Measuring Income Mobility with Dirty Data

We examine the performance of measures of mobility when allowance ismade for the possibility of data contamination. We find that “singlestage”indices – those that are applied directly to a sample from amultivariate income distribution – usually prove to be non-robust in theface of contamination. However, “two-stage” models of mobility –where the distribution is first “discretised” into income intervals andthen a transition matrix or other tool is applied – may be robust if thefirst stage if appropriately specified.[...]