From Wave 2 to 12, the jbsoc00 question about the occupation of the main current job (an open-ended description of the job title and tasks and duties, subsequently coded to the Standard Occupational Classification (SOC)) was preceded by the jbsoc00chk question. Note that the jbsoc00 question was:
jbsoc00: “What was your main job last week? Please tell me the exact job title and describe fully the sort of work you do”.
The jbsoc00chk question was asked of all respondents who had a job, and whose answer to the jbsoc00 in the previous wave was codable. In the jbsoc00chk question, the respondent was asked if the description of their occupation given previously (the actual text description provided by the respondent was displayed in full on the screen) was still accurate.
jbsoc00chk: Is [ff_JBSOC00] still an accurate description of your occupation in your main job?
If the respondent answered ‘yes,’ then the jbsoc00 question was not asked, and the occupation SOC code from the previous interview was copied over (fed-forward). If the respondent said ‘no,’ (which ranged between 12 – 18 %) then the jbsoc00 question was asked. Therefore, for around 80 – 90 per cent cases, the previous occupation code was copied over. Using this method of dependent interviewing for the main job occupation question, an occupation change was registered at around 10 per cent of respondents in continued employment. [1]
The way we asked about current employment was changed in Wave 13. Instead of asking about the main job and the second job in two separate modules, respectively, “Current Employment ” and “Second Jobs”, respondents were asked about all jobs they had (up to 13) in the modified “Current Employment” module (see Changes to employment questions from Wave 13). The jbsoc00chk question could not be asked in Wave 13 because, unlike in prior waves [2], we asked a series of jbsoc00 questions in which the respondent was prompted to describe the occupation of each of the jobs they reported:
“Please describe fully the sort of work you do for your job as a/an [JOBCODE] {if MULTIJOBSTOTAL > 1}.”
Although we could identify which of these was the main job, because of the changes in wording, we could not ensure that prior waves matched Wave 13. Moreover, given that we were asking about all other jobs in the same manner (which had not been done before) we wanted consistent measurement within waves across all jobs as well. Also note that in Wave 13, the jbsoc00 questions (one for each job reported) only asked to describe the job while the job titles were collected separately in the questions alljobstitle. So, for each job reported, respondents were asked:
alljobstitle: “What is the exact job title for your job”
jbsoc00: “Please describe fully the sort of work you do for your job as a/an[jobcode]”.
In other words, every employed respondent provided a description of their current main job occupation (the job title, tasks, and duties). These answers were then coded independently of the previous description, or the previous occupation SOC code reported in the previous wave. We found an unusually high percentage of occupation changes in the main job occupation SOC codes among the sample in continued employment (compared to previous waves). There were three reasons for this. First, there was an accidental omission of parts of the available information used in coding the occupations – only the description of the task and duties (jbsoc00) was coded, excluding the job title (alljobstitle) information. This led to lower overall precision of coding. Second, this procedure was also inconsistent with the pre-wave 13 method, where both the job title and the task description were coded together. Third, the increase in occupation changes was due to the variability inherent to coding answers to open-ended occupation questions.
Two steps were taken to address these issues. First, all jbsoc00 verbatim responses about the job description were concatenated with the job titles reported in alljobstitle and coded using SOC coding frame. This eliminated one source of error and reduced the proportion of job changes observed between earlier waves and Wave 13, from 60 to 45 per cent of the sample in continued employment. Since this was still much higher than the percentage of changes observed pre-wave 13, a further step was taken. For all cases where a change of SOC code was observed, a comparison of the fed-forward occupation verbatim responses (i.e. the title and description of the job collected in the last interview) with the occupation verbatim response collected in wave 13 was undertaken. Here, we aimed to emulate the logic of the pre-wave 13 questionnaire. Although the jbsoc00chk question was not asked, meaning there was no respondent’s self-assessment of how accurately their last description reflected their current job, we still had the last and the current descriptions. By comparing the two descriptions, we attempted to assess how each respondent would have responded to the jbsoc00chk question had they been asked. If the verbatims were similar and described the same occupation, the fed-forward SOC code was assigned. If the verbatims were not similar and thus signalled an occupation change, the new SOC code was assigned. The revision was applied to all previously released versions of the main current job occupation variables: m_jbsoc00, m_jbsoc00_cc, m_jbsoc10, m_jbsoc10_cc.
As a result, 70 per cent of the occupation changes were assessed as spurious (the verbatims were similar), with the remaining 30 per cent assessed as genuine (the verbatims were dissimilar). This translated into an overall rate of change of 15 per cent in the sample in continued employment.
Table: Example (fictitious) answers to the jbsoc00 question and the comparison result.
| Fed-forward jbsoc00 verbatim | Fed-forward jbsoc00 verbatim | Comparison result |
| a farmer who farms land | farmer, farming land | similar |
| serving customers in a bar | I work in a nursery looking after children | dissimilar |
[1] This percentage could be lower than the share of ‘no’ answers in the jbsoc00chk, as some respondents who answered ‘no’ provided a description similar enough to the previous one to be coded with the same SOC code as before.
[2] There were two separate questions about occupation, main job: jbsoc00, second job: j2soc00.
