Predictive Decision Synthesis for Portfolios: Betting on Better Models

TL;DR

This paper introduces Bayesian predictive decision synthesis (BPDS) for portfolio analysis, enabling improved sequential learning and decision-making under model uncertainty in financial time series forecasting.

Contribution

It develops a novel Bayesian framework for predictive decision synthesis, enhancing portfolio rebalancing strategies with a practical case study on exchange rates.

Findings

BPDS improves decision outcomes over traditional Bayesian methods

Case study demonstrates practical advantages in financial forecasting

Enhanced sequential learning for portfolio management

Abstract

We discuss and develop Bayesian dynamic modelling and predictive decision synthesis for portfolio analysis. The context involves model uncertainty with a set of candidate models for financial time series with main foci in sequential learning, forecasting, and recursive decisions for portfolio reinvestments. The foundational perspective of Bayesian predictive decision synthesis (BPDS) defines novel, operational analysis and resulting predictive and decision outcomes. A detailed case study of BPDS in financial forecasting of international exchange rate time series and portfolio rebalancing, with resulting BPDS-based decision outcomes compared to traditional Bayesian analysis, exemplifies and highlights the practical advances achievable under the expanded, subjective Bayesian approach that BPDS defines.