Dynamisk aktivallokering gennem konjunkturfaser: En empirisk analyse af aktivklassers performance og anvendelsen af Black-Litterman modellen baseret på OECDs CLI-indikator

Abstract

This thesis investigates how business cycles affect the return and risk profiles of seven asset classes over the period 1990–2024, and whether this knowledge can be applied in a dynamic portfolio strategy based on the Black-Litterman model. Utilizing the OECD’s Composite Leading Indicator (CLI), four economic phases — expansion, slowdown, contraction, and recovery — are identified and used to analyze asset class performance and inform the construction of dynamic portfolio weights. The asset classes include equities, investment grade and high yield bonds, commodities, gold, bitcoin, and leveraged loans. The analysis reveals that asset class performance varies significantly across the business cycle. For example, equities perform best during expansion, while investment grade bonds dominate in contraction. High yield bonds demonstrate resilience across both business cycles and interest rate environments, appearing particularly attractive during recovery. Non-interest-bearing assets such as gold and commodities perform relatively better in low-interest rate environments. However, the observed patterns may also be influenced by structural market changes rather than interest rates alone. Using CLI-based views within the Black-Litterman model, two dynamic portfolios are constructed for the 1990–2024 period and compared against a static benchmark. One portfolio is constrained to “long only” allocations, while the other is subject to stricter interval-based weight constraints. To assess the robustness of the models, the dataset is divided into in-sample and out-of-sample periods. The results indicate that both dynamic strategies outperform the static benchmark across all performance metrics in both the in-sample and out-of-sample periods. The portfolio’s extension from 2014 to 2024 to include bitcoin and leveraged loans particularly highlights challenges related to overfitting and unstable Black-Litterman weights, especially during contraction, where the bitcoin allocation results in negative returns. Methodological limitations of the thesis include the use of revised CLI data, the omission of transaction costs, and simplistic estimation methods for investor views. It is concluded that while the CLI indicator, in combination with the Black-Litterman model, succeeds in outperforming the static benchmark across all performance measures in the primary analysis period (1990-2024), the practical applicability of the strategy is constrained by several uncertainties.