© 2016, © Emerald Group Publishing Limited. Purpose – For over two decades numerous studies have provided evidence on the predictive ability of the yield spread for real economic growth. While all this large literature has focussed on how well the spread helps predict real activity, none of these has given an answer on why the spread predicts. The purpose of this paper is to deal with this issue by trying to find an answer on the reason and the economic conditions under which the spread proves to be so powerful predictor of economic activity. Design/methodology/approach – The authors examine whether the explanation of spread’s predictive ability lies behind interest rate volatility supposing that the economy oscillates between high- and low-volatility regimes. For this reason the authors nest GARCH models into Markov regime switching models. Findings – When the authors assume that the economy simply oscillates between different regimes, interest rate volatility does not explain the spread’s predictive ability. However, the authors obtain a very interesting result when the authors augment the conditional variance with a level effects term. This ensures that in an environment with high levels of interest rates – in which the rational agents expect the economy to slow down – there is a greater possibility for the economy to switch to a high-volatility regime. Under these economic conditions, interest rate volatility appears to be the reason of spread’s predictive power from one up to three years. Originality/value – This study contributes to the relevant literature by providing an explanation on the reason and the economic conditions under which the spread proves to be so powerful predictor of economic activity.