Background: Social capital can be viewed as a societal process that works toward the common good as well as toward the good of the collective based on trust, reciprocity, and solidarity. Our study aimed to present two multivariate statistical analyses to examine the formation of latent classes of social capital using the IQ-SC and to identify the most important factors in building an indicator of individual social capital.
Findings: A cross-sectional study was conducted in 2009 among working adolescents supported by a Brazilian NGO. The sample consisted of 363 individuals, and data were collected using the World Bank Questionnaire for measuring social capital. First, the participants were grouped by a segmentation analysis using the Two Step Cluster method based on the Euclidian distance and the centroid criteria as the criteria for aggregate answers. Using specific weights for each item, discriminant analysis was used to validate the cluster analysis in an attempt to maximize the variance among the groups with respect to the variance within the clusters. "Community participation" and "trust in one's neighbors" contributed significantly to the development of the model with two distinct discriminant functions (p < 0.001). The majority of cases (95.0%) and non-cases (93.1%) were correctly classified by discriminant analysis.
Conclusions: The two multivariate analyses (segmentation analysis and canonical discriminant analysis), used together, can be considered good choices for measuring social capital. Our results indicate that it is possible to form three social capital groups (low, medium and high) using the IQ-SC.