The state-dependent time variation in the value premium

It is well documented that asset and strategy returns are generally exposed to identifiable risk factors. Moreover, the exposure to these systematic risk factors tends to be time varying. Examples of strategies that exhibit such regime-dependent variability include value investing in equity markets and carry in foreign exchange markets. The literature proposes several macro-based explanations for this time variation such as liquidity or cyclicality risk depending on the strategy and horizon under consideration. This project relies on developments in the academic literature in the area of state-dependent asset pricing models (see, for example, survey by van Dijk et al, 2002). Under such a framework, a strategy return is modelled using a multi-factor asset pricing equation where the risk sensitivities (or betas) to the pre-defined factors are allowed to vary according to one or multiple state variables. The approach considered here uses a logistic smooth transition regression methodology applied to the value strategy, where the state variables have an economic meaning (see Christiansen et al, 2011 application to foreign exchange carry strategy). By distinguishing between low and high state regimes (in a continuous space), the aim is to first investigate improvements in the explanatory power of the model relative to the linear (no-state dependency) approach. In particular, we study the performance of the value factor during turbulent periods, and identify economic variables driving its time-varying behaviour. We also present an algorithm that allocates dynamically to a long–short value overlay portfolio conditioned on regime transitions. We show that the performance of this dynamic portfolio is superior to a portfolio with a static value overlay.