Static tax-benefit microsimulation models (MSMs) are widely used and well-regarded tools for public policy analysis, but it is essential to use them very carefully. This paper focuses on the analysis of MSM output, suggesting the use of non-parametric methods as a useful, informative and relatively straightforward complement to detect effects not always captured by measures often used to present MSM results. Non-parametric methods are used here to analyse the output of an MSM applied to the 1998 Italian personal income tax reform, the main change in which concerned the tax schedule: the first tax rate was increased from 10 per cent to 18.5 per cent and the top one was reduced by 4.5 percentage points. Non-parametric methods highlight that the effects of this reform were very different for different types of households, with low-income pensioner households among the main losers. Results are checked for robustness by standard statistical methods and compared with empirical results obtainable using quintile histograms.