Datt-Ravallion Decomposition

The residual arises because the growth and redistribution effects interact: the poverty impact of a given growth depends on the level of inequality, and vice versa, so when both move, the change cannot be split exactly using a single reference date. The residual equals the difference between evaluating the components at the initial versus the final period. To eliminate it, use a symmetric formulation — average the initial-base and final-base decompositions, equivalent to applying the Shapley decomposition — which assigns the interaction evenly and leaves no residual.

It is the main input to that judgment. If the growth component accounts for most of an observed poverty reduction and the redistribution component is small or adverse, growth was the dominant but not distribution-improving driver. Pro-poor growth analysis builds on this by comparing the actual poverty reduction with what distribution-neutral growth would have delivered, or by examining the full growth-incidence curve. The decomposition itself is descriptive: it quantifies the two channels but does not by itself define a pro-poor threshold.

Yes, to any poverty measure that can be written as a function of the ratio of the poverty line to mean income and the Lorenz curve, which includes the entire FGT family (headcount, poverty gap, squared gap) and the Watts index. The counterfactual experiments — shifting the mean with the Lorenz curve fixed, and shifting the Lorenz curve with the mean fixed — are well defined for all such measures, so the same growth/redistribution split can be reported for each.