Recoding occupational status variable from multiple countries

Hi folks,

I hope to recode the raw occupational status (occ) variable to ISCO-08, so that I can further convert it to the ISEI score. Upon closer examination of the occ variable, I realized this variable is coded very differently in different censuses. I understand a few censuses do not have detailed occ variables and only provide 2 digits ones, which I’m happy to just add two zeros to make them 4 digits ISCO variables. But I feel like there is so much inconsistency that doing so would probably be wrong. I’m wondering how I should proceed to recode the variable.

The picture above is a screenshot of the raw occ variables from the countries I hope to analyze. Could you tell me what the variable label (2 digits/ 3 digits) mean, besides the descriptive meaning? In other words, do the occ variables with the same variable label use the same coding standards?
I know there are programs within Stata that will automatically convert ISCO-88 and ISCO-08 variables. One thing I tried to figure out is whether a raw occ variable uses the ISCO-88 or ISCO-08 standards, or something else. The following link indicates this information for some of the countries, but not all of them.
Finally, are there differences in coding due to time? Most of the censuses I hope to use are from the 2010 round, but some are from the 2000 round. Does this impact the coding of the occ variable? If so, how can I go around it?

I’m sorry I have so many questions. I would appreciate your response!



The comparability tab for OCC will tell you whether a sample is coded using ISCO 1988, 2008, or using a separate method. For instance, both Armenia and Belarus use the ISCO 1988 classification and are generally comparable with each other. In cases where the same coding scheme is used, but the number of digits differ, you may add zeros to make the values comparable. However, this will result in the same occupations having different codes. Two-digit occupation codes do not provide the same level of detail as three-digit codes and therefore it’s unclear which three-digit code individuals with two-digit codes should map to. My recommendation would be to instead remove a digit from three-digit codes in order to run your analysis on this sample. Alternatively, you may use the general recoded variable, OCCISCO, which uses a single-digit ISCO 1988 scheme.

This forum discussion may also be helpful.

1 Like

Thanks for your help!