If we look at how the indices relate to the variables used in this analysis, we find that the index scores naturally have their strongest correlations with the component variables used in their construction, and also with other variables which are very similar (e.g. different definitions of car ownership). Correlations of individual indicators with the indices range from 0.2189 to 0.8721, and from -0.3865 to -0.8753. Median values are 0.5712 and -0.7168. Excluding any direct measures of class from the analysis for the time being, the variables most highly correlated (at 0.6 or higher) are given in Table B.
|Cars per h/h||-0.7977||-0.7715||-0.6612||-0.8753|
|2/+ car h/h||-0.7464||-0.7168||-0.5952||-0.8435|
|No car h/h||0.8376||0.8192||0.7318||0.8721|
|Not owner occ||0.7564||0.8268||(0.5818)||(0.5712)|
|Long term sick||0.6015||(0.5687)||0.6319||(0.5115)|
|Children in low pay h/h||0.8486||0.7596||0.7878||0.7741|
|Lone parent h/h||0.7362||0.6902||(0.5766)||0.7132|
The first three variables in Table B are included as surrogates for low income (Townsend et al., 1992: p. 225). They are essentially three minor variations on one core indicator, and are therefore treated as a single element in some other parts of this paper. The second three are aspects of housing amenities, and the last three relate more to household circumstances. There is some overlap between 'children in low earnings households' (i.e. where neither parent is in paid work or if there is a single parent, she/he is unemployed or in part-time paid work) and 'lone parent households', where the parent may or may not be in employment. We would expect to find strong correlations within each of these sets of variables, as well as their association with the indices as here.
While the three sets of variables can be treated as separate, it is interesting to find that income (and literally possession or non-possession of private transport) and housing (private owner-occupation/local authority rental) both show so strongly. This would support the view that those with low income and few possessions depend on public transport and public housing. In concrete terms, they constitute a consumption category with a different life-style, which goes beyond just using car and house ownership as surrogates for income and wealth as in this analysis. Deprivation is not only a relative lack of things: the deprived are different. Being poor has extensive ramifications: to misquote, deprivation comes not as a single spy but in battalions. Poverty is a way of life.
It is worth noting that had we drawn the arbitrary cut-off (r = 0.6 or higher) at a lower level, the next most highly correlated variable to be identified would have been premature standardised mortality rate. This reinforces the importance of the long term sickness variable in Table B, and is a grim reminder of that most definitive of all deprivations.
Although we have chosen initially to look at the stronger correlations, there is an extensive spread of lower correlations of the order of 0.3 and 0.4 (which would be welcome in many other analyses). We cannot detail them here, but they are included in the factor analysis reported in a later section. As we noted above, it is not the extent of deprivation in itself that most concerns us, but rather understanding its patterns.
|Children in low-pay households||Lone-parent households|
|Children in low pay h/h||1.0000||0.8150|
|Lone parent household||0.8150||1.0000|
|No car h/h||0.7479||0.6339|
|Cars per h/h||-0.7229||-0.6225|
|2/+ car h/h||-0.6971||(-0.5891)|
|Long term sickness||0.6509||(0.4328)|
Apart from the indicators in Table C, only one other pair correlate at higher than 0.6: 'no car households', with 'children in unsuitable accommodation' (0.6121). Correlations between pairs range from 0.1815 to 0.9841 (the 'car' variables) and -0.1701 to -0.9737; median values are 0.4027 and -0.5016. We might therefore reasonably conclude that a major characteristic of deprivation is its location in two distinctive household types, those where there are children and neither parent has a full time job, or where there is a single parent. It is these two types that are most associated with low income (the 'car' variables) and poor housing. Although a case could be made that unemployment or single parent households are in part class-based, the data rather suggest a pattern of social inequality in which having a job, or gender (most single parent households are headed by women) are the stronger features.
This impression is of course based on an analysis that deliberately excluded any direct class variables (that is, always assuming that we are not simply playing word games by calling class factors something different as 'when, in still part-acceptable 'wet' terminology, class inequality is renamed innercity (sic) deprivation, say, or child poverty' [Westergaard, 1995: p. 121]). When direct class indicators are introduced, a somewhat different picture emerges, because two of the six class indicators used (see Annex) correlate strongly with the car/poverty indicators, and to a lesser extent with some of the other variables in the main set. Perhaps most significantly, it is not so much that low social class correlates with indicators of deprivation, but that high social class correlates negatively with deprivation. This indicates the importance of remembering that class is a system of differences, in which advantages and disadvantages only make sense when taken together.
Table D shows first that the four indices do correlate with our indicators of social class. This is most clearly seen for men in the upper SEGs. However, the indices are, as expected, measuring more than just occupational class so that class does not produce the highest correlation values in the table. Whereas the upper SEGs (overall and for males) correlate at over 0.6 with more of these separate indicators of deprivation, SEG 11 overall and for males typically correlates at between 0.4 and 0.5, the exception being the correlations with the Jarman and LWT indices which include class of head of household in their construction.
For females the scores are between 0.3 and 0.4. It is noticeable, however, that the female upper SEGs do not correlate so strongly with absence of deprivation: on the items in Table 5 and generally, their correlations tend to be around 0.3 to 0.5. This may be relevant to current debates about the interaction of gender and class, to which we will return in our final section.
|All SEG 1-5, 13,14||Male SEG 1-5, 13,14||Female SEG 1-5, 13,14||All SEG 11||Male SEG 11||Female SEG 11|
|2/+ cars per h/h||0.6918||0.7158||(0.4380)||(-0.4863)||(-0.4678)||(-0.3092)|
|Cars per h/h||0.6749||0.7054||(0.4150)||(-0.4974)||(-0.4899)||(-0.3072)|
|No car h/h||-0.6223||-0.6577||(-0.3718)||(0.4895)||(0.4953)||(0.2939)|
|Lone parent h/h||(-0.5953)||-0.6186||(-0.3834)||(0.4895)||(0.4296)||(0.3493)|
|Children in low pay h/h||(-0.5812)||-0.6108||(-0.3709)||(0.5079)||(0.4803)||(0.3439)|
Comparing Tables C and D, the correlations between the 'car'/income variables, class, and the highlighted variables of lone parent and children in low earnings households are very similar. The middle class indicators score slightly higher than the lone-parent households variable, but not quite as strongly with low-earnings households having children. Educational qualifications correlate strongly with the middle class but only moderately with the other deprivation indicators. One interesting feature of the less strong correlations, which are not fully reported here, is that the levels associated with our lower-class indicator have a fairly close mirror-image in the levels associated with our measure for the middle class. It seems reasonable to conclude that deprivation does contain a class element, although this new element is most strongly manifested as an absence of the middle classes from deprivation.
Copyright Sociological Research Online, 1996