Linking all waves of BHPS and UKHLS: Inconsistencies?
I've merged all the waves of the BHPS and Understanding Society into one master data file in the long-format (I have one row per person per wave). To check whether this has worked out correctly, I checked whether any respondents had changes in time-invariant variables like their sex. Doing that, I found quite a number of mismatches: Using the variable "sex" by "pidp", there was a change of sex in 15417 rows (and no change in 558476 rows). If I use "sex_dv", there is a change of sex in only 17 rows (no change in 279717 rows; sex_dv has a large number of missing values).
Is it possible that there are that many inconsistencies or is it more likely that I did anything wrong in the process of merging the datasets?