A Research Note Describing a More Inclusive Approach to Identifying Same-Sex Cohabiters in the American Time Use Survey

Abstract

Identification of individuals in same-sex relationships in the American Time Use Survey (ATUS) is of increasing interest to the research community. While the ATUS interviews one person per household, by using information about who else lives in the household, researchers can easily identify respondents in coresident same-sex couple arrangements. Previous research has outlined two approaches to identifying individuals in same-sex relationships in the ATUS that use information on the sex of household members. We extend that work in this research note by using additional information collected from a direct question to identify unmarried cohabiting partners in the Current Population Survey (CPS). We identify 23% more individuals in cohabiting same-sex relationships when we use the CPS direct question information than when we use information from the ATUS alone. We argue that this identification strategy is more inclusive of individuals in same-sex cohabiting relationships.

Introduction

The American Time Use Survey (ATUS) is increasingly used by the research community to study the daily lives of individuals in same-sex couple relationships (Augustine et al. 2017; Fettro 2018; Genadek et al. 2020; Martell and Roncolato 2016; Martin-Storey et al. 2018; Prickett et al. 2015; Schneebaum 2013). The ATUS has been fielded annually since 2003 and is a nationally representative, 24-hour time diary survey of one randomly selected individual per eligible household. Households are sampled from the Current Population Survey (CPS) and ATUS interviews are conducted two to five months following the completion of the CPS. The ATUS collects a household roster that lists every person in the household as well as each person's sex and relationship to the ATUS respondent. Using information about household members, researchers can easily identify respondents in coresident same-sex couple arrangements.

To our knowledge, Prickett and colleagues (2016) were the first to discuss the nuances of identifying ATUS respondents who are in same-sex couple relationships. Their recommendation is to use information about the respondent's and spouse's/partner's sex collected at the time of the ATUS interview rather than during the CPS interview.¹ We agree with this recommendation but also build upon their work and urge researchers to consider additional information from the CPS that leads to an increase in the identification of ATUS respondents in unmarried cohabiting same-sex relationships. The same additional information has been used by the U.S. Census Bureau to improve identification of unmarried cohabiting couples in the CPS (both same-sex and different-sex) (Kreider 2008; Lofquist and Ellis 2011).

Accurate measurement of individuals in same-sex couples is critical given the changing landscape of American families (e.g., Obergefell v. Hodges, 2015) and the growing number of same-sex couples in the United States. This is underscored by recent use of the ATUS to study individuals in same-sex relationships and the rarity of large-scale, representative samples that allow for the identification of same-sex cohabiters. Accordingly, we illustrate the sample size increases achieved by using a second method to identify individuals in unmarried cohabiting same-sex relationships and compare the demographic characteristics and time use of individuals identified using the method recommended by Prickett and colleagues (2016) and our proposed extension.

Background

Beginning in 2007, the Census Bureau improved the way it collects information about cohabitation in the CPS (Kennedy and Fitch 2012; Kreider 2008). Prior to 2007, cohabiting partners were characterized as “householder couples” if a coresident household member was identified as an “unmarried partner” of the CPS respondent (Kreider 2008). “Additional couples” are identified from a direct question that asks “Does [respondent] have a boyfriend/girlfriend or partner in the household?”² (Kreider 2008). If a boyfriend/girlfriend or partner was present, the respondent identified the individual's position in the household roster (i.e., person line number). The direct question improved estimates and resulted in a 21% increase in the number of different-sex cohabiting couples in 2007 compared to the relationship question alone (Kreider 2008). Additional different-sex couples tend to be slightly younger and more likely to include never-married individuals, and a lower percentage have children compared with householder couples (Kreider 2008).

The CPS yields reasonable estimates of same-sex couples compared with the American Community Survey (ACS), which has been the primary source of such estimates, when both householder and additional same-sex couples are considered (Lofquist and Ellis 2011). This improved measurement of cohabiting couples in the CPS may also be leveraged when using the ATUS, because the CPS is the sampling frame for the ATUS. We propose an extension of Prickett and colleagues' (2016) method for identifying individuals in unmarried cohabiting same-sex couple relationships in the ATUS via improved measurement of unmarried cohabiting couples in the CPS (Kreider 2008; Lofquist and Ellis 2011). The inclusion of individuals in “additional couples” should result in more complete representation of, and improved estimates of, time use for individuals in same-sex couple relationships in the ATUS.

Methodology and Analysis

We use information about relationship to the ATUS respondent to identify individuals in householder couples³ and leverage the linkage between the CPS and the ATUS to use the direct question about unmarried partners in the CPS⁴ to identify individuals in additional unmarried cohabiting same-sex relationships in the ATUS (Kreider 2008; Lofquist and Ellis 2011).⁵ The first step in identifying individuals in additional couples is to determine whether the ATUS respondent who is not in a householder couple at the time of the ATUS (on the basis of the variables RELATE in the IPUMS ATUS and TERRP in the original ATUS data from the Bureau of Labor Statistics (BLS)) identified a partner in the household at the time of the CPS. Identification of a partner in the household during the last CPS interview is based on answers to the direct CPS question (considering the variable PECOHAB in both IPUMS ATUS and the original ATUS data from BLS). If the directly identified boyfriend, girlfriend, or partner at the time of the CPS is in the household at the time of the ATUS interview (2–5 months later), we classify the ATUS respondent as being in an additional couple. The second step is to identify individuals in same-sex relationships. We use the sex values from the ATUS for both the ATUS respondent and partner (SEX in IPUMS ATUS and TESEX in data directly from BLS) to code ATUS respondents as being in a same-sex relationship (Prickett et al. 2016).

Using 2007 CPS data, Kreider (2008) identified 17.1% of cohabiting different-sex couples as “additional” couples using the direct question method. Using the CPS direct question, we expect to identify a similar proportion of individuals in the 2007–2019 ATUS who are in “additional” same-sex couple relationships. Table 1 shows the share of ATUS respondents in householder or additional same-sex couple relationships. We include individuals in different-sex couple relationships for comparison purposes. We do not identify any individuals as members of additional couples in 2003–2006 because the direct question was not asked. From 2007 to 2019, the number of individuals in unmarried cohabiting same-sex relationships increases by 131 (for a total of 697) when we include individuals in additional couple relationships. This translates to 18.8% of individuals in an unmarried cohabiting same-sex rela-tionship identified from the direct question, compared with 18.3% of individuals indifferent-sex relationships over the same period, which is slightly higher than Kreider's (2008) estimates from the CPS. The weighted estimates are slightly higher, with individuals in additional couple relationships making up 20.1% and 20.0% of same-sex and different-sex cohabiting couple relationships, respectively. The share of individuals in same-sex additional couple relationships varies across years (Table 2). In 2019, 13.2% of individuals in unmarried cohabiting same-sex relationships were identified using the direct question method, compared with 35.7% in 2008. ATUS respondents whom we classify as being in additional couple relationships on the basis of the direct question from CPS have partners in the following relationships, as reported at the time of the ATUS: 65%, housemate/roommate; 9.9%, roomer/boarder; and 25.1%, other nonrelative.

The share of individuals in same-sex additional couple relationships is in line with expectations (Kreider 2008), but how similar are the individuals in additional couples to those in householder couples? We first compare the demographic characteristics of unmarried cohabiting ATUS respondents in same-sex additional couple relationships (2007–2019) to those in householder couple arrangements in all years (2003–2019) and in 2007–2019 (Table 3). For men and women combined, there are few statistically significant differences between individuals in unmarried cohabiting same-sex relationships identified using the two methods; most differences are among women. Women in additional couple relationships are less likely to be White, are less educated, and are younger than women in householder couples.

We also compare the time allocation for individuals in additional and householder couple same-sex relationships. Table 4 shows the average daily minutes spent in eight broad activities (sleep, grooming, work, household activities, leisure, recreation, caregiving, and other activities).⁶ The raw differences between individuals identified using the additional and householder couple methods are reported in column 3, and the differences net of demographic characteristics (reported in Table 3) and diary day are shown in column 4. Three of the eight time allocation categories show significant differences. The largest difference is in sleep, with individuals in additional couple relationships reporting about 30 more minutes in sleep on average, net of controls. Respondents in additional couple relationships spend 12 fewer minutes in recreation or exercise than respondents in householder couple relationships (column 4) and 10 fewer minutes caring for children (column 3). Men in additional couple relationships spend nearly one hour more sleeping than men in same-sex householder relationships.

Conclusion

We extend previous work on the measurement of individuals in same-sex couples in the ATUS (Prickett et al. 2016) and of cohabiting couples in the CPS (Kreider 2008; Lofquist and Ellis 2011) to identify 131 additional individuals between 2007 and 2019 who are likely in unmarried cohabiting same-sex couple relationships in the ATUS. Improved measurement of unmarried cohabiting couples in the CPS yields a substantial increase in the number of different-sex cohabiting couples (Kreider 2008) and produces similar estimates of same-sex couples as generated from the ACS (Lofquist and Ellis 2011).⁷ Extending this approach to the ATUS yields 23% more (131/566 in Table 2) individuals in unmarried cohabiting same-sex couple relationships in the ATUS across the 2007–2019 period. We show that the individuals in additional couple relationships are quite similar to those in householder couple relationships in terms of demographic characteristics and that their time use is also more similar than different.

Despite the potential of this extension, cautions are in order. Unlike use of the direct question to identify additional cohabiting couples in the CPS (Kreider 2008; Lofquist and Ellis 2011), it is possible that during the two- to five-month gap between the CPS and ATUS the individual identified as a partner in the CPS is no longer the ATUS respondent's partner despite continued coresidence.⁸ There are also occasional discrepancies in the reporting of sex for ATUS respondents and partners between the CPS and ATUS, which could result in misclassification of different-sex couples as same-sex couples or vice versa. Nevertheless, we privilege the ATUS information over the CPS information because since 2014 it has been verified at the time of the ATUS interview and changes are verified by the Census Bureau prior to release of the data.⁹

On the basis of previous research (Kreider 2008; Lofquist and Ellis 2011) and the evidence we have presented here, we encourage researchers to leverage the link between the ATUS and the CPS to identify individuals in additional cohabiting same-sex couple relationships in the ATUS using the direct cohabitation question in the CPS. This should be done cautiously because we cannot say with certainty that the individuals in additional couple relationships remain partnered across the gap between the CPS and ATUS despite continued coresidence. We encourage researchers to perform sensitivity analyses by the method used to indicate individual membership in a same-sex couple. Our analysis of time use yields few differences by same-sex couple identification method, but we selected only eight activity groupings out of the many types of time use that can be analyzed with the ATUS. Given the large difference in sleep for men, we recommend comparison of means for key time use outcomes by couple type separately for men and women. Researchers may also want to ensure that their results are robust to the inclusion of individuals in additional couples in their analysis or include a control for the method of same-sex couple identification in their multivariate analyses.

Application in the ATUS of logic used in the CPS to improve measurement of individuals in cohabiting couples has the potential to produce more robust estimates of the ways that individuals in same-sex cohabiting couples spend their time and to increase the utility of this data source for research on same-sex cohabiters. This is critical for several reasons. First, same-sex cohabiting couples are a small but growing subpopulation; reclassification of single individuals into cohabiting relationships may yield more accurate estimates and knowledge about how cohabiters spend their time. Second, increased sample size may be especially relevant for researchers who want to use ATUS module data that are available only in a selected number of years. Furthermore, accurate measurement of the couple arrangements of individuals in the ATUS has implications that extend beyond individuals in same-sex cohabiting relationships to those in different-sex cohabiting relationships and those who are single. In summary, measurement of relationship type in the ATUS is an issue that requires careful attention and detailed robustness checks but can be enhanced using the relationship information from the CPS.

Acknowledgments

This work was supported by Minnesota Population Center grant P2CHD041023, University of Colorado Population Center grant P2CHD066613, IPUMS CPS grant R01HD067258, and IPUMS Time Use grant R01HD053654, all funded by the Eunice Kennedy Shriver National Institute for Child Health and Human Development.

Notes

References