Builds a bridge from O*NET-SOC codes to the reference SOC codes of an OEWS
weight panel, including the combined codes OEWS publishes in place of some
detailed SOC occupations. Pass the result to the bridge argument of
onet_measure_aggregate() or onet_measure_sensitivity().
Arguments
- occupations
O*NET-SOC codes to bridge: a character vector, a data frame with an
occupation_codecolumn, or an occupation-levelonet_measure(). Usually the occupation scores you are aggregating.- weight_panel
A weight panel from
onet_weight_panel_oews()built from May 2021 or later OEWS estimates, or any data frame withreference_soc_codeandyearcolumns. National, state, metropolitan, and industry panels all work.- occupation_code
Occupation code column when
occupationsis a data frame.
Value
A tibble with one row per O*NET-SOC code and columns
from_onet_soc_code, from_soc_code, reference_soc_code, map_type,
crosswalk_weight, and crosswalk_path. map_type is "direct" when
the occupation's SOC is in the panel, "oews_combination" when OEWS
publishes the occupation inside a combined code that is in the panel, and
"not_in_panel" otherwise. Rows other than "oews_combination" keep the
occupation's own SOC as reference_soc_code, so aggregation treats them as
it would without a bridge. onet_provenance() reports crosswalk_path for
aggregates that use the bridge.
Details
Beginning with the May 2021 estimates, OEWS publishes most detailed 2018 SOC
occupations but combines some of them, either at the broad-occupation level
(for example 31-1120, Home Health and Personal Care Aides, which holds
31-1121 and 31-1122) or as OEWS-specific codes (for example 25-9045,
Teaching Assistants, Except Postsecondary). Without a bridge, O*NET
occupations inside those codes cannot match the panel, and their employment
is left out of covered_employment.
The bridge uses the 12 combined codes, and the SOC occupations each one
includes, listed in the BLS May 2021 OEWS occupation definitions. The May
2023 through May 2025 national files publish the same 12 codes. Because the
list comes from those definitions rather than from the rows of
weight_panel, an occupation missing from a state, metropolitan, or industry
panel, for example because its estimate is suppressed, is never merged into a
neighboring combined code. It stays "not_in_panel".
Occupations inside one combination are averaged with equal weight, the same
way onet_measure_aggregate() averages several O*NET detail codes that
share one SOC. OEWS publishes no employment split among them.
May 2019 and May 2020 OEWS estimates use a hybrid of the 2010 and 2018 SOC
with different combined codes, and earlier estimates use older SOC versions,
so weight_panel must contain only years from 2021 on. Build a bridge by
hand for earlier panels.
Examples
weights <- tibble::tibble(
reference_soc_code = c("29-1141", "31-1120"),
year = 2024L,
employment = c(3000, 4000),
weight_share = c(3, 4) / 7,
source = "OEWS",
source_taxonomy = "2018 SOC",
reference_taxonomy = "2018 SOC"
)
# Stylized scores for illustration only.
scores <- tibble::tibble(
onet_soc_code = c("29-1141.00", "31-1121.00", "31-1122.00"),
measure_score = c(0.2, 0.4, 0.6)
)
bridge <- onet_oews_bridge(scores, weights)
#> Mapped 2 O*NET occupations into 1 OEWS combination code: "31-1120".
bridge
#> # A tibble: 3 × 6
#> from_onet_soc_code from_soc_code reference_soc_code map_type crosswalk_weight
#> <chr> <chr> <chr> <chr> <dbl>
#> 1 29-1141.00 29-1141 29-1141 direct 1
#> 2 31-1121.00 31-1121 31-1120 oews_com… 1
#> 3 31-1122.00 31-1122 31-1120 oews_com… 1
#> # ℹ 1 more variable: crosswalk_path <chr>
onet_measure_aggregate(scores, weights, bridge = bridge, measure_id = "stylized")
#> # A tibble: 1 × 9
#> measure_id aggregate total_employment covered_employment
#> <chr> <dbl> <dbl> <dbl>
#> 1 stylized 0.371 7000 7000
#> # ℹ 5 more variables: employment_coverage_share <dbl>, n_occupations <int>,
#> # n_reference_soc <int>, coverage <list>, provenance <list>