These statements have about as much truth as a heterosexual man claiming not to notice an attractive woman in a miniskirt.

Politicians are obsessed with polls for the wholly logical reason that, despite their faults, they are the best device for measuring public sentiment.

For that reason, anybody else with the remotest interest in the election will follow the polls with quasi-religious devotion. Unfortunately, the processes of conducting and interpreting polls has become extremely complicated. Any judgment on what they are saying has to be predetermined by an arcane interest in methodology.

The main organisations during this campaign will be Populus (predominantly for The Times), ICM (chiefly serving The Guardian), YouGov (primarily acting for The Daily Telegraph), MORI (on behalf of the Financial Times) and NOP (probably mostly with The Independent). They all face common challenges but have settled on different ways to deal with them.

The first challenge is to get voters to talk at all. MORI still employs face-to-face interviews, most of the rest rely on telephone polling, while YouGov solicits a large pool of electors via the internet. It is a matter of long (and, frankly, boring) debate as to how much these different techniques matter. The use of the internet is a source of unusually bitter and intense argument.

What is commonly true, though, is that it is much harder than it used to be to persuade people to offer their opinions.

Worn down by charity “muggers” on the street and cold calls from various companies at home, citizens are more and more inclined to cross the road if they see a man or woman with a clipboard approaching or to slam down the telephone if they hear the words “Hi, I’m Michael from Populus and I wondered . . .” This raises the serious question as to whether the people who agree to take part in polls are inherently unrepresentative.

The second difficulty is how to decide whether those who say that they are going to vote will, in the end, trudge down to the polling station. This is incredibly awkward.

The numbers of “voters” who will openly admit that they have no intention of casting a ballot is comparatively small. Yet more than 40 per cent of the electorate did not show up at the last election. The various polling companies have resorted to separate techniques for assessing the probability of voters voting which, while they are largely similar, can alter the numbers by one or two points.

Finally, there is the fiendishly tricky business of past vote weighting. Among the questions that all pollsters ask their victims, (sorry, interviewees) is how they voted at the last election. This usually produces the answer that a larger proportion claim to have backed Labour in 2001 than the 41 per cent who did. The reasons for this are multiple and uniformly tedious. The consequence is that most pollsters have adopted diverse statistical methods for compensating for this and ironing out the bias.

Once again, while similar, they differ enough to matter at the margins. Peter Riddell and I, along with Andrew Cooper and Michael Simmonds of Populus, have spent dozens of hours discussing the finer points of optimal past-vote weighting — which proves that we are the four saddest people on the planet.

All of which means that those who look at polls have to be careful to avoid comparing apples with pears. As a rule, YouGov’s techniques tend to produce the smallest Labour lead, ICM a larger one and Populus a figure which is quite close to ICM but usually slightly smaller.

To look at an ICM poll on Tuesday with a Labour lead of, say, 9 points, and a YouGov poll on a Thursday with, for instance, a Labour lead of 5 points and to conclude that support for the Government had fallen in the course of 48 hours could be correct, but is probably mistaken. Populus polls should really be compared only with other Populus polls, and so on. Apart from that, polling is pretty simple.

Times Recommends

• Computer warriors pay for far-away mercenaries
• Divorce rate falls to 26-year low
• Boris adviser says Olympians tainted with blood