Paper: arXiv 2307.03391

Authors: Chi-Lin Li, Chung-Han Hsieh

Abstract

This paper introduces a unified framework for adaptive portfolio management, integrating dynamic Black-Litterman (BL) optimization with the general factor model, Elastic Net regression, and mean-variance portfolio optimization, which allows us to generate investors views and mitigate potential estimation errors systematically. Specifically, we propose an innovative dynamic sliding window algorithm to respond to the constantly changing market conditions. This algorithm allows for the flexible window size adjustment based on market volatility, generating robust estimates for factor modeling, time-varying BL estimations, and optimal portfolio weights. Through extensive ten-year empirical studies using the top 100 capitalized assets in the S&P 500 index, accounting for turnover transaction costs, we demonstrate that this combined approach leads to computational advantages and promising trading performances.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper involves advanced mathematical formulations including dynamic Black-Litterman extensions with Elastic Net regularization, factor models, and convex optimization, indicating high mathematical complexity. It also demonstrates empirical rigor through a ten-year backtest on S&P 500 assets with transaction costs, providing performance metrics and a structured algorithmic approach.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Unified Adaptive Portfolio<br>Management Framework"] --> B["Data Source:<br>S&P 500 Top 100 Assets<br>10-Year Empirical Study"]
  B --> C["Methodology:<br>Dynamic Sliding Window<br>Adjusts for Market Volatility"]
  C --> D["Core Modeling Steps:<br>1. Factor Model<br>2. Elastic Net Regression<br>3. Time-Varying Black-Litterman"]
  D --> E["Optimization:<br>Mean-Variance Optimization<br>with Transaction Costs"]
  E --> F["Key Outcomes:<br>Computational Efficiency<br>Robust Performance<br>Error Mitigation"]

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 unified adaptive portfolio‑management framework that combines factor‑based view generation, dynamic Black‑Litterman posterior estimation, EWMA covariance estimation and mean‑variance optimization with a volatility‑responsive sliding window.

Source sentence
This paper introduces a unified framework for adaptive portfolio management, integrating dynamic Black-Litterman (BL) optimization with the general factor model, Elastic Net regression, and mean-variance portfolio optimization, which allows us to generate investors views and mitigate potential estimation errors systema

Assumes 5

  • The prior expected excess‑return vector µtk is Gaussian with mean Πtk and covariance Qtk. statistical
    Source sentence
    The prior distribution for µtk is assumed to be Gaussian; i.e., µtk ∼N(Πtk, Qtk), where Πtk is the prior estimate of expected excess returns, …
  • The view‑error covariance Ωtk is diagonal with positive entries bounded below by a small constant. model
    Source sentence
    We specify Ωtk for parsimony as a diagonal matrix with entries
  • Only data available up to tk‑1 are used to compute factor estimates, BL posterior, and portfolio weights at rebalancing time tk. execution
    Source sentence
    the factor‑view lookback, the EWMA covariance update, the CAPM prior inputs, and the adaptive‑window comparison
  • A risk‑free asset is represented by the U.S. Treasury bill 3‑month rate. market
    Source sentence
    U.S. Treasury bill 3 as the risk‑free asset.
Show 1 more
  • Transaction costs are proportional to turnover with rate ν ≥ 0. execution
    Source sentence
    Costtk = ν Turnovertk V+tk , where V+tk is the account value immediately before paying transaction costs.

Data

  • Top‑100 S&P 500 constituents daily prices · CRSP and Compustat market-prices commercial
    2013‑01‑01 to 2023‑01‑01 · 100 largest‑cap S&P 500 stocks
    Source sentence
    using the top 100 market‑capitalization constituents of the S&P 500
  • Synthetic GBM stress‑test data synthetic unknown
    Source sentence
    We therefore simulate asset returns from a correlated geometric Brownian motion (GBM) model calibrated from the same empirical sample used in Section 3.2.

Recognised sources: Compustat CRSP WRDS

Costs

Modeled Costs are part of the reported headline results.

Proportional cost ν applied ex‑post to turnover; results reported for several cost levels (0 %, 0.01 %, 0.1 %, 1 %).

Source sentence
In the empirical study, transaction costs are applied ex‑post to portfolio turnover after each rebalancing decision. Let ν ≥0 denote the proportional transaction‑cost rate.

Finds 2

  • In the synthetic GBM stress test, Algorithm 1 reduces downside risk relative to optimized MV benchmarks, achieving lower drawdown (22.08 % vs 46.92 %). performance
    Source sentence
    % … reduces downside risk relative to the optimized MV benchmarks.
  • The equal‑weighted 1/n portfolio attains the highest Sharpe ratio (0.76) and excess return (13.09 %) in the sample, confirming its empirical robustness. performance
    Source sentence
    the equal‑weighted benchmark retains the highest Sharpe ratio and annualized excess return in this sample

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 15 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 4 papers in the archive

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.