When is cpsid or cpsidp=0? I'm using data from 2000-2010 and there seems to be many each year with a value of 0.


I’m using data from 2000-2010. I need to create a panel, so I need to use the cpsidp to link individuals over time. However, many of the observations each year seems to have a value of 0. Why is this the case?


I don’t really have an answer but I have a related question. I’ve recently downloaded the 2013 March CPS and the ATUS WB data, with the intention to link the two datasets. I used the Stata code provided to import the data. However, I’ve found a large number of cpsidp=0, over 72k, in the CPS data but not in the WB module. Without a proper cisidp, there’s no way one can merge the two datasets.


The answer to both of these questions is that CPS respondents who are part of the ASEC oversample (ASECOVERP) have CPSIDP values of 0. This is essentially because, at the present time, it is only possible to link ASEC respondents who are also part of the March basic sample to other basic monthly samples. More information about these details can be found here.


Thank you for answering the question so promptly. Your answer seem to suggest that those in the ASEC supplement don’t have a cpsidp, or did you mean only those who appear both in the supplement AND in the regular CPS who don’t have a cpsidp?

My further test seems to suggest the latter. Of the 87,652 cases in my analysis of the March ASEC, 28,766 had cpsidp=0. Once these are dropped, I was able to proceed with the merge with the WB module, resulting in a “small” merged dataset of 1,909 cases.

Also, thanks for the linked working paper, which I’ll read more carefully. The paper suggests that CPS surveys are longitudinal. Along those lines, what proportion of the March ASEC supplements are carried over from year to year? Any idea?

Many thanks.


I’ll try to clarify in more detal. Linking the ASEC with basic monthly samples requires two steps. First, linking the ASEC to the March basic sample. Second, linking the March basic sample to other basic samples. Note that the ASEC sample includes individuals not only from the March basic sample, but also from other month’s basic samples. This linking process implies that it is only possible (at the present time) to link individuals from the ASEC who are also members of the March basic sample. Therefore, individuals who are part of the ASEC oversample (e.g. in the ASEC but not in the March basic sample) have CPSIDP values of 0. The working paper linked above also explains this detail.


Thanks, Jeff, for the clarification, which confirms what I suspected to be the reason. All the best, Tim