The current definitions classify people by occupation - so some 40 per cent of the population may be excluded. Among them are housewives, many of the retired, the long-term sick and the growing numbers not in employment.
Within these groups are some of society's most disadvantaged people, according to a report commissioned by the Office of Population Censuses and Surveys. Their life chances are known to be worse than even those in the lowest class of unskilled workers, yet they can be missed out of the social classification. One of the challenges will be how to include them.
In addition, the explosion in women's employment and the growth of contract, self-employed and part-time working have also contributed to the growing problems in classifying people, according to a study published today by the Economic and Social Research Council.
As a result, the social class definitions have become riddled with anomalies. A high powered secretary - the personal assistant to a company chairman, for example - is in the same social class as the most junior typist or a petrol pump attendant.
"We need a new social classification fit for the 21st century," said Professor David Rose of Essex University, who was academic convener of the report which recommended change.
The social class breakdown is crucial for scientific and medical research, and the allocation of Government money for a whole range of social issues - from the cash health authorities get, to housing grants and local authority allocations. It is also widely used to underpin the classifications used for market research and polling.
"If we get a better, more robust definition, and one which better explains the variations found between social classes, we should as a result get better policies," Professor Rose said yesterday.
The existing social classification was first devised to explain variations in death rates in 1911. It broadly took its present form in 1921. It was based on estimations of the skill needed to do particular jobs.
Professor Rose said: "It was devised when, in most circumstances, families consisted of a single, male breadwinner and, from that, it was much easier to put the whole family into its societal position. The changes in society have simply outpaced the assumptions on which the social classification was first based.
"There are big problems about how you classify people who have never had a job, including, for example, lone parents, in a system which is essentially occupationally based."
In addition, wide numbers of jobs in which people in the past were almost always employees could now be held by the self-employed. Any new system needs to allow for the self-employed to be extracted and examined separately.
The next stage of the study aims to produce a new classification in time for the 2001 census. It seeks to combine into one measure the existing social class definition and the 18 category socio-economic group definitions that the OPCS currently uses. It is likely to have a wide range of categories which can be "collapsed" in to five or six main groupings for ease of use in compiling the data.
How radically the system can be changed may be limited by the need to ensure the new system can be "bridged" back to the old one, Professor Rose said - otherwise decades of long-term studies on health, social mobility and other areas would become worthless.
However, the system has not been static and classifications have changed over the years to reflect changing jobs. Airline pilots, for example, originally came in at social class III, but moved up to social class I. "But there have been no clear rules about how people should be allocated," said Professor Rose.
Other problems result from history. There are three socio-economic groupings for farmers and farm workers who now make up only 2 per cent of the population, but only one for junior non-manual workers who make up 20 per cent.
The new classification will remain occupationally linked as an individual's position in the labour market remains the best determinant of their life chances.
But once the new definition is agreed, the OPCS may commission work on a separate non-occupational classification, based on the extent to which people "control" their own lives, thus avoiding the stress that produces ill health - measures which could include income, educational qualifications, possessions, total household income and the number of earners. That could mean a question on household income will be included in the next census.Reuse content