Weibull Regression of Unemployment Duration Measured With Error Via the EM Algorithm

Several models aim to measure how demographic, professional, and labour market factors influence unemployment duration, which has a compelling social and economic interest. Nevertheless, this quest still appear challenging in both interpretations and solutions, due to various difficulties. Among others, a relevant issue is the accurate measurement of unemployment, which is often biased by the presence of unknown errors. In this paper, we consider the measurement error as a latent variable and derive an ExpectationMaximization (EM) algorithm in the context of a Weibull regression model. The proposed methodology allows to estimate at the same time: (i) the effect of the covariates of interest on unemployment duration and (ii) the measurement error. The reminder of this paper is organized as follows. In section 2, we describe the data, the regression model and the estimation method employed in our analyses. In section 3, we summarize our analyses and compare the findings with those obtained using standard approaches.