Modeling the persistence in expected returns

The major contribution of this paper is to explicitly model the persistence in the time series of expected returns. The series of expected returns is derived from a state space representation of the net present value identity relating expected returns and expected dividend (earnings) growth to the observed price dividend (earnings) ratio. The state space model is adjusted in order to include the possibility of expected returns following an autoregressive fractional integrated (ARFIMA) process which captures the persistence of the process. The new ARFIMA model performs moderately compared to the simple autoregressive process, which may be due to the presence of different regimes and structural breaks. The expected returns series is applied in three instances namely in predictive regressions, analysing the relationship between consumption and discount rates and also in a market timing strategy.