Weight to use at benefit unit level
Hi, I am constructing a dataset for longitudinal analysis at benefit unit level using b_buno_dv. As there is no dedicated weight derived for benefit unit level analysis, would I be better to use the longitudinal household level weight (as benefit units are nested within households)?
Updated by Peter Lynn almost 9 years ago
Yes! The selection probability of a BU is identical to that of the household to which it belongs (as all BUs in a household are included in the sample). The majority of households contain a single BU, so the nonresponse propensity of the BU is also identical to that of the household. The only difference arises with respect to the nonresponse propensity of BUs in multi-BU households. The household weights might miss some relevant factors. But the differences in the weights are likely to be minor.