RDP 1999-11: A Structural Vector Autoregression Model of Monetary Policy in Australia 1. Introduction

This paper examines monetary policy in Australia using structural or identified VAR models. Since the basic principles were put forward by Sims (1980), these models have been used extensively within both closed and open economy environments to model the monetary transmission mechanism.[1] Despite their widespread use elsewhere, it is only recently that these types of models have been used in the Australian context, the most notable example the study by Dungey and Pagan (1998).[2] Our work, which differs in emphasis from Dungey and Pagan, makes a number of contributions. Foremost, it furthers our understanding of monetary policy and the monetary transmission mechanism in Australia. In addition, our results provide useful information more generally within the context of modelling monetary policy in small open economies.

We use as a benchmark the small dimension SVAR model put forward by Kim and Roubini (1999), hereafter KR. These authors use a common structural model to identify the effects of monetary policy for the G7 economies less the United States (which others and we refer to as the G6). Importantly, their model satisfies a set of accepted conditions that underlie most SVAR analysis of monetary policy. For the countries they consider, although with differing degrees of confidence, they find no evidence of the puzzles that have been identified in the SVAR literature: the price, exchange rate, liquidity and forward discount bias puzzles. Resolving the first two of these puzzles is generally accepted as the minimum required to identify monetary policy in these types of models.[3]

Although the KR model is not specifically designed for the Australian economy and is not based explicitly on an accepted macroeconomic model, there are a number of reasons why it might be regarded as a natural starting point. First, the KR study considers a quite disparate set of economies. The fact that their model does reasonably well across these countries suggests that it is robust, at least in some directions, and merits consideration for this fact alone. An additional benefit is that we can systematically compare the results we obtain for Australia with those of the KR study. This allows us to gauge our results against other countries. In addition, by expanding the set of countries considered by KR, we provide further evidence on the suitability of the KR model for modelling monetary policy in an open economy environment.

To develop this final point further, the KR structure, although small in dimension, captures features of a larger dimensioned and more complex open economy model estimated for Canada by Cushman and Zha (1997). Most importantly, both studies argue that a structure allowing for interdependence between domestic monetary policy and the nominal exchange rate is critical for modelling monetary policy in small open economies. Examining these types of models for Australia provides further evidence on this specific issue. To foreshadow our results, we find reasonable support for the KR model, subject to some minor changes, and we find that modelling the interdependence between monetary policy and the domestic exchange rate is critical for a small dimensioned SVAR of the Australian economy.[4]

The other SVAR study for Australia is Dungey and Pagan (1998) and it is useful to relate our work here with their study. In some respects, the purpose and nature of the two studies are similar. We examine a similar time period and similar macroeconomic aggregates and there is common ground in some of the questions we ask of the models. Most importantly, both studies consider the effects of monetary policy on the macroeconomy. There are, however, some important differences between the two studies. First, our focus is much more narrow – we wish to examine how well an existing SVAR model is able to identify monetary policy in Australia and the importance of certain features of SVAR models for Australia. In contrast, Dungey and Pagan consider a much wider range of empirical issues, for example the effects of terms of trade shocks on the Australian economy, with somewhat less focus on the issue of model specification and its relationship to the existing literature.

Second, we pursue quite different modelling strategies in a number of important directions. Dungey and Pagan use a large dimension model (10 or 11 variables) with quite strong restrictions on the interdependence between domestic and foreign variables. In addition, they estimate their model on linearly detrended data. In contrast, the model we consider is small (7 variables), it relies on the interdependence between the domestic and foreign interest rates and the nominal exchange rate, and it is estimated in levels. In addition to having the KR study as a benchmark, this strategy puts us very close to many recent US studies of monetary policy, for example Gordon and Leeper (1994) and Sims and Zha (1998a). As discussed above, we pursue this modelling strategy because we wish to examine and document how well simple SVAR models identify monetary policy in Australia. Because our objectives differ from Dungey and Pagan, and these objectives largely influence the model we consider, we view the models of the two studies as complementary rather than competing alternatives.

The paper proceeds as follows. In the next section, we first briefly review the SVAR framework and discuss in general terms how these models are specified. We then present the KR model and discuss the specific motivations for the specification of the model and the restrictions imposed. In the third section, we present the estimation results for the KR model; as well, we present a variation of this model and discuss why this variation is necessary. We then examine the quality and stability of the results from the model and highlight what we believe to be important features of the structural model. The final section concludes.


Open economy studies include Sims (1992), Eichenbaum and Evans (1995), Cushman and Zha (1997), and Kim and Roubini (1999). Closed economy studies include Sims (1986), Gali (1992), Gordon and Leeper (1994), Christiano, Eichenbaum, and Evans (1996), Bernanke and Mihov (1998), and Sims and Zha (1998a). [1]

These authors provide a survey of related work for Australia. Huh (1999) is another example of a SVAR model for Australia; however, its purpose is primarily to examine the specific predictions of the Mundell-Fleming IS-LM model. [2]

Leeper, Sims and Zha (1996) provides a detailed discussion of SVAR methods, including the resolution of these puzzles, and their relationship with other macroeconometric modelling. [3]

The interdependence refers to a simultaneous relationship between a monetary aggregate, a policy interest rate and the nominal exchange rate. As we demonstrate below, however, what matters is the relationship between the interest rate and the exchange rate. Consistent with previous research for Australia, for example de Brouwer, Ng and Subbaraman (1993), there is no important role for a monetary aggregate in the model we estimate. Despite this, we proceed with the KR model including a monetary aggregate. This allows us to clearly document what is important for model specification. Nevertheless, the predictions of the model are little changed when the monetary aggregate is excluded. [4]