Markov switching causality and the money–output relationship

Abstract

SUMMARYThe causal link between monetary variables and output is one of the most studied issues in macroeconomics. One puzzle from this literature is that the results of causality tests appear to be sensitive with respect to the sample period that one considers. As a way of overcoming this difficulty, we propose a method for analysing Granger causality which is based on a vector autoregressive model with time-varying parameters. We model parameter time-variation so as to reflect changes in Granger causality, and assume that these changes are stochastic and governed by an unobservable Markov chain. When applied to US data, our methodology allows us to reconcile previous puzzling differences in the outcome of conventional tests for money-output causality.

Does money still matter for U.S. output?

“…In fact, we show that money ceases to Granger-cause output around Paul Volcker's chairmanship at the Federal Reserve. This supports arguments made by Friedman and Kuttner (1993) and Psaradakis et al (2005) stressing the importance of sample length and seems to contradict Swanson's (1998) …”

“…2 Both Friedman and Kuttner (1993) and Psaradakis et al (2005) focus on within-sample evidence. The latter use Markov-switching to distinguish between periods with/without Granger-causality of money in a bivariate setting, restricting the underlying models to be the same across time.…”

Testing for Granger (non‐)causality in a time‐varying coefficient VAR model

“…On the other hand, Eichenbaum and Singleton (1986) and Stock and Watson (1989) find that this correlation becomes stronger when data for the 1980s is included in the sample. This structural instability has been addressed in papers such as Psaradakis et al (2005) and Rothman et al (2001). Weise (1999) studies possible nonlinear effects of monetary policies using an LSTR model in which the transition between states depends on inflation, the business cycle or the stance of monetary policy itself.…”

“…However, these models are not designed to detect time-varying Granger-causality and the breaks are assumed to be sharp rather than smooth processes. Models of time varying causality that make use of Markov-Switching models have been developed in Psaradakis et al (2005), Lo and Piger (2005) and Warne (2000). In these models, time-varying causality is determined by a two states unobservable variable governed by a discrete-time, discrete-state Markov stochastic process and hence causality is allowed to change sharply depending on time or the state of the economy.…”

Multivariate contemporaneous-threshold autoregressive models

“…To investigate this issue we adopt a slightly different approach to that of Psaradakis et al (2005) and, instead of inquiring how causality patters change over time, we examine whether the two variables are useful for predicting each other in different economic regimes. Using the C-MSTAR model in Table 6, it can be seen that the off-diagonal elements of A (i) 1 vary significantly across regimes.…”