5.3. The compilation process involves taking data from various data sources and data models to create the standard data components to be published in the balance of payments and international investment statistics. The statistical issues involved include:
- identification of a suitable source of data for each component, which may result in the need to develop a new statistical collection, or the need to use some related data items and form a model through which the derived data items can be estimated;
- assessment of the quality of the data so derived, and working over time to improve its coverage, to align definitions with the international standards, to identify reasons for misreporting and ameliorate them, and to eliminate processing and compilation errors; and
5.4. Most steps in the process may have feedback to earlier steps, and hence the process may be iterative. That is, a later stage may identify errors in earlier processes, modifications to earlier assumptions and procedures, changes to earlier judgements, etc. Users, by analysing and commenting on data, may influence many aspects of future data collection and estimation. It is important to recognise the requirement for judgement in all stages of the compilation process.
5.5. The main requirement is that statistical output be of good quality, in that the statistics provide a reasonable measure of the real world economic events to which they relate. Quality includes aspects such as: definitions consistent with the conceptual framework, accuracy, freedom from major revisions, timeliness, frequency, and level of detail provided. Often these aspects conflict (e.g. accuracy and timeliness), so that the statistician’s skill is in achieving an optimal mix appropriate to meet the needs of economic analysts and policy makers for timely, reliable statistics as the basis for sound decisions. In practice, the resources available to the statistician are invariably limited, and the statistician needs to devise a data collection, estimation, compilation and dissemination strategy to produce an optimal result overall. The factors and judgments in choosing such a strategy will vary from country to country.
5.6. The choice of data sources for a standard component may depend on whether a suitable data source already exists or a specially designed data source is required. In many instances there are data sources which serve a broad statistical or administrative purpose and are also suitable, after modelling, for balance of payments and international investment position statistics. Australian examples include international trade statistics, overseas arrivals and departures, official reserves, and certain Commonwealth Government transactions. Also, the ABS is able to draw on surveys carried out by, or on behalf of, other government agencies which collect data for research and policy purposes. Examples include the International Visitor Survey and the Survey of International Students. While these administrative and external survey data sources may serve a number of purposes, the ABS takes active steps to ensure that their quality is suitable for the balance of payments and international investment position statistics. Where necessary, the ABS may make further refinements to meet its needs.
5.7. However, existing data sources do not cover all statistical requirements, and a number of specially designed collections are necessary. Two historically important collections were established in 1948. The first, the Survey of Companies with Overseas Affiliates, was introduced by the then Commonwealth Bureau of Census and Statistics to measure foreign investment in Australia. The second was the Tickets System introduced by the then Commonwealth Bank to provide measures of selected services, income, transfers and financial transactions. Surveys of international investment have been enhanced over the years and, with the lessening of the relevance and quality of the Tickets System data in a deregulated financial environment, a range of surveys collecting components of international trade in services has been progressively introduced. The resulting Survey of International Investment and Survey of International Trade in Services (both surveys of businesses) are now central to Australian balance of payments, trade in services and international investment position statistics.
5.8. In undertaking these and other statistical surveys of businesses, the ABS considers it necessary that:
- improvement to the timeliness for the source data and, in cases where they cannot be available before publication, development of extrapolation methods of high quality.
- the population of businesses with international trade in services and international investment activity is identified and recorded - many sources of data on such businesses are examined frequently and newly identified businesses surveyed using exploratory forms. The Survey of International Investment also includes other resident entities such as the Reserve Bank, general government entities and persons (although the latter are rare) who have international investment activity;
- the statistical unit (the business unit about which data are reported - it may be different from the business which actually reports the data), particularly for complex business organisations is uniquely defined to avoid omission or duplication in data reporting. In the Survey of International Investment the statistical unit is the enterprise, while in the Survey of International Trade in Services the statistical unit is the management unit. These units are defined in box 5.1;
- survey forms are designed, tested and evaluated to ensure that data requirements are available from business records, and that survey forms are capable of completion and do not place an unnecessary reporting burden on business. There also has to be a regular program of review of the forms to ensure that they remain relevant;
- the collection method is effective and efficient. Here the choice may lie between the use of census, partial coverage and sample survey techniques (discussed below);
- close contact is maintained with business to ensure that forms are completed with sufficient accuracy and on a timely basis, response rates are high (usually in the high 90 per cent range) and unusual transactions are clearly understood and appropriately recorded; and
5.9. Apart from business surveys, the ABS has found it necessary, in the compilation of good quality balance of payments statistics, to conduct appropriate surveys of persons, such as the occasional Survey of Returned Australian Travellers. The same underlying principles are applicable. These include properly defining and identifying returning Australian travellers, good form design and testing, appropriate sample estimation techniques, data analysis and validation, etc.
- survey results are properly tabulated, analysed, validated, disseminated and appropriately explained. This may include the estimation of some components to take account of partial coverage, non-response, reporting errors and other known factors.
|5.1 STATISTICAL UNITS IN ABS SURVEYS OF BUSINESSES: DEFINITIONS |
|The enterprise group comprises one or more legal entities under common ownership and/or control. It covers all operations in Australia of legal entities which are related in terms of the current corporations law. |
A legal entity is an entity which possesses some or all of the rights and obligations of individual persons or corporations. Examples of legal entities include companies, partnerships, trusts, sole proprietorships, government departments and statutory authorities. A legal entity is classified to an institutional sub-sector according to the kind of activity undertaken: non-financial corporations, financial corporations, general government, households, non-profit institutions serving households, and rest of the world. The financial corporations sector is split into sub-sectors which reflect the distinctly differing activities undertaken within the financial sector. These include the central bank, banks, other depository corporations, insurance corporations and pension funds, other intermediaries and financial auxiliaries. These institutional sectors are described in the Standard Institutional Sector Classification of Australia (SISCA).
The enterprise comprises all legal entities (within the enterprise group) which are classified to a single institutional sub-sector. Thus an enterprise group which contains legal entities that belong to more than one institutional sub-sector is split up into two or more mutually exclusive enterprises. The enterprise unit is used in ABS financial surveys, including the Survey of International Investment.
The management unit is the highest level unit within an enterprise group which controls its productive activities and for which a set of management accounts is maintained. In most cases it coincides with the legal entity owning the business. However, in the case of large diversified businesses there are often a number of management units, each coinciding with a division or line of business, which undertake different activities and for which separate and comprehensive accounts are maintained. Therefore, for large diversified businesses, the management unit may be a lower level of unit than the enterprise. The management unit is commonly used in ABS surveys to collect industry or production data, and is the unit used in the Survey of International Trade in Services.
The establishment is the lowest level unit within a management unit. It consists of one or more locations, within a State or Territory of Australia, which controls its productive activities and for which a specified range of detailed data is maintained. The establishment may coincide with the management unit, although in large business structures the establishment may be a lower level of unit than the management unit. While the establishment is used in some ABS collections, it is not used in collections for balance of payments purposes.
5.10. ABS business surveys can take the form of a census, a partial coverage collection or a sample survey:
- a census is a collection of information from all units in the population or a ‘complete enumeration’ of the population. Censuses are used to identify and update information on the known or expected population of businesses with balance of payments or international investment activity. These are referred to as benchmark censuses;
- a partial coverage collection includes units with known relevant activity above certain thresholds. This is used to ensure that the collection covers the big (and volatile) units, which often account for more than 90 per cent of the total activity. Those with known activity below the thresholds are not directly surveyed, but are included in estimated results based on benchmark censuses or other regularly available indicators. A major disadvantage of partial coverage collections is that the activity of the smaller units may be changing quite differently to the activity of big units and may differ from what was recorded in a benchmark census. In the absence of alternative indicator information, benchmark censuses can help in ensuring that all or most units which meet the threshold are identified; and
5.11. The ABS uses occasional benchmark censuses to identify businesses with known or expected international trade in selected services or international investment activity. These censuses provide data for the design of partial coverage or sample collections and, in the case of some partial coverage collections, data for estimating the contribution of statistical units below the threshold. The choice of sample or partial coverage depends on a number of factors, including accuracy, provider load and cost. Sample surveys have been used for some components of the Survey of International Trade in Services and in some components of the Survey of International Investment.
- a sample survey approaches only part of the total population. However, all population members are eligible for selection (unlike in a partial coverage collection). Statistical units are selected according to rigorous sampling procedures, and the data they supply are ‘expanded’ or ‘weighted’ to make inferences about the whole population, either as totals or as averages. As sample surveys are less expensive than censuses, they can be conducted at more frequent intervals. A disadvantage of a sample survey, compared with a census, is the presence of sample error, arising from the fact that information is not known about all units. A disadvantage of sample surveys, compared with partial coverage collections, is that limited resources which could be used to closely analyse very significant and volatile units may be spread more thinly to measure the stable activity of insignificant units.
5.12. An important part of the compilation strategy is the use of estimation methods, which may be described as simple estimation procedures, extrapolation and interpolation techniques, and data models. Simple estimation may involve relatively simple formulae or procedures which may be used to adjust or estimate a source series. For example, international merchandise trade statistics measure the administrative registration of the physical movement of goods across Australia’s customs frontier using administratively determined valuations. Because of a clear understanding of the nature of these administrative by-product statistics, and a close attention to the processes that are followed in their compilation, it is possible to program fairly simple adjustments to re-estimate the data to take account of some of the balance of payments requirements. Therefore, customs valuations become appropriate ‘free on board’ valuations (the appropriate point of valuation for balance of payments purposes), processing variations in the administrative source are removed to better approximate the time of change of ownership, and the coverage of international trade statistics is adjusted to capture those large transactions that do not involve physical movement across a customs frontier.
5.13. Extrapolation is another commonly used method, which involves projecting a particular series forward, taking account of such factors as related statistical indicators, historical developments in the series, the latest available data sources, seasonal influences, appropriate economic and social issues and any other factor that may be considered relevant. Extrapolation is a common method used in data models. For example, where a data source is only available periodically, its results may need to be extrapolated using suitable indicator series. Another common example of extrapolation is to estimate data for the latest period or two until results from a particular data source are available.
5.14. Similarly, when a data source or data model provides data on a less frequent basis than the periodicity of compilation, it is necessary to interpolate data between measurement periods to obtain sufficiently frequent estimates. For example, it may be necessary to interpolate annual data to quarters, and (for some business services) from quarters to months, using the most suitable method.
5.15. The most complex method of estimation is the use of complex data models, which involve bringing together data from two or more sources and making explicit assumptions about their relationship. The distinction between a complex data model and other forms of estimation (simple estimation, extrapolation and interpolation, all of which are data models of varying complexity) depends subjectively on the perceived complexity of the estimation process used.
5.16. The core of the compilation process is the extraction and transfer of data:
1. from a data source to the appropriate standard component or data model; and
Data from sources need to undergo a degree of manipulation (data modelling), either within the data source or after compilation of the source estimates, before transfer to a standard component.
2. from a data model to the appropriate standard component.