Linking CPS and ATUS pre 2004.

I am attempting to merge the CPS with the ATUS/CPS (I am supplementing the already linked dataset) from 2003-2016. I am using the merging guidelines suggested by the ATUS guidebook (i.e. merging on hryear4 hrmonth pulineno hrhhid hrsersuf ). About 3% of my sample in the ATUS does not match to the CPS. All of the unmatched observations come the earlier time period (i.e. before 2004 interview.). Does anyone have any suggestions for why this might be the case? I should also note I am only trying to match individuals in the ATUS to their final CPS interview.

There are several variables, available in IPUMS Time Use (ATUS), that are helpful for this purpose. First, the CPSIDP variable uniquely identifies individuals across CPS samples. All individuals in ATUS, except for thsoe who entered the ATUS household after the final CPS interview, may be located in the final month of the household’s participation in the CPS. This information is indexed by YEAR_CPS8 and MONTH_CPS8. Additionally, the PRESENCE variable identifies ATUS cases that should link to the CPS.