RDP 9102: Indicators of Economic Activity: A Review 5. Summary and Conclusions

The paper has reviewed the performance of a selection of indicators judged to be in common use in analysing the business cycle in Australia. The approach taken does not in any sense attempt to construct optimal forecasting rules using these indicators, but has the more limited aim of assessing which indicators contain information that is useful for short-term forecasting. This question was addressed in two stages, looking first at relationships between major expenditure aggregates, and secondly at the information contained in various partial indicators.

On the first issue, results were found to be very sensitive to the way the tests were set up. The most general VAR specifications showed little evidence for any non-contemporaneous relationships between the variables included. More restricted models did however suggest some significant leading and lagging relationships. For example, the housing sector appeared to lead GDP, while consumption, construction activity, and perhaps imports, lagged. These latter results also found some support in less sophisticated methods such as correlation analysis and visual inspection of turning points in the data, but the overall impression left by the evidence is that these relationships are somewhat unreliable, in the sense that the timing can vary from cycle to cycle.

On the second issue, the statistical results were much clearer. Significant forecasting power was found for partial indicators in all of the four areas studied. Of the indicators considered, local government building approvals and the ANZ job vacancies series appeared to be the most significant in providing forward information about income and spending aggregates.