For its story and graphics showing where Puerto Ricans may have migrated after Hurricane Maria, CNN analyzed two national data sets obtained under the Freedom of Information Act:
- Application data from the Federal Emergency Management System's Individuals and Households Program, which provides housing and financial assistance to eligible disaster victims, and
- Change-of-address orders placed with the US Postal Service.
Individual assistance applications
The FEMA data, which captured 1,057,006 applications through November 11, included ZIP codes for each applicant's current mailing address and "damaged dwelling." An early analysis of the data found thousands of entries where ZIP codes were either invalid or indicated a damaged property location outside Puerto Rico.
FEMA officials told CNN these errors occurred because ZIP codes are "a required field that is entered by the registrant" when they apply for assistance and "it is possible that registrants do not know their ZIP code, and therefore enter any value as the field is required to be entered before they can continue."
While the agency said it has a process in place to correct or validate addresses, "This only occurs if the address is recognized, so it does not catch all cases where an address may have an incorrect entry."
To limit the number of incorrect entries in its analysis, CNN compared each application's damaged dwelling ZIP code against a list of valid Puerto Rican ZIP codes provided by the US Postal Service.
CNN also excluded cases where an applicant listed their "Current Location" as "Damaged Dwelling," which suggested an applicant had either not moved at all or provided a mailing ZIP code that was different from where they were residing. When other data points suggested error, such as one case where the applicant's damaged dwelling was listed as a "Correctional Facility" but listed the current location as "Family/Friends," those records were also excluded.
CNN then calculated the total number of applications by mailing address ZIP code and -- because ZIP codes represent postal delivery routes, not geographic areas -- used geospatial data from the US Census Bureau to match applications to their corresponding ZIP Code Tabulation Area, or ZCTA, a census unit that approximates the geographic areas served by ZIP code routes.
Because ZCTAs and metro areas do not always conform to county -- or even state -- lines, CNN then grouped applications by county and Core-Based Statistical Area based on which counties and CBSAs contained each ZCTA's centroid. CNN calculated state totals by grouping the county-level figures.
CNN obtained data from the US Postal Service listing the number of change-of-address orders originating from Puerto Rico or the US Virgin Islands between October 1 and December 31.
For privacy reasons, the Postal Service would not release change-of-address totals for individual ZIP codes, only at the three-digit prefix level and where the totals were greater than 25. Because the Postal Service data distinguished between three-digit ZIPs where the change-of-address volume was 0 and where the total had been withheld (meaning it was somewhere between 1 and 25), we filled in the minimum possible value: 1.
CNN then analyzed the Census Bureau's geospatial data for ZCTAs, combining shapes that shared a three-digit prefix and joining them with the change-of-address totals to map the address change orders.