I am looking at ACS data to calculate various fertility rates and I’m noticing inconsistencies between FERTYR and NCHILD. From there, I am also noticing that children are in a household without an assigned MOMLOC.
Attached is an example. Using the microdata, what would be the best way to calculate a general fertility rate? Would it be more effective to use YNGCH instead of FERTYR?
The variable FERTYR reports information provided directly by respondents about the number of children they gave birth to in the last year. The universe of the variable is females age 15-50.
The variable NCHILD reports the number of own children residing in the person’s household. The universe of the variable is all persons. NCHILD includes non-biological children, such as step-children and adoptive children. The information reported in NCHILD is not collected directly from respondents. The variable YNGCH reports the age of each person’s youngest own child within their household. The variable MOMLOC is a family interrelationship pointer variable. It reports the person number (PERNUM) of each person’s mother, if they have a mother residing in the same household. The three variables NCHILD, YNGCH, and MOMLOC are, by design, consistent with each other. For example, if a respondent has NCHILD=1, then exactly one person in the household will be identified as the child of that respondent in MOMLOC.
The variables NCHILD, YNGCH, and MOMLOC are all created by IPUMS using algorithms that link individuals to their most likely family members within a household. Our algorithms use age, marital status, sex, and other information to make these links, and the links identify biological and non-biological relationships between parents and children. Note that we do not use FERTYR in our creation of these variables. You can read about the family interrelationship variables created by IPUMS on this page.
FERTYR and NCHILD measure different things—FERTYR is a measurement of recent fertility, while NCHILD provides information about family interrelationships within the household. A person may have given birth in the last year and have no children in the household. Similarly, someone may have a biological, step, or adoptive child of their own in the household, but not have given birth within the last year. It is therefore not possible to determine definitively when or if NCHILD and FERTYR provide conflicting information. You may find the discussion on this forum post about measuring the fertility rate using YNGCH useful.
The family interrelationship variables constructed by IPUMS allow data users to capture additional information about family context beyond what is directly reported; they are accurate, but imperfect, predictions of how members of a household are related. There are certainly instances in which a child lives in a household without their mother (or without their father). However, there are also instances in which the IPUMS family interrelationship algorithm makes an incorrect determination (e.g., indicates there is no probable coresident parent in the household when the child’s parent does reside with the child). In the example household you referenced, it seems likely that the baby (age=0) likely belongs to the 22-year old female with FERTYR=2 (yes).
