Forecasting US interest rates and business cycle with a nonlinear regime switching VAR model

• DOI: http://doi.org/10.1002/for.2458
• Published date: 01 January 2018
• Date: 01 January 2018

Received: 27 November 2015 Revised: 20 July 2016 Accepted: 1 January 2017

DOI: 10.1002/for.2458

RESEARCH ARTICLE

Forecasting US interest rates and business cycle with a nonlinear

regime switching VAR model

Henri Nyberg

Department of Mathematics and Statistics,

University of Turku;and Department of Political

and Economic Studies, University of Helsinki,

Finland

Correspondence

Henri Nyberg, Department of Mathematics and

Statistics, University of Turku,20014 University of

Turku, Turku, Finland.

Email: henri.nyberg@utu.fi

Abstract

This paper introduces a regime switching vector autoregressive model with

time-varying regime probabilities, where the regime switching dynamics is

described by an observable binary response variable predicted simultaneously

with the variables subject to regime changes. Dependence on the observed binary

variable distinguishes the model from various previously proposed multivariate

regime switching models, facilitating a handy simulation-based multistep forecast-

ing method. An empirical application shows a strong bidirectional predictive linkage

between US interest rates and NBER business cycle recession and expansion peri-

ods. Due to the predictability of the business cycle regimes, the proposed model

yields superior out-of-sample forecasts of the US short-term interest rate and the

term spread compared with the linear and nonlinear vector autoregressive (VAR)

models, including the Markov switching VAR model.

KEYWORDS

financial markets, nonlinear vector autoregression, probit model, simulation, turning

points

1INTRODUCTION

Nonlinear econometric modeling has been based heavily on

regime switching mechanisms allowing for parameter coef-

ficients to switch between different states of the world (e.g.,

business cycle recessions and expansions, bear and bull stock

markets, monetary policy regimes, and also some rare events

such as financial crises). The previous literature on multi-

variate models has adopted several differentregime switching

specifications, including Sola and Driffill (1994), Krolzig

(1997), Ang and Bekaert (2002a), (2002b), Guidolin and

Timmermann (2006), Dueker, Psaradakis, Sola, and Spag-

nolo (2011), and Henkel, Martin, and Nardari (2011), among

others. In this literature, the regime switching mechanism

is typically specified as a latent (unobserved) process with

underlying regime probabilities that may be functions of the

lagged endogenous or exogenous variables determining the

economic forces driving the regime switches. However, in

line with nonlinear models in general, the out-of-sample fore-

casting performances of these models have often been found

disappointing; see, for example, the discussion in Dacco and

Satchell (1999) and Clements, Franses, and Swanson (2004).

In this study, we consider a regime switchingvector autore-

gressive (VAR) model, where the regime is determined by an

observed qualitative response (QR) variable predicted simul-

taneously with the variables subject to regime switches, and

hence permitting the method to be implemented in real-time

forecasting. For simplicity, the joint model is referred to

the QR-VAR model. The use of the qualitative response

model yields time-varying regime probabilities between the

observed regimes, making the QR-VAR model much easier

to work with and, in particular, construct forecasts than the

multivariate regime switching models with latent regimes.

Following the large majority of previous studies, we restrict

ourselves to the two-regime case; that is, the qualitative vari-

able is binary throughout this paper. In our empirical appli-

cation, the binary variable is the state of the US business

cycle measured in terms of the official NBER business cycle

turning points. A multinomial case (i.e., multiple regimes) is

a straightforward extension to our model, provided that the

Journal of Forecasting.2018;37:1–15. wileyonlinelibrary.com/journal/for Copyright © 2017 John Wiley & Sons, Ltd. 1