Understanding Society User Support

Dear colleagues,

I am trying to generate a variable that captures job change in a broad sense; that is, that informs of a possible job change within the same employer OR derived from an employer change. My understanding is that, for this purpose, I should rely on 'samejob' ('Check for same job within employer') and 'jbsamr' ('Same employer check').

The universe of the former variable ('samejob': 'Have you been working continuously in the same job since ff_IntDate?') is constituted by all those who fulfil the following conditions: (a) they have been interviewed at a prior wave or have been interviewed previously; (b) have been continuously employed since last interview or furloughed/temporarily laid off at last interview followed by paid work; and (c) are working for the same employer.

The universe of the latter variable ('jbsamr': 'And have you worked continuously for the same employer since ff_IntDate?') is constituted by all those who meet the following conditions: (a) they have been interviewed at a prior wave or have been interviewed previously and (b) have been continuously employed since last interview or furloughed/temporarily laid off at last interview followed by paid work.

If I cross-tabulate these two variables, considering the missing values (for the 14 waves of UKHLS in my data), I obtain the following table:

. tab jbsamr samejob, miss

Same |    Check for same job within
    employer |             employer
       check |       yes         no          . |     Total
-------------+---------------------------------+----------
         yes |   181,677      7,471         54 |   189,202 
          no |         0          0     16,926 |    16,926 
           . |         1          0    431,919 |   431,920 
-------------+---------------------------------+----------
       Total |   181,678      7,471    448,899 |   638,048

In the cross-tabulation above, 16,926 observations correspond to individuals who changed employers. They are marked as missing in 'samejob' because they are not part of the universe of the variable. They do not meet one of the conditions ((c) are working for the same employer) So far, so good.

The table above also made me think that the 431,919 observations that are missing in these two variables correspond to individuals who were not working when they were interviewed. But this is not totally the case, as it appears in the table below. I ask Stata to generate the following table only for the observations that are missing in 'jbsamr' and 'samejob', but there is a substantial number of observations that correspond to 'Paid employment' (74,103)

. tab jbstat if jbsamr==. & samejob==.

Current economic activity |      Freq.     Percent        Cum.
----------------------------------------+-----------------------------------
                        self employed   |     39,006       10.77       10.77
                 Paid employment(ft/pt) |     74,103       20.46       31.22
                        unemployed      |     26,900        7.43       38.65
                                retired |    134,644       37.17       75.81
                on maternity leave      |      1,132        0.31       76.13
                    Family care or home |     28,203        7.79       83.91
                full-time student       |     33,795        9.33       93.24
                    LT sick or disabled |     19,829        5.47       98.71
                   Govt training scheme |        398        0.11       98.82
                Unpaid, family business |        333        0.09       98.92
                      On apprenticeship |        454        0.13       99.04
                            On furlough |        157        0.04       99.08
Temporarily laid off/short term working |        130        0.04       99.12
                                     14 |          2        0.00       99.12
                doing something else    |      3,185        0.88      100.00
----------------------------------------+-----------------------------------
                                  Total |    362,271      100.00

I try to understand how it is possible to be missing in both variables and appear as 'paid employment' (mind you, not self-employed, but paid employment). Could you help me find out about it?

As I said at the beginning of this post, my ultimate goal would be to generate a variable with three categories: (a) same job; (b) a different job with the same employer; (c) a different job with another employer:

Thanks for your attention

And kind regards

Luis Ortiz Gervasi