SMABS 2004 Jena University
SMABS 2004 Home Organization About Jena Sponsors Links Contact SMABS Home

European Association of Methodology

Department of methodology and evaluation research

Jena University


© Webmaster

Contributions: Abstract

The latent class approach to estimating gross flows in the presence of correlated classification errors, with application to data from the French labour force survey

Francesca Bassi Ugo Trivellato
University of Padova
Italy

Classification errors in the observed labour force state can lead to serious biases in the estimation of gross flows from panel data. We focus here on correlated classification errors, which typically emerge when data are collected by retrospective interrogation, because of the effects of memory inaccuracy.

We use latent class analysis to specify and estimate a model relating the true states and observed states at successive points in time. Our approach has the advantage that it does not require external information, either on misclassification probabilites or auxiliary variables affecting transition and classification probabilities.

We apply the approach to data on young people form the French Labour Force Survey, March 1990-March 1992. The model predicts flows in the expected direction: estimated true transitions show higher mobility than observed ones. In addition, the measurement part of the model has signifcant coefficient estimates, and the estimated response probabilities show a clear, interpretable pattern.

We compare our results with those produced by a different approach, which aims at obtaining robust estimates of true transition rates while paying marginal attention to the measurement process.