. Yet that is not at all what the survey suggests. Rather, for a sample of 1,000 people questioned, given a statistically acceptable margin of error, the percent of those planning to vote for the two candidates are acknowledged to be between 26% and 32% for one and 25% and 31% for the other. In other words, the survey cannot settle once and for all the question of who actually outperforms the other.
As for candidates who earn between 2% and 8% of the vote, a sample of 1,000 people is insufficient to predict the actual voting patterns of the entire voter population. A way of limiting abusive survey interpretations would be to force the surveyors to present their results for what they are: very rough estimates. More than thirty years ago in an article for the monthly review Les temps modernes provocatively entitled “Public opinion does not exist,” the sociologist Pierre Bourdieu exposed the techniques surveyors often unknowingly use to “manufacture” opinions. The selection and formulation of survey questions tended to reinforce the status quo and dominant elites. Analyzing the French presidential election of 1995, another sociologist Philippe Champagne wrote that pre-election surveys, especially those like today’s which may affect the outcomes of the real elections, “are not very reliable owing to the fact that they do not collect dynamic opinions, but rather affirmations which are mainly provoked by the survey itself.”
These criticisms are all the more justified given that it is impossible to distinguish when a pre-election survey produces reasonable forecasts or not. The rate of undecided voters, generally very high at the beginning of election campaign (currently around 50%), is more significant than the distribution of votes among those who have already chosen their candidate. To present the results of these surveys as if they were effectively real votes is to uphold (intentionally?) a double error. On the one hand, this assumes rather whimsically that undecided votes will be distributed proportionately to those votes already clearly declared. In addition, this calculation method overvalues the figures from those who have declared their intentions to vote a certain way. To say that such candidate maintains 17% of the vote it is to give an inaccurate representation reality; the true score is in fact only (with a half of undecided) 8,5%.
To prevent the “strategic” behavior to suit the media’s interests, as well as presidential candidates who seek to build their success upon survey results, rather than the issues debated, there is no alternative for the anti-neoliberal left except to confront them head-on with grassroots awareness and recourse to other less-controlled media (such as the Internet). It would nevertheless be naive to think that the current distribution of votes between the three candidates does not pose a problem. In the second round the cumulative total of votes for Marie-George Buffet, Jose Bove, and Olivier Besancenot would add up to around 6% to 12%, percentages comparable with that of anti-neoliberals in Italy or Germany. If the media is allowed to “manufacture” opinion by discrediting anti-neoliberal candidates with few anticipated votes, certain candidates will get less air time and editorial space, diminishing the quality of the debate on economic, social, and political alternatives that the “non” to Europe and the struggles of these past few years has allowed to develop in France. This is a fundamental denial of the democratic process even more serious than the problem of certain candidates’ difficulties in obtaining the 500 signatures needed to make a candidacy official.
Jean-Paul Moatti is a professor and researcher at the Université de la Méditerranée in Marseille, specializing in health-care issues. He is also a member of Marie-Georges Buffet’s National Campaign Council