Return one row per record-county for every PDA record — transform_pda

get_preliminary_damage_assessments() returns one observation per request. This function returns one observation per request-county, drawing the county detail from two sources: the comma-separated county listing that many PDA reports print, and FEMA's own record of the areas a declaration designated. Every record passed in is represented in the output. A record with neither source of county detail–a denial with no matched report, a tribal request–keeps a single row with the county columns empty.

Usage

transform_pda_counties(pda_df, declaration_areas = fetch_declaration_areas())

Arguments

pda_df: A dataframe returned by get_preliminary_damage_assessments().
declaration_areas: A dataframe of FEMA's designated areas, one row per declaration-area, with the columns fetch_declaration_areas() returns. The default fetches them from FEMA, so this function needs a network connection whenever pda_df carries a disaster number column; pass a dataframe to supply them from elsewhere.

Value

pda_df with one row per record-county, and these columns added:

pda_county_name: The county's name as the Census Bureau spells it, or, where the name could not be matched, the name as the report printed it. NA on a row with no county at all.
pda_county_name_reported: The name exactly as the report printed it, kept so that a failed match can be checked against the source. NA on a row that did not come from a county listing.
pda_county_geoid: Five-digit county FIPS code, NA where the place has no county code or its name could not be matched.
pda_geography_type: What kind of place the row describes: "county", "parish", "borough", "census area", "municipality", "municipio", "island", "independent city", "education area", "tribal entity", or "unrecognized" where the name matched nothing and resembles none of these. NA on a row with no county at all.
pda_pa_per_capita_impact_county: The county's per capita impact in dollars, as printed. NA on a row that did not come from a county listing.
pda_county_indicator: TRUE where the report's own county listing named this county.
fema_county_indicator: TRUE where FEMA's designated areas for the declaration named this county.
fema_statewide_request: TRUE where the row came from expanding a statewide designation rather than from a county FEMA named individually.
fema_designated_area: The name FEMA gave the designated area, kept so that tribal lands and other non-county designations stay identifiable. Where more than one designated area falls inside one county, the names are semicolon-separated.
fema_ihp_declared_county, fema_ia_declared_county, fema_pa_declared_county, fema_hm_declared_county: FEMA's record of the programs turned on for this designated area.

Where a request-level column has a county-level counterpart, only the county-level one is returned: pda_pa_per_capita_impact_countywide, the raw text of the county listing, is dropped in favor of pda_pa_per_capita_impact_county, and the declaration-level fema_ihp_declared, fema_ia_declared, fema_pa_declared, and fema_hm_declared are dropped in favor of their _county versions. Problems found with a row are appended to the existing pda_warnings column, semicolon-separated, alongside any the request-level record already carried.

Where the county rows come from

Two columns record which source named a county, and a county named by both is one row with both columns TRUE rather than two rows:

pda_county_indicator is TRUE when the county appeared in the report's own county listing. Only reports that requested Public Assistance and printed a countywide breakdown carry a listing–roughly two thirds of all PDAs.
fema_county_indicator is TRUE when the county appeared in FEMA's designated areas for the declaration, including when it arrived through a statewide designation. Only approvals have designated areas; a denial never receives a disaster number and so has none.

These two are distinct from pda_county_name, pda_county_geoid, and pda_county_name_reported, which say what the county is rather than whether the PDA listing named it.

Because a declaration designates about twenty counties on average, and a statewide designation expands to every county in the state, this function returns many times more rows than the request-level dataset it is given–tens of thousands of county rows for the full archive.

Statewide designations and the vintage of the county list

A designated area that FEMA names "Statewide" covers the whole state rather than one county, and is expanded into one row per county in that state, each with fema_statewide_request set to TRUE. The counties used for that expansion come from a single recent vintage of tidycensus::fips_codes, applied to every year of the archive. For statewide requests, therefore, the county-level observations may differ slightly from the counties that actually existed at the time of the declaration. Connecticut is the clearest case: a statewide Connecticut declaration from 2011 comes back as the nine planning regions that replaced the state's counties in 2022, not as the eight counties that existed on the day of the declaration.

Places that are not counties

The listing mixes counties with the other quasi-county geographies: Louisiana parishes, Alaska boroughs and census areas, Virginia and Missouri independent cities, Puerto Rico municipios. Tribal nations are included when they are subrecipients on state-led requests, as are Alaskan Regional Education Attendance Areas, though neither have county FIPS codes. geography_type provides these details. FEMA's designated areas name tribal lands and other non-county places the same way; those rows keep the name FEMA gave them in fema_designated_area and carry no county code.

Values of zero

A zero is not necessarily a county assessed at no damage. Often, this may reflect when a county-level PDA is ongoing, or perhaps when one was not completed. The value is preserved as printed and the row is flagged in the warnings column so that zeros can be excluded as appropriate.

Identifying an event

Every column of the request-level record is repeated down that record's county rows, so fema_declaration_request_number – FEMA's own identifier for the declaration request, never missing for an approval or a denial – still names the event each row belongs to. Group by fema_declaration_request_number to get back to one row per event.

That column and pda_county_geoid together are the uniqueness key of the returned data, with two exceptions: the repeated county listing described below, and rows that have no county code at all, where several designated areas under one event – two reservations, say – are separate rows told apart by fema_designated_area.

Counties listed twice

A few reports cover two separate incidents and print the whole county list once for each, giving a different per capita figure each time. Both figures are returned as separate rows, since neither supersedes the other, and both rows are flagged in the warnings column.

Examples

if (FALSE) { # \dontrun{
get_preliminary_damage_assessments() %>%
  transform_pda_counties()
} # }