|Page tools: Print Page Print All|
However, several key users and the ABS have become increasingly concerned about the interpretations placed on the monthly balance of payments statistics, given their underlying quality, and in particular the focus on the current account deficit measure. The movements in the monthly series as currently published substantially reflect irregular factors, are subject to occasionally substantial revision, and include extrapolated estimates for the current (or reference) month and for recent past months for many of the component series. As such, the quality of the series cannot support the intense scrutiny that accompanies their publication each month.
The ABS has a number of options to address its concerns about the quality of the monthly balance of payments statistics. These range from, at one end, doing nothing about the underlying compilation of the series but provide users with more information about the quality of the statistics to better inform their use of the data to, at the other end of the spectrum of options, substantial ABS resources could be used, coupled with a substantial increase in the load placed on businesses to provide data, to improve the quality of the data to a level commensurate with their use. In between these extremes, there are a range of options that curtail ABS publication to those monthly series that can be considered of reasonable quality, or which can be improved at a modest cost, and which support most of the needs currently targeted with a more complete balance of payments statement.
In reviewing the options available to it, the ABS has considered:
Based on these considerations, and after consultation with members of the Economic Statistics User Group, the Australian Statistics Advisory Council, and several other users, the ABS is of the view that it can best meet the needs of its users by discontinuing the publication of the monthly balance of payments statistics and replacing them with one of the options identified in Section 4 of this paper. The various options provide users with different monthly data sets for use in their analyses. However, all those options remove from monthly publication the income series, which are based mostly on extrapolation, are most subject to revision, and for which "hard" monthly data never become available. They also all remove the unrequited transfers series which have been largely extrapolated for several years, and the partial capital account that is currently compiled. Of these options, the ABS preference, based on user views expressed to date, is for option (c) in Section 4 of this paper; i.e., withdraw from production of the monthly balance of payments, but continue to publish monthly merchandise exports, merchandise imports, services exports and imports, and the balance on goods and services, all on a balance of payments basis.
Invitation to Comment
The ABS is now seeking to finalise its views on its proposal to discontinue the compilation and publication of monthly balance of payments statistics and to replace it with the statistics outlined in the previous paragraph. We are therefore asking for comment from the wider user community on this Information Paper and the options it presents. Details of the consultation arrangements to be used in reaching a final ABS decision, and contacts for comment or further information are provided in Section 5 of this paper.
Section 2 of the paper provides some background to the issues of balance of payments data quality and quality improvement, and the emerging concerns about the monthly data. Section 3 looks at the quality concerns surrounding the current monthly balance of payments statistics in more detail, and Section 4 looks at the options considered for addressing those concerns.
The quality of Australia's balance of payments statistics has been under review and improvement for many years. When the ABS first published monthly balance of payments statistics in the mid 1970s, it had been known from the 1960s that a key data source, the Reserve Bank of Australia's (RBA) foreign exchange (or "tickets") system(a), was an unsatisfactory source, and was deteriorating. However, no systematic replacement was available in the 1960s, and alternative sources were pursued on a component by component basis. During the 1960s the ABS ceased to use tickets for many income and services measures and moved to ABS business surveys or alternative sources. By the early 1970s the tickets system was still used to measure travel credits and debits, miscellaneous services and royalties, non-official unrequited transfers and Australian portfolio investment abroad.
(a) Introduced in 1948 by the then Commonwealth Bank, this collection was initially designed to cover all transactions through the banking system involving payments to and receipts from non-Sterling countries. It was extended in scope to the Sterling area, but subsequently both the scope and coverage were progressively revised and reduced by the Reserve Bank (as successor to the Commonwealth Bank) to better target BoP information needs; and reduced again by the ABS when it took over responsibility for the collection in 1984.
The international visitor survey (then conducted by the Australian Tourist Commission), together with some administrative and ABS survey sources had replaced tickets for the measurement of travel credits from 1971-72. By the late 1970s new ABS survey sources were being used to measure travel debits, with revisions to published estimates increasing the annual level of the series being estimated three-fold, as a result of the source change, from about $0.5 billion to over $1.6 billion. By then, the tickets system was only being used for the monthly and quarterly current account estimates of unrequited transfers of the non-official sector, and of certain "miscellaneous" services, which together accounted for about 4% of total current account credits and debits.
The ABS publication Balance of Payments, Australia, Concepts, Sources and Methods (ABS Cat No. 5331.0) was first issued in 1981. By documenting in one place all the sources and methods used in balance of payments compilation, it provided both the ABS and users with information on which to assess priority areas for statistical development. In the early 1980s, the ABS reviewed the tickets system once again with a view to either improving this source for the residual items it measured, or replacing it. From 1984 responsibility for the tickets collection passed from the RBA to the ABS and it was again reviewed to determine whether further changes could be made to improve data quality. Substantial improvements were not able to be made, and from 1987-88 the ABS introduced the Survey of International Trade in Services to measure the miscellaneous services. The former tickets system was redeveloped to become the Survey of Foreign Unrequited Transfers, which is currently under review.
Summary concepts, sources and methods publications followed until the 1990 edition of 5331.0, which included a new chapter on data quality. Most recently the ABS followed up on earlier quality analyses, included in 5331.0, by publishing in early 1996, a comprehensive information paper Quality of Australian Balance of Payments Statistics (Catalogue No 5342.0). That publication provides a detailed assessment of the quality of Australia's balance of payments statistics, including reliability as a measure of quality. A copy of that paper is provided with this Discussion Paper to assist readers wishing to pursue in more detail the various aspects of quality assessment.
Other quality improvements in monthly and quarterly sources and compilation methods for the current and capital accounts have been made in a process of continuous improvement. Moving exports compilation onto a shipping date basis in the early 1990s was identified as an important improvement that better aligned exports on a balance of payments basis with other macro-economic statistics. The introduction and enhancement to surveys of returning Australian travellers will improve estimates of travel debits. The recent development of a quarterly source for estimating reinvested earnings of direct investment enterprises has improved not only monthly and quarterly estimation, but provided increased quality control on the annual estimates. The move to introduce comprehensive quarterly sources for all international investment measures of stocks, flows and income is expected to significantly improve the quality of initial estimates of the quarterly financial account. The increasing move to individual security reporting for measuring cross-border financial flows is also expected to significantly improve the coverage and measurement of financial flows in the capital account.
New International Reporting Obligations
In April 1996, the International Monetary Fund (IMF) issued new international reporting obligations with the introduction of the Special Data Dissemination Standard (SDDS), a set of standards for which countries wishing to access international capital markets on more favourable terms are expected to comply. (The Australian Government has subscribed to these standards.) The SDDS requires quarterly rather than monthly balance of payments data, and requires these data to meet the new international statistical standards for balance of payments and international investment position statistics that were published by the IMF and other international agencies in late 1993.
Methodological improvements continue to be pursued by the ABS, and over the past few years, in the area of balance of payments statistics, the focus has been on the implementation of these new statistical standards, and the implications of these developments on statistical quality generally and on the priorities for improvements in the data. Several ABS collections have required substantial alteration, and increased resources have been allocated so that the standards for balance of payments statistics would be substantially met by 1997. Our attention continues to be focused towards the requirements for quarterly data, where reasonable progress can be made to meet international standards with the imposition of reasonable provider load. Meeting the conceptual and methodological requirements of the new standards for data of monthly frequency would significantly increase provider load and ABS costs.
Subjective Assessments of Quality
In the context of implementing new international statistical standards and new methods in national accounts compilation, users consulted in 1995 about priorities for, and issues, in national accounting generally, noted concern about the quality of the monthly balance of payments statistics. The completion of the quality Information Paper (5342.0) in early 1996 further highlighted, for the ABS, the user concerns about the quality of the monthly balance of payments, particularly in light of the impact that their release often engenders. The following table is reproduced from 5342.0 and provides the subjective assessments of the accuracy of initial balance of payments estimates.
The accuracy ratings in Table 1 refer to initial transactions estimates. That is, the "D" rating is provided for the initial monthly current account balance - the key focus at each monthly release. Such a low rating is of concern. However, it should be noted that this rating of the estimate of the level of the deficit does not necessarily reflect the quality of estimates of period-to-period movements.
Table 1: Subjective Accuracy Ratings of Balance of Payments Current Account Aggregates(a)
(a) The accuracy ratings used relate to the following approximate margins of error; A less than 5%; B less than 10%; C less than 15%; D 15% and greater. It should be noted that these accuracy ratings relate to estimates of the value of aggregates, not estimates of the movements in these aggregates.
QUALITY OF MONTHLY BALANCE OF PAYMENTS STATISTICS
The Context of Data Use
In assessing the quality of any statistical series, the purposes to which they are to be put determines to a large extent the framework for that quality assessment. An approximate answer to the right question will, for many issues, be a suitable measure. The monthly balance of payments statistics are being assessed by some users in ways that take account of the various quality limitations inherent in some component series and in the overall aggregates. The series are viewed in the longer term, with limited account placed on any one month's figures. However, other user reactions appear to lack this balanced assessment of quality for the applications to which the data can be put.
The most obvious area of apparent overreaction to monthly balance of payments statistics has been in the financial markets, which have in the past responded quite strongly to one month's numbers and out of proportion to the quality of the data to support such reactions. Analysts generally perceive volatility in the financial markets to be exacerbated by the reaction to each month's volatile performance in the balance of payments numbers. Such volatility has real economic consequences for the nation. Dr Steven Kates, Chief Economist, Australian Chamber of Commerce and Industry, has noted that the volatility in the monthly numbers has had a destabilising effect on the markets, and that the consequent fluctuations in interest rates have reduced business confidence and made investment decisions more tentative. The following table provides indicative information on the scale of that reaction, and generally supports Dr Kates' observations regarding interest rate volatility.
Table 2: Market Reaction to Monthly Balance of Payments Releases
(a) The percentage change is calculated as the percentage difference between the 9.00 am yield on the day of release of the monthly balance of payments and the 9.00 am yield on the day following release.
Comparing rates at the start of trading on the day of release of the monthly balance of payment with rates 24 hours later provides some measure of the impact of the monthly release. No attempt has been made to consider other causative factors, and the results, while suggestive, cannot therefore be considered conclusive. Over the past 19 months, on seven occasions the seasonally adjusted current account deficit has exceeded general market expectations. On all but one of those occasions, bond yields rose on the day of release. And on four of those occasions the higher than expected result has been followed by one or more months where the result was at or below expectation and bond yields fell on the day of release.
To put the observations on rate movements on the day of release into perspective, of the 492 trading days covered in the 19 month period used in Table 2, the rate fell on 52% of occasions, rose on 41% of occasions, and was steady for the remaining 7%. Looking at some of individual releases, the rise on the day of release of the "bad" January 1995 data followed 4 days of decline. The May 1995 result saw the rate fall the day before release, and fall for 5 days after release - only on the day of release of the "bad" data did it rise. For the September 1995 release, the rate rose before and on release day, but fell for three days after that. The rise on the May 1996 release day was preceded by falls for 6 days prior to release. On the other two occasions of 'bad" results and rising rates the pattern around the release dates was mixed. Generally the number of rises (6 out of 7) on the back of "bad" BOP releases seems to be a probable reaction to these BOP data.
The following graphs of the main components and balances in the balance of payments are intended to illustrate their relative magnitudes and volatility. The data are presented in seasonally adjusted form (except where indicated on the graph). The first page of graphs shows the balance on current account, and the four component balances that comprise it. The second page of graphs takes Net Income and shows various component balances for it.
Balance on Current Account - Seasonally Adjusted
Balance on Merchandise Trade - Seasonally Adjusted
Net Services - Seasonally Adjusted
Net Income - Seasonally Adjusted
Net Unrequited Transfers - Seasonally Adjusted
Net Income - Seasonally Adjusted
Balance on Labour and Other Income - Original
Net Other Property Income - Original
Net Investment Income - Original
Balance on Official Investment Income - Original
Balance on Non-Official Investment Income - Original
Irregularity in Monthly Balance of Payments Series
The volatile nature of the monthly balance of payments series can also be considered in terms of the contribution of irregular factors to movements in the series. Table 3 below presents the contribution of the irregular to movements in various seasonally adjusted items of the current account of the balance of payments. It shows, for example, that on average 79% of the change in the seasonally adjusted monthly balance on current account is due to irregulars. Most of the other monthly items and balances have similar, or nearly so, proportions of irregularity. The implication is that most of the monthly movement is "noise" and very little of it is "signal". On average (mean), the trend estimate for the reference month for the past 5 years has been subsequently revised by 15%. It is thus difficult to estimate and make use of the signal (ie, the trend) when it is contaminated by so much noise, except with hindsight.
Table 3: Median Contribution(a) of the Irregular to Movements in the Seasonally Adjusted Series
Having noted in the previous two sub-sections the volatility in, and irregularity of, the monthly balance of payments series, the next issue to be addressed is the reliability of the data. Can a user take the latest published statistics and be reasonably confident that the estimates will persist in the short to medium term as the best view of what was happening in Australia's external economic transactions for that month? Will the story be changed through revisions?
The ABS Information Paper Quality of Australian Balance of Payments Statistics (Catalogue No 5342.0), published in early 1996, provides a detailed assessment of the quality of Australia's balance of payments statistics, including reliability as a measure of quality. The following graph taken from that publication illustrates the size of revisions that were made within the year following initial publication to the balance on current account for each month of the period 1986 to 1993.
One-Year Revisions to Initial Monthly CAD Estimates
Initial estimates of the monthly current account deficit have been subject to a median revision of about $30 million (or 2%). However, as can be seen in the above graph, the swings above and below the initially estimated level can on occasion be quite large. The median revision without regard to sign is much higher, at about $124 million (or 11%). The mean absolute revision was a little higher still, at $134 million (or 13%).
Accuracy as a Measure of Quality
The above three sub-sections reviewed the users' reaction to initial monthly balance of payments estimates in light of the data's volatility, irregularity and experience with occasionally sizeable revisions. It is also useful to analyse the compilation practices used to estimate the various components of the monthly balance of payments. Again, the ABS Information Paper Quality of Australian Balance of Payments Statistics (Catalogue No 5342.0) provided a quite detailed assessment of the accuracy of published balance of payments statistics and the components. In this Discussion Paper the focus is placed on the sources and methods used in the compilation of components of the statistics and their suitability for the purpose.
The ABS publication Balance of Payments, Australia, Summary of Concepts, Sources and Methods, 1996 Edition (Catalogue No 5351.0) includes an appendix listing all the balance of payments items together with their sources of data and methods of estimation. The following Table 4 summarises that information in relation to the goods and services components. The size and percentage contribution of monthly sources is also shown.
Table 4. Goods and Services, Summary Sources and Monthly Coverage
(a) Data as published in the June quarter 1996 issue of Balance of Payments, Australia (Cat No 5302.0).
In excess of 80% of the value of the goods and services components of the monthly balance of payments are derived from monthly source data. In many cases (e.g., merchandise trade, shipment debits) the monthly data are available in time for initial estimates for the reference month. In a few cases, such as passenger fares, the monthly data are collected with a lag of 3 months, and extrapolations, based on good monthly indicators in ABS overseas arrivals and departures statistics (OAD) and in passenger movements statistics from the Department of Transport and Regional Development are used.
In addition to monthly source data there is a further 12% of the value of the goods and services components that is estimated from quarterly or less frequent source data, with both extrapolations and interpolations based on good monthly indicators. These indicators are ABS OAD data, passenger movements statistics, and merchandise trade statistics. The series affected by these estimation processes are the quarterly "other transportation" and "travel" series.
With 95% of monthly goods and services estimates captured in timely monthly source data, extrapolated from less timely monthly data with good indicators, or both extrapolated and interpolated from quarterly data using good indicators, the overall quality of the initial monthly estimates of the goods and services series is considered satisfactory for most purposes to which the data might be put. Nevertheless, there are occasionally substantial revisions to "hard" data such as for merchandise.
The only component of goods and services for which there is no monthly source, and no prospect of a monthly indicator, is "other" non-official services. This component accounts for about 5% of both goods and services credits and debits. However, these series are relatively stable quarterly. The quarterly movement in the net series has not exceeded $60m (implying very small monthly movement) and in recent years the quarterly movement in the credit and debit aggregates for these series has been less than half of one percent of the aggregates for goods and services credits and debits. In the light of these observations, an even monthly allocation of quarterly estimates is reasonable and unlikely to contribute to a misleading estimate of the monthly balance on goods and services.
The same conclusion cannot be drawn for the monthly income and unrequited transfers series. Table 5 presents information similar to Table 4 but for the income and transfers items.
Table 5: Income and Unrequited Transfers, Summary Sources and Monthly Coverage
With only 20% of the value of the monthly income and transfers estimates ever sourced from monthly data, and with no good monthly indicators currently available, it is difficult to see a clear purpose served by their continued publication.
The monthly income and transfers series are generally estimated at one third of the quarterly source data. For reinvested earnings (24% of the total value of the series) extrapolations are required for up to 9 months until quarterly data become available. A further 8% of the series value (most of non-official transfers) are currently extrapolations from monthly surveys undertaken some years ago.
There are estimation methodologies that can be employed to improve the monthly estimation of some components of income. The adoption of full accrual accounting for income, commencing with the introduction of new international statistical standards from late 1997, should provide the basis for better estimating the interest income on a monthly basis. However, for much of income (reinvested earnings, dividends, and the distributed profits of branches, which together account for over half of investment income) there is no ready source or indicator for monthly income measures.
A monthly source is currently being evaluated for non-official unrequited transfers credits and debits. Should it prove successful, there would be a further (small) improvement in the coverage of monthly sources in monthly estimates of income and transfers. However, even with improvements in the data sources for these two components, it is unlikely that the total coverage would increase beyond 50% of the total income and transfer transactions values.
With all options, the balance of payments on a quarterly basis would continue to be published with at least the same level of detail and timeliness as at present. Avenues to continue to improve these data would also be followed. The options considered by the ABS to better align the quality of monthly data on external transactions with the range of uses to which the data can reasonably be expected to support are:
(a) maintain the status quo;
(b) continue to publish the monthly balance of payments, but put additional effort into user education on the nature and quality of the data and emphasis on trend figures, and where possible consider taking advantage of information from collections being introduced for improved quarterly balance of payments estimates;
(c) withdraw from production of the monthly balance of payments, but continue to publish monthly merchandise imports, merchandise exports, services imports and exports, and the balance on goods and services, on a balance of payments basis;
(d) withdraw from production of the monthly balance of payments, but continue to publish monthly merchandise imports and merchandise exports, and the balance on merchandise trade, on a balance of payments basis;
(e) withdraw from the production of any monthly data on a balance of payments basis and continue to publish separately the monthly imports and exports data on an international trade basis; or
(f) apply additional ABS resources, and increased provider load, to develop a reasonable coverage of monthly income and transfers measures so that reasonable quality monthly balance on current account estimates, commensurate with the uses to which they are put, could be published.
Option (a) is not considered viable. The increasing importance of income over the past decade, for which monthly sources currently provide very low coverage, prevents the status quo being a reasonable approach to providing statistics of sufficient quality to meet the range of needs to which users are now applying these data.
With option (b) (a slight variant on (a)), there is the opportunity, were it warranted, for some improvement over the next two years in the quality of the income measures as a result of collection changes to support better quality quarterly data and revised international standards, including reporting data on an accruals basis. However, these methodological improvements (and any others that might be considered) will not improve the monthly coverage of income data sources beyond about 50%. In light of the significant gap that would remain, the costs of improving some parts of income estimation are not considered by the ABS to be warranted.
Option (c) is the ABS preference and offers the users well based monthly estimates that will support analyses of the impact of net exports on growth and of demand for imports. While the series are volatile, with a high degree of irregularity in the movements in seasonally adjusted series, users are getting data that are at least 80% "hard", with up to 95% based on extrapolation techniques from monthly series with monthly indicators. The major uses of the monthly data, as far as ABS can determine from the advice given from time to time during user consultations, are in the general monitoring of the economy, and as an indicator of potential movements in the exchange rate and/or interest rates due to Australian economic conditions. Informed users look at the various components of the balance of payments to develop considered views of economic conditions rather than just focusing on the bottom-line balance on current account. Option (c), in conjunction with the full set of quarterly accounts, should suit most of these applications.
It would be reasonable under option (c) to limit the presentation of monthly data on a balance of payments basis to one publication of goods and services, and discontinue the separate release of a publication for imports on a balance of payments basis. Of course, the detailed goods imports data would continue to released on at least as timely a basis as at present, and may be able to be released more quickly without the requirement for balance of payments adjustment, etc.
Option (d) moves further down the line towards limiting monthly publication to those broad aggregates that are fully sourced from "hard" data. This approach denies users much of the well based monthly series on services that are currently collected and compiled, and extrapolated in a reasonably rigorous fashion from available sources.
Option (e) represents a retreat from compilation of any balance of payments components on a monthly basis. This option continues to support the user requirement for detailed data on goods commodities, but does not facilitate significant macro-economic analysis. This position is the one taken by many other countries, which provide monthly merchandise trade data and quarterly balance of payments data (and is consistent with the IMF's SDDS).
Option (f) would be very expensive for the ABS and would impose a very large increase in provider load to try to meet the objective. It is doubtful that providers would co-operate with the ABS in a monthly collection of investment income (requiring monthly profits estimates to be reported), nor in a monthly collection of "other" services. It is also likely that an ABS attempt in this direction would impact adversely on provider cooperation with existing data collections in support of quarterly balance of payments statistics. The ABS therefore does not consider option (f) as viable.
Most countries do not produce monthly BOP. Significant exceptions are Japan and several European countries, which generally use banking records for the purpose, and which produce a monthly balance of payments including a complete capital account. It is possible that more European countries will be required to compile monthly BOP when monetary union is implemented. However, other countries are unlikely to commence compilation of monthly BOP, since monthly BOP statistics are not part of the International Monetary Fund's SDDS.
he ABS proposes, subject to further consultation with users, to discontinue the monthly balance of payments publication in its current form and replace it with option (c). The last issue of 5301.0 is expected to be the December 1996 issue due for release on 31 January 1997. January 1997 would be the first issue of any replacement service.
The quarterly balance of payments publication (Catalogue No. 5302.0) will continue to be produced, and the results will continue to be widely disseminated. Changes in concepts and presentation of quarterly balance of payments statistics, in line with recently revised international standards, will be introduced from the September quarter 1997, as previously announced.
The ABS invites written comments from users of balance of payments statistics on the issues raised and options identified in this Information Paper. Any other options would also be considered. Such comments should be made by 24 January 1997, and be addressed to:
The Australian Statistician
Attention: Director, Balance of Payments
PO Box 10
Belconnen ACT 2616
or faxed to (06) 251 6009
The Australian Statistician expects to make a decision on the future of the monthly balance of payments statistics, and on any replacement services, in the week beginning 27 January 1997.
For further information on the detail of compilation of the balance of payments, or on other issues raised in this Information Paper, contact Ivan King on (06) 252 6038.
These documents will be presented in a new window.