Understanding Society User Support

Hi both. I am facing the same problem. I don't really understand why this universe is applied in this way as jboff is a variable that identifies those who had no work last week but they have already a paid job, so I don't see the reason these people are not assigned with a NSSEC value. Then when someone cross-tabulate jboff and jbhas with jbstat then results do not make sense as noone knows then in what state respondents are. This is very confusing. In my case, I am trying to add the unemployed/inactive as a separate category within the jbnssec_dv variable and I cannot find any valid way of doing that without having clashes with other variables that indicate a different state. Long story short, I am very worries about the reliability of the NSSEC variable.

Hi Roberto

Thanks for your reply. I am still a bit confused though as I don't really understand how then people who have identified themselves as not in paid employment or being self-employed (like Retired, F-T students etc) still have valid values of NS-SEC. Is this because they identify themselves as performed some sort of work the last week even though they also reported themselves as not being in paid employment or being self-employed in the jbstat variable? And if this is the case, then which of the two responses are more valid when someone looks on participants employments status in relations to their NS-SEC value? Is there any accurate way for me to construct a NS-SEC variable where I can get only NS-SEC values using the values in the jbstat variable for the self-employed, in paid employment, on maternity leave, in Govt training scheme, on apprenticeship, on furlough, on shared parental leave, on adoption leave and then use the rest as a distinct "non-employed" category? Thanks

Hello Theocharis,

You are correct that some respondents who report their main activity in jbstat as “Retired”, “Full-time student”, etc., may still have a valid NS-SEC classification (jbnssec*_dv). This occurs because:
• jbstat captures self-reported main activity status.
• jbnssec*_dv is derived from current job information, based on occupation and employment characteristics.
• The valid universe for jbnssec*_dv includes those who:
o Did paid work last week (jbhas = 1), or
o Did not work last week but have a job to return to (jboff = 1).

In practice, some individuals may identify their main activity as “Retired” (or similar), yet still report having done some paid work in the reference week (or having a job). In those cases, they fall within the valid universe for NS-SEC classification. This is not an inconsistency in the derivation of NS-SEC, but rather reflects the fact that main activity status (jbstat) and labour market behaviour in the reference week (jbhas/jboff) measure related but distinct concepts. NS-SEC is occupation-based and therefore applies to those with a current job, irrespective of how they describe their primary status.

There is no single “more valid” variable, it depends on your analytical objective. jbnssec*_dv relates to occupational class, jbstat relates to self-defined labour market status, and current employment behaviour in the reference week is defined by jbhas and jboff. Constructing an NS-SEC variable with a “non-employed” category, you would first need to define what you consider to constitute employment status. Among respondents who meet that definition, you would retain their jbnssec*_dv value. For all remaining observations, you would assign a separate “non-employed” category. However, the appropriate approach ultimately depends on your specific research objective.

I hope this information is helpful

Best wishes,

Roberto Cavazos
Understanding Society User Support Team

Hi Roberto

Many thanks for clarifying this. However, looking on the data what I can see is that although for jboff valid NSSEC data exists only for joboff==1, when I crosstab jbhas I can see many valid observations for both 1 (Yes) and 2 (No) jbhas values. As far as I underdstand, this is different compared to how you described the universe under which NSSEC valid observations exist. In terms of my specific query about constructing a "valid" NSSEC variable with the out-of-work respondents in, I am planning to create that category by including those with jbstat values other than in self-employment and paid employment, in maternity, shared parental, adoption leave or on furlough but with only those with no valid NS-SEC data (apart from missing and proxy, which will be excluded from analysis anyway).

As Roberto said, NNSEC is computed using information about current job which is asked if JBHAS = 1 OR (JBHAS=2 & JBOFF=1). So, you will have some cases with valid NSSEC where JBHAS=1 or 2. You can create a flag of those who should have valid NSSEC data. If using Stata,

generate valid=(jbhas==1|(jbhas==2 & jboff==1))
replace valid=-9 if inlist(jbhas,-1,-2,-9)
ta jbnssec_dv valid

Those with valid=1 should have jbnssec_dv>0 and those with valid=0 or -9 should have jbnssec_dv=-8
Across 15 waves, among all person-wave observations (N=601,795)
(1) for those with valid=1, 2.9% have NSSEC=-9, rest have jbnssec_dv>0.
(2) for those with valid=0, 3 cases have jbnssec_dv>0, for the rest jbnssec_dv=-8
(3) for those with valid=-9, 1 case has jbnssec_dv>0, for the rest jbnssec_dv=-8

The 2.9% cases with valid=1 but jbnssec_dv=-9 is due to insufficient information to code NSSEC
The 4 cases in (2) & (3) with valid jbnssec_dv values are an error.