I would like to observe individuals (CPSIDP) transition between occupations (OCC) across surveyed months (MISH). When I attempt on Stata “reshape wide occ, i(cpsidp) j(mish)”, I encounter the problem of repeated mish values within cpsidp. Same problem for declaring panel data. How can I fix this or otherwise approach the observation of worker flows?
There are a few possible reasons for this error. One is that you may have selected both ASEC and basic monthly samples in your extract. If that is the case, individuals who are in both of those samples could have duplicated CPSIDP+MISH values. You will want to drop either the basic monthly or the ASEC respondents in these cases, or make a new extract only including one of the types of sample. Another possibility is that you are including individuals with CPSIDP==0, which means these people cannot be linked across samples. You should drop these before the reshape.
Occupational transitions are tricky. I would advise two things: 1) occupations (especially at the detailed level - 4 digit codes) are measured with lots of error, and when you measure changes in occupations the error is magnified, because if either one or both of the time periods has a different code due to reporting error, you will get a false transition - so the transition rate will be inflated. This is especially the case in CPS because often a different household member answers the questions in different months. 2) You can look at transitions across years using ASEC data (using OCC and OCCLY) and this is somewhat better because these are reported by the same person at the same time. But these variables are not entirely comparable, with OCCLY asking about the longest job in the previous calendar year and OCC giving the occupation of the current primary job (usually in March). I suggest reading this paper for more info on the issues with measuring occupational transitions with various data sources.