User Note - Family Interrelationship Variables
By design, the NHIS is a household survey, enabling the use of the NHIS data to estimate associations between family structure or characteristics of other household members and health-related outcomes of interest. For example, NHIS has been used to study topics like racial disparities in self-rated health for interracial couples; health outcomes for children of same-sex versus different-sex co-residential parents; and the role of living arrangements in quality of life and psychological distress among older adults.
Users interested in exploring the relationships among household members should note the existence of a set of family interrelationship variables created by IPUMS staff that impose a consistent and transparent approach to the identification of a person's (probable) coresident spouse/partner and/or parent(s). These family interrelationship variables are used by our extract system to allow users to seamlessly attach the characteristics of other household members -- characteristics like spouse's race or employment status, parental health status, or cohabiting partner's health insurance status -- as new variables on the extract. For instructions on how to attach the characteristics of other household members as new variables to your extract, please see this blog post or video example.
IPUMS NHIS family interrelationship variables vs. NHIS relationship to reference person
The original NHIS data includes a variable (RELATE) that shows how an individual household member is related to a reference person. However, relationships among other household members described by this variable may be ambiguous. For example, in a household with a "householder", two "children", and one "grandchild," it is not immediately clear who the parent of the "grandchild" is. In contrast, IPUMS NHIS uses additional information--including age, sex, and marital status--to infer relationships between persons who are not the family reference person thereby reducing ambiguity about these relationships.
These IPUMS NHIS family interrelationship variables are available for all years of IPUMS NHIS and comparable across years (as well as across US IPUMS datasets). IPUMS NHIS family interrelationship variables identify the location in the household of each person's spouse (SPLOC), mother (MOMLOC), and father (POPLOC). Additional variables (MOMLOC2 and POPLOC2) are used to identify a second parent of the same-sex. IPUMS NHIS also provides detailed rule variables for each these spouse/partner (SPRULE) and parent location variables (MOMRULE, POPRULE, MOM2RULE, POP2RULE), which report how a spouse/partner and parent location variables were inferred and the level of ambiguity around that inference.
There are a few key differences between the IPUMS NHIS family interrelationship variables and those provided by NCHS in the original public use files:
- Types of relationships included
- Attaching characteristics of a person's spouse/partner and parents
The IPUMS NHIS variables are available for all years, and are broadly comparable over the time series. In contrast, NCHS only offers the location of a person's spouse/partner (SPOUSEPNUM) for 2006-forward, and parental locations (MOMPNUM and DADPNUM) for 1997-forward. NCHS uses self-reported family member location in the household. IPUMS NHIS uses a two-stage protocol that uses information on age, sex, marital status, and relationship to householder to identify a person's probable spouse/partner and parents; this makes the IPUMS family interrelationship variables available for all years of data. The two-stage protocol first prioritizes links by clarity of relationship, and then selects the most appropriate spouse/partner or parent location when choosing among multiple potential links. Other than changes in the values that can be reported for the relationship to the household reference person and changes to NCHS editing procedures , the IPUMS NHIS family interrelationship variables are comparable over time. The logic used to assign the links is consistent over time, and clearly communicated via the corresponding rule variable. IPUMS NHIS does not use the self-report variables in assigning family interrelationship links, and instead uses them to check the validity of the IPUMS linkages. IPUMS NHIS assignments for spousal location match self-reported values in over 99.99% of cases ; match self-reported mother location 99% of the time, and match self-reported father location over 98% of the time .
Types of relationships included
IPUMS NHIS protocol for assigning family interrelationships accounts for same-sex and cohabiting couples. Among those identified as same-sex couples by IPUMS in the 2006-2014 NHIS survey years, 100% of IPUMS spouse/partner assignments match those self-reported values provided by NCHS. Additionally, IPUMS NHIS allows for easier identification of same-sex couples prior to 2006. Please see User Cautions below for important information on changes in the NCHS editing procedure that limit the validity of same-sex spouse/partner links in the 2004-2007 NHIS files. Many non-matched parent locations between NCHS and IPUMS NHIS assignments are not the assignment of a wrong parent, but the assignment of an additional parent. One other notable difference between NCHS and IPUMS NHIS family interrelationship variables is that IPUMS NHIS identifies social relationships instead of exclusively biological relationships. For example, if a "householder" lives with an "unmarried partner" and that partner's children, IPUMS links both the "unmarried partner" and the "householder" as the parent to these children, while the NCHS-provided variables will only link the "unmarried partner" as a parent.
Finally, users may use the IPUMS NHIS "Attach Characteristics" feature to add include information about a person's spouse or parent as a variable on the individual person record. Users simply add the variables and sample years of interest to their IPUMS NHIS data cart and proceed to creating their data extract as per usual. However, from the Extract Request page, users can choose to "Attach Characteristics." Selecting this button will bring the user to a page where they create new variables that attach the characteristics for a person's spouse, mother, or father for each variable that is already in the user's data cart. The original NHIS data files include variables about coresident mother's and father's education for each individual (in 1998-forward), and users can use these parental or spousal location variables to programmatically attach other characteristics in years where these variables are offered by NHIS. However, the IPUMS NHIS feature allows users to automatically include these variables in the extract and is available for all years of data.
When identifying same-sex married or cohabiting couples in IPUMS NHIS, users should be aware of changes to NCHS editing procedures that may affect the validity of these same-sex couple links in a limited number of years. Different-sex couples where one member of the couple's sex was misclassified can comprise a large proportion of same-sex couples (NCHS, 2015; Black, Sanders, and Taylor, 2007; Kreider and Lofquist, 2015). Prior to 1997, there is limited documentation about any editing done for same-sex couples in the NHIS. Beginning in 1997, NHIS staff initiated more systematic checks of same-sex couples (described below). However, from 2004 through the second quarter of 2007, NCHS did not further review couples identified as same-sex. An informal review by NCHS staff of data files for 2004 through the second quarter of 2007 using an approach described below find that 77% of couples identified as same-sex were likely different-sex couples where one member's sex code was incorrect (NCHS, 2015). Analyses of same-sex couples in 2004 through the second quarter of 2007 using these data are discouraged. Analyses of same-sex couples prior to 1997 should note the limited documentation available about any potential editing.
Additional information on specific NCHS editing procedures for same-sex couples.
Prior to 1997, there is limited documentation about any editing of same-sex couples. In 1994, the field representative's manual notes: "If two persons of the same sex (two males or two females) consider themselves as married, show the relationship as 'same sex spouse'." In 1995 and 1996, the field representative's manual notes that the sponsor doesn't recognize same-sex marriages and encourages the enumerator to conduct separate interviews as they would for unrelated persons living in the same household, but may conduct the two interviews together if more efficient. The 1995 manual also specifies that same sex spouses will be considered unrelated for supplements and both persons will be selected as a sample adult and require separate interviews. The 1996 manual notes that a person's legal marital status should be reported and explicitly indicates that those identified as a same-sex spouse who report being married should either leave marital status blank and include a note if the same-sex partners insist they are married to each other, or, if the same-sex spouse is still legally married to a different-sex person, recorded as "Married spouse not in HH" or "Separated." Beginning in 1997, NCHS staff flagged all same-sex married couples. In 1997-2003, staff reviewed the confidential data files containing household members' names, and tried to confirm the sex of each member of the couple based on names. If one person's sex was believed to be incorrectly coded based on this information and the person's "correct" sex could be determined with "high probability" based on their name, the sex of the person and other correlated variables were changed. Data were not modified for persons with gender neutral names.
In an effort to streamline review of records and release data more quickly, same-sex couples were not manually reviewed because they comprise a small percentage of the whole sample in 2004 through the second quarter of 2007. In quarter 3 of 2007, manual review of all same-sex couples was reinstated and modified. First, enumeration notes are reviewed to determine if there was any sex-specific language that could be used to determine the person's sex. If enumeration notes cannot be used to determine the sex, the specific enumerator was asked if they recalled the sex of both people. The third step is to call the respondent directly. If none of these steps yielded information on the couple members' correct sexes, their names are reviewed as described in the 1997-2003 editing process. Additionally, in 2008, a verification screen was added, allowing the enumerator to review and confirm the sex and relationship for members of same-sex couples.
An NCHS report with additional details on these editing procedures is available here. There is an MPC working paper on the development of the new family interrelationship variables for US IPUMS data projects. A version of this paper was also published in the Journal of Economic and Social Measurement in 2017.
Gorsuch, M. M., & Williams, K. C. W. (2017). Family matters: Development of new family interrelationship variables for US IPUMS data projects. Journal of Economic and Social Measurement, 42(2), 123-149.
Kreider, R. M., & Lofquist, D. A. (2015). Matching survey data with administrative records to evaluate reports of same-sex married couple households. Washington, DC: US Census Bureau. SEHSD Working Paper, (2014-36).
National Center for Health Statistics - Centers for Disease Control and Prevention. (2015). Changes to data editing procedures and the impact on identifying same-sex married couples: 2004-2007 National Health Interview Survey.