Unknown eligibility whilst weighting for non-response: The puzzle of who has died and who is still alive?

A challenge for longitudinal surveys is to update eligibility status over time for sample units, including those who cease participating. Ineligible units should be excluded from the base for weighting, as weighting would otherwise increase the influence of responding sample units that are similar to the ineligible ones. This paper estimates likelihood of eligibility of the British Household Panel Survey (BHPS) sample members by gender and age using population mortality data. This likelihood is then used to adjust non-response weights. Analysis is conducted using adjusted and unadjusted weights. Results show that the adjustment affects the significance level of some variables.