Job Changes and Wage Changes: Estimation with Measurement Error in a Binary Variable*

Many studies of labour market dynamics use survey data so it is valuable to know about the quality of the data collected. This paper investigates job transitions in Ireland over the period 1995 to 2001, using the Living in Ireland Survey, the Irish component of the European Community Household Panel. In applied work on job mobility, researchers often have to rely on self-reported accounts of tenure to determine whether or not a job change has taken place. There may be measurement error in these responses and consequently observations may be misclassified. The paper finds that there are substantial inconsistencies or measurement error in the responses used to determine job changes so there is a risk of misclassifying cases as being job changes when truly they are job stays and vice versa. The paper explores the impact of misclassification in a model of job change using an estimator developed by Hausman, Abrevaya and Scott-Morton (1998). It finds that ignoring misclassification may substantially underestimate the true number of job changes and it can lead to diminished covariate effects. The paper then investigates the relationship between job mobility and wage growth. Misclassification in a binary explanatory variable causes attenuation in OLS estimates. A two-step approach to controlling for misclassification in job changes is adopted to estimate the wage effects of job mobility. The paper finds that controlling for misclassification has a substantial impact on the estimated effect of changing jobs on wage growth