Forecasting covariance for optimal carry trade portfolio allocations
Matthew Ames, Guillaume Bagnarosa, Gareth W. Peters, Pavel V. Shevchenko, Tomoko Matsui
Abstract
Modelling and forecasting of asset volatility and covariance is of prime importance in the construction of portfolios. In this paper, we present a generalised multi-factor model that incorporates heteroskedasticity and dependence in the idiosyncratic error terms. We apply this model to forecasting the time-varying covariances in a basket of high interest rate and a basket of low interest rate carry trade currencies and then utilise these forecasts for portfolio optimisation. We compare traditional Markowitz portfolio optimisation to the more recently popular risk-based portfolio optimisation. Our model is shown to provide superior risk-adjusted returns for a currency carry trade strategy over the period 1999 – 2014.
BibTeX
@inproceedings{icassp2017_forecastingcovar,
title = {Forecasting covariance for optimal carry trade portfolio allocations},
author = {Matthew Ames and Guillaume Bagnarosa and Gareth W. Peters and Pavel V. Shevchenko and Tomoko Matsui},
booktitle = {ICASSP 2017},
year = {2017}
}