Data sources
Budget forecasts used to construct historical forecasting errors have been sourced from Budget Papers from the 1998–99 Budget through to the 2013–14 Budget.
Actual fiscal outcomes used to compare against Budget estimates have been sourced from Final Budget Outcome documents spanning 1998–99 to the latest published outcome. Fiscal variables relative to GDP are calculated using original (non-seasonally adjusted) series with adjustments to allow for revisions in data concepts (see footnote 12). Note that for all the data and analysis in this paper, receipts exclude GST and include Future Fund earnings, payments exclude GST and the underlying cash balance excludes Future Fund earnings. Hence receipts less payments measured in this way will not exactly equal the underlying cash balance.
For nominal and real GDP, the forecasts are compared to seasonally adjusted financial year average outcomes of the following series:
Australian Bureau of Statistics, 2013. Table 1—Key National Accounts: Seasonally Adjusted Gross Domestic Product: Chain Volume Measure. National Accounts: National Income, Expenditure and Product, June Quarter 2013. ABS Catalogue 5206.0. Australian Bureau of Statistics: Canberra. 4/9/2013.
Australian Bureau of Statistics, 2013. Table 1—Key National Accounts: Seasonally Adjusted Gross Domestic Product: Current prices. National Accounts: National Income, Expenditure and Product, June Quarter 2013. ABS Catalogue 5206.0. Australian Bureau of Statistics: Canberra. 4/9/2013.
Table A1 summarises some features of the forecast errors which were described in the main text.
Table A1: Summary statistics of forecast errors
| Forecast error | Bias (forecast - actual) |
Serial correlation | Normality |
|---|---|---|---|
| Real GDP (average annual growth rate in percent) | |||
| Year about to end | -0.4** | -0.1 | 1.2 |
| Budget year | -0.3 | -0.3 | 0.4 |
| Subsequent year | -0.1 | -0.3 | 1.3 |
| Nominal GDP (average annual growth rate in percent) | |||
| Year about to end | -0.6** | 0.4 | 0.1 |
| Budget year | -0.7* | -0.1 | 0.1 |
| Subsequent year | -1.0* | -0.1 | 0.2 |
| Receipts (as share of GDP in percent) | |||
| Year about to end | 0.1 | -0.2 | 1.1 |
| Budget year | 0.1 | 0.5* | 1.1 |
| Subsequent year | -0.2 | 0.7** | 0.7 |
| Payments (as share of GDP in percent) | |||
| Year about to end | 0.4** | 0.3 | 0.4 |
| Budget year | 0.5* | 0.2 | 0.4 |
| Subsequent year | 0.8* | 0.1 | 1.1 |
| Underlying cash balance (as share of GDP in percent) | |||
| Year about to end | -0.3** | 0.1 | 0.5 |
| Budget year | -0.5 | 0.3 | 1.7 |
| Subsequent year | -0.9 | 0.4 | 1.5 |
| Receipts (no-GDP-error approach — level error as a percentage of GDP) | |||
| Year about to end | 0.0 | -0.2 | 1.4 |
| Budget year | -0.2 | 0.4 | 0.9 |
| Subsequent year | -0.7 | 0.4 | 2.0 |
| Payments (no-GDP-error approach — level error as a percentage of GDP) | |||
| Year about to end | 0.2** | 0.3 | 3.5 |
| Budget year | 0.2* | 0.4 | 1.3 |
| Subsequent year | 0.1 | 0.5 | 1.0 |
| Underlying cash balance (no-GDP-error approach — level error as a percentage of GDP) | |||
| Year about to end | -0.3** | 0.1 | 0.5 |
| Budget year | -0.4 | 0.4 | 1.6 |
| Subsequent year | -0.8 | 0.4 | 1.4 |
Note: Bias is the average forecast error (expressed as forecast less actual so a positive number indicates overprediction), where significance is tested using White heteroskedasticity-consistent standard errors. Serial correlation is the first order autocorrelation of the mean-corrected forecast errors with the statistical significance based on the Ljung-Box Q-statistic for first order autocorrelation. For normality, the Jarque-Bera statistic (which has a chi-squared distribution with 2 degrees of freedom under normally distributed errors) is reported for the forecast errors. The sample is based on forecasts from the 1998-99 Budget onwards (see footnote 11). * indicates statistical significance at the 5 per cent level and ** at the 1 per cent level.
Source: Treasury.