I'm personally not a big fan of households as a unit of analysis, at least not in isolation of any context. As far as I'd understand it, they hide meaningful is

I'm personally not a big fan of households as a unit of analysis, at least not in isolation of any context. As far as I'd understand it, they hide meaningful issues. For example, if everyone had high personal income and an abundance of space, they'd likely live privately. Whereas if you've got no income, you risk homelessness, you may live with someone you'd rather not. e.g. bad marriages, domestic violence etc is often tolerated due to a lack of financial power to move out. As such you may find the median household income overstates wealth. It may be considerable simply because multiple people live there, who are too poor to live by themselves. In this sense, household income may hide poverty and overstates affluence. Particularly when you consider that there's a negative correlation between income and birth rates, you end up seeing large families (and possibly considerable household income) who are poor, in the data. But just reporting on household income doesn't do that justice. Similarly, this creates an apples/oranges comparison with median/average household income. The median household in the US today is 2 people. In the 1960s it was 3 people. (close to 4, even!) But it's different from the average household size, which has dropped too. As such, you may just be observing that the median household relative to the average household size has been dropping. That could have nothing to do with polarization, it may just mean there's a shift in living preferences. I'm not asserting anything here, just that there are alternative explanations for the observation you make, and that just looking at these averages without context, controlling for confounding variables or any theoretical basis may lead you to incorrect conclusions.