|Page tools: Print Page RSS Search this Product|
5: Increase the quantity of output
Performance Measure 5.2: Innovative outputs
In addition to the new statistics released in the publications listed in Performance Measure 5.1, the ABS has conducted research which has led to the release of innovative new estimates, classifications and publications. Also released were a number of information papers informing users about conceptual issues, new methodologies and pending changes to current ABS collections.
Some of the innovative outputs released by the ABS in 2002-03 included:
6: Improve the quality of outputs
Performance Measure 6.1: Achieve or exceed timeliness, statistical reliability, response rates and accuracy objectives through:
The timeliness of ongoing series is measured by the gap between the reference period and the date of publication of results. The ABS continues to adhere to preannounced release dates and make improvements, where possible, to the timeliness achieved. Table 8 presents information on timeliness for ABS monthly and quarterly publications for Main Economic Indicator (MEI) statistics, and other general releases.
The average number of days elapsed between the end of the reference period and the release of data has improved for both MEI publications and other releases in 2002-03. The timeliness of all categories of releases, except monthly MEI publications, increased last year due to the consolidation of several publications to provide a more comprehensive and coherent picture of business economic activity. The publications that were discontinued had a very short lead time, and their exclusion led to an apparent increase in the average time between the end of the reference period and the release of data.
Table 8: TIME BETWEEN END OF REFERENCE PERIOD AND RELEASE OF DATA
(average number of elapsed days)
One measurable component of reliability is revisions to data. Revisions are generally measured by their size and frequency over time. The ABS aims to minimise revisions as much as possible through effective sample and methodological design. It is also ABS policy to inform users of any significant revisions and where appropriate to revise past time series and advise users accordingly.
A new sample for labour force statistics, based on updated geographic information from the 2001 Census, was implemented progressively from November 2002 to June 2003. Although results were subject to some initial volatility, common sample estimates were produced to assist analysts during the changeover period. Labour force estimates have since settled down.
The table below describes the revisions to quarterly Gross Domestic Product (GDP) over the past several years. In particular, it shows the difference between the first estimate of GDP, and that estimate one year later in terms of the mean revision and the mean absolute revision, expressed as percentage points. The figures show that revisions to quarterly GDP in recent years remain relatively small (mean absolute revision) and largely offsetting (mean revision).
Table 9: REVISIONS TO GROSS DOMESTIC PRODUCT, PERCENTAGE CHANGE(a)
(b) Figures based on three quarters of GDP data.
Table 10 shows the revisions to the quarterly current account deficit averaged over the financial year for 1997-98 to 2001-02. Similar to the GDP table, the mean revision shows the percentage difference between the first estimate of the current account deficit, and that estimate one year later, averaged over the four quarters for the year. The mean absolute revision shows the average absolute values of the mean revision. The revisions to the current account deficit are expressed in percentage terms however, rather than percentage points as is the case with the revisions to GDP.
The figures show that revisions to the quarterly current account deficit are typically around 4-5% which is not considered significant. Comparisons between the mean revision and the mean absolute revision indicate that there has been in recent years a degree of offsetting revisions to the current account deficit figures.
Table 10: REVISIONS TO CURRENT ACCOUNT DEFICIT, LEVEL ESTIMATES(a)
(b) Figures based on three quarters of data.
The ABS has consistently had very high response rates. Past international benchmarking studies have shown these response rates compare favourably with other international statistical organisations.
Since response rates vary little over time, especially for sub-annual surveys, it is more appropriate to report against target response rates rather than changes in the rates from year to year. Table 11 shows that response rates for selected economic collections either achieve or exceed the target response rates set by the ABS. Response rates for both the Australian Industry survey and the Manufacturing survey conducted in 2002-03 are not yet final. However, past experience has shown that these figures will rise when the surveys are finalised.
It is important to note that in regard to the business surveys, follow up procedures tend to focus on the more significant businesses, that is, those with typically high sales or employment relative to the rest of the industry. For example, the response rate for businesses in the manufacturing survey might be 88%, but the businesses that have responded may comprise 96% of the total employment in the industry.
Table 11: ABS RESPONSE RATES FOR SELECTED ECONOMIC INDICATORS
(b) Annual surveys. The response rates refer to surveys conducted in the year listed, but for which the reference period is one year prior.
(c) The response rate for this survey is based on the Economic Activity Survey direct collection as the tax component does not have response rate. The publication is expected to be released in December 2003.
(d) Response rates were not finalised as at the end of June 2003.
The response rates for selected social surveys are listed in table 12 below. The social surveys listed generally have high response rates, apart from the latest National Crime and Safety Survey which has a relatively low, although acceptable, response rate.
Table 12: ABS RESPONSE RATES FOR SELECTED SOCIAL SURVEYS
(b) The Child care Survey was previously conducted in 1999, while the National Crime and Safety Survey was previously conducted in 1997-98. The General Social Survey and Indigenous Social Survey were first conducted in 2002-03.
(c) The General Social Survey, Australia (cat. no. 4159.0) publication is expected to be released in September 2003, while the Indigenous Social Survey (cat. no. 4714.0) publication is expected to be released in December 2003.
While all ABS outputs maintain high levels of accuracy in all tables, graphs and text, two types of error are possible in estimate based sample surveys: sampling error and non-sampling error. Sampling error occurs because a sample rather than the entire population is surveyed. One measure of the likely difference resulting from not including all units in the population in the survey is given by the standard error. It is ABS policy that standard errors are included in survey publications, along with the descriptions of other types of errors to which outputs may be subject. Non-sampling error arises from inaccuracies in collecting, recording and processing the data and can occur in any statistical collection. The ABS ensures non-sampling error is minimised by careful design of questionnaires, intensive training and supervision of interviewers, and efficient data processing and editing procedures.
The following example illustrates the use of the standard error in quantifying sampling error for unemployed persons in the labour force survey. The published original level estimate for unemployed persons in Australia in June 2003 was 620,000. The calculated standard error for this estimate was 9,300. The standard error is then used to interpret the level estimate. For example:
The magnitude of standard errors varies between collections due to factors such as the sample size and the value of the estimate in question. Therefore, it is impossible to compare standard errors between different surveys, or even between variables within the same survey. The relative standard error, obtained by expressing the standard error as a percentage of the estimate to which it refers, is a much more useful measure in that it provides an immediate indication of the percentage errors likely to have occurred due to sampling, and thus avoids the need to refer also to the size of the estimate. The relative standard error is also a good indicator for comparing the accuracy of estimates between surveys.
The tables below only present a summary view of the accuracy for key aggregates from a number of major ABS publications, as expressed by their relative standard errors. The low relative standard errors in both tables highlight the accuracy of ABS’ statistical collections. More details are available from the publications, or the concepts, sources and methods publications associated with the collections themselves.
Table 13: RANGE OF RELATIVE STANDARD ERRORS FOR SELECTED ECONOMIC INDICATORS(a)
(b) The relative standard errors refer to surveys conducted in the year listed, but for which the reference period is one year prior.
(c) The publication is expected to be released in December 2003.
Table 14: RANGE OF RELATIVE STANDARD ERRORS FOR SELECTED SOCIAL SURVEYS
Performance Measure 6.2: Conduct ongoing research and reviews of quality, and implement their recommendations:
Outline of ABS statistical reviews
The ABS reviews its statistical collections regularly to ensure that its statistics are of good quality. Some reviews cover all aspects of a particular collection, from user requirements, through to data collection, processing, analysis and dissemination. Other reviews focus on particular elements of collections, such as the methodology.
In reviews of statistical collections, external users are widely consulted and, in some instances, external users assist the review team. Internal reviews cover both the effectiveness and efficiency of various ABS activities.
A number of reviews of statistical collections and programs commenced, or were completed, in 2002-03, including:
The household income review found that the quality of data reported had been impacted on by a deterioration in the coverage of benefit recipients in the sample estimates and by an increasing complexity of the government benefits scheme itself. As a result changes were made to the way income distribution data are presented and revisions to earlier periods data were published in the June 2003 Australian Economic Indicators (cat. no. 1350.0). A redeveloped Household Income and Expenditure survey will collect some new information on assets and liabilities to both improve the quality of income distribution statistics and provide for analysis of the household distribution of wealth.
The review of the preliminary 1996-2001 intercensal discrepancy led to the implementation of revisions to final estimated resident populations for 30 June 2001.
The outcome of a review of the Producer Price Index ‘stage of production’ indexes was an updated index structure consistent with the 1996-97 input-output tables and an improvement in the coverage of prices within the indexes. During 2002 a program of quality review also commenced for a number of service industry producer price indexes resulting in a number of improvements for the indexes.
The ABS completed a review of the experimental composite leading indicators of the Australian business cycle in 2002-03 and, as a result have decided to terminate the publication of the indicators from September 2003. The review suggested that while the composite leading indicators performed adequately in the early 1990s, they have lost their ability to provide consistent early signals for turning points of the reference business cycles.
Outcomes and recommendations from reviews are often not implemented in the year that the review was conducted. Two of the improvements to the statistical collections in 2002-03 stemming from quality reviews conducted in recent years were:
Innovative practices - improvements to existing collections as a result of research and development
Apart from the ongoing reviews of ABS statistical collections, research and development in subject matter areas continues to result in innovative practices for the collection and compilation of data. The results of innovative practices is twofold. Firstly, it will lead to more reliable and accurate statistics. And secondly, it may lead to reduced provider load.
The way in which the ABS processes its business statistics was significantly changed as a result of the implementation of the Business Statistics Innovation Program (BSIP) from July 2002. With the use of innovative technologies and methodologies the ABS set about improving the quality and relevance of its business statistics while at the same time increasing the efficiency of processes which benefit both the ABS and its providers. Other benefits of BSIP include the potential to respond to emerging statistical demand as a result of increased capacity and, the ability to provide a better national statistical service through strong statistical leadership.
The ABS planned and conducted two major field tests of computer assisted interviewing (CAI) for the monthly population survey during 2002-03. The tests were designed to fully test field and office systems, procedures, training, hardware and occupational health and safety. These tests proved successful against set objectives and approval (subject to successful final system test) was given to phase in CAI for the monthly population survey between October 2003 and May 2004. CAI is expected to deliver substantial improvements in data quality, survey efficiency and management arrangements.
Following a review conducted in 2001-02 to identify a range of statistical collections which could benefit from the incorporation of taxation data, the ABS implemented strategies for a number of surveys (such as mining and manufacturing), taking advantage of tax and other administrative data to supplement or replace existing data. The use of administrative data has not only reduced collection costs and minimised the reporting load on businesses, it has allowed the scope of some surveys to increase to cover non-employing businesses. Previously, a number of surveys were limited to employing businesses only.
7: Achievement of cost effective outputs
Performance Measure 7.1: Conduct efficiency reviews and audits, and implement their recommendations
Efficiency reviews and audits may be initiated by senior management and by the ABS Audit Committee so as to assess whether resources are being used effectively and efficiently to achieve the ABS’ objectives.
Major reviews addressing efficiency issues were completed during 2002-03 in the following areas:
These reviews resulted in:
A number of other efficiency reviews are in progress for a range of statistical collections and corporate service functions. These reviews aim to achieve efficiency gains via improvements to collection strategies, methodologies, use of information technology, use of contractors, and/or substituting/complementing existing data with administrative data sources.
Performance Measure 7.2: Test operating efficiencies of statistical activities by benchmarking internally and externally
Benchmarking is a key part of the ABS strategy to assess the value for money of its statistical and non-statistical outputs, to understand and learn from best practice and to improve performance. The ABS views the process of benchmarking as an ongoing exercise, enabling the organisation to achieve continuous improvement across a variety of its outputs.
Benchmarking currently being undertaken at the ABS includes comparisons between ABS statistical collections; comparisons between the operations of other international statistical agencies and the ABS; and comparisons of corporate service functions between other Australian government agencies and the ABS. These are outlined below.
Comparisons between ABS statistical collections
During 2002-03 internal benchmarking was conducted for all areas in the Economic Statistics Group (ESG). The benchmarking exercise was conducted in respect of 2001-02 and concentrated solely on the financial framework due to the organisational changes stemming from the BSIP. The aim of this exercise was to identify areas of efficiency through cost savings and modifications in the cost structure for each area in the economic statistics group in order to provide the efficiencies necessary to meet ongoing resource constraints.
As part of the financial framework reporting, areas indicated how salary and technology application costs were split across the BSIP organisational structure. Future internal benchmarking of collections in the Economic Statistics Group will be assisted by the requirement for all collections to report basic financial and performance information on an annual basis.
Comparisons between the operations of a number of international statistical agencies and the ABS
The ABS continues to benchmark its functions and services against a range of international agencies. The ABS has found that the most effective mechanism for benchmarking functions and services arises from bilateral discussions with particular agencies. Detailed comparisons of particular statistical or non-statistical work programs often result from such discussions, and following more detailed comparisons, the ABS reviews practices and methods as necessary with the aim to achieving the best outcomes and outputs possible from its resource allocation. Bilateral discussions with Statistics Canada and Statistics Sweden are planned during 2003-04.
International benchmarking activities undertaken in 2002-03 included comparisons of the monthly retail survey; corporate governance; and data and metadata available on the web site.
Comparisons of corporate services functions between other Australian government departments
The ABS continues to conduct benchmarking studies with other Australian government departments.
In 2002-03, the ABS has been involved in two benchmarking activities conducted by the Australian National Audit Office (ANAO). These were:
The ABS has provided input into these studies in the form of survey participation and by providing comments on drafts. The studies have been finalised and reports have been issued by the ANAO.
Performance Measure 7.3: Market test a number of non-statistical activities to identify possible outsourcing opportunities
The ABS has outsourced a number of key functions including:
The ABS will continue to investigate additional outsourcing opportunities as they arise.
Performance Measure 7.4: Minimise provider load
In order to fulfil its mission of providing information to support discussion, debate and decision making, the ABS is required and empowered to collect information from businesses. At the same time the ABS is conscious that the needs of users for information must be balanced against the load placed on businesses in providing that information.
The following table shows the provider load (measured in thousands of hours taken to complete statistical forms) imposed on businesses by the ABS for 1995-96 and from 1999-2000 to 2002-03. These figures have been adjusted to ‘smooth’ the series to account for large irregular surveys and censuses (e.g. the Agricultural census).
Table 15: PROVIDER LOAD IMPOSED ON BUSINESSES BY THE ABS
The ABS continues to focus on reducing provider load on businesses. This is consistent with government policy, the ABS Corporate Plan and the recommendations of the Commonwealth Government’s 1996 Small Business Deregulation Task Force. The task force found that the ABS only accounted for about 1% of total business compliance costs. However since 1995-96 total ABS provider load on businesses decreased by around 40%, and for small businesses the fall has been almost 50%.
Two of the major initiatives which led to a continued reduction in provider load during 2002-03 were:
Although the ABS focuses on reducing provider load on businesses, it is also taking steps to reduce respondent burden on households and individuals. One significant initiative in this respect has been the move to a biennial Survey of Income and Housing Costs. The survey, previously conducted annually, will have a sample size of 11,000 households, an increase of 4,000 households. Despite the increase in households surveyed the total respondent burden over a two year period will decrease substantially.