Paper: arXiv 2403.02500

Authors: Yilun Wang, Shengjie Guo

Abstract

In recent years, the dynamic factor model has emerged as a dominant tool in economics and finance, particularly for investment strategies. This model offers improved handling of complex, nonlinear, and noisy market conditions compared to traditional static factor models. The advancement of machine learning, especially in dealing with nonlinear data, has further enhanced asset pricing methodologies. This paper introduces a groundbreaking dynamic factor model named RVRAE. This model is a probabilistic approach that addresses the temporal dependencies and noise in market data. RVRAE ingeniously combines the principles of dynamic factor modeling with the variational recurrent autoencoder (VRAE) from deep learning. A key feature of RVRAE is its use of a prior-posterior learning method. This method fine-tunes the model’s learning process by seeking an optimal posterior factor model informed by future data. Notably, RVRAE is adept at risk modeling in volatile stock markets, estimating variances from latent space distributions while also predicting returns. Our empirical tests with real stock market data underscore RVRAE’s superior performance compared to various established baseline methods.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper involves advanced deep learning architecture combining variational inference and recurrent neural networks, resulting in high math complexity; empirical rigor is high due to the use of real stock market data, backtesting against baselines, and detailed methodology for implementation.

Research Flowchart

  flowchart TD
  A["Research Goal: Predict Stock Returns using Dynamic Factors"] --> B{"Methodology"}
  B --> C["Data: Real Stock Market Data<br>(Input for Modeling)"]
  C --> D["Computational Process: RVRAE Model<br>Combines Dynamic Factor Model + VRAE<br>Uses Prior-Posterior Learning"]
  D --> E["Key Findings/Outcomes"]
  E --> F["1. Superior Prediction Performance<br>vs. Baseline Methods"]
  E --> G["2. Effective Risk Modeling<br>Estimates Variances from Latent Space"]
  E --> H["3. Handles Nonlinear & Noisy<br>Temporal Market Data"]

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 RVRAE, a dynamic latent factor model that combines recurrent neural networks with a variational autoencoder to extract factors from noisy market data and predict cross‑sectional stock returns.

Source sentence
We introduce RVRAE as an innovative dynamic latent factor model tailored to extract key factors from noisy market data efficiently.

Assumes 5

  • The prior distribution of the latent variable z_t is a standard Gaussian. model
    Source sentence
    In vanilla VAE, pθ (zt) is assumed as a standard Gaussian distribution.
  • Factor exposures β_t can vary over time (dynamic betas). model
    Source sentence
    However, challenges arise with dynamic betas, indicating that factor exposures may change over time as firms evolve.
  • Monthly individual stock returns from CRSP are representative of the US equity market. data
    Source sentence
    We collect monthly individual stock returns from CRSP The returns dataset starts from January 2000 to December 2020.
  • The 46 firm characteristics capture relevant information for factor extraction. data
    Source sentence
    According to [20] and [21], we choose 46 firm characteristics.
Show 1 more
  • Transaction costs are a flat 30 basis points per trade. execution
    Source sentence
    Following [9], we consider two situations: Sharpe Ratio without transaction costs and Sharpe Ratio with 30 bps transaction costs.

Data

  • CRSP monthly individual stock returns · CRSP market-prices commercial
    2000-01 to 2020-12 · 5682 US stocks
    Source sentence
    We collect monthly individual stock returns from CRSP The returns dataset starts from January 2000 to December 2020. The number of stocks is 5682.
  • Firm characteristics fundamentals unknown
    Source sentence
    According to [20] and [21], we choose 46 firm characteristics.

Recognised sources: CRSP Kenneth French Data Library

Costs

Modeled Costs are part of the reported headline results.

30 basis points per trade are subtracted when reporting Sharpe ratios.

Source sentence
Following [9], we consider two situations: Sharpe Ratio without transaction costs and Sharpe Ratio with 30 bps transaction costs.

Finds 2

  • All models’ Sharpe ratios decline when transaction costs are added, but RVRAE still outperforms the others. performance
    Source sentence
    It is interesting to notice that all models perform worse when adding transaction costs, but models with Attention mechanism (ALSTM, Trans) have the smallest difference...
  • RVRAE consistently outperforms baselines across robustness tests with omitted stocks (higher Rank IC and Rank ICIR). statistical
    Source sentence
    It reveals that our RVRAE model outperforms other baseline methods for all values of ’m’.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 2 papers in the archive (2 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.