Is there anyone who feels they’ve settled on really robust logic for aggregating members into households that maintain some reasonable level of integrity over the time dimension? If so, do you mind sharing some of the key principles that guide your business logic?
To date, I’ve mostly just generated a field that concatenates member zip code with the address lines of their residence and used that as a key for aggregating spouses and children who all live together as a logical financial unit. Some MCIF systems I’m familiar with also look for a last name match when forming households, but I’ve tended to steer clear of that additional element. Why? While it helps to solve the issue of boarders — folks who have independent finances who happen to live at your exact same address — it potentially creates even more significant errors when wives keep maiden names or take hyphenated last names, causing them to be householded apart from their husbands and/or children — not to mention unmarried couples living together, with mostly integrated finances.
My methodology is probably analytically sufficient for a particular moment in time, but it gets a bit clumsier when looking at data sets that stretch over longer periods of time. First, when households change addresses, one needs to track those timestamped changes to maintain continuity. Otherwise, the current residents at an address might inherit the financial attributes of other members who previously lived at the same address. Yikes. Furthermore, over the course of time members of households come and go. Divorces happen. Remarriages happen. Grown children leave for college and might reasonably be considered distinct (financially) from the parents’ household. Or not so much. Then they sometimes come back to live at home and the financial independence is as muddy or muddier. Is there nuance we should bring to these cases?
Even more challenging is this — when we’re wondering about the accounts and activities of a household five years ago, are we implicitly asking “what accounts and activities did the current members of a particular household have five years ago when we aggregate them all together, regardless of whether they were all a household back then?” Or are we asking, “who were all the members of a household five years ago, and what did they collectively own and do?” One can quickly picture how those two often do not generate the same outcomes. Either might be a reasonable household formulation for informing decisions we’re trying to make about households in the present moment, but I confess I’m not sure, a priori, if one way of thinking has clear advantages over the other.
The householding rabbit hole goes even deeper, if we dare plunge down it.
Are there any among you who goes still further to consider other variables when aggregating members into households? Do you consider joint ownership of accounts or co-borrowing of loans? Do you consider named beneficiaries on accounts at all? Or do you look to the transaction level and say that, even if someone lives at the same address as someone else, they’re their own household if they have a checking account that has activity that suggests independence?
Do any take the radical step of inviting members a chance to give input or revise who is or isn't included within your concept of a household?
Appreciating financial relationships in a whole-family context is obviously often critical to broadly appreciating member needs and interests. Maybe my concatenated zips/addresses is perfectly sufficient, even looking over time, admitting some inevitable errors. But if others have had notable success with some more sophisticated logic, I’d be grateful — and I daresay the CULytics community would be similarly thankful — for what insights you’d share.
STCU R&D Strategist
CULytics Community Chair