6 Concluding remarks
In this paper, we constructed a composite index of leading indicators of New Zealand employment that was then used in an indicator model to forecast quarterly changes in employment. The indicator model is able to explain about 67 percent of the quarterly variation in employment, in sample, and correctly predicted the direction of next period employment almost 80 percent of the time, out of sample. The composite index of leading indicators thus appears to be a good predictor of employment changes.
A VAR model including changes in the composite index and employment produces results only slightly worse than the indicator model. Thus, to produce out-of-sample forecast for more than one quarter ahead, the indicator model, which can be interpreted as a restricted VAR, could be re-estimated as a VAR (using seemingly unrelated regression estimation). The estimation of the indicator model as a restricted VAR is left for future work.
