Addi van Bergen

Chapter 3 58 Measurement properties Content validity To examine the degree to which the indices cover the multidimensional concept of social exclusion, we compared, for each dimension, the items in the constructed indices with those in the SCP index. The dimension ‘limited social participation’ of the SCP index comprises items on social isolation and on participation in social networks. From Table 3 we can see that all three constructed indices included items on social isolation, but only Index3 contained an item on participation in social networks i.e. contacts with neighbours. In the dimension ‘material deprivation’ the SCP index includes items on the financial situation of the household and on the lack of basic goods and services. The three constructed indices did contain 2 items on the financial situation of the household, but items on the lack of basic goods and services, were absent in all three indices. In the dimension ‘inadequate access to basic social rights’ the SCP index contains aspects of good living environment and access to health care. Index2 and Index3 contained similar items on good living environment, but only Index3 contained additional items on access to healthcare. These items however, referred to the need for information or assistance and not the actual lack of access, as does the SCP questionnaire. As floor or ceiling effects may limit the content validity [14], we examined the frequency distributions of the three indices. All three distributions were right-skewed, which corresponds well with the distribution of the SCP index and is consistent with the expectation that a large part of the population is not excluded, while the degree of exclusion at the right end of the scales varies widely. Internal consistency Table 4 shows the findings on the internal consistency of the indices and subscales. The canonical correlations of three constructed indices ranged from 0.35 (Index1) to 0.44 (Index3), which is sufficient. Index2 and Index3 had even higher canonical correlations than the SCP index (r d =0.38). The Cronbach’s alphas of the transformed subscales were good for dimensions 1 and 2. For dimension 3 Cronbach’s alphas were 0.68 (Index2) and 0.65 (Index3). In the SCP study Cronbach’s alphas were not calculated. Internal structure Table 5 provides the correlations between the subscales and the general indices and between the subscales themselves. The correlations between the subscales and the general indices ranged from 0.68 to 0.82, and were similar to those of the SCP index. As expected, the correlations between the subscales were weaker than with the general indices. They ranged from 0.33-0.55, which is in line with the internal structure of the SCP index.