Fed-forward employment status at W3
I'm slightly concerned about the'fed-forward' employment status variables from wave 2 to wave 3, because they don't seem to match up with current activity from the wave before.
In wave 3, the variable c_ff_jbstat doesn't match up with the current employment status reported at wave 2, at least not for the BHPS sample (to which my analysis is restricted).
I'm aware that there are two separate accounts of employment status at W2, which already to some extent conflict with each other: 1) b_jbstat, in answer to 'which of these best describes your current employment situation?', and 2) what you get from the annual events history module, i.e. the fed-forward activity from the wave before if that is still current (b_ff_jbstat if b_empchk==1 | b_notempchk==1), or one of the subsequent activity spells up to the 10th if that activity spell was reported as still current as of the wave 2 interview (i.e., if b_currjob==1 | b_currstat==2).
HOWEVER, c_ff_jbstat doesn't match up with either of these for thousands of people. Even is a simple case, restricting to the BHPS sample people who at wave 2 said they were still doing the wave 2 fed-forward activity from the last wave of the BHPS, b_ff_jbstat!=c_ff_jbstat for almost two thousand people.
I then looked at the equivalent problem for wave 2 - i.e. whether the fed-forward employment status at wave 2 for the BHPS sample actually matches up with current employment status as reported at wave 18 of the BHPS. Again, it doesn't for 3,924 of the 14419 people present at both waves.
In both cases (BHPS 18/UKHLS W2, and UKHLS W2/UKHLS W3) many of the discrepant cases are where there is data for one variable but not the other - but how could this happen if it was genuinely fed-forward? Also in both cases, there are discrepant cases which are both valid values but incompatible, for example 'employed' and 'retired'.
Given all of this, could you explain how at both UKHLS W2 and UKHLS W3 the 'fed-forward' activity was actually derived?