Are that many married homosexual men not living together? - American Community Survey

Dear IPUMS team,

I am using a sample from the American Community Survey (ACS) as provided over IPUMS.

I am specifically interested in analyzing the wage gap of male homosexual couples. To do so I created a variable (this variable is called VAR_CH going forward) to indicate if an individual is cohabitating with a same sex partner. I created this variable after the following logic:

Var_CH: Two individuals are homosexual and cohabitating if they are linked by the SPLOC variable and share the same gender.

To check the accuracy of this variable I did a cross tabulation with the variable that indicates a same sex marriage (SSMC) for the years 2013+. I noticed that of the married ~45.000 homosexual male individuals in the sample ~10.000 individuals did not show up as cohabitating. In other words Var_CH did not identify them as two men living together but SSMC did indicate that these men were married to another men.

I assume this is the case because these ~10.000 married men do not live together with their partner. Is this assumption correct?

Is there a question in the ACS that would document a same sex marriage even if the men don’t live together?

Furthermore if the answer to these questions above is two times yes (i.e. these 10.000 married men don’t live together with their partner & SSMC captures a homosexual marriage even if the partners don’t live together) my follow up questions concerns the ratio. The share of men that are married but don’t live together seems rather large (i.e. 10.000/45.000 ~ 20%).

If I understand the information provided on IPUMS correct this ratio is probably even more skewed. Because married homosexual couples that live together show up in the sample twice compared to their counterparts that don’t live together. In other words, if two married men live together they are both interviewed and will show up as two homosexual men living together. If a men indicates that he is in a same sex marriage and the partner is not living in the same place then the partner is not interviewed and not in the sample. Thus, married homosexual couples that live together are counted double compared to married homosexual couples that don’t live together. Thus, the ratio is not 10.000/45.000 but actually rather 20.000/55.000 (here the 10.000 men that are married and don’t live together are counted double because the married men that live together are counted double as well). This ratio of 40% seems very large.

So my question here is: Did I understand the information on IPUMS correctly? Is my logic above correct? Is the rate of homosexual married men that don’t live together actually this large (i.e.40%) ?

P.: I did not used any weights for the ratios above, and thus they are just a first approximation.

Best regards,

Maximilian Schiele

It seems as though there may be a bit of confusion regarding some of the spouse/relate/marriage variables.

By definition the SPLOC variable identifies a spouse’s location in the household; therefore, if your logic for creating Var_CH relies on SPLOC then in all cases, Var_CH should indicate the men as cohabitating.

The RELATED variable (the detailed variable for RELATE, which is automatically included in your extract when you add RELATE) could help you identify same sex couples that may not necessarily be a spouse by using the code for “Unmarried partner” (1114).

Unfortunately, there is not currently a way for you to identify same sex couples that are married but do not live together. With the use of MARST, SSMC, SPLOC, and RELATE, you can identify same sex married couples/same sex partners, however, there is no information collected on the sex of someone who is married but the spouse is absent from the household, we simply know if their marital status.

It should also be noted that SSMC is a household-level measure which indicates if the householder and spouse are a same sex married couple. When working with a rectangularized extract (the default), this will put the SSMC indicator on every individual in the household, but it really only represents the relationship of the householder and their spouse.

I hope this helps.

Thank you so much for your reponse!

I’m still a bit confused to be honest. To make it concrete, how would i correctly identify homosexual couples?

a) that are married

b) that are not-married

Thank you in advance,


First, it should be noted that it is very difficult to identify same-sex couples because there are no explicit questions about it. The Census Bureau has some resources on the subject, including this paper on measuring same-sex couples, as well as this fact sheet. With that said, there are several ways in which identifying same-sex married couples could be accomplished using data from IPUMS USA:

  1. Use SSMC (>1) to identify households with same-sex married couples. Because SSMC is a household level variable, keep in mind that you will want to remove other members of the household that are not spouses. Note that SSMC is a variable created by the Census Bureau using restricted information. Additionally, it is only available in the 2013 ACS onward. As noted in the description for SSMC, prior to the 2013 ACS/PRCS, same-sex married couples were recoded by the Census Bureau from married to unmarried partners. Furthermore, it should be noted that same-sex married couples are only identified by SSMC if the householder (RELATE = 01) is a part of the same-sex married couple.

  2. Use SPLOC along with the attach characteristics option to get SEX_SP (sex of spouse). SEX_SP==SEX will identify same-sex couples. You can then use MARST to explicitly identify those same-sex couples who are married. In contrast to SSMC, SPLOC is an IPUMS-constructed variable; it was created based on inference from the public use microdata only. Additionally, SPLOC allows you to identify same-sex couples that do not include the householder. Detailed information about SPLOC can be found here.
    As noted in my original response, there is not currently a way for you to identify same-sex couples that are married/unmarried but do not live together. However, you can identify same-sex non-married cohabitating couples by the following:

  3. As listed above for identifying married same-sex partners, you could use SPLOC along with the attach characteristics option to to get SEX_SP (where, again, you would want SEX_SP=SEX) and RELATED_SP (detailed relationship of spouse) to identify same-sex unmarried partners.
    In general, your analysis will most likely be easiest if you limit your data to just heads of household and their spouses. Another note, depending on your analysis and other variables of interest, you may also want to look into using IPUMS CPS. In addition to SPLOC and MARST, IPUMS CPS has a variable specifically for cohabitating partners, PECOHAB. This variable is available from 2007 onward. Unlike the inferred variables SSMC and SPLOC, it is generated from a direct question to unmarried respondents in households with unrelated adults.