Overview of Subfamilies and Family Interrelationships
On this page:
- What are subfamilies, and how does the IPUMS measure them?
- How do subfamilies differ from the "families" as measured by the IPUMS?
- How do IPUMS subfamilies differ from Census Bureau subfamilies?
- What weights should I use when studying subfamilies?
- What else should I know about subfamily measurement?
What are subfamilies, and how does the IPUMS measure them?
Subfamilies are simply families that live in the household of someone else. The Census Bureau defines two main types of subfamilies. Married-couple subfamilies comprise a husband and wife living with or without children. Parent-child subfamilies comprise an unmarried parent living with one or more of their own, never-married children under 18 years of age. Subfamilies can be further divided into related subfamilies (whose members are related to the household head by birth, marriage, or adoption) and unrelated subfamilies (whose members are not related to the household head). Unrelated subfamilies (which used to be called secondary families) are no longer measured by the Census Bureau, but the IPUMS includes them for users' convenience.
The IPUMS provides four variables concerning subfamilies:
|SFTYPE||Person||Type of subfamily (if any) to which each person belongs||---|
|SFRELATE||Person||Relationship within subfamily (if any)||---|
|SUBFAM||Person||Identifies the subfamily (if any) to which each person belongs||FAMUNIT||Identifies the family to which each person belongs|
|NSUBFAM||Household||Number of subfamilies in household||NFAMS||Number of families in household|
For more information, see the individual variable descriptions.
The following table gives examples of these variables in two hypothetical households:
|Example and person number||Relationship to head||Spouse location||Mother location||Father location||Age||Sex||Marital Status||SFTYPE (Subfamily type)||SFRELATE (Relationship within subfamily)||SUBFAM (Subfamily number)|
|6||Other relative||0||5||0||12||M||Never married||4||3||2|
In Example 1:
- Persons 3 and 4 are relatives of the head and are married, but do not have any children in the household. They are therefore classified as a married-couple related subfamily without children (SFTYPE = 2) and have been coded as the first subfamily in the household (SUBFAM = 1), with the husband as the reference person (SFRELATE = 1) and the wife as the spouse (SFRELATE = 2).
- Person 5 is also a relative of the head and has a child (person 6) who is under 18 and has never been married. They are therefore classified as a mother-child related subfamily (SFTYPE = 4) and have been coded as the second subfamily in the household (SUBFAM = 2), with the mother as the reference person (SFRELATE = 1) and her son as the child (SFRELATE = 3).
- Persons 7 and 8 are not relatives of the head, but they are married and have a child (person 9) who is under 18 and has never been married. They therefore constitute a married-couple unrelated subfamily with children (SFTYPE = 5) and have been coded as the third subfamily in the household (SUBFAM = 3), with the relationships (SFRELATE) as described above.
In Example 2:
- Person 2 is the father of the head and has a child (person 4) who is under 18 and has never been married. The two are thus classified as a father-child related subfamily (SFTYPE = 3), the first subfamily in the household (SUBFAM = 1). Note that person 3 is also the child of person 2, but is at least 18 years old and is not included in the subfamily.
- Person 5, who is unrelated to the head, is the father of person 6. They constitute a father-child unrelated subfamily (SFTYPE = 7), the second subfamily in the household (SUBFAM = 2).
Finally, note that all members of the household who are not members of a subfamily are coded as 0 for all subfamily variables, whether or not they are related to the head.
How do subfamilies differ from the "families" as measured by the IPUMS?
SUBFAM is analogous to FAMUNIT in that it identifies membership in unique family units within each household, but the specific related unit measured by each is different.
FAMUNIT groups all members of households into extended family units. The household head and all relatives are coded 1. Non-relatives are grouped into family units when possible; otherwise, each non-relative is assigned a unique FAMUNIT code.
In SUBFAM, household heads are never in subfamilies and are always coded as 0. The head's relatives may be grouped into subfamilies with one another as described above, but are never grouped with non-relatives. Non-relatives who are related to other non-relatives can be grouped into subfamilies with one another. Non-relatives who are not related to anyone in the household are not in subfamilies and are coded as 0.
In example 1, persons 1-6 are thus included as members of the head's family (FAMUNIT = 1), even though this group also includes two subfamilies. The same is true for persons 1-4 in example 2, which includes one subfamily. Additionally, FAMUNIT counts as separate family units all individuals who are unrelated to the head and who live without a spouse or children (persons 10 and 11 in example 1; persons 7-9 in example 2); SUBFAM does not and codes them as 0. However, members of all unrelated subfamilies (persons 7-9 in example 1; persons 5 and 6 in example 2) receive unique codes in both FAMUNIT and SUBFAM.
This difference carries over to NSUBFAM (the total number of subfamilies in the household) and NFAMS (the total number of families in the household). Example 1 includes three subfamilies (so NSUBFAM = 3), and example 2 includes two subfamilies (NSUBFAM = 2). NFAMS is more complicated. Example 1 includes one group of people related to the head, one unrelated subfamily, and two unrelated individuals, for a total NFAMS of 4. Example 2 includes one group of people related to the head, one unrelated subfamily, and three unrelated individuals, for a total NFAMS of 5.
|Example and person number||Relationship to head||Spouse location||Mother location||Father location||Age||Sex||Marital Status||SUBFAM (Subfamily number)||FAMUNIT (Family number)||NSUBFAM (Number of subfamilies)||NFAMS (Number of families)|
|6||Other relative||0||5||0||12||M||Never married||2||1||3||4|
How do IPUMS subfamilies differ from Census Bureau subfamilies?
Users should note that the IPUMS subfamily variables are not based on the original Census Bureau subfamily variables. Subfamilies, whether constructed by the Census Bureau or by the IPUMS, are not based on explicit survey items about how one is related to others in one's household. Rather, they are educated guesses based on other variables, and the IPUMS and the Census Bureau differ in the variables and procedures they use to construct subfamilies.
IPUMS subfamilies are based on the IPUMS family interrelationship variables SPLOC, MOMLOC, and POPLOC, which in turn contain information from other variables that considerably improves the accuracy of subfamily identification.
The Census Bureau's procedures for classifying subfamilies have varied considerably over time, and prior to the more recent ACS samples were often inaccurate. While the Census Bureau does construct their subfamilies using surname similarity--an item not publicly available after 1950--it requires that subfamily members' surnames match under some conditions, which may exclude subfamilies where one spouse kept their given name. Additionally, the Census Bureau does not measure unrelated subfamilies. For more information, see this article and this conference paper.
For these reasons, and because IPUMS subfamily measures are fully consistent with other IPUMS variables, the IPUMS recommends that researchers use the IPUMS subfamily measures when analyzing multiple years of data. We also make available analogous subfamily measures as classifed by the Census Bureau from 1960 onwards: CBSFTYPE, CBSFRELATE, CBSUBFAM, and CBNSUBFAM. The Census Bureau subfamily variables are useful mainly for comparing to published estimates such as those found inthe Census Bureau's summary files.
What weights should I use when studying subfamilies?
Complex household structures naturally create questions about what weights to use. In general, the IPUMS advises the following:
- Use the PERWT of individuals when studying the characteristics of individuals within subfamilies. Examples include the ages of children in subfamilies or the incomes of wives in married-couple subfamilies.
- Use the HHWT when studying the characteristics of subfamilies themselves. Examples include the total number of parent-child subfamilies or the average total income of subfamilies. (Note: Before May 2011, IPUMS advised using the PERWT of the subfamily's reference person in these situations. However, the Census Bureau uses HHWT when obtaining counts of subfamilies, so HHWT should be used for comparing PUMS figures to published estimates.)
- Use the HHWT of the household when studying the characteristics of households containing subfamilies. Examples include the proportion of households containing subfamilies or the average number of people in households containing subfamilies.
What else should I know about subfamily measurement?
Users should be aware that the IPUMS does not link unmarried partners as spouses. This means that many people who could be considered part of the head's family (if they are a partner of the head) or part of a married-couple subfamily (if they are a partner of another household member) will be classified instead as an unrelated individual (if they do not have children living with them) or as the reference person in a parent-child subfamily (if they do have children living with them).
IPUMS plans to link unmarried partners as spouses in the future; until then, users should be aware that the IPUMS subfamily variables may overestimate the prevalence of parent-child subfamilies.
In a small number of households, subfamilies will overlap. For example, there will be a parent-child subfamily in which the child is also the parent to another child. Rather than coding these as two separate subfamilies, the IPUMS classifies all these individuals as a single parent-child subfamily.
Finally, the subfamily variables do not apply to group-quarters cases, all of whom receive codes of '0' on all subfamily variables.