Does forecasting improve portfolio management returns? Empirical results from strategy backtesting

Abstract

This paper aims to determine whether forecasts from two popular macroeconometric models are useful to improve portfolio returns. The paper begins by estimating two large macro models namely Global Vector Autoregressive (GVAR) and Factor-Augmented Vector Autoregressive (FAVAR). The forecasts from these models are then used in a backtester, simulating a trading rule. In the first empirical test with a simple single position test, perfect forecast performed best but the highest return came from a strategy that uses GVAR forecast although it has a lower Sharpe ratio. The result from the second backtest with multiple positions is more in line with expectation as a strategy using the perfect forecast outperformed GVAR in all scenarios. The evidence from this paper shows how investment returns are driven by forecast accuracy but also heavily on portfolio management criteria.