bd_CPS_id.ipynb skips months prior to May 1995. I could try to generate IDs for the 1989-93 data with the current code. I could also try to customize the code to handle Jan 1994-April 1995 without trying to match to May 1995 (or later).

Pretty confident that I've figured out the 1993-1994 match process (since variables differ). If I match by HHID, STATE, and match the last digit of HHID2 (1994-onward) to H-HHNUM (1993), the results look to be correct.

UPDATE: There's more to this--several cases have a second to last digit of HHID2 that is not 0 and therefore likely has a counterpart variable in the 1989-93 data. Perhaps these are cases where one unit gets split into two or similar. I'm not sure I'll find a counterpart variable in the pre-1994 data.

For the example that I picked, all of cases work out if I append HHID2 % 100 to the end of HHID and compare it to pre-1994's: HHID | 0 | HHNUM. I still think the 0 is a variable somewhere within the pre-1994 data, but haven't found it yet.

Because the pre-1995-break data are not going to change (whereas the CPS_id is going to be generated for new months), I could create a new cell in bd_CPS_id.ipynb that adds checks if the pre-1995 IDs are available, and, if not, generates them, then adds them to the dictionary containing the 1995-onward IDs.

Part of the problem here (which I run into often) is that I do not outline (in words) what is done by bd_CPS_id.ipynb meaning how the IDs are generated. This is covered in issue #171 and should be given some consideration as I make updates.