Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Latent Variable Estimation in Bayesian Black-Litterman Models

Thomas Y. L. Lin, Jerry Yao-Chieh Hu, Paul W. Chiou, Peter Lin

TL;DR

Addresses the reliance on subjective investor views in the Black-Litterman framework by treating $q$ and $\Omega$ as latent variables to be inferred from market data within a single Bayesian network, yielding a closed-form posterior for $\theta$ and predictive returns $\widetilde{r}$. The approach unifies feature integration with parameter learning and introduces two latent-view configurations (SLP-BL and FIV-BL), with a Mixed-effect BL (M-BL) serving as a bridge when views are observed. Empirically, on 20- and 30-year Dow Jones and SPDR Sector ETF data, the method increases Sharpe ratios by about $50\%$ and reduces turnover by about $55\%$ relative to Markowitz and market indices, demonstrating a fully data-driven, view-free, coherent Bayesian framework for portfolio optimization.

Abstract

We revisit the Bayesian Black-Litterman (BL) portfolio model and remove its reliance on subjective investor views. Classical BL requires an investor "view": a forecast vector $q$ and its uncertainty matrix $Ω$ that describe how much a chosen portfolio should outperform the market. Our key idea is to treat $(q,Ω)$ as latent variables and learn them from market data within a single Bayesian network. Consequently, the resulting posterior estimation admits closed-form expression, enabling fast inference and stable portfolio weights. Building on these, we propose two mechanisms to capture how features interact with returns: shared-latent parametrization and feature-influenced views; both recover classical BL and Markowitz portfolios as special cases. Empirically, on 30-year Dow-Jones and 20-year sector-ETF data, we improve Sharpe ratios by 50% and cut turnover by 55% relative to Markowitz and the index baselines. This work turns BL into a fully data-driven, view-free, and coherent Bayesian framework for portfolio optimization.

Latent Variable Estimation in Bayesian Black-Litterman Models

TL;DR

Addresses the reliance on subjective investor views in the Black-Litterman framework by treating and as latent variables to be inferred from market data within a single Bayesian network, yielding a closed-form posterior for and predictive returns . The approach unifies feature integration with parameter learning and introduces two latent-view configurations (SLP-BL and FIV-BL), with a Mixed-effect BL (M-BL) serving as a bridge when views are observed. Empirically, on 20- and 30-year Dow Jones and SPDR Sector ETF data, the method increases Sharpe ratios by about and reduces turnover by about relative to Markowitz and market indices, demonstrating a fully data-driven, view-free, coherent Bayesian framework for portfolio optimization.

Abstract

We revisit the Bayesian Black-Litterman (BL) portfolio model and remove its reliance on subjective investor views. Classical BL requires an investor "view": a forecast vector and its uncertainty matrix that describe how much a chosen portfolio should outperform the market. Our key idea is to treat as latent variables and learn them from market data within a single Bayesian network. Consequently, the resulting posterior estimation admits closed-form expression, enabling fast inference and stable portfolio weights. Building on these, we propose two mechanisms to capture how features interact with returns: shared-latent parametrization and feature-influenced views; both recover classical BL and Markowitz portfolios as special cases. Empirically, on 30-year Dow-Jones and 20-year sector-ETF data, we improve Sharpe ratios by 50% and cut turnover by 55% relative to Markowitz and the index baselines. This work turns BL into a fully data-driven, view-free, and coherent Bayesian framework for portfolio optimization.
Paper Structure (33 sections, 14 theorems, 110 equations, 11 figures, 6 tables)

This paper contains 33 sections, 14 theorems, 110 equations, 11 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Methodology
  4. Bayesian Black-Litterman Network
  5. Bayesian Feature-Integrated Black-Litterman Network
  6. General Scenario: Features with Observed Views
  7. Scenario: Features with Latent Views
  8. Proof-of-Concept Experiments
  9. Discussion and Conclusion
  10. Hyperparameter Selections
  11. Related Work
  12. Bayesian Portfolio Optimization
  13. Data-Driven Black-Litterman Model
  14. Supplementary Theoretical Backgrounds
  15. Prior and Likelihood of the Black-Litterman-Bayes model
  16. ...and 18 more sections

Key Result

Theorem 2.1

Let $r \in \mathbb{R}^m$ be the vector of asset returns with covariance $\Sigma \coloneqq \mathrm{Cov}[r]$. Let $P \in \mathbb{R}^{k \times m}$ be the portfolio weight matrix for $k$ specified portfolios, and $(q, \Omega) \in \mathbb{R}^k \times \mathbb{R}^{k \times k}$ represent investor views and where $G_{\tau} \coloneqq (\tau\,\Sigma)^{-1} + P^\top\,\Omega^{-1}\,P$. Moreover, the predictive d

Figures (11)

  • Figure 1: Black-Litterman network $(\theta, r, q, \Omega)$.
  • Figure 2: Feature-Integrated BL network with features and observed views $(\theta, r, q, \Omega, F, \Omega^F)$.
  • Figure 4: Cumulative Return on SPDR Sectors ETFs Dataset.
  • Figure 5: Cumulative Return on Dow Jones Index Dataset.
  • Figure 6: Asset Allocation of traditional MV model (rolling window: 100 days) on SPDR Sectors ETFs Dataset.
  • ...and 6 more figures

Theorems & Definitions (45)

  • Definition 2.1: Unconstrained Risk-Adjusted Mean-Variance Optimization
  • Theorem 2.1: Black-Litterman (BL) Formula and Predictive Estimation, Theorem 1 of satchell2007demystification
  • Definition 2.2: BLB Model $(\theta,r,q,\Omega)$, Modified from Definition 1 of kolm2017bayesian
  • Lemma 2.1: Estimations by BLB Model
  • proof
  • Definition 3.1: $\theta\leftrightarrow F$ Linear Model
  • Definition 3.2: $q\leftrightarrow F\leftrightarrow \theta$ Linear Model
  • Remark 3.1: Rationale
  • Definition 3.3: Mixed-effect Black-Litterman (M-BL) Model $(\theta,r,q,\Omega,F,\Omega^F)$
  • Theorem 3.1: Parameter Estimation of the M-BL Model
  • ...and 35 more