I am computing some estimates for a specific PUMA using the 5-year 2021 file. When running the data weighted I get a total of 27,184 households in the sample compared to 27,105 persons (I am restricting the analyses to persons ages 60 and older living in community). Is it possible to have more households than individuals? I ran the sample with the same restrictions without weighting and the size of the sample is the same for a hh and an individual variables . Thank you Allen Glicksman
It is not improbable for the weighted number of households to be slightly greater than the number of persons. Part of this discrepancy is due to rounding of person-level weights. Another part is the fact that person-level weights, unlike household weights, are stratified to match totals by demographic characteristics on the county level. You can find more about ACS sampling and weighting in the ACS Design and Methodology Report.