Rural Dataverse

cori.data.qcew

Access and analyze Bureau of Labor Statistics Quarterly Census of Employment and Wages (QCEW) data at the county, state, and national level from 1990 to present. Covers total employment, average annual pay, sector composition, and employment concentration using CORI's tradable goods / tradable services / local services classification.

Drew Rosebush 2026-08-31

Overview

The cori.data.qcew package provides access to Bureau of Labor Statistics Quarterly Census of Employment and Wages (QCEW) data at the county, state, and national level from 1990 to the present. Derived from unemployment insurance records rather than a survey, QCEW covers roughly 95% of all U.S. employment. It is the standard source for local labor market analysis and more reliable than survey-based estimates in the small rural counties where sample sizes are thin.

Beyond access, the package adds two analytical layers over the raw federal source. The first is CORI’s supersector classification — a three-way grouping of BLS’s 11 NAICS supersectors into Tradable Goods, Tradable Services, and Local Services, loosely following the methodology in Eckert (2018). This classification distinguishes industries that bring external income into a community from those that primarily serve local demand, which is central to understanding a rural economy’s long-run wage trajectory. The second is an employment concentration measure (HHI) that reveals how economically diversified a county is across sectors — a structural indicator that employment headcounts alone cannot provide.

All functions return tidy long-format data consistent with the cori.data ecosystem: geoid | year | variable | value, with an agg_var employment-weight column on rate and share variables for correct aggregation across geographies of different sizes.

Installation & quick start

remotes::install_github("ruralinnovation/cori.data.qcew")

library(cori.data.qcew)

# See all available variables and which functions return them
get_qcew_codebook()

# Total employment — all counties, all years
get_employment(geography = "county")

# Average annual pay — specific counties
get_wage_salary(geoids = c("33009", "33011"), years = 2010:2023)

# Employment by CORI supersector
get_sector_employment(geography = "county", years = 2018:2023,
                      sector_type = "CORI", value_type = "share")

# Employment concentration (HHI)
get_employment_concentration(geography = "county", years = 2023)

Check the current data vintage:

latest_qcew_vintage()

What this data tells us about rural America

Employment is the most fundamental indicator of a local economy’s health and one of the most easily misread in a rural context. Aggregate employment counts can grow while wages stagnate, or decline while productivity rises. In CORI’s community research, many rural counties show GDP per capita growth alongside employment that has not recovered to prior peaks — a pattern that reflects a smaller, more productive workforce rather than a stagnant one. The headline number and the underlying story can point in different directions.

What the QCEW data adds beyond simple employment counts is sector composition. CORI’s three-way classification — tradable goods, tradable services, and local services — is particularly useful for rural analysis. Local services (retail, healthcare, food service) meet local demand and scale with population; they are essential to community function but do not drive wage growth or export earnings. Tradable goods and services — manufacturing, agriculture, finance, tech, professional services — bring external income into a community and shape its long-run wage trajectory. Rural counties with strong tradable sectors reflect economies whose historical asset base remains economically active. Those shifting toward tradable services are often communities building new competitive advantages alongside inherited ones.

Industry concentration is a structural feature of many rural economies, reflecting the historic dominance of particular assets — a major employer, an agricultural commodity, a natural resource. Concentration can be a strength (scale, specialization, deep supply chains) and a source of vulnerability (exposure if that industry contracts). The QCEW data makes this composition visible at the county level and trackable over time, supporting both the identification of strengths and the analysis of diversification opportunities.

Average annual pay reflects both the composition of employment and wages within sectors. Understanding the wage structure of a rural economy — not just its size — is essential for reading its trajectory and its capacity to retain talent.

Introduction

Adapted from the Introduction vignette on GitHub.

What is this data?

The Bureau of Labor Statistics Quarterly Census of Employment and Wages (QCEW) covers virtually all U.S. jobs — roughly 95% of all employment — through the unemployment insurance system. Unlike surveys, QCEW is a near-census of covered employment and wages, making it the standard source for local labor market analysis.

Data is available at the county and state level from 1990 to the present.

Sector breakdowns

Two sector classification systems are available via sector_type:

  • "BLS" (default) — 11 NAICS supersectors from the source data
  • "CORI" — 3 custom supersectors: Tradable Goods, Tradable Services, and Local Services
# BLS sector employment counts
bls_emp <- get_sector_employment(
  geography   = "county",
  years       = 2018:2023,
  sector_type = "BLS",
  value_type  = "count"
)

# CORI super-sector employment shares
cori_share <- get_sector_employment(
  geography   = "county",
  years       = 2018:2023,
  sector_type = "CORI",
  value_type  = "share"
)

# Average pay by CORI super-sector
cori_pay <- get_sector_wages(
  geography   = "county",
  years       = 2018:2023,
  sector_type = "CORI"
)

Note on CORI counts: CORI supersector employment counts are not stored directly. To derive approximate counts from shares, multiply value by agg_var:

library(dplyr)

cori_share |>
  mutate(approx_employment = value * agg_var)

Filtering by geography

# Specific counties
emp <- get_employment(geoids = c("33009", "33011"), years = 2010:2023)

# Specific states
pay <- get_wage_salary(geoids = c("33", "50", "23"), years = 2015:2023)

Rural vs. non-rural comparison

library(dplyr)

county_emp <- get_employment(geography = "county", years = 2010:2023)

rural_definition <- ruraldefinitions::cbsa_2023 |> select(geoid, is_rural)

county_comparison <- county_emp |>
  left_join(rural_definition, by = "geoid") |>
  filter(!is.na(is_rural), year == 2023) |>
  group_by(year, is_rural) |>
  summarize(total_employment = sum(value, na.rm = TRUE), .groups = "drop")

print(county_comparison)

Understanding agg_var

Functions that return rates or shares include an agg_var column — the employment weight used for proper aggregation across geographies. To compute an employment-weighted average pay across a set of counties:

pay <- get_wage_salary(geography = "county", years = 2023)

pay |>
  summarize(
    weighted_avg_pay = sum(value * agg_var, na.rm = TRUE) /
                       sum(agg_var, na.rm = TRUE)
  )

Adapted from the Rural vs. Nonrural Employment Trends vignette on GitHub.

This vignette compares employment trends between rural and nonrural counties since the Great Recession using get_employment(), the ruraldefinitions package, and cori.charts for visualization.

library(cori.data.qcew)
library(cori.charts)
library(ruraldefinitions)
library(ggrepel)
library(dplyr)
library(ggplot2)

load_fonts()

Rural vs. nonrural employment since the Great Recession

Pull county-level annual employment for all years since 2007:

emp_raw <- get_employment(
  geography = "county",
  years     = 2007:2024
)

Keep only 5-digit county geoids and attach the rural/nonrural flag from ruraldefinitions::cbsa_2023:

emp_co <- emp_raw |>
  filter(nchar(geoid) == 5) |>
  left_join(
    cbsa_2023 |> select(geoid, is_rural),
    by = "geoid"
  )

Index employment to 2007 levels within each group to show relative change:

chrt_dta <- emp_co |>
  filter(!is.na(is_rural)) |>
  group_by(is_rural, year) |>
  reframe(tot_emp = sum(value)) |>
  group_by(is_rural) |>
  mutate(
    index_value = tot_emp / tot_emp[year == 2007],
    year        = as.numeric(year)
  )

Quick stats for the most recent year-over-year change:

latest_yr <- max(chrt_dta$year)

chrt_dta |>
  filter(year >= (latest_yr - 1)) |>
  arrange(is_rural, year) |>
  group_by(is_rural) |>
  mutate(pct_change = (tot_emp - lag(tot_emp)) / lag(tot_emp)) |>
  filter(year == latest_yr) |>
  select(is_rural, pct_change)

label_df <- chrt_dta |> filter(year == max(year))

# NOTE: chart title uses "fragile" — revisit framing before publishing
fig_line <- chrt_dta |>
  ggplot(aes(x = year, y = index_value, color = is_rural, group = is_rural)) +
  annotate("rect", xmin = 2007, xmax = 2009, ymin = .9, ymax = 1.2,
           fill = "grey90", alpha = 0.5) +
  geom_line(linewidth = 1.5) +
  geom_text_repel(
    data          = label_df,
    aes(label     = is_rural),
    nudge_x       = 1.5,
    direction     = "y",
    hjust         = 0,
    segment.color = NA,
    size          = 5
  ) +
  scale_y_continuous(
    labels = scales::label_percent(accuracy = 1),
    limits = c(.9, 1.2),
    expand = expansion(mult = c(0, 0.05))
  ) +
  scale_x_continuous(
    expand   = expansion(mult = c(0, 0)),
    n.breaks = 5,
    labels   = scales::label_number(accuracy = 1)
  ) +
  scale_color_manual(
    values = c(
      "Rural"    = cori_palettes$ctg2ruralnonrural[[1]],
      "Nonrural" = cori_palettes$ctg2ruralnonrural[[2]]
    )
  ) +
  theme_cori() +
  theme(
    legend.position = "none",
    axis.ticks.x    = element_line(color = "#d0d2ce", linewidth = .25),
    plot.title      = element_text(color = cori_colors[["Dark Green"]])
  ) +
  labs(
    title    = "Rural employment set back twice since the Great Recession",
    subtitle = "Relative change in employment since 2007",
    x        = NULL,
    y        = NULL,
    caption  = paste0(
      "Source: CORI analysis of BLS Quarterly Census of Employment and Wages.\n",
      "Note: Indexed to 2007 levels. Rural = nonmetro counties per CBSA 2023 definition."
    )
  )

fig_line

Regional breakdown

To see how rural employment trends differ by Census region:

region_xwalk <- tigris::states() |>
  sf::st_drop_geometry() |>
  select(GEOID, REGION) |>
  mutate(
    region = case_when(
      REGION == "1" ~ "Northeast",
      REGION == "2" ~ "Midwest",
      REGION == "3" ~ "South",
      REGION == "4" ~ "West"
    )
  )

regional_rural <- emp_co |>
  filter(!is.na(is_rural), is_rural == "Rural") |>
  mutate(st_fips = substr(geoid, 1, 2)) |>
  left_join(region_xwalk, by = c("st_fips" = "GEOID")) |>
  filter(!is.na(region)) |>
  group_by(region, year) |>
  reframe(tot_emp = sum(value)) |>
  group_by(region) |>
  mutate(
    index_value = tot_emp / tot_emp[year == 2007],
    year        = as.numeric(year)
  )

region_colors <- c(
  "West"      = cori_palettes$ctg4mid[[1]],
  "South"     = cori_palettes$ctg4mid[[2]],
  "Midwest"   = cori_palettes$ctg4mid[[3]],
  "Northeast" = cori_palettes$ctg4mid[[4]]
)

label_df <- regional_rural |> filter(year == max(year))

regional_rural |>
  ggplot(aes(x = year, y = index_value, color = region)) +
  annotate("rect", xmin = 2007, xmax = 2009, ymin = -Inf, ymax = Inf,
           fill = "grey90", alpha = 0.5) +
  geom_hline(yintercept = 1, color = "grey70", linetype = "dashed") +
  geom_line(linewidth = 1.5) +
  geom_text_repel(
    data          = label_df,
    aes(label     = region),
    nudge_x       = 0.5,
    direction     = "y",
    hjust         = 0,
    segment.color = NA,
    size          = 4.5,
    show.legend   = FALSE
  ) +
  scale_color_manual(values = region_colors) +
  scale_x_continuous(
    limits = c(2007, max(regional_rural$year) + 3),
    expand = expansion(mult = c(0.02, 0)),
    labels = scales::label_number(accuracy = 1, big.mark = "")
  ) +
  scale_y_continuous(
    labels = scales::label_percent(accuracy = 1),
    expand = expansion(mult = c(0, .1))
  ) +
  theme_cori() +
  theme(
    legend.position = "none",
    axis.ticks.x    = element_line(color = "#d0d2ce", linewidth = .25),
    plot.title      = element_text(color = cori_colors[["Dark Green"]])
  ) +
  labs(
    title    = "Rural employment by Census region: West leads recovery",
    subtitle = "Relative change in rural employment since the Great Recession",
    x        = NULL,
    y        = NULL,
    caption  = paste0(
      "Source: CORI analysis of BLS Quarterly Census of Employment and Wages.\n",
      "Note: Indexed to 2007 levels. Rural = nonmetro counties per CBSA 2023 definition."
    )
  )

Industry concentration: HHI

The Herfindahl-Hirschman Index (employment_hhi) measures how concentrated a county’s employment is across industries. Higher values indicate fewer industries dominating local employment — a sign of lower diversification. Comparing rural and nonrural counties in the Midwest illustrates how this plays out within a single region.

library(ggbeeswarm)

midwest_fips <- c("17", "18", "19", "20", "26", "27", "29",
                  "31", "38", "39", "46", "55")

emp_raw <- get_employment(geography = "county", years = 2024) |>
  filter(nchar(geoid) == 5, substr(geoid, 1, 2) %in% midwest_fips) |>
  select(geoid, employment = value)

hhi_raw <- get_employment_concentration(geography = "county", years = 2024) |>
  select(geoid, employment_hhi = value)

midwest_co <- emp_raw |>
  left_join(hhi_raw, by = "geoid") |>
  left_join(cbsa_2023 |> select(geoid, is_rural), by = "geoid") |>
  filter(!is.na(is_rural), !is.na(employment_hhi))

fig_swarm <- midwest_co |>
  ggplot(aes(x = employment_hhi, y = is_rural,
             size = employment, color = is_rural)) +
  geom_beeswarm(cex = 0.6, groupOnX = FALSE, alpha = 0.4) +
  scale_x_continuous(
    labels = scales::label_comma(),
    expand = expansion(mult = c(0.02, 0.02))
  ) +
  scale_color_manual(
    values = c(
      "Rural"    = cori_palettes$ctg2ruralnonrural[[1]],
      "Nonrural" = cori_palettes$ctg2ruralnonrural[[2]]
    ),
    guide = "none"
  ) +
  scale_size(
    range  = c(0.5, 6),
    labels = scales::label_comma(scale_cut = scales::cut_short_scale()),
    guide  = guide_legend(
      title          = "Employment",
      title.position = "left",
      title.theme    = element_text(margin = margin(r = 8), size = 11),
      label.theme    = element_text(margin = margin(r = 5), size = 11)
    )
  ) +
  theme_cori() +
  labs(
    title    = "Rural Midwest counties have more concentrated economies",
    subtitle = "Employment HHI by rural status, Midwest counties (2024)",
    x        = "Employment HHI (higher = more concentrated)",
    y        = NULL,
    caption  = paste0(
      "Source: CORI analysis of BLS Quarterly Census of Employment and Wages.\n",
      "Note: Each dot is one county, sized by total employment. ",
      "Rural = nonmetro per CBSA 2023 definition."
    )
  )

fig_swarm
NOTE

Data attribution: QCEW data is produced by the U.S. Bureau of Labor Statistics. When citing analysis using this package, reference: U.S. Bureau of Labor Statistics, Quarterly Census of Employment and Wages, www.bls.gov/cew.