We analyse the main determinants of the transition from a job into unemployment by using individual administrative data from Germany. Although the sample size is large and the information that is used for operations is often highly accurate, variables which are not required by the administration but used for the statistical analysis are subject to considerable measurement error. We show that the high degree of misclassification can even persist after comprehensive logical editing and imputation rules have been applied. We find that the measurement error has a sizable effect on our estimation results. Long tenure rather than a higher educational qualification appears to be the key ingredient for a safe job in Germany.
|Journal||Journal of the Royal Statistical Society, Series A (Statistics in Society)|
|Publication status||Published - 2012|