A Cap–Axis Integral Diagnostic of Factor Models

By Useong Shin

Rating

1654
Battle Count: 54

Relevance

5/10
The paper is primarily an academic diagnostic tool for factor model evaluation rather than a direct trading strategy. However, it is relevant to quantitative trading in several ways: (1) it helps practitioners assess whether a chosen factor model adequately prices assets along the size dimension, which affects portfolio construction; (2) the finding that q5's daily distortion attenuates under lead-lag correction while FF/Carhart distortions emerge monthly informs frequency-dependent model selection; (3) the factor-coordinate scan showing cap-axis magnitude is distinct from Sharpe gain helps in factor selection for multi-factor strategies; (4) understanding where pricing errors localize along the cap-rank axis can inform risk management of factor exposures. The relevance is moderate as it is a model evaluation tool rather than a direct alpha-generating methodology.

Implementation Complexity

7/10
Implementation requires: (1) constructing an investable CRSP universe with hysteresis-based screening rules; (2) building whole-stock bridge returns on a 201-point grid with buy-and-hold accounting and dividend reinvestment; (3) joint OLS estimation across the grid; (4) multivariate Newey-West long-run covariance estimation; (5) HAC-Gaussian-process simulation with 50,000 antithetic draws for nonlinear functional p-values; (6) residual-block bootstrap calibration with six-month circular blocks; (7) lead-lag correction across multiple window sizes; (8) rank-area portfolio construction and boundary-quantization auditing. The computational infrastructure for the 201-point grid with daily data over 58 years is substantial, though the matrix algebra is standard. The main complexity lies in careful portfolio accounting and the multi-layered inference procedure.

Reproducibility

4/5
The paper uses publicly available data sources (CRSP, Fama-French Data Library, Global-q.org, Global Factor Data/Jensen et al. 2023). The methodology is fully specified with explicit formulas for bridge construction, grid implementation, HAC-GP inference, and residual-block calibration. Annual Jul-Jun formation, buy-and-hold accounting, and all parameter choices (grid of 201 points, 50,000 Gaussian draws, 21-day/6-month HAC lags) are clearly stated. However, no code repository is provided, and the full implementation details of the CRSP investable universe construction with hysteresis rules would require careful replication.

About this paper

Methodology: Cap-Axis Zero-Alpha Diagnostic. Problem types: Model Evaluation and Testing, Asset Pricing Diagnostics, Statistical Hypothesis Testing, Factor Model Comparison.

The interactive Everscope explorer (charts, battles, favorites) loads below.