By Christine Jeavans BBC News Online

But this has led to accusations officials have "made people up" to complete the statistics.

Not according to the Office of National Statistics, which says it is so sure that 2% of the population is out there - just not on the census - that it has come up with details including age, sex and addresses for a million or so "individuals".

Statisticians "found" them by looking at the data gathered at the time of the census and working out how many people had been missed out.

Nobody is made up... these are the real people in the country who we have been unable to find using the census John Pullinger Census director

They then used advanced statistical techniques to estimate the personal details of each one.

But the number-crunchers strenuously deny that these are figments of the census compilers' imaginations.

"No census anywhere gets everyone to respond - and we haven't," said John Pullinger, Director of Social Statistics and Census.

But "nobody is made up," he insisted.

"These are the real people in the country who we have been unable to find using the census and the census survey."

Jigsaw

The census will be invaluable to local authorities, health services and others who need information to plan policies - and accuracy is essential.

Officials describe the 2% as like missing pieces in a jigsaw - if you can see all the pieces around a gap, you will have a good idea of what the bit the cat has eaten looks like.

Completing the picture

88% sent back forms 6% gave over forms on doorstep Basic info on a further 4% gathered by census takers 2% "imputed" after separate survey

They say many of the missing people are from demographic groups which are notoriously hard to count, especially young men living in inner London and other large cities.

Others include: houses split into several flats or bedsits; people who do not speak English as a first language; students in shared houses; babies; the very old; and people in the armed forces who move base frequently.

The census takers knew from the start that they were going to have problems getting everyone to return their details.

In an effort to simplify things, people were allowed to post their forms back - and 88% of households did so.

Then the 70,000 strong census team went out in search of those forms which had not been sent back and got hold of another 6% - bumping the total of forms received up to 94%.

But that still left another 6% of the population outstanding, many in the hard to count groups.

Risking fines

Some people simply didn't want to fill in the forms.

One census taker delivering forms in east London told BBC News Online several householders were suspicious of what they saw as the government checking up on them.

Hard to count

Young men in inner cities Students Babies - parents often forget to add them Very old Non-English speakers Armed forces personnel Unemployed people

Others told him they could not be bothered to complete the paperwork - despite risking a �1,000 fine.

"I'd say about 30% of my forms were still not sent in, even after I went back and checked twice," he said.

The census takers were told to use their common sense to fill in basic information on households where there was no response.

They judged for themselves how many flats a house had been split into and the likely number of occupants.

Cross-checking

At the same time as the census takers with their yellow satchels were patrolling the doorsteps, another team with a much lower profile was conducting a back-up survey.

This Census Coverage Survey was carried out across 330,000 people living in 16,000 different postcodes - focussing on those with a high proportion of hard to count households.

The results from the census and the survey were then cross-checked - and from this huge sample, researchers could work out how many people across the land had been missed in the census.

Nameless

Finally, the tricky part: using the knowledge of the kind of households and people who had not responded to the census, a panel of experts estimated the details of the missing million.

They then patched the "imputed" data about individuals onto addresses where no information had been gathered.

Others were added to existing households where it was suspected individuals may have been missed off the form.

And so back to the somewhat existential question of whether these people are real.

One thing is for certain, the Office of National Statistics has stopped short of giving them names - although we would be none the wiser until 2101 anyway.

For reasons of confidentiality all censuses are locked away for 100 years.