irf23/canadian-parliamentary-expenditures

Canadian House of Commons Parliamentary Expenditures Dataset

This dataset contains detailed expenditure records from the Canadian House of Commons, spanning from 2021 Q2 to 2025 Q4, with 1,219,648 total expenditure records across 450 parliament members.

Dataset Structure

parliamentary_data_hf/
├── data/
│   ├── train/                    # Training split (2021-2024)
│   │   ├── expenditures-2021-q2.parquet
│   │   ├── expenditures-2021-q3.parquet
│   │   ├── ...                   # All quarters from 2021-2024
│   │   ├── expenditures-2024-q4.parquet
│   │   └── members.parquet       # Parliament member records
│   └── test/                     # Test split (2025)
│       ├── expenditures-2025-q1.parquet
│       ├── expenditures-2025-q2.parquet
│       ├── expenditures-2025-q3.parquet
│       ├── expenditures-2025-q4.parquet
│       └── members.parquet       # Parliament member records
├── analytics.json                # Analytics metadata
└── metadata.json                 # Dataset metadata

Data Statistics

• —Total Records: 1,219,648 expenditures
• —Total Members: 450 parliament members
• —Time Period: 2021 Q2 to 2025 Q4 (19 quarters)
• —Parties: Liberal (534,383 records), Conservative (426,037), NDP (82,187), BlocQuebecois (153,319), Independent (17,437), Green (6,285)
• —Categories: Travel, Hospitality, Contract
• —File Sizes: From 474B (2019 Q2) to 67MB (2024 Q4)

Data Format

Expenditure Records

Each expenditure record contains:

• —Id: Unique identifier
• —MemberId: UUID of the member
• —MemberName: Full name of the member
• —Constituency: Electoral district
• —Party: Political party (Liberal, Conservative, NDP, BlocQuebecois, Green, Independent)
• —Category: Type of expenditure (Travel, Hospitality, Contract, Other)
• —Amount: Dollar amount
• —Description: Description of the expenditure
• —Location: Location of expenditure
• —Supplier: Vendor/supplier name
• —PeriodYear: Fiscal year
• —PeriodQuarter: Fiscal quarter (1-4)
• —ReportingPeriodYear: Reporting period year
• —ReportingPeriodQuarter: Reporting period quarter
• —DateIncurred: Date when expense was incurred
• —ClaimId: Claim reference number
• —CreatedAt: Record creation timestamp
• —UpdatedAt: Record update timestamp

Members Data

Each member record contains:

• —MemberId: Unique identifier
• —Name: Full name
• —Constituency: Electoral district
• —Party: Political party affiliation
• —Province: Province code
• —IsActive: Whether currently serving
• —CreatedAt: Record creation timestamp
• —UpdatedAt: Record update timestamp

Data Format

The dataset uses Parquet format, which provides:

• —Efficient columnar storage with excellent compression
• —Fast loading and querying performance
• —Native support in pandas, PySpark, and other data tools
• —Individual quarterly files for selective loading
• —A combined file (expenditures.parquet) with all records
• —Strong typing and schema preservation

Usage

Loading Parquet Data

import pandas as pd
from pathlib import Path

# Load all training data
train_files = list(Path('data/train').glob('expenditures-*.parquet'))
train_df = pd.concat([pd.read_parquet(f) for f in train_files])
print(f"Training records: {len(train_df)}")

# Load test data
test_files = list(Path('data/test').glob('expenditures-*.parquet'))
test_df = pd.concat([pd.read_parquet(f) for f in test_files])
print(f"Test records: {len(test_df)}")

# Load members data
members_df = pd.read_parquet('data/train/members.parquet')
print(f"Total members: {len(members_df)}")

# Load a specific quarter
q1_2024 = pd.read_parquet('data/train/expenditures-2024-q1.parquet')
print(f"Q1 2024: {len(q1_2024)} records")

Working with Analytics Data

import json

# Load analytics metadata
with open('analytics.json', 'r') as f:
    analytics = json.load(f)

# View summary statistics
print(f"Total records: {analytics['total_records']}")
print(f"Date range: {analytics['date_range']['start']} to {analytics['date_range']['end']}")

Analysis Examples

# Total spending by party (using training data)
party_spending = train_df.groupby('Party')['Amount'].sum().sort_values(ascending=False)

# Top spenders across all data
all_df = pd.concat([train_df, test_df])
top_spenders = all_df.groupby('MemberName')['Amount'].sum().nlargest(20)

# Spending trends over time
quarterly_spending = all_df.groupby(['PeriodYear', 'PeriodQuarter'])['Amount'].sum()

# Category breakdown
category_breakdown = all_df.groupby('Category')['Amount'].agg(['sum', 'count', 'mean'])

# Join with members data
merged_df = train_df.merge(members_df, on='MemberId', suffixes=('', '_member'))

Loading with Hugging Face Datasets

from datasets import load_dataset

# Load the complete dataset
dataset = load_dataset('irf23/canadian-parliamentary-expenditures')

# Access expenditures and members
expenditures = dataset['expenditures']
members = dataset['members']

Data Quality Notes

• —Some records may have negative amounts (adjustments/corrections)
• —Party affiliations are now correctly mapped from the members data (previously all showed as "Other")
• —All amounts are in Canadian dollars (CAD)
• —Date fields use ISO 8601 format with timezone information
• —A data entry error where year 2024 was recorded as 224 has been corrected

License

This dataset is compiled from publicly available government data from the House of Commons of Canada. The original data is in the public domain.