Linked representative samples "realprop_1" variable


I am working with the 1850-1880, 1860-1880, and 1870-1880 linked representative samples. I am working with the “realprop_1” variable that gives the reported value of real estate owned by the respondent in the first of the two linked years (i.e., in 1850, 1860, or 1870 - the variable was not recorded in the 1880 census and so “realprop_2” does not exist).

Normally, according to your own documentation, IPUMS records missing/blank values of real estate property as zero. This is consistent with the 1850-1880 linked representative sample - most values of realprop_1 are 0 as expected. But the 1860 and 1870 values of realprop_1 appear to record missing values as “999,999” instead. This seems to be an inconsistency, and there is nothing in the realprop documentation on IPUMS to explain what else “999,999” might refer to. I can only guess that “999,999” refers to missing/blank values and should actually be recorded as zero, because there are no zeroes recorded in the 1860 and 1870 values of realprop_1 otherwise. The smallest real estate value in 1870, for example, is 25 - which doesn’t seem right. This value should be mostly zeroes, but none exist for 1860 or 1870.

I guess this is less of a question and more of a comment, unless someone can correct me regarding what “999,999” refers to in the linked representative samples for the value of “realprop_1”. And thanks in advance if anyone can offer more information or correct me on this point.

The linked representative samples were released about 10 years ago, but haven’t been updated alongside other enhancements to the 1880 data. The result is that there are differences between the IPUMS USA documentation and these files. We are looking into our contemporaneous documentation to try and track down an answer to your specific question–my apologies that I don’t have the information just yet. I want to note that our historical data team is working on allowing users to link full count data in the near future (as a replacement to the linked representative samples). While we think that users will be excited about more data, I do want to note that we aren’t going to include the 19th century data initially.

Hi Kari,

Thank you very much for your reply. I’m pretty sure the 999,999’s recorded in the 1860-1880 and 1870-1880 samples should indeed be recoded as zero, but it would be nice to have a confirmation of that (and perhaps a link to some documentation?), since I didn’t see any documentation online. This is for a scholarly article and I just want to fully satisfy the reviewers on this point.

Thanks again for looking into this and take care,

I agree that it seems likely that most of these should be 0s, but another option is that some of them should be 999,9997 or 999,9998. There are a small number of cases with 999,999 values for realprop_1 in the 1850-1880 linked file too (but way fewer than in the 1860-1880 and 1870-1880 linked files, which do not have 0 values for realprop_1 as you note). Thank you for context for clarifying this issue; I will continue to follow up with my colleagues who worked on the linked files to track down an answer (and hopefully documentation) for you to provide to reviewers.

I am hopeful that the IPUMS USA revision history will be sufficient for your reviewers. On November 4, 2009, we made a change to the non-LRS version of REALPROP:

In the 1850-1870 samples, blanks on the REALPROP and PERSPROP items are now coded as 0. Some of the samples previously coded blanks as 999999.

I suspect this is where the online documentation and values in the LRS data became out of sync. The initial LRS release was in August 2008, and a final version of the data were released in June 2010; however, neither the revision history nor the LRS final release page document changes to REALPROP for LRS data. Furthermore, the note from November 2009 provides support for the theory that the 999999 values are actually blanks. Based on all of this, your decision to treat these as 0s seems perfectly appropriate.

Thank you very much for your follow up. I think you are right that it is safe to assume the LRS samples were not changed and use the 2008 realprop codes.

I wasn’t aware of the revision history document but will rely on and cite it in this project and in the future. Thanks again!

1 Like