RDP 2005-07: The Australian Business Cycle: A Coincident Indicator Approach 1. Introduction

This paper constructs coincident indicators of Australian economic activity and uses them to explore several features of the business cycle. These coincident indicators extract the common component from a large number of series using techniques recently developed by Stock and Watson (1999, 2002a, 2002b) and Forni et al (2000, 2001). These techniques have been used to construct coincident indices for the US (the Chicago Fed's CFNAI index) and Europe (the EuroCOIN index published by the CEPR).

There is a long-standing debate in the academic literature, dating from the seminal work of Burns and Mitchell (1946), as to whether the business cycle should be measured using GDP or some average of individual economic series. While GDP by definition measures the total output of the economy, there are several arguments as to why coincident indicators may be a useful alternative measure of the state of the economy. GDP, like other economic series, is estimated with noise. An index that uses statistical weights to combine a large number of economic series may be able to abstract from some of this noise. Assessing the business cycle based only on aggregate GDP may also obscure important developments relating to different sectors of the economy. For example, estimates of GDP may at times be driven by temporary shocks to one part of the economy, for example short-lived shocks to the farm sector or to public spending, that are not representative of developments in the broader economy. A further advantage of coincident indicators is that they can be constructed with monthly data, and if they are produced on an ongoing basis they may be more timely than GDP because many economic series are published with a shorter lag than GDP. Coincident indicators could potentially be less prone to the revisions experienced by GDP, in part because they can be constructed from series that either are not revised or are subject to smaller revisions.

Both the Stock and Watson (hereafter SW) and Forni, Hallin, Lippi and Reichlin (FHLR) techniques assume that macroeconomic variables – or more specifically, growth rates in most macroeconomic variables – can be expressed as linear combinations of a small number of latent ‘factors’. The SW and FHLR techniques use large panels of individual data series to estimate these unobserved factors, which are common to the variables in the panel. These factors can be used to produce coincident indices of the common economic cycle in the variables (Altissimo et al 2001; Federal Reserve Bank of Chicago 2000, 2003; Forni et al 2000, 2001; and Inklaar, Jacobs and Romp 2003). They can also be used to forecast macroeconomic variables (for example see Artis, Banerjee and Marcellino 2005; Bernanke and Boivin 2003; Boivin and Ng 2005, forthcoming; Forni et al 2005; and Stock and Watson 1999, 2002a, 2002b) and to identify shocks (for example in a VAR framework by Bernanke, Boivin and Eliasz 2005 and Forni and Reichlin 1998).

The remainder of this paper proceeds as follows. Section 2 discusses coincident indices and the intuition of factor models. Section 3 more formally explains the SW and FHLR techniques. Section 4 briefly discusses the panel of data we use. The estimated quarterly and monthly coincident indices are presented in Sections 5.1 and 5.2. In Section 6 these coincident indices are used to investigate the changing volatility and structure of the Australian business cycle, the length of economic expansions and contractions, and its correlation with the US business cycle. We conclude in Section 7.