Dynamic Portfolio Management with Views at Multiple Horizons
Applied Mathematics and Computation, Volume 274, 1 February 2016, Pages 495-518
33 Pages Posted: 24 Mar 2015 Last revised: 10 Dec 2017
Date Written: April 16, 2015
Abstract
We introduce Dynamic Entropy Pooling, a quantitative technique to perform dynamic portfolio construction with discretionary, non-synchronous views. With Dynamic Entropy Pooling, the portfolio manager can embed in the allocation process signals with life spans ranging from minutes to years, calendar views, autocorrelation stress-testing, and the traditional views on expectations, correlations and volatilities.
After introducing the theoretical framework for Dynamic Entropy Pooling, we show how to solve the respective portfolio construction problem by means of dynamic programming with time-dependent coefficients. To understand the optimal exposures ensuing from Dynamic Entropy Pooling we analyze a variety of relevant sub-cases and we present two case-studies.
Keywords: dynamic strategies, discretionary allocation, mean-reversion, multivariate Ornstein-Uhleneck, Kullback Leibler divergence, relative entropy, optimal policy, dynamic programming, calculus of variations
JEL Classification: C1, G11
Suggested Citation: Suggested Citation