French poll findings can no longer be published before the final round of voting in France's legislative elections this Sunday, but the polling institutes' resident political scientists are still working overtime.
Opinion polls, banned from publication one week before the first round of voting, are still being commissioned by political parties hungry for clues to their prospects. Academic consultants are gearing up to produce estimates which will be run by television and radio from 8pm on the dot on Sunday.
Although no fewer than six major polling institutes compete in the tricky field of election forecasts - an indication of the insatiable appetite for polls in France - their experts all hail from the same place, the Centre d'Etudes de la Vie Politique Francaise.
The centre is run jointly by the Fondation Nationale des Sciences Politiques and the Centre National de la Recherche Scientifique. A dozen of the centre's 40 researchers work on opinion polls and the big six institutes all have consultants at the centre.
Political scientist Daniel Boy, who works for the Sofras Institute and will do its Sunday-night estimates for the high-profile TFI television channel, acknowledges the set-up is somewhat incestuous.
"We are all from the same milieu, researchers and pollsters. Everyone knows everyone. The centre dominates the field and really has no competition inside France," he said.
Using academics from the same centre does not mean the institutes come up with the same forecasts. Typically, two institutes which do come up with the same percentage of support for one party will then forecast widely differing gains or losses due to the use of different weighting systems.
The opinion polls came in for a lot of criticism when they yo-yoed erratically during the campaign, first giving the right a firm lead, then placing the left neck-and-neck before putting the right ahead again a few days later.
Mr Boy and colleague Jean Chiche explained to Liberation that if the media had looked at the findings for shifts in voting intentions rather than forecasts for losses or gains of seats, then "there is no more shift to the right today than there was a shift to the left a week ago".
The experts have to account for vote transfers between first and second rounds, and the likelihood of two-way or three-way run-offs.
Even with the more straightforward presidential election in 1995, they got it memorably wrong, giving Jacques Chirac a three to seven point lead in the first round which in the event he lost to Lionel Jospin by 2.5 points.
In the face of such results, criticism is sharp.
"The institutes try to translate something which cannot be translated. They hand out horoscopes in the name of science," said sociologist Patrick Champagne at the Ecole des Hautes Etudes en Sciences Sociales.
Mr Boy defends his field stoutly. "Ours is one of the rare areas of the social sciences whose findings are submitted to the acid test of truth. It is because our work is measured against outcomes that it goes on improving. We are under an enormous obligation to increase the quality of our polls," he argued.
According to Mr Boy, the main difference between British and French polling lies in media standards. "I've seen a very precise analysis of poll methodology in the British press, but never here. The more intellectual papers don't ask us to analyse the 1995 failure, they just write acid little articles.
"There is an intellectual laziness in our media, which are not very professional. French journalists call you with a simple question it would have taken them ten minutes to find out in a library and have no idea of asking more complex questions."
But what effect does it have to belong to a single research centre which has no competition, works for all the leading polling institutes and gets no challenging questions from the media?
"Even if we didn't call ourselves into question enough, we are attacked from outside every day," maintains Mr Boy. "Also, the centre is not uniform, there are differences of approach, the centre includes partisans of qualitative analysis who challenge our quantitative approach."
"Our centre has signed a contract with Liberation to write on the outcome, analysing who voted for whom. It's never happened before and should give the centre a higher media profile," he added.
said Mr Boy.
"The Pierre Bourdieu school of sociologists will always be hostile. They throw mud at us but have nothing to say themselves about the vote. Elections fundamentally do not interest them, yet they will say we understand nothing about elections," he commented.