In Business Analyst desktop, you can set you preferences for how the data is aggregated. The most accurate method is to use the "Block Appointment" method. This uses 6+ million Census block points to apportion the number of households within the Census block group that is intersecting your drive time area. This allows you to model a situation where the distribution of households is not homogeneous throughout the block group area that is within the drive time area. Thus, if the area of the block group is split 50/50 inside/outside the drive time area but there is a 90/10 split for the number of households, you'll get the 90% of households for that block group and that percentage can be used to apportion other household based data.
In ArcGIS Pro Business Analyst and Business Analyst Web App, I think the "Enrich Layer" option and the BA reports use block apportionment, but I have not checked that. For performance, it might also use the hybrid approach.