How to safely do ZIP code in exports?

[colinxfleming]

Thanks for creating an issue! Please fill out this form so we can be sure to have all the information we need, and to minimize back and forth.

• What are we trying to do?

A fund has a standard practice of only collecting zip (instead of city/state), but our exports only spit out city/state, which makes it difficult-to-impossible for them to do any kind of geo analysis.

I think we've talked about including zip before and I'd like to revisit that - it might be a decision that's outlived its usefulness at this point? Are we still worried about this?

• What feature or behavior is this required for?

fund reporting

• How could we solve this issue? (Not knowing is okay!)

I'm not sure! Here are some opts:

• yolo it and put ZIP in exports as is

• have some kind of toggle where you can either export city/state OR zip (probably more trouble than it's worth, but an option)

• maybe something else?

• Anything else?

nah

[colinxfleming]

cc @lomky @hkang06

[hkang06]

we've been talking about zip vs. city/county and state in the NNAF OSU data steering committee conversations, and came to consensus that ZIP code + other basic demographic info would be too identifying in many rural areas. So I'd be a little wary of the yolo-export-ZIPs-regardless option looking to the future, but given future scope, maybe that's a bit too farsighted?

How much effort would it take to add an admin config toggle on the geographic preference to display/export?

(I'm sneaking in the question about "display" here because in Baltimore, we have a lot of trouble with the difference between Baltimore City and Baltimore County, and folks simply entering in "Baltimore" in either city/county field, where that doesn't actually give us the info we need. We got a request on breaking out our callers by Baltimore City, Baltimore County, elsewhere in MD, out-of-state, and couldn't really do it since the data are so muddled. Tannis at BRAF also mentioned some of the Virginia-specific weirdness around how to identify location with city vs. county. Just food for thought.)

[lomky]

I am still not onboard with the idea of exporting ZIP codes - they were not created with anonymous data analysis in mind and don't guarantee anonymity at all when you add in any other demographic information. The US Census doesn't use them for this reason, and HHS has guidance on when data exports are allowed to use even the first three of a zip in collections of data, let alone for individuals.

My first thought is can we some sort of lookup on export to set an equivalent county/city for a given zip code?

[elimbaum]

Maybe we can use the canonical lookup! – a la https://tools.usps.com/zip-code-lookup.htm?citybyzipcode

Though I would still be a bit concerned (say, in rural areas) where USPS "city name" might be map to a tiny post office. Really want you want here is some population-constant geographic area which... isn't strictly available in general.

[colinxfleming]

I think @hkang06 and @lomky make compelling enough points here that we shouldn't do this. I'll message this with the fund that's thinking about this stuff

[colinxfleming]

I should have read these comments a little closer, ha. let's keep open and see if canonical citystate might work for a zip code as an alt.