Count of same-sex couples in 1990


I was trying to get the count of same-sex couples in 1990 at the tract level using NHGIS, which at the best of my knowledge is available.

I used the geographic unit filter Tracts, the year filter 1990 and the topics filter Sex and Relationship (I also tried OR as well), but I couldn’t find the source table. I was able to find the same counts for 2000 and 2010, but I couldn’t for 1990. Could you help me finding the table if it is available?


Yes, there are a few 1990 tables providing counts of same-sex couples, and yes, it is a little difficult to find these.

First, I recommend using the “Marriage” topic in this case. This topic includes tables that use a classification of “Presence and Sex of Unmarried Partners”.

You should then be able to find table NPB12 in the 1990_STF4b dataset.

This dataset doesn’t provide data for the standard Tract geographic level, but it does provide data for tract parts, which mostly match up with whole tracts. When selecting the level, you’ll have to look for one of these “compound” levels. (Find and select the SHOW COMPOUND LEVELS option.)


Thanks for this inquiry. It reminds me that it would be helpful for NHGIS to include a “same-sex couples” topic. We’ll keep that in mind as a potential improvement for a future update.

1 Like

Thanks! It helped a lot.

Can I ask a clarifying question on how to use this tract parts data? I have downloaded the dataset, and when I sorted it by state, county, and tract number, I could see there are multiple tracts (parts of tracts) with same information on state, county, and tract number. I am assuming that to make the tract level information, I just need to sum up those data points with same information on state, county, and tract. Is that correct?


I am sorry for posting multiple questions in a row before I got answers, but I encountered some problems while trying to do what I described in my clarifying question.

To be specific, I was making tract id by concatenating state, county, and tract ids making it 11 characters, adding 0 as buffers. What I just found was that in the tract parts data in 1990 I downloaded, the tract number system is pretty different from other years’ data set.

For example, Tract 1 has tracta value of just 1, Tract 53.02 has value of 5302. I found that in the data set, Tract 101 has tracta value 101 and Tract 1.01 has also 101. If this is the case, I think it is pretty hard to make tract level id out of this data set unless I use anpsadpi variable that is describing tracts more accurately. I tried ignoring this problem and making tract id and merge with other data set, but the tract ids didn’t really match. Many of them were left without being matched.

Would there be other way that I don’t know that can use tracta to make comparable tract id with other data sets? Thanks!

In answer to your first question, yes, you can compute counts for whole tracts by summing counts for the corresponding tract parts.

To construct unique identifiers for whole tracts, you’re right that cases like 101 and 1.01 can be ambiguous. However, these codes are not actually ambiguous in the original data file that you get from NHGIS. That file includes leading zeros that distinguish, e.g., 00101 from 0000101. The problem is that many software packages read these codes as numbers and drop the leading zeros. I’m guessing that’s what’s happened in your case.

One option is to ensure that your software is reading these codes as text and not as numbers. (How to do that depends on which software you’re using.) Another option is to extract the state, county, and tract codes from the GISJOIN field, which provides unique identifiers that concatenate all of those codes as well as the codes for other units in the geographic level hierarchy.

1 Like

I see. Thank you so much!