Zip Code Overlap

[woodhull]

How are you representing ZCTA that map to more than one CD in this file? When you say you de-dupe the data, does that mean that you remove ZCTAs that span CDs removing those ZCTA entirely or just choose one?

I've been considering A/B testing a zip code lookup option and then falling back on full address if there is not an exact match for VisitThem.org.

We currently always ask for full address and then do the lookup via polygon.

[jlev]

I deduplicate only repeated zipcode->CD matches (of which there are many in the county_rel file). But if a zipcode is part of multiple congressional districts or crosses state lines, there are multiple rows in the zccd.csv file.

I have tested this for a known set of zipcodes that cross geographic boundaries and the output appears correct. http://gis.stackexchange.com/questions/53918/determining-which-us-zipcodes-map-to-more-than-one-state-or-more-than-one-city

Lookup by polygon is always going to be better. This is just faster, and really useful for when you can't request full address (i.e., when prompting location by phone)