The purpose of the thesis is to investigate to which extent factor models can be used to explain returns in the US corporate bond market. Several risk factors have been suggested as relevant descriptors in the cross-section of contemporaneous returns in this market.
Comparison between the articles proposing the factors is difficult, since many choices of factor construction methods, data-filtration, cleaning, and methodology for testing the models differ considerably between the articles.
By consolidating the research in this area, recreating the experiments in the articles, breaking down the entire factor model construction, holding all else equal but the filtration processes of the bespoke articles, the thesis provides a more comparable analysis of the proposed factors, and thus provides guidance as to which factors are the most relevant.
The recreation of the articles’ filtration processes shows that the core characteristics of the filtered datasets vary to some extent. The results of this indicate that the choice of filtration indeed makes a difference in the perceived significance of the factors, both respectively and combined, and that the significance of some factors is eliminated when controlling for the other proposed factors, as well as when controlling for classic stock-market based factors and interest-rate based control variables.
Our results show that the dataset with the most steps in the filtration process, also contains the most non-investment grade bonds, and vice versa, and that the more detailed filtration process also gives rise to undesirable levels of idiosyncratic risk in portfolio returns, as a result of less diversification. We conclude that factor models are relevant for explaining returns in the US corporate bond market, and that credit-risk and downside-risk are both historically and cross-sectionally significant, and that these results are robust across different filtration processes. Momentum has been a profitable investment strategy in the US corporate bond market, but the associated returns are not significant in the cross-section of contemporaneous returns in this market. Finally, the aggregate volatility factor was neither profitable nor significant, both historically and in regression-based tests.
|Educations||MSc in Finance and Accounting, (Graduate Programme) Final Thesis|
|Number of pages||155|