Hugging Face's logo

Datasets:
tidy-finance
/
factor-library-grid

Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
100K - 1M
Tags:
finance
asset-pricing
factor-models
portfolio-sorts
empirical-finance
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
factor-library-grid / README.md
voigtstefan's picture
voigtstefan
Update dataset card
0d530c0 verified
preview code
|
Raw
History Blame Contribute Delete
6.08 kB
metadata 
pretty_name: Tidy Finance Factor Library Specification Grid
license: cc0-1.0
language:
  - en
tags:
  - finance
  - asset-pricing
  - factor-models
  - portfolio-sorts
  - empirical-finance
size_categories:
  - 100K<n<1M

Tidy Finance Factor Library: Specification Grid

Lookup table mapping specification IDs to portfolio sorting configurations. Use this dataset together with the Portfolio Returns dataset to identify the methodological choices behind each factor return series.

Dataset Details

Dataset Description

The dataset contains approximately 960,000 unique specification paths for constructing long-short portfolio returns. Each row defines a complete set of preprocessing and sorting choices (sample exclusions, lagging convention, breakpoint definition, weighting scheme, rebalancing frequency). The id column links to the corresponding return series in the Portfolio Returns dataset.

  • Curated by: Christoph Frey (Lancaster University), Christoph Scheuch (Tidy Intelligence), Stefan Voigt (University of Copenhagen), Patrick Weiss (Reykjavík University)
  • Funded by: Danish Finance Institute
  • License: CC0 1.0

Dataset Sources

Uses

Direct Use

  • Joining with the Portfolio Returns dataset to filter or group factor returns by specific methodological choices.
  • Robustness and sensitivity analysis: selecting subsets of specifications to study how preprocessing decisions affect factor premia.
  • Replication: documenting the exact configuration behind a reported result.

Out-of-Scope Use

  • Standalone analysis. The grid contains no return data and must be joined with the Portfolio Returns dataset via the id column.

Dataset Structure

The dataset consists of a single Parquet file with 14 columns and approximately 960,000 rows.

Column Type Description
id int32 Unique specification identifier, foreign key to the Portfolio Returns dataset
sorting_variable string Sorting characteristic (e.g., sv_ag for asset growth, sv_bm for book-to-market)
min_size_quantile double Minimum NYSE size quantile for sample inclusion: NA (none) or 0.2 (bottom 20th percentile excluded)
exclude_financials bool Whether financial firms (SIC 6000-6799) are excluded
exclude_utilities bool Whether utility firms (SIC 4900-4999) are excluded
exclude_negative_earnings bool Whether firms with negative earnings are excluded
sorting_variable_lag string Lagging convention: 3m, 6m, or ff (Fama-French)
rebalancing string Rebalancing frequency: monthly or annual (July)
n_portfolios_main double Number of quantile portfolios for the primary sort: 3, 5, or 10
sorting_method string Sorting method: univariate, bivariate-dependent, or bivariate-independent
breakpoints_min_size_threshold double Minimum NYSE size quantile used when computing breakpoints: NA (none) or 0.2
n_portfolios_secondary double Number of quantile portfolios for the secondary sort (size): 2, 5, or NA for univariate sorts
breakpoints_exchanges string Exchanges used for breakpoint computation: NYSE or AMEX|NASDAQ|NYSE
weighting_scheme string Portfolio weighting: EW (equal-weighted), VW (value-weighted), or capped VW

Dataset Creation

Curation Rationale

Factor construction involves many subjective methodological choices. Rather than committing to a single specification, we enumerate all valid combinations to enable systematic robustness analysis and transparent reporting.

Source Data

Data Collection and Processing

The grid is generated programmatically from the full factorial combination of preprocessing choices, with invalid configurations removed (e.g., univariate sorts have no secondary breakpoints; market equity is excluded from bivariate sorts where size is the secondary variable; earnings-to-market excludes configurations that allow negative earnings). See code/01_define_portfolio_sorts_grid.R in the companion repository for the exact generation logic.

Who are the source data producers?

The grid is a methodological artifact created by the dataset authors. No external data sources are involved.

Personal and Sensitive Information

The dataset contains no personal or sensitive information. All columns describe portfolio sorting configurations.

Bias, Risks, and Limitations

  • The grid reflects the authors' choice of specification dimensions and does not cover all possible methodological variations (e.g., alternative industry classifications, different minimum listing requirements, or alternative risk-free rate definitions).
  • Some specifications may produce portfolios with very few stocks in certain months, particularly for smaller sorting variables or restrictive exclusion criteria.

Recommendations

Always join with the Portfolio Returns dataset via the id column. When reporting results, cite the specific id or the full set of column values to ensure reproducibility.

Citation

BibTeX:

@article{frey2026transparent,
  title={A Transparent Financial Risk Factor Library},
  author={Frey, Christoph and Scheuch, Christoph and Voigt, Stefan and Weiss, Patrick},
  year={2026},
  journal={Working Paper}
}
Dataset Card Authors
Christoph Frey, Christoph Scheuch, Stefan Voigt, Patrick Weiss

Dataset Card Contact
Stefan Voigt (stefan.voigt@econ.ku.dk), Patrick Weiss (patrickw@ru.is)



---