Hi. Can anybody explain why the 2023 ACS published table
estimates 37.8 million households with 1 or more people under 18, but if i filter for relate=1, IPUMS NCHILD variable seems to there are 49 million households where NCHILD is 1 or greater… thanks (is it because NCHILD includes adult children of the householder?)
I am going to answer your question on how to estimate the number of households in the US with children under 18 years old in three parts.
First, NCHILD includes children of any age, and not just those under the age of 18. When isolating your analysis to RELATE=1, you will only be looking at the children of the corresponding head of household (and will not include any persons under 18 not related to the head of household). Note also that NCHILD is a person-level variable, meaning that the value may differ between members within the same household.
Second, to replicate the linked table, you will need to create a variable that counts only households with at least one person under 18. The online analysis tool is unable to generate household-level estimates using records of multiple people within a household. You can instead run your analysis using a stats package (we provide support for Stata, R, SAS, and SPSS) and request a customized data extract that includes the 2023 ACS variables of interest as well as the variables SERIAL, AGE, and HHWT.
And third, I created an extract and read into Stata, created a variable named youngest which represents only the youngest person in each household (bysort serial (age): replace youngest = age if _n == 1) and ran a weighted tab. I was able to match the estimate you linked (37,772,894 vs 37,780,108).
