From PAPI to CAPI: consequences for data quality on the British Household Panel Study