Investigating the trend characteristics of the Australian business cycle

Publisher Website
Google Scholar
Abstract
The non-stationarity tests of Phillips-Perron (1988) (PP tests) suggest that Australian macroeconomic output possesses a stochastic rather than a deterministic log-linear trend. Kwiatkowski et al. (1991) argue that such tests have low power and propose the KPSS test, in which the null is stationary. However, the KPSS test results reinforce the PP findings for Australia. Cochrane (1988) variance ratio (VR) tests further suggest that there may be a very strong random-walk component in the Australian business cycle. Rappoport and Reichlin (1989), however, argue that all such tests are biased in favour of the stochastic trend alternative if there are trend breaks in the data. Following up on this point, the paper finds that, in Australia's case, the stochastic trend alternative is statistically dominated when the data are allowed to be characterized by a probabilistic, regime-switching, segmented trend specification. Therefore, to the extent that real business-cycle theories of the business cycle gain support from stochastic trends in real output, this latter piece of Australian evidence does not strengthen their case.