Support #767

Work experience in UKHLS

Added by Sabine Hübgen over 6 years ago. Updated over 6 years ago.

Start date:
% Done:



Dear support team,

I have a question regarding work experience in UKHLS.

I try to generate a work experience variable based on the employment history files (a_empstat and e_empstat).
My problem is to define the beginning and the end of episodes:

- there are only starty and startm, but no endy or endm. This is annoying, but would not be a big problem if we had the start information of all episodes per person.

- but: There is one spelltype 0 "current status reached, no further change", for which starty/m is always missing. This spelltype applies to 22% of all episodes.

Is there any possibility to identify the beginning of this "current status"??

Thank you,
best regards
Sabine Hübgen


Updated by Victoria Nolan over 6 years ago

  • Status changed from New to Feedback
  • Assignee set to Sabine Hübgen
  • % Done changed from 0 to 80
  • Private changed from Yes to No

Dear Sabine,

I'm not sure exactly which variables you're trying to use, but below is some essential information for working with annual employment history data which hopefully should answer your questions.

The main things to be aware of are that:

1. At every UKHLS wave, there are two accounts of the ‘current’ job or non-work activity; one is the report from the standalone item of jbstat, and one is the end point of the series of information on past spells since the last interview (the series stops when the participant declares a spell to be the current one). Usually these match up, but they don’t always, and it is the decision of the individual analyst how to handle that discrepancy.

2. This is also the case for the start date of the current job – this was asked about in a standalone question, but there is an alternative account in the series of information on all work and non-work spells since the past interview, one is which is reported as current. I use the account from the former activities series, because that has always been more appropriate for my research questions. Perhaps that may be fruitful for you here.

3. Except for where people reported a ‘don’t know’ or similar, you can compute the start date information for all episodes in the series, current or ended, from the end of the preceding episode in the sequence. The exceptions are a) if it is a current spell according to jbstat which appears not to match the last one in the series of past activities, then you may need to make a guess or exclude those people; and b) if the episode you are interested in was fed-forward from the wave before (i.e., was current at the N-1 interview), then you need to merge in the previous wave (sometimes more than one wave previously) and look for the relevant date information for the changeover point there (i.e. the start date for the last reported spell in that wave’s series of past spells).

4. People only get sent through the questions about past spells if they were present and reported jbstat at the previous interview (note, this is NOT the end point of the equivalent series, but the standalone report of current status which doesn't always match up). This means there will be missingness in all the information for people not present at the previous wave.

5. Each person’s series of past spells begins with either empchk or notempchk, which is the check for whether their work or non-work activity, respectively, reported as current at the previous year’s interview was still current. There are a small group of people who were sent through both branches of the loop because they gave conflicting information the previous year between jbstat and jbhas; again, it is up to you what to do with them. Throughout the whole series, information on work spells and non-work spells are kept separate, such that there are essentially two parallel tracks of information for every person, one of which should always have an ‘inapplicable’ code where the other has usable info.

6. Within the blocks of information on employment history since the last wave's interview, the naming conventions require care. The series starts with the fed-forward activity, which has its own currentness flags, date information etc, then comes the 'next' activity (whose type is stored in the variable nxtst for work spells, and nxtstelse for non-work spells, with date info in the variables with equivalent parts to the names), and then there begins a numbered series of information blocks, with suffixes from 1 up to 8 (at least at some waves, but it depends how many new spells people have had that year), corresponding to spells which happened after the first spell which itself came after the fed-forward activity. When I use these ones, I derive new variables for activity type and date info on the nxtst and nxtelse activities with names which are consistent with the variables relating to these later spells, but with a -0 suffix. That allows you to manipulate them all in the same loop using a numbered series starting at 0.

7. For each of the activities in the series, there is a currentness flag - or rather, there is a pair of currentness flags, one for work spells and one for non-work spells These are called, for instance, currjob5 and currstat5; if the -5 activity was a work spell, currstat5 will be ‘inapplicable’, vice versa for a non-work spell. Crucially, the CODING IS REVERSED for these two currentness flags, such the -5 activity is current is currjob5==1 but if currstat5==2.

I hope that helps!

Best wishes

On behalf of the data user support team.


Updated by Sabine Hübgen over 6 years ago

Dear Victoria,
many thanks for your quick and detailed answer! This is already very helpful.

What you describe above refers to work experience since the last interview - right? This mainly covers the years of 2009-2015.

I try to build full work experience for respondents, i.e. starting with the first spell after leaving ft education and ending with the most recent interview date.

To this end I use the two available employment history spell data sets (a_empstat and e_empstat). The variable 'empstat' indicates the type of the spell, 2 "being FT employed" and 0 "having reached current status" for instance. By now I merged the following variables from 'indresp' w1 and w5: jbstat, jbft_dv, jbbgm and jbbgy. With their help I try to identify the "current status" (empstat=0) and its respective starting year/month.

Is this appropriate?

Many thanks and best regards,

Sabine Hübgen


Updated by Alita Nandi over 6 years ago

  • % Done changed from 80 to 90

Yes, you are correct. The start date of the current employment status is asked every wave (w_jbbgd w_jbbgm w_jbbgy in w_indresp). So, you can use these to replace w_leshem w_leshsy4 from the employment history files (a_empstat e_empstat) for the current employment status.

I am guessing you mean "w_leshst" and not "empstat" in "The variable 'empstat' indicates the type of the spell,"?

Best wishes,


Updated by Victoria Nolan over 6 years ago

  • Status changed from Feedback to Closed
  • % Done changed from 90 to 100

Also available in: Atom PDF