Choosing the Bandwidth for Decomposing the Redistributive Effect: Evidence from Milan Using AMeRIcA Data

When close equals groups are used for decomposing the redistributive effect of taxation, it arises the issue related to the choice of a proper bandwidth. Recent literature (Kim and Lambert, 2008; Urban and Lambert, 2008) suggests to set the bandwidth for close equals groups to maximize the vertical effect measure in accordance with van de Ven, Creedy and Lambert’s suggestion (van de Ven et al., 2001). But the close equals methodology proposed by Urban and Lambert differs from the van de Ven, Creedy and Lambert’s one. Then the choice of the bandwidth still involves a sensitivity analysis of the decomposition results. In this work we recall the various decomposition methods and, following Vernizzi and Pellegrino (2007) approach, we select the bandwidth by comparing the behaviours of the alternative vertical components over a wide bandwidths range. We consider a micro data set derived from AMeRIcA Data Warehouse which collects pre-tax and post-tax incomes of residents in the Milanese area. Our findings show that for the selected bandwidth all the alternative measures of vertical effect explain a potential redistribution.