This is the title of a working paper by Bianchi, Buchner and Tamoni (2019). There have been many attempts in the academic literature to predict government bond returns, with (successful) factors such as momentum, value and carry, but also with (much less successful) macroeconomic information. In a crude way you could summarize existing approaches as “linear”, where relatively simple input is directly used to predict bond returns. Machine learning techniques are more likely to come up with “non-linear” relationships. The paper finds that when complex non-linear features are introduced via ensembled deep neural networks macroeconomic information does have substantial out-of-sample forecasting power for bond returns. This echoes the conclusion of Gu, Kelly and Xiu (2019) who also claim that machine learning offers an improved description of expected return behavior due to the allowance of nonlinear predictor interactions that are missed by other methods.
As quantitative investors, we are interested to learn whether these claims are indeed true in that they can improve our forecasting models. And whether these results extend to international markets which provides a true out-of-sample test for the claims made on U.S. data. Finally, a crucial test for us is whether we can back up empirical results with a solid economic rationale. Bianchi, Buchner and Tamoni investigate which variables are the main drivers of their predictions, but we would like to dig deeper and thoroughly understand and explain those non-linear relationships.
Robeco’s Global Dynamic Duration fund has a size of around €3.5 billion and its outperformance is driven since 1998 by Robeco’s oldest quantitative model. A successful outcome of the research could lead to a new addition to this model.
The internship candidate should have a strong quantitative background including programming skills and a keen interest in machine learning.
 Bianchi, Daniele, Matthias Buchner, and Andrea Tamoni, 2019. Bond risk premia with machine learning. (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3232721)
 Gu, Shihao, Bryan T. Kelly, and Dacheng Xiu, 2019. Empirical asset pricing via machine learning. (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3159577)