ABSTRACT
In this article, we develop a nonparametric functional data analysis (NP-FDA) model to forecast the term-structure of Brazil, Russia, India, China and South Africa (BRICS). We use daily data over the period of January 1, 2010 to December 31, 2016. We find that, while it is in general difficult to beat the random-walk model in the shorter-horizons, at longer-runs our proposed NP-FDA approach outperforms not only the random-walk model, but also other popular competitors used in term-structure forecasting literature. In addition, the NP-FDA model is also found to produce economic gains, besides statistical gains, over the random-walk model. Our results have important implications for both policymakers aiming to stabilize the economy, and for optimal portfolio allocation decisions of financial market agents.
Acknowledgments
The authors would like to thank two anonymous referees and the subject editor, and former and current editors, Professor Ali M. Kutan and Professor Paresh K. Narayan respectively, for many helpful comments. However, any remaining errors are solely ours.
Notes
1. For further details about the method, see Ferraty and Vieu (Citation2006) and Ramsay and Silverman (Citation2005).
2. For other estimators of this type see Caldeira and Torrent (Citation2017)
3. For details, see sub-section 3.1 of Ferraty and Vieu (Citation2006).
4. We have also estimated the models using an expanding window. However, the results obtained were qualitatively similar to those presented here, and are available upon request from thee authors.
5. See the appendix for details on the FS-test.
6. We follow the strategy of Rapach and Zhou, (Citation2013) and estimate the variance of bond returns using the sample variance computed from a one-year (-obs) rolling window of historical returns.