Abstract
Survey evidence shows that beliefs about income inequality and own income rank often differ systematically from population statistics. Social sampling provides a natural starting point: people learn about society from comparison sets that are local rather than representative. A recurring feature of subjective rank data is a middle-class bias, with reported positions compressed toward the centre of the income distribution. We use this evidence to motivate rank-symmetric comparison windows centred on the observer and ask what inequality such samples contain. We show that the curvature of the log income profile determines whether inequality in a fixed-width rank window rises or falls with rank, with the main result holding in the Lorenz sense. We then derive a finite-sample approximation for the local Gini and apply the framework to standard income distributions. Finally, we show that local sampling can reverse inequality rankings across societies. For lognormal and Pareto incomes, the local ranking changes at a unique rank cutoff, and global and local rankings can disagree for a majority of rank-local observers. The results isolate a distributional channel through which the same locally centred sampling rule can generate heterogeneous inequality signals.