Errori di misura in dati longitudinali e analisi dei flussi lordi

The problem of analysing dynamics using panel data possibily affected by measurement error is outlined. The paper focuses on the problem of estimating gross flows when classification errors are present. We agree that the modified LISREL approach provides a convenient general framework. We show how the log-liner model with latent variables can be appropriately specified and estimated to correct observed gross flows for possibly correlated classification errrors. We apply the approach to two real data sets from the SIPP and the Itlain Labour Force Survey respectively, and get reasonably satisfactory results.