The Treasury

Global Navigation

Personal tools

4  Conclusion

This study extends Treasury’s current forecast accuracy work by making direct comparisons of Treasury’s forecast performance against other forecasters. When compared across all evaluation periods (calendar years 1996-2005), Treasury’s GDP forecast performance was ranked in the middle at seventh place out of 16 on the average relative rank basis. Treasury forecasters misjudged the impact of the Asian financial crisis and droughts on economic activity in 1998, resulting in large forecast errors at that time which had a material impact on Treasury’s overall forecast performance. This is reflected in a relatively large RMSE compared to other forecasters. When compared only against other individual forecasters across all evaluation periods (ie, excluding Consensus, Median and Mean), Treasury’s GDP forecast performance is in the top half, and better than other major forecasting groups. However, Treasury’s CPI forecast performance ranks in the bottom half, out-performing only two other forecasting groups. Large forecast errors for CPI relating to the 1998-2000 period were a large contributor to the poor forecast performance.

Treasury’s forecast performance was better when evaluating only the current year Budget forecasts, placing fourth for GDP and sixth for CPI. This suggests that Treasury is better at forecasting the current year than the year ahead. Compared to other individual forecasters, Treasury’s Budget current year GDP forecast was bettered only by one other forecaster.

Consistent with international studies, no single forecaster consistently outperforms the Consensus, and all forecasters find it difficult to pick recessions and turning points. Large forecasting groups, on average, tend to perform poorly at forecasting GDP compared to private sector forecasters on the average relative rank basis. This finding is similar to Blix et al (2001). But their performance is comparable on the RMSE ranking, suggesting that forecasting groups tend to be more conservative in their forecasts, which may not result in a closer forecast to the actual, but lessens the chance of a large forecast error. Another possible explanation is that major forecasting groups tend to produce comprehensive forecasts at certain times during the year, which do not coincide with the dates chosen for this study. Their responses to Consensus Economics could therefore be quick updates which reduce their forecast performance. It is difficult to quantify and resolve this timing issue which may affect forecasting performance as measured in this comparison.

Page top