The cyclical dynamics and volatility of Australian output and employment

In this paper we examine the volatility of aggregate output and employment in Australia with the aid of a frequency filtering method (the Butterworth filter) that allows each time series to be decomposed into trend, cycle and noise components. This analysis is compared with more traditional methods based simply on the examination of first differences in the logs of the raw data using cointegration-VAR modelling. We show that the application of univariate AR and bivariate VECM methods to the data results in a detrended series which is dominated by noise rather than cyclical variation and gives break points which are not robust to alternative decomposition methods. Also, our conclusions challenge accepted wisdom in relation to output volatility in Australia which holds that there was a once and for all sustained reduction in output volatility in or around 1984. We do not find any convincing evidence for a sustained reduction in the cyclical volatility of the GDP (or employment) series at that time, but we do find evidence of a sustained reduction in the cyclical volatility of the GDP (and employment) series in 1993/4. We also find that there is a clear association between output volatility and employment volatility. We discuss the key features of the business cycle we have identified as well as some of the policy implications of our results.