ABSTRACT
In this study, we looked for a strategy that would rotate between an equity-based portfolio and a bond-based portfolio given the properties of hidden market regimes. We used end-of-the-day price data for the SPY (S&P 500) and AGG (U.S. Bond) ETFs and MarketPsych sentiment data as features. The evaluated period was from January 1998 to September 2021. In-sample results (1998–2005) indicated that states with a negative SPY expected return are, on average, only associated with a realised negative 1-day forward return when the contemporaneous sentiment is also negative. We translated this finding in the validation period (2006–2015) with a strategy that rotates from 100% long SPY into 100% long AGG when in such state. The out-performance of the strategy (vs the SPY buy& hold) in the validation period was confirmed in the out-of-sample period (2016–2021). The back-tested portfolio yielded a theoretical Sharpe ratio of 1:3 (versus a 0.9 of the long-only SPY).
Keywords: SPY, AGG, sentiment data, hidden regimes, GMM
