Longitudinal attrition in ASHE

Forth, John; Phan, Van; Stokes, Lucy; Bryson, Alex; Ritchie, Felix; Whittard, Damian; Singleton, Carl

Abstract

The Annual Survey of Hours and Earnings (ASHE) provides many of the UK’s official earnings statistics. The survey operates on an annual 1% sample of employee jobs. However, the method of sampling - based on the final two digits of an employee’s National Insurance number – means that records are linkable longitudinally. Many
government and academic studies have utilised the dataset in this way. However, the longitudinal integrity of the ASHE sample has been the subject of little prior investigation, with the panel sample generally assumed free of any attrition biases that might compromise longitudinal analysis. We explore the validity of this assumption by comparing rates of year-on-year sample retention in ASHE with rates
of employment retention estimated from a reference dataset (the Longitudinal Annual Population Survey). Our analysis confirms the existence of systematic patterns of longitudinal attrition in ASHE, which have the potential to introduce bias into longitudinal analyses of these data. We go on to construct longitudinal weights
that correct for estimated attrition biases over adjacent years in ASHE. In an illustrative analysis, the application of these weights brings about a small widening of the distribution of individual wage growth.

• Working Paper Type: Working Paper
• Deposit Date: Oct 24, 2022
• Publicly Available Date: Oct 24, 2022
• Public URL: https://uwe-repository.worktribe.com/output/10105383