Forty per cent of British women who go on a holiday to Spain have sex with a stranger within the first five hours of their arrival. Believe this? Then you are ready to follow this week's local election results.

Nearly one quarter of young drug users have smoked cannabis together with their parents. Convinced? Then you are missing your vocation. A job awaits you as a political reporter.

The cod survey is ubiquitous. Britain's favourite flavour of yoghurt as determined by 132 respondents on the Yoghurt Council website; the ten top musical acts of the last millennium (winner Robbie Williams, runner-up Mozart) as nominated by listeners to LBC; the news that 38 per cent of people prefer their vacuum cleaner to their spouse (according to a survey of vacuum cleaner retailers).

Most of these start their life on the PC of a hopeful PR executive and end it in the news in brief column of a free circulation newspaper. But imagine if a cod survey were to dominate political debate for a week, become the lead item on the news, create a massive storm that swirls around the feet of the Prime Minister. Well, imagine no longer because that is precisely what is about to happen.

At the end of this week, millions of voters will go to the polls and elect thousands of local government representatives. And the results will matter to those who live in the areas involved. But they will also be reported as providing a useful guide to the state of national opinion, a verdict on the Prime Minister and a glimpse of the result of the next general election. Such reporting is a preposterous error.

Ask a politician to respond to an opinion poll finding and chances are you will get back this cliché: “I don't believe in polls, I believe in real votes in real ballot boxes.” Michael Heseltine even gave this answer once when he was asked about the views of Polish people. Journalists pretend to be cynical when they hear this cliché. But really it represents our view too. Surely real voters going out to cast their ballot in proper elections are giving us a clearer view of the political picture than opinion polls?

No, they aren't.

Let me start with the question. One of the hottest topics in economics at the moment concerns the way choices are framed. Recently the Nobel prize was awarded on the back of research showing how people make different decisions depending on the way, and even the order, in which alternatives are put to them.

And any pollster will tell you that tiny changes in wording can make a big difference to the outcome of a survey. If, for instance, you remind people that the Liberal Democrats exist before asking about voting intention, the number who say they are Liberal Democrats increases.

But here we are not talking about tiny differences. We're talking about trying to find out who you want to be prime minister by asking who you want, say, to be mayor of london. It is hardly surprising if this does not prove very enlightening.

Next, there is the nature of the sample. This week's local elections do not take place in every part of the country, so one can only guess at what might have happened in the places that are not going to the polls. But there is an even more serious problem with the sample - it is self-selecting group of enthusiasts.

The surveys that I mentioned at the beginning of this article - the holiday sex poll and the cannabis poll - were real examples. They came from magazines that asked their readers to write in with their drug and holiday experiences. Naturally respondents were people happy to share details of their love life and drug use and sufficiently motivated to put pen to paper. Not exactly representative.

Those who vote in local elections are a similar bunch. Not sex addicts and drug users, but unusually enthusiastic about their own peculiar vice - picking their local councillor. Angry protesters and political junkies form a much bigger proportion of the local vote than they do in a general election. The sample in local elections seems so much better than that provided by pollsters, because it is so much larger. But unfortunately it isn't properly balanced.

And finally - a minor point this, but worth noting - there is the fact that not every seat is contested by every party. Quite often voters will go to the polls to find that there is no Labour candidate, say.

None of these is merely a theoretical point. The proportion of votes cast in local elections and those cast in general elections are wildly different. But there are analysts who will play with the figures and claim they can tell you what it all means. They will dominate the coverage next weekend and the Prime Minister's future may be depend on what they had to say.

For them I have two questions. The first is - how? How can you really adjust properly for the different question voters were asked and for the unrepresentative enthusiasm of the sample?

The other question is - why? Why would you bother? Who would construct a poll by, to start with, asking millions of people in only some parts of the country to go down to their local school if they want to and opine on an entirely different matter; and then, reverse engineer the results?

Surely you just would ask 1,000 or so voters in a representative sample what their general election voting intention is? You would use a poll, with all its imperfections.

But that would mean admitting that anything we will find out on Thursday, we already know more accurately now. And where's the fun in that?

daniel.finkelstein@thetimes.co.uk

Times Recommends

• We have kit and leadership. Let’s go
• Behind the scenes at the biggest deal
• Conferences — what a waste of money