Polls built from straw?

parisdailyphoto

Each time the temptation arises to be sniffy about opinion polls, I take care to remind myself that in a past life, I had regular dealings with pollsters.

Once a month, Anthony King, gentleman, scholar and highly entertaining company, would call at the Telegraph offices in Canary Wharf with a couple of executives from YouGov to discuss future polls.

Guests and assorted hacks alike would be offered champagne on arrival for lunch, and good wine was then served in reasonable though hardly excessive quantities throughout the convivial meal that invariably followed. Ideas would be tossed around until some firm plan evolved.

In an austere new world, a stop was soon put to the champagne and wine, of course, though I am glad to say that Britain and I had parted company by then.

The conviviality, I am sure, survived the prissy change and the lunches will doubtless have remained as businesslike and productive as before, though I am prepared to bet they are not more so.

Professor King - or better, Tony - takes a highly methodical approach to his preparation of polls, and the analysis of their findings.

Viewers will know all about that if they have tuned in on election night and seen him assessing the exit polls and early declarations from places like Sunderland South (honest - they're always first or nearly first) and Billericay.

With his photographic memory and grasp of detail, Tony is a sort of Canadian Leslie Welch of politics and psephology.

He was a great fan of YouGov, which conducts its surveys via the internet.

I was always a little less convinced by this method of monitoring public opinion; the cyber age was not at that point so firmly established as it is now and whole swathes of voters were necessarily excluded.

When I suggested a poll of British Muslim attitudes on such issues as citizenship, loyalty and terrorism in the aftermath to Sept 11 and military action in Afghanistan, the problems of creating a viable sample very nearly proved fatal to the exercise.

In the event, the YouGov findings - though based on a sample as low as it could get before the level at which Tony would have advised against going ahead - mirrored those of subsequent polls on the same subject carried out by more conventional means and with many more respondents.

Pollsters know the science of their métier, and will usually argue that it gives a very reasonable prospect of an accurate reflection of public opinion.

But in the case of my Muslim poll, I felt very strongly at the time that the most valuable part of the project was the work of two reporters of Asian background who were sent off to canvass views in towns with large Muslim populations.

One of them turned in a weighty tome running to thousands of words which I had to whittle down to a few hundred. Useful as his research was in shaping our coverage, I remember hoping that he would find a home for an uncut version of his extraordinarily detailed epic.

My doubts about polls do not end with the internet since there are other ways in which the process of choosing a sample is flawed, a view reinforced by a small article seen in Le Canard Enchainé on the train from Toulon to Paris (a visit that may explain my silence here since Monday).

If the old Chained Duck is to be believed, and often it can, French polling institutions routinely exclude 31 per cent of the French from their telephone surveys, for the simple reason that those taking part are selected from listed, fixed-line subscribers.

Yet there is an army of people out there that has no phones of any kind, makes all or virtually all its calls on mobiles or uses internet phone services such as Skype. Young voters and people on modest incomes are said to be particularly likely to fall into one of these categories.

Add to that the obvious fact that not everyone likes being bothered by cold callers - and as many as 30 per cent refuse to take part - and you begin to see the fault lines developing.

The margin of error surely gets wider if you take account of such drawbacks, and there are at least three other complications affecting the French presidential elections:
* large numbers of apparently undecided voters
* the emergence of a strong third contender (Le Pen in 2002, Bayrou this time)
* the exclusion of French citizens - 2.3 million of them - living in the Dom-Toms, overseas departments and territories.

Does it matter? Well, yes it does. I see weight to the argument that polls have an ability to shape - in other words distort - the public view, or to encourage tactical voting that might not otherwise have occurred.

It is just as well, as I said in a piece at the Guardian Comment Is Free site, that France bans opinion polls in the last week of campaigning.

But, repeating from that item what I take to be the logical follow-up question, would it not make the democratic process a shade more democratic if we were to impose a much longer ban?

We are within six weeks of the first round of the presidential elections, and François Bayrou's success as Third Way candidate is inspiring all sorts of mathematical as well as political calculation.

How much healthier it would seem if we knew there wouldn't be another poll from a week on Saturday until the real votes were cast on April 22.