Some working-age families in Britain experience combinations of disadvantage which mean that they are almost certain to have no work. But their poor prospects can be explained largely in terms of the cumulative effects of each of their specific disadvantages - the number of their problems does not seem to be an issue in its own right. Richard Berthoud of Essex University's Institute for Social and Economic Research (ISER) has undertaken a detailed analysis of the job-chances of more than half a million men and women. The study found:

• One-sixth of British adults aged 17 to 59 do not have either a job or a working partner. Those at high risk of non-employment are: men and women without partners (especially lone parents); disabled people; those with low qualifications and skills; those in their 50s; those living in areas of weak labour demand; and members of certain minority ethnic groups.
• Only 4 per cent of individuals with none of these disadvantages are non-employed. The more disadvantages, the greater the risk: more than 90 per cent of people with all six disadvantages are non-employed.
• Some specific combinations of two or three disadvantages carry a higher risk of non-employment than might have been expected; other combinations showed an unexpectedly low risk.
• Nearly one-tenth of adults have characteristics which increase their risk of non-employment to more than 50 per cent.
• The pattern of non-employment risks is not as complicated as some have argued. This analysis largely justifies the common assumption that variations in the risk of non-employment can on the whole be explained just by adding the effects of each disadvantage together. The study does not endorse the idea that disadvantages are exponential - with the risk of non-employment rising faster and faster as the number of disadvantages increases.
• This 'additive' pattern suggests that addressing the hindrances to employment associated with one kind of disadvantage will yield dividends without having to worry too much about its links with all possible other disadvantages.

This study is based on detailed analysis of 550,000 adults, collected from a nine-year sequence of Labour Force Surveys. The research focused on the characteristics associated with 'non-employment', defined as men and women who:

• are not working at least 16 hours per week, nor in full-time education; and
• do not have a working partner.

'Non-employment' is a broader term than 'unemployment', because it includes people (especially lone parents and disabled people) who are not seeking work and are therefore 'economically inactive'. Because the definition takes account of partners' working status, most non-employed families depend mainly on social security benefits, and a high proportion are poor.

17 per cent (around one-sixth) of British adults are without earnings, according to this definition. Only 4 per cent of those with none of the disadvantages described in Table 1 are non-employed.

Adding these detailed measures together provides quite an accurate analysis of the probability that any particular individual is non-employed.

• additive: the effects of each disadvantage can just be added together;
• combinations: specific combinations of disadvantage have effects which increase or decrease risk, compared with the additive hypothesis;
• independent: every combination of characteristics has its own pattern of risks, without regard for any other combination;
• exponential: the risk of non-employment rises faster and faster as the number of disadvantages increases;
• logarithmic: the risk of non-employment rises less and less rapidly as the number of disadvantages increases;
• class: having any of these disadvantages imposes a high risk of non-employment; extra disadvantages make no further difference.

Combinations of disadvantages
Specifying every possible combination of disadvantages - from single items, through pairs and triplets up to the combination of all six - as a distinct option revealed that the risk of non-employment associated with specific combinations of four, five or six disadvantages is not significantly different from what would be expected on the basis of their component parts. But eight of a possible 68 triplets, and 20 out of a possible 38 pairs, do have significant effects. To take two of the most important examples:

• Lone parents of Caribbean or African descent face a lower risk of non-employment (55 per cent) than would have been predicted on the basis of their family structure and ethnic group (68 per cent).
• Older Pakistanis and Bangladeshis with low qualifications and skills have an even higher risk of non-employment (82 per cent) than might have been expected from adding up the influences of those three characteristics (71 per cent).

In general, though, pairs and triplets have relatively little influence on the distribution of non-employment, compared with the separate influences of the six primary characteristics. Thus there is some support for the combinations hypothesis, but it is not as strong as the additive assumption.

Number of disadvantages
Two-thirds of adults in the age-range under analysis have at least one of the characteristics associated with disadvantage. Nearly a tenth have at least three. But only 1 in 5,000 (106 members of the sample) has a full set of six disadvantages. As might be expected, the more disadvantages facing any individual, the more likely s/he is to be non-employed. The range of divergent risks is surprisingly wide, though - from a risk of just 4 per cent among those with no disadvantages, to 91 per cent among those with six (see Figure 1). The simple additive model comes close to predicting these variations accurately, but there are some signs that the level of risk may be slightly lower than expected for people with multiple disadvantages. This latter finding provides weak support for the logarithmic hypothesis.

Cumulative disadvantage
Once the effects of combinations have been taken into account, the analysis is extremely effective at estimating the probability that any individual will be non-employed - at very high levels of risk as well as at the lower end of the distribution. Of course, most individuals have a low risk. But the study strikingly identified individuals with very high levels of risk - nearly one-tenth of the population have characteristics which give them a risk in excess of 50 per cent, including a small number with risks well into the 90s. These people's chances of having either a job or a working partner are close to zero.

This is a fairly straightforward conclusion. The pattern of non-employment risks is not as complicated as some have argued. This is convenient for analysts, whose common assumption of a straight additive model has been largely justified. It is also helpful to policy analysts, who can be reassured that addressing the hindrances to employment associated with one kind of disadvantage will yield dividends without having to worry too much about its links with all possible other disadvantages. Some specific combinations do require special attention though.

Perhaps the most striking finding of the research is the huge disparity in risks - between the 'typical' figure for non-disadvantaged individuals of about 4 per cent, through the 'average' figure for the population as a whole of 17 per cent, and on to the high levels of 50 or even 90 per cent. People with very high risks of non-employment probably spend long periods without earnings, and their difficulties cry out for policy initiatives. The positive news, though, is that high levels of risk are sensitive to changes in the economy, and this may imply that they are susceptible to changes of policy.