Value stocks tend to have higher returns on average. Their performance is particularly stronger when the value spread, defined by differences in B/M ratios, between value and growth stocks is wider. In this paper, we show that this predictability becomes even stronger when we account for the spread in growth, measured by short-term expectations, long-term expectations, and past growth. We use analyst expectations on individual firm s earnings to construct a range of proxies for earnings growth expectations. We find that adding the growth spread greatly increases the predictive power also in out-of-sample tests.