Significance Editing in the Survey of Employment and Earnings

Significance editing is a statistical technique which is used to prioritise and control the amount of input editing in a survey. The technique works on the premise that only those units which fire edit queries that are considered to be significant need to be edited. An edit query for a unit is considered to be significant if it is assigned a score above a prespecified cut-off value. The score is based on the expected effect on survey estimates caused by changing the unit's reported data to some value determined by the edit rule. The technique ensures that the bias due to not editing some of the survey forms is less than 10% of the variance of the estimate at the state by industry division level. The introduction of significance editing in the survey of Average Weekly Earnings (AWE) was very successful, resulting in negligible effects on survey estimates and resource savings of between three and four staff years.