Modelling errors in survey and administrative data on employment earnings: Sensitivity to the fraction assumed to have error-free earnings

• Published in: Economics letters Vol. 192; p. 109253
• Main Authors: Jenkins, Stephen P., Rios-Avila, Fernando
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.07.2020; Elsevier Science Ltd
• Subjects: Earnings; Employment; Errors; Kapteyn–Ypma​ model; Labor market; Labour earnings; Measurement error; Misclassification error; Polls & surveys; Robustness; Sensitivity analysis; D31; Labour earnings; Misclassification error; Kapteyn–Ypma​ model; C81; Measurement error; C83
• ISSN: 0165-1765; 1873-7374
• DOI: 10.1016/j.econlet.2020.109253

Kapteyn and Ypma (Journal of Labour Economics 2007) is an influential study of errors in survey and administrative data on employment earnings. To fit their mixture models, Kapteyn and Ypma assume a specific fraction of their sample have error-free earnings. Using a new UK dataset, we assess the sensitivity of model estimates and post-estimation statistics to variations in this fraction and find some lack of robustness. •Sensitivity analysis of Kapteyn and Ypma (Journal of Labor Economics 2007) study.•New UK dataset linking survey data and administrative record data on earnings.•Identification requires an assumption about the fraction with error-free earnings.•Kapteyn and Ypma assume the fraction is 14.8%.•We use eight fractions ranging from 0.25% to almost 17%.•Some model features are robust to the choice of fraction, but others are not.•One non-robust feature is the size of error variances.•Another is the distribution of mixture group membership probabilities.