RDP 2015-07: A Multi-sector Model of the Australian Economy 1. Introduction

Dynamic stochastic general equilibrium (DSGE) models are one of the most widely used tools for empirical and theoretical research in macroeconomics. They represent a general class of macroeconomic models that emphasise households' and firms' intertemporal decisions in a general equilibrium setting. In addition to their popularity in the academic literature, DSGE models are also widely used by central banks for quantitative policy analysis and forecasting.

This paper presents a DSGE model of the Australian economy developed by staff at the Reserve Bank of Australia (RBA). The paper outlines the core features of the model, its estimation and its mechanisms. We then show how the model can be used for scenario analysis in the policy process and also use the model to decompose the Australian business cycle in terms of the economy's underlying driving forces.

The model is part of a set of macroeconomic models maintained by the Economic Research Department at the RBA. These models complement, but do not substitute for, the more detailed sectoral analysis and judgement-based projections. While it is often the case that many identified inconsistencies between model-based and judgemental analysis can be justified with information from outside the model, the process of reconciling the two can be informative.

Within the set of internal models, the DSGE model is most often used for scenario analysis rather than forecasting. As discussed by Smets and Wouters (2003), DSGE models are particularly suited to counterfactual scenario analysis because they provide a coherent theoretical framework with which to understand the potential impact of a variety of policy actions or the realisation of various risks. In this regard, their advantage over simpler reduced-form time series models, such as vector autoregressions (VAR), is that they make the economic mechanisms at work within the model transparent and account for forward-looking behaviour.

The DSGE model described in this paper is designed to be detailed enough to answer questions of particular relevance to a small open commodity-exporting economy like Australia, while still being simple enough to make the economic mechanisms at work within the model transparent and straightforward to communicate. We have therefore chosen to keep the number of real and nominal frictions to a minimum. An additional advantage to this approach is that it can be readily augmented to include additional features required to answer specific questions in a targeted manner. The expectation is that this model can be used as a baseline, to which additional features – such as a housing or financial sector – can be added or subtracted, depending on the specific policy question at hand.

Overall, this emphasis on simplicity represents something of a departure from the approach adopted in the development of several other central bank models, which have tended to emphasise data coherence, particularly with respect to dynamics and forecasting.[1] However, as discussed by Edge and Gürkaynak (2010), forecasting ability is not always a good criterion for judging a model's success. For example, if monetary policy responds aggressively to deviations of inflation and output from their target levels, the predictability of macroeconomic outcomes should be low (Benati and Surico 2008). In such an environment, all models are likely to produce poor unconditional forecasts. But a DSGE model may still be able to provide plausible counterfactual scenarios that describe how the economy will behave conditional on the outside influences affecting it.

This paper contributes to the existing literature that has estimated DSGE models of the Australian economy. Early work in this area includes Buncic and Melecky (2008) and Nimark (2009), who estimated small-scale models largely for the purpose of examining the dynamic effects of monetary policy shocks. More recently, Jääskelä and Nimark (2011) constructed a medium-scale DSGE model that, along with other models, was used for forecasting and scenario analysis at the RBA. Relative to the model in this paper, Jääskelä and Nimark incorporated a larger number of shocks and frictions but featured a less detailed production structure. For instance, Jääskelä and Nimark did not include a non-tradeable goods sector. Moreover, the objectives of the two models are different. Whereas Jääskelä and Nimark (2011) was developed as a tool for forecasting, the model in this paper is intended primarily for use in scenario analysis.

The rest of the paper is organised as follows. The next section describes the main features of the model. Section 3 outlines the estimation strategy and Section 4 explores the dynamics of the estimated model. Readers who wish to avoid the technical details of the model may safely skip these sections and instead focus on the remainder of the paper, which shows the model in action. Section 5 provides an example of how the model can be used to construct a scenario involving an extended period of lower resource prices. Section 6 uses the model to uncover the sources of Australian business cycle fluctuations over recent decades. Section 7 concludes.


The relative forecasting performance of DSGE models compared to less structural modelling approaches remains an open question. Del Negro and Schorfheide (2013) survey the existing literature and conclude that DSGE model forecasts are comparable to standard AR or VAR models, but can be dominated by more sophisticated univariate or multivariate time series models. However, Gürkaynak, Kisacikoğlu and Rossi (2013) compare the real-time forecasting accuracy of the Smets and Wouters DSGE model against several reduced-form time series models and conclude that none of the forecasting models unambiguously outperforms any other over all horizons and samples. [1]