Dataset Card for Risk-Free Rate Dataset (tidy-finance/risk-free)

Pre-processed U.S. risk-free rates constructed from Treasury bill yields, designed for asset pricing, empirical finance, and reproducible workflows.

Dataset Details

Dataset Description

This dataset provides daily and monthly U.S. risk-free rates derived from FRED Treasury bill series:

• TB3MS / DTB3 (3-month T-bill)
• DTB4WK (4-week T-bill)

The series is spliced at 2001-07-01 to ensure continuity across maturities and is updated automatically each month.

Methodology

Source data are annualized bank discount rates (360-day basis). These are converted to holding-period returns (HPR):

HPR = d * (n / 360) / (1 - d * (n / 360))

and scaled to the target frequency:

(1 + HPR)^(target/source) - 1

Splicing

• Pre-2001:

  • Monthly: TB3MS (n = 90), exponent = 1/3
  • Daily: DTB3 (n = 90), exponent ≈ 1/63

• From 2001:

  • DTB4WK (n = 28)
  • Monthly: last observation per month, exponent = 365 / (28 * 12)
  • Daily: exponent ≈ 1/20

Daily gaps (e.g. holidays) are forward-filled.

Coverage

• Monthly: from 1934-01-01

• Daily: from 1954-01-04

• Curated by: Tidy Finance

• License: CC0-1.0

Dataset Sources

• Repository: https://huggingface.co/datasets/tidy-finance/risk-free
• Source: FRED (Federal Reserve Bank of St. Louis)

Uses

Direct Use

• Asset pricing and excess return construction
• Backtesting and portfolio analysis
• Financial econometrics
• Teaching and replication

Out-of-Scope Use

• Intraday or high-frequency applications
• Non-U.S. risk-free rates
• Real-time or unrevised data use

Dataset Structure

Two tables:

• daily
• monthly

Each is a tibble with:

• date: observation date
• risk_free: risk-free rate

Dataset Creation

Curation Rationale

Provides a standardized, reproducible risk-free rate series consistent with common empirical finance practice.

Source Data and Processing

• Data retrieved from FRED
• Discount rates → holding-period returns → target frequency
• Series spliced at 2001-07-01
• Missing daily values forward-filled
• Updated monthly via automated pipeline

Personal and Sensitive Information

None. The dataset contains only public aggregate financial data.

Bias, Risks, and Limitations

• Depends on specific maturities and fixed splicing rule
• Approximate trading-day conventions introduce small errors
• Forward-filling smooths short gaps
• FRED data may be revised
• U.S.-only

Recommendations

Ensure consistency with return frequency and be aware of construction choices when comparing to alternative risk-free proxies.

Citation

BibTeX:

@book{scheuch2023tidy,
  title = {Tidy Finance with R},
  author = {Scheuch, Christoph and Voigt, Stefan and Weiss, Patrick},
  year = {2023},
  publisher = {Chapman and Hall/CRC},
  doi = {10.1201/b23237}
}

APA:

Scheuch, C., Voigt, S., & Weiss, P. (2023). Tidy Finance with R. Chapman and Hall/CRC. https://doi.org/10.1201/b23237