Can Return Forecasts Enhance International Asset Allocation? Evidence from the Sum-of-Parts Approach

Abstract

We examine whether real-time return forecasts can be valuable to an investor allocating their
portfolio internationally.We expand the Sum-of-Parts (SoP) method for forecasting stock returns
by adding foreign exchange returns as an additional component. We use two dierent methods
to calculate the forecasts. The rst method, Empirical Mode Decomposition (EMD), analyses
each part into locally independent sub-signals, while the second method combines historical
averages and predictive regressions. We then compare the performance of various mean-variance
optimal portfolia under the SoP and historical average return forecasts, with rebalancing taking
place every period. SoP forecasts deliver substantial portfolio performance and economic gains
to an international investor compared to historical average forecasts, especially when EMD is
implemented. Our results remain robust under a battery of tests, variations and dierent investor
domicile, and are driven by an increase in the forecast performance of each part, most notably
the foreign exchange return.

Keywords: Return forecasting, Sum of the Parts, Global asset allocation, EMD, International portfolio optimisation

JEL Classification: G10, G11, G17, G12, G14