Testing for sample-selection bias due to location in the labour-market behaviour of respondents from the British Household Panel Survey

A five-state origin and destination competing risk labour-market model is formulated and applied to some individual-level labour-market data obtained from the British Household Panel Survey (BHPS). The BHPS used a multistage stratified clustered sample design for selecting households and individuals. The primary sampling unit was the postcode sector. 250 out of a population of 8951 postcode sectors were selected. A test for the informative nature of the primary sampling units for labour-market behaviour is developed and applied to the BHPS labour-market data. The results suggest that the sample design is not completely ignorable for male and female labour-market behaviour.