Modelling yields at the lower bound through regime shifts

Focus

Following the 2008-09 crisis, a number of central banks cut monetary policy rates to levels close to the effective lower bound. We propose a regime-switching approach to deal with the lower bound in dynamic term structure modelling.

Contribution

Our model allows the drivers (state variables) of yields to evolve according to separate laws of motion in normal times and at the lower bound. The model is able to capture important differences in the way the economy operates when it is at the lower bound, compared with normal times, as well as the resulting implications for bond pricing. Another feature of the model is that the probability of switching from one regime to the other depends on the level of the short-term interest rate. These features of the model allow for a very gradual normalisation of monetary policy following a lower bound episode.

Findings

We apply our model to US data and show that it does well in fitting key properties of yields at the lower bound. Moreover, after interest rates "lift off", the model is able to capture the slow pace of monetary policy normalisation that is evident from the recent US experience. This is because the possibility of returning to the lower bound regime continues to influence the early phases of normalisation, pulling expected future interest rates downwards. Finally, a reliable yield-forecasting performance results from the model's ability to account for different state dynamics in and out of the lower bound period, coupled with its ability to allow for a high probability of switching back to the lower bound regime after having exited it.

Abstract

We propose a regime-switching approach to deal with the lower bound on nominal interest rates in dynamic term structure modelling. In the "lower bound regime", the short term rate is expected to remain constant at levels close to the effective lower bound; in the "normal regime", the short rate interacts with other economic variables in a standard way. State-dependent regime switching probabilities ensure that the likelihood of being in the lower bound regime increases as short rates fall closer to zero. A key advantage of this approach is to capture the gradualism of the monetary policy normalization process following a lower bound episode. The possibility to return to the lower bound regime continues exerting an influence in the early phases of normalization, pulling expected future rates downwards. We apply our model to U.S. data and show that it captures key properties of yields at the lower bound. In spite of its heavier parameterization, the regime-switching model displays a competitive out-of-sample forecasting performance. It can also be used to gauge the risk of a return to the lower bound regime in the future. As of mid-2018, it provides a more benign assessment than alternative measures.

JEL codes: E31, E40, E44, E52, E58, E62, E63.

Keywords: zero lower bound, term premia, term structure of interest rates, monetary policy rate expectations, regime switches