Tables 3.2, 3.3, 3.4 and 3.5 present the odds ratios from the logistic regression models. The modelling results for the different clusters for each independent variable are discussed below.
In accord with previous research, the non-adjusted odds ratios reveal a moderately elevated risk of material deprivation for women in all clusters. After adjusting for the effects of the other covariates, the statistically significant female disadvantage (1.12 times) persists only in the Eastern cluster. In the Northern and Western clusters, gender makes no significant difference to the odds of material deprivation. By contrast, the Southern cluster features an inversion of the gender gradient: according to the adjusted model, women are less deprived.
Our results indicate considerable diversity in the age pattern of material deprivation across clusters. With regard to the non-adjusted estimates, the Northern and Western clusters exhibit no cross-cutting change in material deprivation in relation to age. The observed pattern is curvilinear with an increase in the odds ratio from 50–64 year-olds to 65–79 year-olds followed by a decrease among the 80+ age group. As a result of these opposing shifts, the non-adjusted deprivation risks among the youngest and oldest age groups are similar. By contrast, the Southern and Eastern clusters feature a systematic age-related increase in material deprivation to markedly high levels. Similar to the findings for gender, adjustment for the effects of other covariates produces a substantial transformation of the pattern. In the Northern and Western clusters, adjustment leads to the emergence of an inverse relationship between age and deprivation. Among the 80+ age group, the odds of material deprivation are 0.59 and 0.64 times lower compared with 50–64 year-olds, respectively. However, the Eastern and Southern clusters show no statistically significant association in the adjusted model between advanced age and the odds of deprivation. Among the 65–79 age group, differences across welfare clusters are smaller.
The association between household context and material deprivation is strong and relatively uniform. In all clusters, living as a couple markedly reduces the odds of deprivation relative to living alone. In the adjusted model, the reduction appears largest in the Northern cluster (0.45 times) and smallest in the Eastern cluster (0.67 times). Interestingly, the contrast between living as a couple and living alone peaks in the Northern cluster. Considering that the welfare systems in the Nordic countries are the least familistic, one might have expected a different result.
In most clusters, couples living with others are also better protected against material deprivation than older individuals living alone. The only exception is the Southern cluster in which the presence of other family members in the household is associated with the same risk of material deprivation as experienced by those living in one-person households. Finally, the highest odds of material deprivation are found among single persons living with others. In the Western and Southern clusters, their risk of deprivation significantly exceeds that of the reference group. The similarity of the adjusted and non-adjusted estimates suggests that the relationship between living arrangements and material deprivation is relatively independent of the other factors considered in the analysis.
Number of Children
In most clusters, childlessness and having one child are associated with elevated risks of material deprivation relative to the reference group (individuals with two children) in the non-adjusted models. However, after adjustment, moderate excess risks persist only in the Eastern cluster, and to a limited extent in the Western cluster (only for those with one child). In the Southern cluster, childlessness is associated with lower odds of deprivation. However, having a large family distinctly increases the risks of material deprivation in all clusters. In the Western, Southern and Eastern clusters having five or more children is associated with a 1.95 to 2.21 increase in the adjusted odds of deprivation. Only in the Northern cluster does the excess risk appear somewhat smaller, plausibly reflecting the capability of Nordic welfare systems to bolster economic inequalities arising from family circumstances.
In all clusters, individuals with a medium or low education exhibit substantially higher risks of deprivation compared with those with high education. In the adjusted models, the odds ratio of deprivation ranges from 1.23 to 1.76 for medium-educated older persons, and from 1.68 to 3.04 for those with low education. Plausibly supported by generous welfare systems and lower economic inequality, differences in material deprivation according to the level of education appear smallest in the Northern cluster. By contrast, the largest differences are found in the Southern and Eastern clusters.
Labour Market Status
The non-adjusted estimates show that being employed reduces the risk of material deprivation: with the exception of the Western cluster, the difference from the reference group (retirees) is statistically significant. However, after adjustment for the effects of the other covariates, the protective effect associated with employment loses significance in most clusters. This suggests that retirement in itself does not involve a significant increase in the risks of material deprivation. The opposite may hold true in the Northern cluster, although only to a limited extent.
The association between homemaking and material deprivation exhibits more variation. In the Northern and Western clusters, homemakers do not show any excess risk of deprivation. According to the adjusted estimates, the odds of being deprived are as much as 0.76 times lower for homemakers relative to the reference group. In the Eastern and Southern clusters, however, homemaking is related to a significant excess in risk of deprivation, ranging from 1.64 to 1.75. Individuals in the residual category feature substantially elevated risks of material deprivation, but, unlike for homemakers, the pattern is similar across clusters.
Chronic Diseases and Activity Limitations
Having multiple chronic diseases and activity limitations adds substantially to the risk of deprivation. In all clusters but one (the Eastern), a significant association between deprivation and chronic diseases persists after the inclusion of the other covariates in the model. The effects of activity limitations are significant in all clusters. The effect appears more pronounced in Southern and Eastern clusters and more moderate in Northern and Western clusters. This suggests that welfare systems in the Northern and Western clusters are more supportive of the economic needs of older persons in poor health.
Area of Residence
Area of residence makes only a limited difference in the risks of material deprivation. In the Northern and Western clusters, differences in the odds of deprivation associated with area of residence are not significant. In the Southern cluster, living in a city entails a reduction of 0.73 times in the odds of deprivation relative to rural residence. In the Eastern cluster, the largest advantage relates to living in suburbs. This finding is not surprising, as many countries of Eastern Europe experienced a tide of suburbanisation among the more affluent strata of the population after the fall of state socialism. Overall, in both the Southern and Eastern clusters, the results indicate a disadvantage for rural residents that is not counterbalanced by the welfare system [see Vidovićová et al. this volume for the consequences of such a disadvantage for care provision].
Although arrival in the host country usually occurs relatively early in the life course, the disadvantage associated with immigrant origin does not disappear but persists well into old-age. Our results show that higher risks of deprivation among immigrants can be found in all clusters. However, there is a considerable variation in the odds ratios of deprivation for immigrants, ranging from 1.38 in the Southern cluster to 2.44 in the Eastern cluster in the adjusted model. We think that the observed differences stem not only from contrasts between host societies but also from the diverse origins and characteristics of immigrants across clusters.