This paper illustrates two empirical approaches to the measurement of multidimensional inequality. The first approach is based on the analysis of the independent distribution of monetary and nonmonetary welfare attributes. The second approach considers pair-wise joint distributions of those attributes, hence allowing for differences in the various distributions, as well as possible correlations between the attributes. The analysis is based on household survey data from Brazil for 1996. We focus on inequalities in income, education, health and political participation outcomes. We calculate the extent of vertical and horizontal monetary and non-monetary inequalities, examine the determinants of both types of inequality and analyse their impact on household welfare. Our results show that economic analyses based solely on the distribution of income variables will not portray fully the degree of socio-economic and political inequalities in Brazil. In fact, traditional analysis of inequality may overestimate the extent of inequality, as education and other non-monetary welfare attributes appear to be more equally distributed in Brazil than income.