Paper: arXiv 2306.14004

Authors: Alain-Philippe Fortin, Patrick Gagliardini, Olivier Scaillet

Abstract

We develop a pseudo maximum likelihood method for latent factor analysis in short panels without imposing sphericity nor Gaussianity. We derive an asymptotically uniformly most powerful invariant test for the number of factors. On a large panel of monthly U.S. stock returns, we separate month after month systematic and idiosyncratic risks in short subperiods of bear vs. bull market. We observe an uptrend in the paths of total and idiosyncratic volatilities. The systematic risk explains a large part of the cross-sectional total variance in bear markets but is not driven by a single factor and not spanned by observed factors.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Research Flowchart

  flowchart TD
  A["Research Goal<br>Latent Factor Analysis in Short Panels<br>Without Sphericity or Gaussianity"] --> B["Methodology<br>Pseudo Maximum Likelihood (PML) +<br>Asymptotically Uniformly Most Powerful Test"]
  B --> C["Data Input<br>Large Panel of Monthly<br>U.S. Stock Returns"]
  C --> D{"Computational Process<br>Apply PML to Short Subperiods<br>Bear vs. Bull Markets"}
  D --> E["Key Findings<br>1. Uptrend in Total & Idiosyncratic Volatilities<br>2. Systematic Risk Dominates in Bear Markets<br>3. Systematic Risk Not Spanned by Observed Factors"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a pseudo maximum likelihood method for latent factor analysis in short panels without requiring sphericity or Gaussianity, and derives an asymptotically uniformly most powerful invariant test for the number of factors.

Source sentence
We develop a pseudo maximum likelihood method for latent factor analysis in short panels without imposing sphericity nor Gaussianity. We derive an asymptotically uniformly most powerful invariant test for the number of factors.

Assumes 5

  • Errors are not assumed to be spherical. statistical
    Source sentence
    We develop a pseudo maximum likelihood method for latent factor analysis in short panels without imposing sphericity nor Gaussianity.
  • Errors are not assumed to be Gaussian. statistical
    Source sentence
    We develop a pseudo maximum likelihood method for latent factor analysis in short panels without imposing sphericity nor Gaussianity.
  • Weak cross‑sectional dependence and general heteroskedasticity in idiosyncratic errors are allowed. statistical
    Source sentence
    In addition, we accommodate weak cross‑sectional dependence and general forms of heteroskedasticity (individual and common across assets) in idiosyncratic errors
  • Errors are independent Gaussian vectors in the Gaussian‑error special case. statistical
    Source sentence
    Let us consider the case where the errors εi ind ∼N(0, σiiVε) are independent Gaussian vectors.
Show 1 more
  • Fixed‑T short panels with large cross‑section (n) are the asymptotic framework. model
    Source sentence
    short panels without imposing sphericity nor Gaussianity

Data

  • CRSP panel of monthly U.S. stock returns · CRSP market-prices commercial
    U.S. stocks
    Source sentence
    On a large panel of monthly U.S. stock returns, we separate month after month systematic and idiosyncratic risks in short subperiods of bear vs. bull market.

Recognised sources: CRSP

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Total and idiosyncratic volatilities show an uptrend over time. statistical
    Source sentence
    We observe an uptrend in the paths of total and idiosyncratic volatilities.
  • Systematic risk explains a large part of cross‑sectional variance in bear markets and is not captured by a single factor. theoretical
    Source sentence
    The systematic risk explains a large part of the cross‑sectional total variance in bear markets but is not driven by a single factor and not spanned by observed factors.
  • Imposing sphericity (PCA) leads to an increase of 1–2 (or up to 7) estimated factors, indicating massive over‑rejection. statistical
    Source sentence
    We have an average increase of 7 under sphericity (PCA) as expected from the massive over‑rejection reported in Section 5.3
  • A higher number of latent factors is observed during recessions, robust to different significance levels. statistical
    Source sentence
    Setting α = 0.5% yields an average of 8 factors, while α = 0.05% yields an average of 6 factors. In all cases, the pattern of a higher number of factors during recessions remains unchanged.
  • Observed factor models (FF5+MOM, etc.) struggle to span the latent factor space, especially in market downturns, with only marginal additional explanatory power from less traditional factors. theoretical
    Source sentence
    Observed factors struggle spanning latent factors since their associated linear space is of a dimension r smaller than the one of the latent factor space

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 26 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.