Create a PUMS Weight Panel

Usage

onet_weight_panel_pums(
  pums,
  year,
  socp = "SOCP",
  weight = "PWGTP",
  group = NULL,
  replicate_weights = NULL,
  reference_taxonomy = "2018 SOC",
  source_taxonomy = NULL,
  crosswalk = NULL
)

Arguments

pums: ACS PUMS microdata or already-filtered person records.
year: ACS PUMS data year.
socp: Occupation column, commonly SOCP. Do not pass Census OCCP; it is not an SOC field.
weight: Person weight column, commonly PWGTP.
group: Optional columns for demographic or geographic cells.
replicate_weights: Optional replicate-weight columns.
reference_taxonomy: Reference SOC label.
source_taxonomy: Optional source taxonomy label.
crosswalk: Optional source-to-reference crosswalk.

Value

A normalized employment-weight panel. When group is supplied, weight_share is computed within each group cell.

Details

onet_weight_panel_pums() expects ACS PUMS records that have already been filtered to the employment universe used by the analysis. For employment weights, a common ACS starting point is employed civilians, ESR %in% c(1, 2), often restricted to age 16 or older. Use SOCP as the occupation field. OCCP is a Census occupation recode, not an SOC field.

ACS SOCP can include aggregate codes with trailing X characters. Those rows cannot be matched directly to O*NET or OEWS SOC codes, so they are dropped with a warning unless the caller supplies a crosswalk that maps them to valid reference SOC codes.

Replicate-weight standard errors use sqrt((4 / R) * sum((theta_r - theta)^2)). For ACS PUMS, pass the full 80 PWGTP1 through PWGTP80 columns. Passing a smaller subset produces a warning because the resulting standard error is not survey-valid.

The year argument means the 1-year ACS PUMS data year. 5-year PUMS files pool responses across vintages and can mix SOCP code versions within one file; SOC-vintage labeling is not reliable for them. Prefer 1-year files for weight panels, or resolve codes explicitly with a crosswalk.

Examples

pums <- tibble::tibble(
  SOCP = c("151252", "151252", "291141"),
  PWGTP = c(50, 50, 300),
  state = c("WA", "WA", "WA")
)
onet_weight_panel_pums(pums, year = 2024, group = "state")
#> # A tibble: 2 × 8
#>   reference_soc_code state  year employment weight_share source source_taxonomy
#>   <chr>              <chr> <int>      <dbl>        <dbl> <chr>  <chr>          
#> 1 15-1252            WA     2024        100         0.25 PUMS   2018 SOC       
#> 2 29-1141            WA     2024        300         0.75 PUMS   2018 SOC       
#> # ℹ 1 more variable: reference_taxonomy <chr>