Choosing Employment Weights: OEWS versus PUMS

Weights answer a population question. OEWS is usually the best choice when the target is official occupation employment and wage context. PUMS is useful when the target is a custom sample or demographic cell.

Start with the Same O*NET Score

abilities <- onet_archive_read(
  "30.3",
  "Abilities",
  path = archive_303,
  release_date = "2026-05-01"
)

score <- abilities |>
  filter(element_id == "1.A.1.a.1") |>
  transmute(onet_soc_code, measure_score = data_value)

score |>
  knitr::kable(digits = 3, align = "l")

onet_soc_code	measure_score
15-1252.00	4.35
29-1141.00	4.71
11-1011.00	4.50
41-1011.00	4.15

OEWS Weights

oews_weights <- onet_weight_panel_oews(
  onet_oews_national(2023, path = oews_path),
  year = 2023
)
#> Dropped 2 OEWS aggregate rows; keeping "detailed" occupations.

oews_result <- onet_measure_aggregate(
  score,
  oews_weights,
  measure_id = "oral_comprehension_fixture"
)

oews_weights |>
  knitr::kable(digits = 3, align = "l")

reference_soc_code	year	employment	weight_share	source	source_taxonomy	reference_taxonomy
11-1011	2023	211230	0.040	OEWS	2018 SOC	2018 SOC
15-1252	2023	1847900	0.353	OEWS	2018 SOC	2018 SOC
29-1141	2023	3175400	0.607	OEWS	2018 SOC	2018 SOC

oews_result |>
  select(-coverage, -provenance) |>
  knitr::kable(digits = 3, align = "l")

measure_id	aggregate	total_employment	covered_employment	employment_coverage_share	n_occupations	n_reference_soc
oral_comprehension_fixture	4.574	5234530	5234530	1	4	4

PUMS Weights

pums <- tibble::tibble(
  SOCP = c("151252", "151252", "291141", "291141", "111011", "111011"),
  PWGTP = c(120, 80, 90, 110, 10, 30),
  ESR = c("1", "1", "1", "1", "2", "2"),
  sex = c("F", "M", "F", "M", "F", "M")
)

pums_weights <- onet_weight_panel_pums(
  pums,
  year = 2022,
  group = "sex"
)

pums_f <- onet_measure_aggregate(
  score,
  pums_weights,
  measure_id = "oral_comprehension_fixture",
  cell = list(sex = "F")
)
pums_m <- onet_measure_aggregate(
  score,
  pums_weights,
  measure_id = "oral_comprehension_fixture",
  cell = list(sex = "M")
)

pums_weights |>
  knitr::kable(digits = 3, align = "l")

reference_soc_code	sex	year	employment	weight_share	source	source_taxonomy	reference_taxonomy
11-1011	F	2022	10	0.045	PUMS	2018 SOC	2018 SOC
11-1011	M	2022	30	0.136	PUMS	2018 SOC	2018 SOC
15-1252	F	2022	120	0.545	PUMS	2018 SOC	2018 SOC
15-1252	M	2022	80	0.364	PUMS	2018 SOC	2018 SOC
29-1141	F	2022	90	0.409	PUMS	2018 SOC	2018 SOC
29-1141	M	2022	110	0.500	PUMS	2018 SOC	2018 SOC

Raw ACS PUMS needs a universe filter before this step. Employment weights usually start with employed civilians, ESR %in% c(1, 2), and age 16 or older. Use SOCP; OCCP is a Census occupation recode, not a six-digit SOC code.

pums <- tidycensus::get_pums(
  variables = c("SOCP", "PWGTP", "ESR", "AGEP"),
  year = 2023,
  survey = "acs1"
) |>
  filter(ESR %in% c("1", "2"), AGEP >= 16)

Compare the Answers

comparison <- tibble::tibble(
  population = c("OEWS national", "PUMS-style F", "PUMS-style M"),
  aggregate = c(oews_result$aggregate, pums_f$aggregate, pums_m$aggregate),
  coverage = c(
    oews_result$employment_coverage_share,
    pums_f$employment_coverage_share,
    pums_m$employment_coverage_share
  )
)

comparison |>
  knitr::kable(digits = 3, align = "l")

population	aggregate	coverage
OEWS national	4.574	1
PUMS-style F	4.504	1
PUMS-style M	4.550	1

If your claim is about the national labor market, OEWS is the natural default. If your claim is about a subgroup, geography, or survey-defined population, build a PUMS weight panel and state the cell explicitly.