Global value and momentum: a balancing act

Abstract

The global financial landscape is rapidly transforming due to technological advances, shifting geopolitical tensions, and macroeconomic disruptions. In this dynamically evolving environment, traditional asset pricing models – long tasked with explaining the cross-section of returns – are increasingly being challenged, no longer adequately being served by frameworks that isolate individual risk factors or focus on narrow market segments. Asness et al. (2013) addressed these limitations, proposing a two-factor framework combining value and momentum – two of the most empirically validated characteristics in financial economics – highlighting a strong comovement structure shared across global markets, with funding liquidity risk emerging as a key explanatory factor. However, their analysis ends in July 2011, preceding a period marked by substantial liquidity shocks – from oil price collapses and expansive quantitative easing to pandemic-related disruptions, rapid rate policy shifts, and heightened geopolitical instability. In light of these shifts, there is a clear need to reassess the robustness of the two-factor framework. As liquidity conditions and risk transmission mechanisms evolve, so too must our understanding of the shared drivers of asset returns.

This thesis makes three key contributions to the asset pricing literature. First, it revisits and extends the influential Asness et al. (2013) two-factor framework using a larger, more comprehensive dataset covering global markets and asset classes through to 2024. The extended empirical investigation provides new evidence on the persistence and robustness of the comovement between value and momentum, even amid the liquidity shocks and regime shifts of the post-2011 period. Notably, the results suggest that market liquidity, rather than funding liquidity, now plays a central role in linking value and momentum. Additionally, the explanatory power of the three-factor global asset pricing model has shifted in comparison with the other models considered, indicating changes in the underlying risk drivers.

Second, this research tests for similar comovement structure and latent risk factors within arguably the most prominent novel asset class of the post-2011 era –cryptocurrencies. Applying the same value-momentum framework, our results indicate that while value, as defined by Asness’ uniform measures, persists in cryptocurrencies, traditional cross-sectional momentum does not hold. However, when smoothing the underlying return data with a 200-day moving average, we observe strong momentum but no value in cryptocurrencies, likely due to the masking effect of extreme short-term volatility and microstructure frictions. Additionally, we observe significant exposure of cryptocurrency momentum strategies to funding liquidity shocks from conventional financial markets, pointing to emerging cross-market linkages through liquidity channels. This contribution not only extends the Asness et al. (2013) methodology to a new domain but also tests its universality in an asset class defined by decentralization, volatility, and innovation.

Third, this research explores how machine learning – particularly neural network modelling – can enhance asset allocation strategies based on the identified common risk structure. While the original framework used equal weighting for global value and momentum exposures in a combined portfolio, this research relaxes that assumption and proposes a time-varying allocation informed by changing risk realizations. This research develops a flexible portfolio construction framework that demonstrably enhances the capture of global risk premia. By allowing a neural network to dynamically rebalance between global value and global momentum, we show that the model reliably identifies predictable structure in factor returns and delivers substantial improvements in both total and risk-adjusted performance.

Overall, the findings of this research offer practical value to a wide spectrum of stakeholders in financial markets. For institutional investors and asset managers, the insights into the persistence and adaptability of value and momentum factors –particularly under shifting liquidity regimes – can support more robust and informed portfolio construction across both traditional and digital assets. Academics and researchers will benefit from the extended empirical validation of a unified asset pricing framework and its applicability to emerging markets like cryptocurrencies, as well as from the integration of advanced machine learning techniques in factor allocation. Regulators and policy-makers will also find the results relevant, particularly in understanding the systemic implications of liquidity-driven risks that span asset classes and market structures. Ultimately, by bridging theory with data-rich, real-world applications, this research contributes to a more adaptive, cross-asset understanding of return drivers in an increasingly complex and interconnected financial system.

Date of Award	2026
Original language	English
Supervisor	Adrian Gepp (Supervisor), Christopher Bilson (Supervisor), Geoffrey Harris (Supervisor) & Bruce James Vanstone (Supervisor)