The FED model: Is it still with us?

• Reconsider the existence of a stable relation between equity and bond yields. • Evidence supports level-shifts (breaks) in the relation. • Accounting for shifts improves point and directional returns prediction. • Shifts arise due to changes in monetary policy and the nominal-real relation. The Fed model, a stable relation between equity and bond yields, with accompanying predictive power for subsequent stocks returns has proved a controversial idea within empirical finance. This paper re-examines the equity-bond yield relation and its predictive power in the light of potential breaks or level-shifts. We argue that while a positive relation exists between the two yields it does not fluctuate around a single stable point. Notably, we demonstrate switching behaviour between high and low values of the Fed series, which correspond inversely to movements in the real interest rate and is linked to observed changes in the inflation-growth relation. Accounting for such shifts leads to an improvement in relative predictive power over a set of baseline models, including a linear, non-regime varying, approach. This occurs both in terms of point prediction and directional accuracy, including crash prediction. We also provide evidence in favour of bond return predictability, although predominantly at longer-horizons. The results here reveal that the equity and bond yield interaction is informative for investors, but only when accounting for shifts in behaviour.