3.62. As noted in the previous subsection, analysts want to understand the variability they are likely to see in revisions to estimates over the short term. The direction of revisions discussed in that subsection gives a perspective on whether the oneyear revisions have been predominately upward or downward. The magnitude of oneyear revisions discussed in this subsection provides a perspective on how large these revisions have been. Table 9, presents measures of the magnitude of oneyear revisions to initial monthly, quarterly and annual estimates. The magnitude of revisions refers to the spread of revisions above and below the initial estimates without regard to sign and thus complements measures of the direction of revisions. The format of Table 9, is similar to that in Table 6; again the median value is used as the average and revisions after one year are used. Because measures of magnitude are expressed in terms of absolute revisions, the values are generally larger than the equivalent amounts in Table 6. If all revisions were in the same direction the scaled and unscaled results in both tables would be the same. Conversely, the larger the differences between the direction and the magnitude measures for a particular aggregate, the greater the variability in the direction of revisions.
3.63. As an example of the use of data in Table 9, consider income debits. After one year, the median absolute revision to the initial monthly estimate was $98 million or 7.3%. When compared with the equivalent direction of revision statistics in Table 6 ($54 million or 3.9%), it is clear that the predominantly positive revisions to this aggregate must be partly offset by some negative revisions. It can also be said that revisions are dispersed to some degree between increases and decreases and that, as a result, there is no evidence of any strong tendency for revisions to be in one direction within the first year.
3.64. Examination of Table 9 shows that the magnitude of revisions to the initial estimate is greatest for the capital account aggregates for both quarterly and annual statistics. This is consistent with the patterns of bias in Table 7. Of interest in the current account is the fact that the median absolute value of revisions to the annual estimate of income debits ($397 million) is equal to the equivalent measure of bias in Table 6. However the median scaled value for the magnitude of revisions statistic (5.9%) is larger than the equivalent direction of revisions statistic (2.3%). This indicates that, on average, revisions to this aggregate are positive but that there are some negative revisions present and that, in absolute terms, these negative revisions are individually equal to or smaller than the value of the median direction of revisions measure. Equal values are also shown for the unscaled direction and magnitude of revisions measures for revisions to annual estimates of nonmerchandise debits and general government sector transactions (both for FIA and AIA, but not for the net result). These outcomes reflect, in part, the relatively few annual observations used in the analysis. No similar equivalence between the unscaled direction and magnitude of revisions measures occurs with either monthly or quarterly estimates.
3.65. One further feature displayed by Table 9 is the much larger revisions, on average, after one year to the initial quarterly and annual estimates of the balance on capital account than to the corresponding estimates of the balance on current account. The oneyear revisions to the balance on capital account are on average nearly seven times larger than those to the balance on current account for both quarterly and annual series. This largely reflects the timing of availability of comprehensive data on international investment flows from the annual collections within the ABS survey of Foreign Investment. The initial estimates of these flows are based on partial data available from the monthly and quarterly collections within this Survey and significant revisions frequently occur when the more complete annual data become available about six months after the end of the reference year.
3.66. Appendix 2 illustrates the magnitude of revisions in another way. It contains graphs which plot the revisions, after one year, to the initial estimates of a balance of payments aggregate for the three periodicities monthly, quarterly or annual statistics. Zero on the vertical scale represents the initial estimate for each period plotted. These graphs complement the information in Table 9 by giving an indication of the magnitude of revisions over the periods covered by the analysis and the way this pattern may have changed. They also give an indication of the direction of revisions, complementing Tables 6, 7 and 8. It should be borne in mind that the amounts plotted are in dollars and therefore may be affected by inflationary effects and by changes in the size of the aggregate being amended. It should also be noted that, because the scale is not the same on every graph, care is needed in comparing one graph with another.
3.67. The picture conveyed by Appendix 2 is broadly consistent with the information presented so far. Within the current account, the largest magnitude of revisions measures (in dollar terms) shown in Table 9 are for nonmerchandise debits, mainly due to the magnitude of revisions to income debits and services debits. This pattern is also evident in the appropriate Appendix 2 graphs. Of interest for income debits is the increasing magnitude of revisions in monthly statistics (Graph B.6) and the shift in the direction of revisions to monthly statistics from being strongly positive up to about March 1990 to a more balanced pattern until March 1993, then turning positive again in later periods. The switch from positive to a more balanced pattern from 1990 is also reflected in Table 7. This pattern of magnitude and direction in revisions to the monthly series for income debits is also evident in the Appendix 2 graph for quarterly series (Graph B.20). Graph B.39 is the equivalent graph for revisions to the annual income debits series. It shows a predominantly positive direction of revisions. This suggests that some large monthly and quarterly revisions offset each other and are not evident in annual statistics.
3.68. The very substantial upward revision to 199293 annual statistics of income debits (Graph B.39) is due principally to the incorporation of revised estimates of reinvested earnings on foreign direct investment in Australia. Information for these estimates is collected annually in the Survey of Foreign Investment, with data becoming available about six months after the end of the reference year. Prior to survey data becoming available, estimates are based on extrapolations made during the reference year. In times of rapidly changing economic conditions, such as in 199293, it proves particularly difficult to reliably extrapolate this item.
3.69. A similarly large upward revision was made to 199293 annual statistics of income credits (Graph B.38). Again this is principally attributable to revisions to estimates of reinvested earnings, in this case on Australian direct investment abroad. Essentially the same methodology is used in estimating these earnings as for those included in income debits but with the added complexity that, in this case, extrapolations have to take into account expectations of economic conditions and business profitability in those foreign countries in which Australia has direct investment.

9: MAGNITUDE OF REVISIONS TO INITIAL ESTIMATES, MEDIAN ONEYEAR REVISIONS (a)  
     Unscaled median oneyear absolute revisions  Scaled median oneyear absolute revisions (b)  
     Monthly  Quarterly  Annual  Monthly  Quarterly  Annual  
     $m  $m  $m  %  %  %  
  
 Current account        
  Merchandise exports  33  54  95  0.8  0.4  0.3  
  Merchandise imports  10  11  24  0.2  0.1  0.1  
  Balance on merchandise trade  40  61  142  n.a.  n.a.  n.a.  
  Nonmerchandise credits  63  234  303  3.9  4.5  2.4  
  Nonmerchandise debits  122  305  532  4.6  3.6  2.4  
  Net nonmerchandise  92  170  436  n.a.  n.a.  n.a.  
  Services credits  48  141  210  5.4  5.3  4.1  
  Services debts  42  107  457  3.8  3.4  2.7  
  Net services  57  146  336  n.a.  n.a.  n.a.  
  Income credits  38  134  261  9.9  13.7  13.2  
  Income debits  98  198  397  7.3  5.0  5.9  
  Net income  79  219  402  n.a.  n.a.  n.a.  
  Unrequited transfers credits  7  16  22  2.2  1.9  0.6  
  Unrequited transfers debits  3  8  18  1.9  1.2  0.9  
  Net unrequited transfers  8  23  31  n.a.  n.a.  n.a.  
  Balance on goods & services  73  141  377  n.a.  n.a.  n.a.  
  Balance on current account  110  191  327  9.1  5.8  3.4  

Capital account        
  Official sector transactions  151  346  338  n.a.  n.a.  n.a.  
   General government transactions  149  346  338  n.a.  n.a.  n.a.  
    Foreign investment in Australia  n.a.  428  625  n.a.  43.6  21.4  
    Australian investment abroad  n.a.  52  50  n.a.  42.2  37.7  
   Reserve Bank transactions  0  0  0  n.a.  n.a.  n.a.  
    Foreign investment in Australia  n.a.  0  0  n.a.  0.0  0.0  
    Australian investment abroad  n.a.  0  0  n.a.  0.0  0.0  
   Nonofficial sector transactions  n.a.  935  1 015  n.a.  n.a.  n.a.  
    Foreign investment in Australia  n.a.  1 140  3 367  n.a  25.4  19.7  
    Australian investment abroad  n.a.  927  1 661  n.a  56.7  20.8  
  Balance on capital account  n.a.  1 239  2 295  n.a.  n.a.  n.a.  

Balancing item 
n.a.

1 221

2 228

n.a

n.a

n.a
 

(a) For monthly statistics, from January 1986 to December 1993. For quarterly statistics, from the March quarter 1986 to the December quarter 1993. For annual statistics, generally from 198182 to 199394 except for the following series for which analysis commences in 198586: direct investment capital; the net series for services income and transfers; and balance on goods and services.
(b) Scaled revisions to net aggregates may be misleading. They are omitted here with the exception of revisions to balance on current account which have been applied to an aggregate which has been consistently negative over the whole period.  
3.70. The behaviour of revisions to the monthly income debits series is reflected to a large extent in the pattern of revisions to the monthly balance on current account shown in Graph B.11. From about June 1990 there is a noticeable shift in the direction of revisions to the monthly balance on current account from strongly negative, involving increases in the deficit, to a more balanced pattern but with a positive tendency. That is, over the period from June 1990 to December 1993, oneyear revisions tend to reduce the current account deficit. This change is also evident in Table 7 where the unscaled median revision to this balance in subperiod 2 is positive after being negative in the previous subperiod. This pattern is not apparent in quarterly and annual statistics (Graph B.25 and B.44).
3.71. Table 9 shows that, over the 13 year period from 198182, the size of revisions made to initial annual estimates of the balance on current account after one year average 3.4%, rising to 5.8% for quarterly estimates (over the period March quarter 1986 to December quarter 1993) and 9.1% for monthly estimates (over the period January 1986 to December 1993). The graphs in Appendix 2 (Graph B.11)., (Graph B.24)., (Graph B.44). show that the size of individual revisions after one year varied considerably over the period, and that these revisions were generally in a negative direction (implying upward revisions to the deficit) particularly for the quarterly and annual series. It should be recalled from the earlier analysis based on Appendix 1, however, that the revisions after one year are not the final revisions to estimates for particular periods. As shown in Graphs A.11, A.25 and A.44, the generally upward revisions after one year take the estimate of the current account deficit further from the final estimate than the initial estimate was in some cases. This emphasises the need for care in using the reliability indicators presented in this paper.
3.72. Within the capital account, revisions to monthly statistics are dominated by revisions to net capital transactions of the general government sector (Graph B.13). These transactions are predominantly borrowing by Australia. From about December 1987 the magnitude of these revisions has increased noticeably. This timing coincides with a substantial increase in the value of foreign borrowings by the State Governments and an associated increase in the size of revisions. The monthly data are collected shortly after the end of the reference period to meet monthly timetables and hence they are not as reliable as the quarterly and annual data. As Graph B.26 and Graph B.45 show the magnitude of revisions is still substantial in quarterly and annual statistics, but the netting across time means the magnitude is not quite as marked as in the monthly series.
3.73. The magnitude of revisions to quarterly capital transactions of the nonofficial sector for both FIA and AIA (Graphs B.30 and B.31) is substantial and has been so throughout the period for which data are analysed. The greater proportional magnitude exhibited by revisions to quarterly AIA statistics (57% for AIA compared with 25% for FIA) is a reflection of the difficulties experienced in maintaining satisfactory coverage of quarterly AIA transactions. Preliminary quarterly statistics are often subject to large revisions, particularly as annual survey results become available. Graphs B.49 and B.50 plot the equivalent revisions to annual statistics. As would be expected, these are also widely dispersed and show the positive direction indicated in Table 6. Table 9 measures the average proportional magnitude of revisions to annual FIA and AIA estimates after one year as 19.7% and 20.8%, respectively.