Poverty dynamics corrected for measurement error

We use latent class models to correct measurement error in estimates of the dynamics of relative income poverty in ten EU countries measured over four waves of the European Community Household Panel. A latent mover-stayer Markov model gives an acceptable fit to all ten transition tables. We focus in more detail on four countries – Denmark, the Netherlands, Italy and the UK – and show that mobility in poverty transition tables is over-estimated by between 25 and 50 percent if measurement error is ignored. In addition, once error is corrected, poverty rates show less cross-national variation.