Report reporting buckets that sum more than one territory
Source:R/polity_folds.R
polity_bucket_coverage.Rdpolity_area_code is an aggregation bucket, not an identity: FABIO folds
several FAOSTAT reporting areas into one numeric code, and WHEP's builds sum
them under it. reporting_polity_code is then resolved from that bucket
code, so a bucket folding more than one live territory can end up labelled
with a polity that covers only part of what the value covers, or that covers
the whole of it but in a period that has ended.
This lists every (polity_area_code, year) that folds more than one polity
and classifies whether the bucket's own label covers the fold:
"aggregate": the bucket resolves to an aggregate polity (Rest of World, Belgium-Luxembourg, the FAOSTAT combined-reporting entities), whose name and polygon already mean the union of its members. Honest."predecessor": the bucket is labelled with a polity that has ended, and whose publishedsuccessorset is exactly the set of polities the bucket folds. The extent is right — that predecessor's territory is the union of its successors — but the period is not, so a consumer filtering polities by span drops the rows."partial": the bucket sums several territories but is labelled with a polity covering only part of them, so the value and its polity describe different extents. This is the worst case, and no bucket is in it today."unlabelled": the bucket code resolves to no polity, so rows carryNAand the gap is at least visible rather than wrong.
An area counts as a member only in the years it reports: its polity must
be in span, and the upstream FAOSTAT map must report the area that year. A
year-aware lookup answers every (area_code, year) pair regardless, standing
in with the nearest period, so asking it about an area that does not report
in that year invents a member. FAOSTAT reports area 206 for 1961-2011 and
areas 276/277 for 2012-2024, never in the same year, so counting the stand-ins
reported bucket 206 as a three-way fold in all 65 years rather than a two-way
fold in the 14 it is one (whep#414).
Bucket 206 is the one fold reported today, and it is "aggregate" for
2012-2025: it sums FAOSTAT areas 276 Sudan and 277 South Sudan and resolves
to F206-2011-2025 "Sudan and South Sudan (combined reporting)", the
aggregate whep#860 wired in. It read "predecessor" before that, labelled
with SUD-1956-2011, a polity that had ended.
So the labelling is settled and only the regionalisation is still open:
whether to keep folding the two successors at all is whep#680, and
options(whep.unfold_predecessor_bucket = "all") is the switch that
promotes them. It is off by default because it is not mass-neutral — see
folded_reporting_areas() for what it withdraws and why.
.aggregate_to_polities() warns when it builds such a bucket; set
options(whep.warn_polity_folds = FALSE) to silence that warning.
The polity reported here as the bucket's own is also the area label the
builds attach to the summed row, and the one the reporting columns resolve.
A bucket carries one label whatever its members resolve to, because area
is a join key and a bucket under two labels stops summing (whep#563).
Value
A tibble with one row per folded (polity_area_code, year), with
the folded member polities, the polity the bucket itself resolves to, and
the coverage classification. Zero rows means no bucket folds more than
one polity in the requested years.
Examples
polity_bucket_coverage(years = 2015L)
#> # A tibble: 1 × 10
#> polity_area_code year n_member_polities member_polity_codes member_area_codes
#> <int> <int> <int> <chr> <chr>
#> 1 206 2015 2 SDN-2011-2025, SSD… 276, 277
#> # ℹ 5 more variables: bucket_polity_code <chr>, bucket_polity_name <chr>,
#> # bucket_mapping_status <chr>, bucket_polity_type <chr>, coverage <chr>