5342.0 - Balance of Payments Statistics, Information Paper on Quality , 1996  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 20/02/1996   
   Page tools: Print Print Page Print all pages in this productPrint All  
Contents >> Chapter 1. Data quality concepts >> Factors influencing quality


1.15. Apart from the fact that data may only be practically available on a slightly different basis from that called for in the conceptual framework, there are a number of other factors that have a bearing on the perceived quality of balance of payments statistics.

      • Timeliness
      • Collection methodology
      • Time of recording
      • Valuation
      • Coverage
      • Methodological deficiencies
      • Conceptual and methodological changes
      • Revisions policy


1.16. There is an important trade-off between the accuracy and reliability of the estimates and the timeliness of their release. To meet the decision-making needs of users there is pressure to produce statistics on as timely and frequent a basis as possible. To the extent that complete or accurate data may not be available with the desired timing or can only be obtained at unacceptably high resource costs, there will be compromises in data quality. Generally, significant improvements in timeliness can only be made at the expense of detail, accuracy, reliability or additional resources.

1.17. The ABS releases monthly, quarterly and annual balance of payments statistics. ABS policy is to release initial monthly estimates as soon as the ABS has sufficient data to provide a broad indicator of current balance of payments trends; in practice, this currently means about 21 working days after the end of the month, once data that are sourced monthly are available on merchandise trade, official reserve assets, most general government sector transactions, and some non-official public sector investment transactions. Other components of the accounts are sourced quarterly, annually or at longer time intervals, and initial monthly estimates are extrapolations based on the trend in recent periods and adjusted using currently available indicator information.

1.18. Initial estimates for a quarter, for the restricted range of series released in monthly publications, will be available with the release of estimates for the last month of that quarter. However, the quarterly balance of payments publication is not currently released for a further 21 working days; that is, until about 42 working days after the end of the quarter. The extra time is required to generate constant price estimates for trade in goods and services, to incorporate significant additional detail becoming available from quarterly survey results, and to process revisions for earlier periods arising from those quarterly sources. As a result of the more comprehensive data sources available quarterly, initial quarterly statistics are generally more accurate than initial monthly statistics, but care is still required in interpreting statistics for the more recent quarters.

1.19. The initial estimates for a full financial year are available with the release of monthly data for June of that year. Such annual estimates will contain a mixture of relatively accurate data for the full year compiled from timely monthly sources; somewhat less accurate data from quarterly sources for the earlier quarters of the year and for which significant revisions could still be expected; and extrapolations for periods where either quarterly or annual survey results are not yet available.

Collection methodology

1.20. Where estimates are based on the results of a sample survey, rather than complete enumeration, they are subject to sampling error. Samples are designed to produce estimates of varying accuracy at various levels of detail, with standard errors lowest at the highest levels of aggregation. For some detailed estimates, relative standard errors can be high.

1.21. For those estimates based on a partial collection, in which data are collected only from units with activity above a certain threshold, estimates of the activity of the smaller units which are not approached are based on their contribution to estimates for an earlier benchmark period. The activity of these units is checked periodically and, where it exceeds the threshold, they are added to the collection. However, this methodology may lead to inaccuracies if the activity of the units that are not approached changes significantly from the contribution they made to earlier benchmark estimates.

Time of recording

1.22. Balance of payments statistics are sensitive to the time of recording of transactions, and imbalances will be present if the credit and debit entries for a single transaction are recorded in different time periods. Such time of recording differences arise particularly where different data sources are used for capturing the two sides to any transaction. For example, merchandise imports information will generally be recorded (as a debit) in the period in which customs documents are lodged in Australia. While this time is generally a good practical approximation to the change of ownership of merchandise imports, some goods may change ownership at an earlier or later point in time. In that case, while the associated financing (payment, trade credit or other settlement) would be correctly reported for that earlier or later period by either the owner or the financial intermediary involved, the imports will be understated in the earlier or later period and overstated when the recording occurs, with consequent effects on the aggregates and balances in both periods. While efforts are made to detect and correct for any significant recording departures from the time of change of ownership in compiling the principal components of the accounts, it is not possible to compensate for them all.


1.23. As with the time of recording of transactions, if different valuations are used in measuring the credit and debit aspects of a transaction, there will be both inaccuracies and imbalances in the accounts. Although market valuation is the standard valuation principle for the accounts (often approximated by transactions values), and a standard principle for converting from foreign currencies is prescribed, some source data may not conform. For example, if a non-resident buys shares in an Australian enterprise from a resident, the financing of the transaction may be captured through intermediaries at the actual transactions prices involved but the change in ownership of the shares may have to be approximated based on changes over time in the stock of such shares held by non-residents multiplied by an average share price for the period. While efforts are made to get as close as possible to the preferred basis of valuation, this will not always be feasible and there will be valuation errors in the statistics.


1.24. Undercoverage of balance of payments transactors typically stems from the difficulties in identifying all the entities engaged in particular types of international transactions. For example, since the removal of most types of restrictions on international capital flows, there has not been a comprehensive source which the ABS can use to maintain its coverage of entities having international capital flows. Instead, the ABS has had to rely on several partial sources, none of which is comprehensive. On occasions, this has meant that the existence of transactors with international capital flows has only come to notice well after the flows have occurred. In such circumstances, revisions to past periods may be necessary some time after statistics for those periods have been first released. In addition, for identified transactions, there may be deficiencies in coverage of these transactions as a result of misreporting in ABS collections.

Methodological deficiencies

1.25. Even where the population of transactors with international flows can be identified, the only data readily available from those transactors or from administrative sources may indistinguishably include domestic and international transactions. Although the ABS may apply an adjustment to isolate the international component, this can only be an approximation based on some earlier estimate of the relative proportions of domestic and international transactions. Other deficiencies can arise because the distinctions embodied in definitions and classifications are not made in the source data. Also, in cases where the existing methodology relies on assumptions and approximations, or on indirect measures, some level of inaccuracy is unavoidable.

Conceptual and methodological changes

1.26. The concepts and methods used in preparing ABS balance of payments statistics are regularly reviewed and periodically changes are made to the basis of compilation of an item in order to more closely approximate the concept described in the international standards. An example of a change in concept is the change (made in March 1992) for recording merchandise exports for a period, from primarily one based on the date of processing through Customs to one based on the date of shipment from Australia. The latter is a better practical approximation to the desired change of ownership concept. Methodological changes that change the basis of compilation result in one-off revisions to statistical series that may be substantial. Similarly, other one-off revisions to statistical series from methodological changes that improve coverage are also quite often substantial. Both types of change can be considered to improve the accuracy of the statistics but should not be seen to detract from their reliability. However, for practical reasons, only the impact of the change to the concept for recording merchandise exports has been removed prior to undertaking the reliability analysis described later in this paper.

Revisions policy

1.27. Frequency and timing of the application of revisions to data series may influence the apparent nature of errors and user perceptions of quality. For example, in some countries, revisions may be stockpiled for many periods and then applied simultaneously. Between these major updates, there will be an artificial impression of reliability in the statistics since the last major update that may be misleading for users unfamiliar with the revisions practice. On the other hand, some stockpiled revisions may offset each other which will tend to lessen the impact of the revisions when they are applied, but this will be at the cost of releasing less accurate statistics between updates.

1.28. The general policy adopted for the Australian balance of payments is to take every practical opportunity to incorporate more accurate information into the estimates for every period. However, to minimise disruption to historical statistical series, revisions to periods prior to the current financial year for the investment income and capital account series, and prior to the previous financial year for all other series, are generally only made twice a year. These revisions are made in the July and January issues of the monthly publication and, consistently, in the June and December quarterly publications that are produced shortly afterwards. Exceptions are made to this practice where significant revisions are identified; these are incorporated into the next available issue.

1.29. As a practical matter, and generally reflecting the quarterly periodicity of many of the data sources used in balance of payments compilation, revisions to series for the current (and in many cases the previous) financial year are incorporated at quarterly intervals - in each quarterly issue and in the July, October, January and April issues of the monthly publication. Other monthly issues generally carry revisions for the latest one or two months only. Again, exceptions are made where significant revisions are identified.

Previous PageNext Page