Using counterfactual microsimulations, Shapley decompositions of time change in inequality and poverty indices make it possible to disentangle and quantify the relative effect of tax-benefit policy changes, compared to all other effects including shifts in the distribution of market income. Using this approach also helps to clarify the different issues underlying the distributional evaluation of policy reforms. An application to the UK (1998-2001) confirms previous findings that inequality and depth of poverty would have increased under the first New Labour government, had important reforms like the extensions of income support and tax credits not been implemented. These reforms have also contributed to substantially reduce poverty among families with children and pensioners.