and Challenges”

Social indicators (of poverty or inequality) are known to be potentially sensitive to the presence of data contamination and to extreme incomes at either or both tails of the income distribution. EU-SILC being an important source for the estimation of such in-dicators in Europe, it is important to assess how much indicators derived from it are sensitive to alternative adjustments meant to control the impact of extreme incomes. The paper presents the results of a large scale sensitivity analysis considering both simple, classical adjustments (trimming, winsorizing) and a more sophisticated approach based on modelling parametrically the tails of the income distribution. Reassuringly, ordinal comparisons of countries are found to be robust to varying data adjustment procedures. Cardinal comparisons may however be more markedly sensitive. Parametric tail mod-elling is found to be a potentially promising method