15.1. To be of most benefit to users, statistics need to be a reasonable and timely measure of the real world economic events to which they relate. There are a number of dimensions of high quality statistics. They need to be:
- not subject to large revisions;
- relevant (i.e. measure the concepts in which users are interested);
15.2. This chapter considers each of these aspects of quality, and then assesses the balance of payments and international investment statistics against them.
- comprehensive in coverage; and
Accuracy and revisability
15.3. Ideally, all balance of payments and international investment position statistics should be compiled by summing the notional foreign accounts of all the economic units (governments, businesses and households) in Australia. Alternatively, every individual transaction between residents and non-residents could be accessed, classified and summed. If either approach could be undertaken, the true measure of the balance of payments would be obtained. In practice neither approach is feasible. Economic units do not maintain ‘foreign accounts’ (i.e. many units are not interested in identifying their counterparties according to whether they are resident or non-resident), nor do they compile accounts for a subset of their counterparties. In addition, many of the accounting practices adopted by businesses and others do not conform with the requirements of national and international statistics. Also, of course, the costs of accessing every individual transaction would be huge. Instead, the estimates are derived by combining available data from a wide variety of sources (including administrative processes, sample surveys and censuses), carrying varying degrees of accuracy, frequency, detail and timeliness.
15.4. Accuracy is concerned with the proximity of an estimate of a data component to the ideal or true value of that component. Revisability refers to the differences between the initial estimate for a particular period or point in time, any intermediate estimates for that period or point in time and the final estimate for that period or time point.
15.5. In the extreme case, estimated data may be completely inaccurate, but subject to no revisions, i.e. although the estimate bears no resemblance to the true value, because of the absence of better information the initial estimate is never revised. Alternatively, highly accurate estimates may only be producable after long time lags, while early estimates are poor and subject to considerable revisions. Generally, as the estimate for a particular period passes through a sequence of revisions, the size of revisions becomes smaller (the statistic becomes less revisable) and the estimate moves closer to the true value (the statistic becomes more accurate). In balance of payments and international investment statistics this is often the case, but not always.
15.6. The accuracy of the statistics is affected by a number of influences, including:
- data collection errors (inability of survey providers to report on the correct basis; mistakes in reporting or late reporting by providers; errors introduced during recording and processing of data reported);
- methodological errors (arising from misinterpretation of the nature and extent of transactions and the residency status of transactors, and also from shortcomings in data sources and estimation methods);
- inconsistencies in the time of recording and valuation of the corresponding credit and debit entries and the corresponding stocks;
- undercoverage (due to difficulties in identifying all the businesses involved in international economic activity during a period);
- the application of a revisions policy, in the production of estimates, which has the effect that data are not always revised as soon as they could be, although the impact is considered to be small (see paragraph 4.12); and
15.7. The revisability of these statistics is also affected by a number of influences, including:
- the use of sample surveys rather than complete enumeration.
- the need to extrapolate series prior to the availability of source data;
- errors in reported data which are detected and corrected at a later date. Such errors sometimes arise from transactors’ attempts to meet timeliness requirements
- errors in ABS estimation and compilation processes which are subsequently detected and corrected; and
- the introduction of improved methodologies and data sources which better approximate the concept underlying a series, but which necessarily result in revisions to previous estimates for that series.
15.8. Where data are estimated from a sample survey rather than from a fully enumerated survey, the estimates are subject to sampling error. Sampling procedures are used in some of the survey sources contributing to the balance of payments and international investment position statistics, such as:
- the Survey of International Trade in Services;
- the Survey of International Investment;
- Overseas Arrivals and Departures;
- the International Visitor Survey;
- the Survey of Returned Australian Travellers; and
15.9. The relative sampling errors for the main aggregates from these surveys are generally small; for the more detailed aggregates, relative sampling errors may be somewhat higher.
- the Survey of International Students’ Expenditure.
15.10. A range of factors (listed in paragraph 15.6) other than those caused by the use of sample surveys, contribute to errors in the estimates. These are referred to collectively as non-sampling errors.
15.11. Timeliness refers to the lag between the end of the reference period and the initial publication of statistics for that period. There is an important trade-off for many users between accuracy, revisability and detail on the one hand, and timeliness of the release of statistics on the other. Frequently the data available to compilers in these statistical fields come from a wide variety of sources which reflect varying frequency, detail and timeliness, and for which valuation and coverage may only approximate a desired concept. Generally, within given resource levels, significant improvements in timeliness can only be made at the expense of detail, accuracy and revisability. One user undertaking long-term analysis may well prefer statistics which are available only annually and released with a considerable time lag, but which provide very accurate measures of transactions and stocks for quite detailed cross-classifications. Another user may prefer very timely and frequent estimates of the ‘broad brush’ type that indicate current directions.
15.12. The timing of quarterly and annual balance of payments and international investment position publications is based on a number of compromises. For example, the first published quarterly estimates are made available about 43 working days after the end of the reference quarter. In this timeframe it is not possible for all the surveyed enterprises and other data sources to provide detailed figures on their foreign activities. For those service and income components for which full data for the latest quarter are not available in time for inclusion in the quarterly publication, estimates are made by extrapolating historical series using the best available indicators of current activity. The latest estimates for the quarter are subsequently revised as more data become available. Similarly, in the financial accounts and international investment position, only broad preliminary estimates are included for the latest quarter (more complete data are available with the later release of Cat. no. 5232.0 - see paragraph 4.10), with the detailed standard components being released with a one quarter lag. While care is needed in the use and interpretation of the preliminary estimates for the quarter, they are considered useful for those analytical purposes which require a broad indication of current developments in the external accounts. (The same principle applies to the compilation of monthly goods and services data, which are published in the monthly publication (Cat. no. 5368.0) within 21 working days of the reference month, i.e. many services and some goods components need to be extrapolated initially.)
15.13. The annual publication, Cat. no. 5363.0, provides detailed estimates which are, as far as possible, not subject to many of the sources of revision that arise in the trade-off between timeliness and accuracy in the more frequent estimates, but are still subject to revision due to methodological improvements. The annual publication incorporates data for the reference year from all major data sources. It is not issued until detailed figures from the various contributing surveys become available for the reference year. Timeliness is given a lower priority, in favour of providing detailed statistics that are more accurate.
Relevance and links to other series
15.14. An important aspect of statistical quality is that the concepts, definitions and classifications used should be relevant to and understandable by users, be internally coherent, and support relatability of these statistics to others. ABS balance of payments and international investment statistics are produced within an integrated statistical framework, based on international standards (BPM5) which are fully compatible with those for the national accounts (SNA93). These standards have been developed on the basis of the needs of key users involved in macroeconomic analysis and policy formation. To maintain the relevance of the statistics, the conceptual framework is augmented from time to time to deliver additional, relevant domestic perspectives to accommodate economic and financial developments. The sources and methods used are reviewed regularly, and periodically changes are made to the compilation of components to approximate more closely the concepts being measured.