mliliu/dime-recipients

DIME Recipients Database with Campaign Finance Ideology Scores

Dataset Description

This dataset contains comprehensive information about political recipients (candidates and committees) in the United States from 1980-2024, including their campaign finance-based ideology scores from the Database on Ideology, Money in Politics, and Elections (DIME).

Key Features

• 479,502 recipients across 1980-2024
• Campaign Finance (CF) ideology scores for ideological positioning
• Multiple office levels: Federal (House, Senate), State, Local
• Party affiliations with cleaned coding
• Financial data: Receipts, contributions, expenditures

Dataset Source

• Original Source: Stanford University - Adam Bonica
• Website: https://data.stanford.edu/dime
• Primary Citation: Bonica, Adam. 2014. "Mapping the Ideological Marketplace." American Journal of Political Science 58(2): 367-386.

Dataset Structure

Basic Statistics

• Total Records: 479,502
• Unique Recipients: 216,371
• Time Coverage: 1980-2024
• Total Receipts: $1,114,933,155,977
• Individual Contributions: $573,053,829,718

Key Columns

Identifiers

• bonica.rid: Unique recipient identifier (primary key)
• name: Recipient name (candidate or committee)
• bonica.cid: Contributor identifier (for matching with contributions)

Political Information

• party: Party code (100=Democrat, 200=Republican, 328=Independent)
• recipient.cfscore: Campaign Finance ideology score (-2 to +2, negative=liberal, positive=conservative)
• nimsp.office: Office sought (house, senate, state:lower, local:council, etc.)
• state: State abbreviation
• district: Congressional district (for House candidates)

Financial Data

• total.receipts: Total money raised
• total.indiv.contribs: Individual contribution amounts
• total.pac.contribs: PAC contribution amounts
• num.givers: Number of contributors

Additional Scores

• recipient.cfscore.dyn: Dynamic CF score (time-varying)
• dwnom1: DW-NOMINATE score (for legislators)
• composite.score: Composite ideology measure

Data Distribution

Party Distribution

• I: 173,888 (36.3%)
• D: 154,373 (32.2%)
• R: 149,643 (31.2%)
• O: 1,036 (0.2%)
• L: 226 (0.0%)

Office Distribution

• : 169,451
• house: 115,899
• senate: 42,089
• local:other: 16,436
• local:council: 15,877

CF Score Distribution

• Mean: 0.171
• Std Dev: 1.020
• Range: -6.864 to 6.714
• Median: 0.074

Usage Examples

Basic Loading

from datasets import load_dataset
import pandas as pd

# Load full dataset
dataset = load_dataset("mliliu/dime-recipients")
df = dataset['train'].to_pandas()

print(f"Dataset shape: {df.shape}")
print(f"CF scores available: {df['recipient.cfscore'].notna().sum():,}")

Filtering Examples

# Recent federal candidates only
federal_recent = df[
    (df['cycle'] >= 2016) & 
    (df['nimsp.office'].isin(['house', 'senate'])) &
    (df['recipient.type'] == 'cand')
]

# Major party candidates with CF scores  
major_parties = df[
    df['party'].isin(['100', '200']) &  # Dem/Rep
    df['recipient.cfscore'].notna()
]

# Senate candidates by ideology
senate_liberal = df[
    (df['nimsp.office'] == 'senate') &
    (df['recipient.cfscore'] < -0.5)
]

Ideology Analysis

import matplotlib.pyplot as plt

# Plot ideology distribution by party
dem_scores = df[df['party'] == '100']['recipient.cfscore'].dropna()
rep_scores = df[df['party'] == '200']['recipient.cfscore'].dropna()

plt.hist(dem_scores, alpha=0.7, label='Democrats', bins=50)
plt.hist(rep_scores, alpha=0.7, label='Republicans', bins=50)
plt.xlabel('CF Score (Liberal ← → Conservative)')
plt.ylabel('Frequency')
plt.legend()
plt.show()

Pre-processed Versions Available

This dataset has been optimized and filtered into several versions:

• dime_recent.parquet: Records from 2010+ (277,297 rows)
• dime_federal_candidates.parquet: House + Senate candidates (157,988 rows)
• dime_house_candidates.parquet: House candidates only (115,899 rows)
• dime_senate_candidates.parquet: Senate candidates only (42,089 rows)
• dime_major_parties.parquet: Democrat + Republican only (304,016 rows)

Data Quality Notes

Missing Data Rates

• cf_score: 0.0% missing
• party: 0.0% missing
• state: 0.0% missing
• district: 0.0% missing

Data Cleaning Applied

• Party codes standardized (100→D, 200→R, 328→I, etc.)
• CF scores converted to numeric format
• Office types extracted from NIMSP data
• Decade groupings added for temporal analysis

Methodology: Campaign Finance Scores

The CF scores are estimated using a Bradley-Terry model applied to campaign contribution patterns:

1. Contributors make donations reflecting ideological preferences
2. Recipients receive donations from ideologically-aligned contributors
3. Scaling algorithm positions recipients on liberal-conservative dimension
4. Scores range from -2 (very liberal) to +2 (very conservative)

Key advantages:

• Covers candidates, PACs, and committees
• Available for all time periods
• Not dependent on roll-call votes
• Captures fundraising-based ideology

Licensing and Citation

Usage Rights

• ✅ Academic Research: Permitted
• ❓ Redistribution: Contact original authors
• ❓ Commercial Use: Requires permission

Required Citation

@article{bonica2014mapping,
  title={Mapping the ideological marketplace},
  author={Bonica, Adam},
  journal={American Journal of Political Science},
  volume={58},
  number={2},
  pages={367--386},
  year={2014}
}

Additional References

For methodology details:

• Bonica, Adam. 2016. "Avenues of influence: on the political expenditures of corporations and their directors and executives." Business and Politics 18(4): 367-394.

Technical Details

File Formats

• Parquet: 37.2 MB (recommended for analysis)
• CSV.gz: 28.8 MB (human-readable)
• Sharded: Available for distributed processing

Performance Benchmarks

• Loading time: ~3-5 seconds for full dataset
• Memory usage: ~500MB RAM for full dataset in pandas
• Query performance: Optimized with column indices

Contact

For questions about this dataset preparation:

• Dataset processing: Created for academic research
• Original data: Contact Adam Bonica (Stanford)
• Usage questions: See DIME project documentation

---

Data card generated on 2025-08-03 Processing pipeline version: 1.0