sorry for a maybe somewhat dumb question, but I’m not really familiar with this kind of micro data and relabeling. I need average wage levels for sectors given through SIC/NAICS codes between 1960/1990.
I keep confusing different specifications etc, really lost hopelessly. I have downloaded some data with IND1950, IND1990, and then compared the labels from the do-file with the left column in https://usa.ipums.org/usa/volii/indcr… - they don’t match. What IND codes do I need to use that crosswalk to arrive at NAICS?
The crosswalk table you linked to describes the relationship between the 2000 Census codes (stored in the variable IND for 2000 and ACS/PRCS samples) and the 2000 NAICS codes (stored in the variable INDNAICS). The variable IND1990 was created by IPUMS-USA as a consistent measure of Industry across time.
The Codebooks for the 1960 and 1970 samples include a crosswalk of Census codes to the 1970 SIC codes. Similarly, the Codebooks for 1980 and 1990 include crosswalks between the Census and SIC codes from their respective years. If you would like to create a consistent NAICS coding system using the table you linked to, I would recommend downloading the 2000 5% sample, making sure to include the variables IND1990 (or IND1950 if it seems more applicable to your work) and INDNAICS. Once you have the extract downloaded, keep only one respondent per INDNAICS code and drop all variables besides IND1990 and INDNAICS. The resulting dataset should function as a crosswalk between the INDNAICS variable from 2000 and the IND1990 variable that appears in all of the samples. Once you have transformed IND1990 into it’s INDNAICS 2000 equivalent, you can use the labels from the crosswalk you linked to.
Since I have no idea on how to post a followup, here goes a second answer and a report (please forgive me abusing the system).
I would like to now merge the data with compustat - so, to go from INDNAICS to the current NAICS. Is there any file containing these matches? Copying/writing from the linked table seems pretty tedious.
There is no direct translation between INDNAICS and NAICS codes since INDNAICS codes can encompass multiple or partial NAICS codes (as noted at the bottom of the table you linked to in your previous post). With that in mind, I have attached two CSV exports, one with the 2000 INDNAICS to 1997 NAICS translations and one with the 2003 INDNAICS to 2002/2007 NAICS translations. These files will likely need some minor reformatting to be used as a proper crosswalk, but I hope this helps as a start.
This looks like exactly what I need: have you got a csv already? No worries if not: I can figure it out. I actually only need IND1990 → NAICS 2 digit cross walk, but one to 3 digit will of course work too.
We do not have a crosswalk that associates the harmonized IPUMS variable IND1990 variable with NAICS codes directly; however, because you only require NAICS codes to the second digit, I think you could create a data extract that includes IND1990 and INDNAICS for the year of interest in whatever format you prefer to work with the data (fixed width, Stata-formatted, .csv, etc.). You can derive the two-digit NAICS codes by creating a new variable that is a substring of the first two columns of INDNAICS.
I will note that because IND1990 is designed to increase comparability across time, it occasionally collapses two or more individual industry codes (IPUMS variable IND) into a single code. I suspect that won’t be problematic at the two-digit NAICS level because these are fairly broad groupings; however, it is theoretically possible that an IND1990 code could be associated with more than one NAICS code because of this aggregation.
To turn the microdata file into a crosswalk, you would then need to collapse the file to one observation per IND1990 code (or per unique combination of IND1990 and your new two-digit NAICS code to make sure you are able to at least view any cases where an IND1990 code is associated with more than one NAICS code).