VRscores: Bulk downloads

Use is subject to VRscores Terms of Use: noncommercial only; no governmental/quasi-governmental or political activity. Terms of Use

This beta release contains bulk VRscores panels at the employer, occupation, industry, and metro (MSA) levels covering 2012–2024, published on Harvard Dataverse. Panels are constructed from the 2024 ingest and span the full 2012–2024 period. Read our methodology for field definitions and sourcing.

Jump to panel downloads

Workforce scale

24.5M workers

Unique employees tracked over a decade.

Employer coverage

534K+ employers

Unique employers (VRIDs; VRscores employer identifiers).

Industry depth

1,000+ industries

Six-digit coverage with NAICS2 rollups supplied in exports.

Time span

2012–2024

Annual, unbalanced panel.

Available panels

Download ready-to-use parquet files or preview them on Harvard Dataverse.

VRID (employer) × year

Employer-Year Panel

Beta

2012-2024 employer (VRID) panel. The full dataset contains 6.26M employer-year observations, measuring the partisanship of 24.5M unique workers and more than 534K unique employers.

Years: 2012-2024
Rows: 6.26M employer-years (≈534K unique employers)

View codebook

SOC (occupation) × year

Occupation-Year Panel

Beta

2012-2024 occupation aggregates spanning 256 6-digit SOC groups (4,979 rows).

Years: 2012-2024
Rows: 4,979 occupation-years

View codebook

NAICS × year

Industry-Year Panel

Beta

2012-2024 NAICS (6-digit) panel with 13,131 rows.

Years: 2012-2024
Rows: 13,131 NAICS-years (≈1,010/year)

View codebook

Metropolitan statistical area × year

MSA-Year Panel

Beta

2012-2024 metro (MSA) panel across 366 MSAs.

Years: 2012-2024
Rows: 4,758 metro-years (366/year)

View codebook

Quick start: preview files before downloadingExpand

Panel files are sizeable (employer-year parquet slices approach ~200 MB each). Inspect the headers before streaming the full payload:

# Check size & content type
curl -I https://dataverse.harvard.edu/api/access/datafile/FILE_ID

# Peek at a small sample (text output)
curl "https://dataverse.harvard.edu/api/access/datafile/FILE_ID" | head

Replace FILE_ID with the numeric identifier listed on Dataverse. The same ID works for Parquet (format=default) and CSV (format=original) downloads.

DuckDB (Parquet)

duckdb <<'SQL'
INSTALL httpfs;
LOAD httpfs;
SET enable_progress_bar = true;
-- Inspect a sample without downloading the full file
SELECT * FROM read_parquet('https://dataverse.harvard.edu/api/access/datafile/FILE_ID', hive_partitioning = 0) LIMIT 10;
-- Persist locally if you need everything
COPY (SELECT * FROM read_parquet('https://dataverse.harvard.edu/api/access/datafile/FILE_ID')) TO 'employer_year.parquet';
SQL

Python + Polars (Parquet)

import polars as pl

scan = pl.scan_parquet("https://dataverse.harvard.edu/api/access/datafile/FILE_ID")
print(scan.limit(10).collect())  # preview

# Download selected columns locally
subset = scan.select([
    "vrid",
    "company_name",
    "pct_dem_cws_emp",
    "pct_rep_cws_emp",
])
subset.collect(streaming=True).write_parquet("employer_year_subset.parquet")

CSV access for tabular ingests

# Dataverse stores CSV uploads as .tab; append ?format=original
curl -L 'https://dataverse.harvard.edu/api/access/datafile/FILE_ID?format=original' -o msa_year.csv

Dataverse exports use the same guardrails as the explorer (employers with fewer than 25 matched workers are excluded unless noted otherwise).

Codebooks

Preview the column dictionaries for each dataset before downloading. Expand a panel to see the available fields and definitions.

Employer-Year PanelExpand

Full column dictionary for the VRID-year aggregate (2012-2024).

Column	Type	Description
vrid	string	VRscores employer identifier (VRID).
year	integer	Calendar year.
company_name	string	Name of employer.
employee_count	integer	Number of unique workers in the company-year (minimum 5).
avg_match_quality	float	Average match quality across all matched workers (higher is a higher probability match).
dem_workers_raw	float	Workers with Democratic affiliation based solely on L2 voter registrations (raw counts).
rep_workers_raw	float	Workers with Republican affiliation based solely on L2 voter registrations (raw counts).
other_workers_raw	float	Matched workers whose L2 record is neither Democratic nor Republican (raw counts).
party_known_workers_raw	float	Workers with any L2 party assignment (Democratic, Republican, or other).
dem_workers_imp	float	Imputed Democratic worker count (we are able to impute most, but not all, independents/unknown).
rep_workers_imp	float	Imputed Republican worker count (we are able to impute most, but not all, independents/unknown).
other_workers_imp	float	Workers who are not registered Democrat or Republican and for whom we could not confidently impute partisanship (`max(employee_count − dem_workers_imp − rep_workers_imp, 0)`).
democrat_pct_raw	float	Democratic share using raw L2 counts (`dem_workers_raw / employee_count`).
republican_pct_raw	float	Republican share using raw L2 counts.
nonpartisan_pct_raw	float	Nonpartisan share using raw L2 counts.
democrat_pct_two_party_raw	float	Democratic share among two-party workers from raw L2 counts (`dem_workers_raw / (dem_workers_raw + rep_workers_raw)`).
republican_pct_two_party_raw	float	Republican share among two-party workers from raw L2 counts (`rep_workers_raw / (dem_workers_raw + rep_workers_raw)`).
two_party_margin_raw	float	Republican minus Democratic share among two-party workers (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / (dem_workers_raw + rep_workers_raw)`).
overall_margin_raw	float	Republican minus Democratic share of the full workforce (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / employee_count`).
democrat_pct_imp	float	Democratic share after imputing independents/unknown (`dem_workers_imp / employee_count`; registered partisans remain unchanged).
republican_pct_imp	float	Republican share after imputing independents/unknown (`rep_workers_imp / employee_count`; registered partisans remain unchanged).
nonpartisan_pct_imp	float	Nonpartisan share after imputing independents/unknown (`other_workers_imp / employee_count`).
democrat_pct_two_party_imp	float	Democratic share among imputed two-party workers (`dem_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
republican_pct_two_party_imp	float	Republican share among imputed two-party workers (`rep_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
two_party_margin_imp	float	Imputed two-party margin (`(rep_workers_imp − dem_workers_imp) / (dem_workers_imp + rep_workers_imp)`).
overall_margin_imp	float	Imputed overall margin (`(rep_workers_imp − dem_workers_imp) / employee_count`).
political_diversity_raw	float	One minus the sum of squared raw partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
political_diversity_imp	float	One minus the sum of squared imputed partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
effective_parties_raw	float	Effective number of parties based on raw shares (inverse Herfindahl index).
effective_parties_imp	float	Effective number of parties based on imputed shares (inverse Herfindahl index).
latest_processed_at	string	ISO 8601 timestamp indicating when the company-year was last processed.

Occupation-Year PanelExpand

SOC-level column dictionary for the occupation-year panel (2012-2024).

Column	Type	Description
onet_code	string	O*NET SOC occupation code at the occupation-year aggregation level.
onet_title	string	Occupation title corresponding to the O*NET SOC code.
employee_count	integer	Number of unique workers in the occupation-year (minimum 5).
rep_workers_raw	float	Workers with Republican affiliation based solely on L2 voter registrations (raw counts).
other_workers_raw	float	Matched workers whose L2 record is neither Democratic nor Republican (raw counts).
party_known_workers_raw	float	Workers with any L2 party assignment (Democratic, Republican, or other).
dem_workers_imp	float	Imputed Democratic worker count (we are able to impute most, but not all, independents/unknown).
rep_workers_imp	float	Imputed Republican worker count (we are able to impute most, but not all, independents/unknown).
other_workers_imp	float	Workers who are not registered Democrat or Republican and for whom we could not confidently impute partisanship (`max(employee_count − dem_workers_imp − rep_workers_imp, 0)`).
avg_match_quality	float	Average match quality across all matched workers (higher is a higher probability match).
democrat_pct_raw	float	Democratic share using raw L2 counts (`dem_workers_raw / employee_count`).
republican_pct_raw	float	Republican share using raw L2 counts.
nonpartisan_pct_raw	float	Nonpartisan share using raw L2 counts.
democrat_pct_two_party_raw	float	Democratic share among two-party workers from raw L2 counts (`dem_workers_raw / (dem_workers_raw + rep_workers_raw)`).
republican_pct_two_party_raw	float	Republican share among two-party workers from raw L2 counts (`rep_workers_raw / (dem_workers_raw + rep_workers_raw)`).
two_party_margin_raw	float	Republican minus Democratic share among two-party workers (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / (dem_workers_raw + rep_workers_raw)`).
overall_margin_raw	float	Republican minus Democratic share of the full workforce (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / employee_count`).
democrat_pct_imp	float	Democratic share after imputing independents/unknown (`dem_workers_imp / employee_count`; registered partisans remain unchanged).
republican_pct_imp	float	Republican share after imputing independents/unknown (`rep_workers_imp / employee_count`; registered partisans remain unchanged).
nonpartisan_pct_imp	float	Nonpartisan share after imputing independents/unknown (`other_workers_imp / employee_count`).
democrat_pct_two_party_imp	float	Democratic share among imputed two-party workers (`dem_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
republican_pct_two_party_imp	float	Republican share among imputed two-party workers (`rep_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
two_party_margin_imp	float	Imputed two-party margin (`(rep_workers_imp − dem_workers_imp) / (dem_workers_imp + rep_workers_imp)`).
overall_margin_imp	float	Imputed overall margin (`(rep_workers_imp − dem_workers_imp) / employee_count`).
political_diversity_raw	float	One minus the sum of squared raw partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
political_diversity_imp	float	One minus the sum of squared imputed partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
effective_parties_raw	float	Effective number of parties based on raw shares (inverse Herfindahl index).
effective_parties_imp	float	Effective number of parties based on imputed shares (inverse Herfindahl index).
latest_processed_at	string	ISO 8601 timestamp indicating when the occupation-year was last processed.

Industry-Year PanelExpand

NAICS-level column dictionary for the industry-year panel (2012-2024).

Column	Type	Description
naics_code	string	Primary NAICS code.
naics_desc	string	Text description for the NAICS code.
employee_count	integer	Number of unique workers in the NAICS-year group (minimum 5).
dem_workers_raw	float	Workers with Democratic affiliation based solely on L2 voter registrations (raw counts).
rep_workers_raw	float	Workers with Republican affiliation based solely on L2 voter registrations (raw counts).
other_workers_raw	float	Matched workers whose L2 record is neither Democratic nor Republican (raw counts).
party_known_workers_raw	float	Workers with any L2 party assignment (Democratic, Republican, or other).
dem_workers_imp	float	Imputed Democratic worker count (we are able to impute most, but not all, independents/unknown).
rep_workers_imp	float	Imputed Republican worker count (we are able to impute most, but not all, independents/unknown).
other_workers_imp	float	Workers who are not registered Democrat or Republican and for whom we could not confidently impute partisanship (`max(employee_count − dem_workers_imp − rep_workers_imp, 0)`).
avg_match_quality	float	Average match quality across all matched workers (higher is a higher probability match).
democrat_pct_raw	float	Democratic share using raw L2 counts (`dem_workers_raw / employee_count`).
republican_pct_raw	float	Republican share using raw L2 counts.
nonpartisan_pct_raw	float	Nonpartisan share using raw L2 counts.
democrat_pct_two_party_raw	float	Democratic share among two-party workers from raw L2 counts (`dem_workers_raw / (dem_workers_raw + rep_workers_raw)`).
republican_pct_two_party_raw	float	Republican share among two-party workers from raw L2 counts (`rep_workers_raw / (dem_workers_raw + rep_workers_raw)`).
two_party_margin_raw	float	Republican minus Democratic share among two-party workers (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / (dem_workers_raw + rep_workers_raw)`).
overall_margin_raw	float	Republican minus Democratic share of the full workforce (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / employee_count`).
democrat_pct_imp	float	Democratic share after imputing independents/unknown (`dem_workers_imp / employee_count`; registered partisans remain unchanged).
republican_pct_imp	float	Republican share after imputing independents/unknown (`rep_workers_imp / employee_count`; registered partisans remain unchanged).
nonpartisan_pct_imp	float	Nonpartisan share after imputing independents/unknown (`other_workers_imp / employee_count`).
democrat_pct_two_party_imp	float	Democratic share among imputed two-party workers (`dem_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
republican_pct_two_party_imp	float	Republican share among imputed two-party workers (`rep_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
two_party_margin_imp	float	Imputed two-party margin (`(rep_workers_imp − dem_workers_imp) / (dem_workers_imp + rep_workers_imp)`).
overall_margin_imp	float	Imputed overall margin (`(rep_workers_imp − dem_workers_imp) / employee_count`).
political_diversity_raw	float	One minus the sum of squared raw partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
political_diversity_imp	float	One minus the sum of squared imputed partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
effective_parties_raw	float	Effective number of parties based on raw shares (inverse Herfindahl index).
effective_parties_imp	float	Effective number of parties based on imputed shares (inverse Herfindahl index).
latest_processed_at	string	ISO 8601 timestamp indicating when the NAICS-year was last processed.

MSA PanelExpand

Metro-level column dictionary for the MSA panel (2012-2024).

Column	Type	Description
year	integer	Calendar year.
msa	string	Combined CBSA/MSA identifier derived from position-level data.
employee_count	integer	Number of unique workers in the MSA-year (minimum 5).
dem_workers_raw	float	Workers with Democratic affiliation based solely on L2 voter registrations (raw counts).
rep_workers_raw	float	Workers with Republican affiliation based solely on L2 voter registrations (raw counts).
other_workers_raw	float	Matched workers whose L2 record is neither Democratic nor Republican (raw counts).
party_known_workers_raw	float	Workers with any L2 party assignment (Democratic, Republican, or other).
dem_workers_imp	float	Imputed Democratic worker count (we are able to impute most, but not all, independents/unknown).
rep_workers_imp	float	Imputed Republican worker count (we are able to impute most, but not all, independents/unknown).
other_workers_imp	float	Workers who are not registered Democrat or Republican and for whom we could not confidently impute partisanship (`max(employee_count − dem_workers_imp − rep_workers_imp, 0)`).
avg_match_quality	float	Average match quality across all matched workers (higher is a higher probability match).
democrat_pct_raw	float	Democratic share using raw L2 counts (`dem_workers_raw / employee_count`).
republican_pct_raw	float	Republican share using raw L2 counts.
nonpartisan_pct_raw	float	Nonpartisan share using raw L2 counts.
democrat_pct_two_party_raw	float	Democratic share among two-party workers (`dem_workers_raw / (dem_workers_raw + rep_workers_raw)`).
republican_pct_two_party_raw	float	Republican share among two-party workers from raw L2 counts (`rep_workers_raw / (dem_workers_raw + rep_workers_raw)`).
two_party_margin_raw	float	Republican minus Democratic share among two-party workers (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / (dem_workers_raw + rep_workers_raw)`).
overall_margin_raw	float	Republican minus Democratic share of the full workforce (raw L2 counts; `(rep_workers_raw − dem_workers_raw) / employee_count`).
democrat_pct_imp	float	Democratic share after imputing independents/unknown (`dem_workers_imp / employee_count`; registered partisans remain unchanged).
republican_pct_imp	float	Republican share after imputing independents/unknown (`rep_workers_imp / employee_count`; registered partisans remain unchanged).
nonpartisan_pct_imp	float	Nonpartisan share after imputing independents/unknown (`other_workers_imp / employee_count`).
democrat_pct_two_party_imp	float	Democratic share among imputed two-party workers (`dem_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
republican_pct_two_party_imp	float	Republican share among imputed two-party workers (`rep_workers_imp / (dem_workers_imp + rep_workers_imp)`; imputations only affect independents/unknown).
two_party_margin_imp	float	Imputed two-party margin (`(rep_workers_imp − dem_workers_imp) / (dem_workers_imp + rep_workers_imp)`).
overall_margin_imp	float	Imputed overall margin (`(rep_workers_imp − dem_workers_imp) / employee_count`).
political_diversity_raw	float	One minus the sum of squared raw partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
political_diversity_imp	float	One minus the sum of squared imputed partisan shares (1 − ∑pᵢ²); higher values indicate more partisan diversity.
effective_parties_raw	float	Effective number of parties based on raw shares (inverse Herfindahl index).
effective_parties_imp	float	Effective number of parties based on imputed shares (inverse Herfindahl index).
latest_position_enddate	timestamp	Most recent position end timestamp among workers contributing to the MSA-year.