70 years of evidence on our dynamic duration model

We used a deep historical dataset that was recently made available to backtest the duration model used for the duration positioning of Robeco’s Global Dynamic Duration strategy. This backtest demonstrates that our model works well over 70 years of data. The model has generated 0.7% annualized alpha in the 23 years of live trading, and it performed well in the original backtest. Together, these two periods are marked by the secular decline in yields. 

This longer backtest shows that the model worked equally well in the 1950s, 1960s and 1970s, when yields were predominantly in a rising trend. The model has typically added most value in periods when bond yields moved most, including recessions and weak periods for risky assets like equities.

Although the model performs well on average, and the positive returns are on average larger than the negative returns, the long backtest also confirms that model performance can be volatile. Around 15% of the annual performances are below -2%, and even over 3-year periods the model performance can be negative. 

The model’s average outperformance and especially the performance when markets move most – both when yields rise and when they fall strongly – makes the Global Dynamic Duration strategy valuable in a portfolio context. But strong hands are a requirement. In order to reap its full benefits, investors need to hold on to the strategy also through weaker periods.