Estimating Bond Fund Duration for Government Bond Timing

Summary

The study estimates duration for medium- and long-term pure bond funds using rolling regressions of fund returns on bond-index returns across maturity segments. It compares constrained linear regression, stepwise selection, and LASSO, selecting LASSO to reduce estimation error and improve robustness. Fund duration is inferred by combining estimated maturity-bucket exposures with the durations of the corresponding indices. The authors report close agreement with disclosed duration figures in their tested sample.

The analysis treats changes in average or median fund duration as signals of managers’ interest-rate views, while cross-sectional duration dispersion measures disagreement. It develops a government-bond timing strategy from these measures and also studies market responses to central-bank policy announcements. Adding duration and policy signals to an earlier multi-factor model improved the reported historical results through January 2022. These findings are based on historical Chinese bond-market data; the document cautions that model performance may weaken when market conditions change, and its reported backtests do not establish future performance.

Key ideas

• Fund duration is estimated from rolling regressions of fund returns on maturity-segment bond-index returns.
• LASSO variable selection is chosen after comparison with constrained linear and stepwise regression methods.
• Changes in fund duration are interpreted as a signal of managers’ rate outlook, while duration dispersion indicates disagreement.
• The study combines duration and monetary-policy announcement signals in government-bond timing models.
• The reported results are historical backtests and may not persist under different market conditions.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.