Appendix 2. One-year revisions to initial estimates
1. Graphs in this appendix illustrate, over the period January 1986 to December 1993, the direction (bias) and magnitude (dispersion) of revisions made to initial monthly and quarterly estimates of balance of payments aggregates after one year. For annual estimates, the series under analysis generally commence in 1981-82.
2. Each graph plots, for an aggregate, the difference between the initial estimate for a period and the estimate for that same period 1 year later (that is, after 12 monthly revision cycles, 4 quarterly cycles or one annual cycle). Graphs are included for monthly, quarterly and annual statistics.
3. Revisions to gross series in this appendix (that is, for the components of the current account) can be readily interpreted as magnitude changes. Movements above the origin mean an estimate of greater magnitude - a credit series is a larger credit or a debit series is a larger debit - and movements below the origin mean an estimate of smaller magnitude. For example, a movement above the origin in Graph B.1 (merchandise exports) indicates that exports have been revised upward (a larger credit) while movements above the origin in Graph B.2 (merchandise imports) indicates that imports have been revised upward (a larger debit). Movements below the line for either series would mean an estimate of smaller magnitude. Similarly for all net series other than balances (that is, all the capital account component series) movements above the origin mean increased credits or increased debits. For the balances on current and capital accounts, movements are not unambiguously movements in magnitude. Movements above the origin mean increased surpluses or decreased deficits, and movements below the origin mean smaller surpluses or larger deficits.
4. Care should be used in interpreting the graphs presented in this appendix because not all revisions are in place after one year. As shown in Appendix 1, for many series significant revisions remain to be processed several years after the initial estimates have been made before final (or latest) estimates are reached.
5. It should be noted in interpreting these graphs that not all graphs use the same scale. Quarterly and annual revisions can be multiples of the monthly revisions experience, requiring wider scales for effective presentation. Also, in the case of balances (the balances on current and capital accounts), the aggregated revisions estimates may be augmentations of those shown for components (if the net revisions do not offset), requiring a scale that has a larger range of values either above or below the origin to capture the direction of the cumulative effect of these revisions. For the net series shown for the capital account, much wider scales are used to capture the magnitude of the swings experienced in revising estimates of the net effect of the very large gross flows involved.
6. Also, because revisions in this appendix are expressed in millions of dollars at current prices, no account is taken of inflationary effects or any change in the size of the aggregate being amended.
7. Graph B.6 provides an example of the interpretation of bias in revisions shown in the graphs in this appendix. The preponderance of observations above the origin up until June 1990 indicates that revisions after one year generally increased the size of the debit being measured (a positive bias in revisions, or an understatement in initial estimates), followed by a few years of negative bias (up until December quarter 1992) with revisions reducing the debits measures and correcting overstatements in the initial magnitudes being measured. From March quarter 1993, positive bias in revisions re-emerges. This revisions experience largely reflects a fall in the size of reinvested earnings of direct investment enterprises in Australia (due to falling profits) on a scale greater than initially forecast and corrected only when annual survey data became available.
8. Graph B.1 for merchandise exports provides an example of the interpretation of the dispersion of revisions shown in this appendix. Up until September quarter 1989, the magnitude of revisions to merchandise exports is somewhat muted, reflecting the stable nature of the administrative by-product source used in the estimation procedure. From that point on, the dispersion increases with swings above and below the origin largely reflecting increased estimation variability with a new administrative data source.
Graphs B.1 to B.52
The attached file contains a printable copy of pages 89 to 102 of the paper publication containing graphs B.1 to B.52.