Cardiff - Wales 16 – 21 October 2005
Session No 7 – Implementing Classification Changes
Robin Slater & Dennis Robson, Australian Bureau of Statistics, Australia
Implementation of a New Industry Classification on the ABS Business Register
Introduction
1.The ABS and Statistics New Zealand have developed a new classification of industrial activity, the Australian and New Zealand Standard Industrial Classification (ANZSIC) 2006, which will replace the existing classification, ANZSIC 1993. This paper describes the changes to the classification at a very broad level, provides an overview of the strategy for implementing the new classification into ABS statistical series, and provides some detail of how the new classification is being incorporated into the ABS Business Register. This paper also reports on progress to date.
ANZSIC 2006
2.ANZSIC 2006 was developed to provide a more contemporary industrial classification system. Issues such as changes to the structure and composition of the economy, changing user demands and comparability with other major industrial classification standards were taken into account. The conceptual framework adopted for the development of ANZSIC 2006 used supply-side based industry definitions and groupings. More detail about the rationale and objectives for the development of the new classification, conceptual framework and principles behind the redeveloped classification, and significant changes to the classification can be found in the information paper ANZSIC 2006 Development (ABS Catalogue no. 1294.0), available on the ABS website (
3.One of the impacts of the redevelopment of ANZSIC was to increase the number of industries at each level of the hierarchy within the classification. The ANZSIC 2006 division structure includes 19 divisions compared with 17 in ANZSIC 1993. The divisions for ANZSIC 1993 and ANZSIC 2006 are shown in Attachment 1. The number of ANZSIC subdivisions increased from 53 in 1993 to 86 in 2006, which will lead to improved comparability between ANZSIC and other major international industrial classifications. Units on the ABS Business Register are classified at the 4-digit industry class level. There was a direct correspondence between the 1993 and 2006 ANZSIC for close to two thirds of these industry classes.
Overall Implementation Strategy
4.Implementing a new industry classification is one of the most significant change programs undertaken by the ABS. The last major change in industrial classification (from ASIC to ANZSIC) occurred in 1993 and the transition was inconsistent and in some cases inadequate. It involved high level workloads in concentrated periods for some areas, especially late in the process, which actually delayed the release of Australia's National Accounts. This particularly affected those areas with a broader obligation and perspective such as the Business Register, National Accounts and Methodology. Consequently, for the current ANZSIC 2006 implementation, the ABS has adopted a far more strategic approach to planning, within a formalised Project Management Framework.
5. Primary lessons learned from the 1993 ANZSIC implementation were the importance of providing explicit and adequate funding for implementation, the need to get early buy-in and accountability of all stakeholder areas, and the critical importance of centralised coordination and management of implementation. In response to these lessons, the ABS has ensured that the change from ANZSIC 1993 to ANZSIC 2006 was more comprehensively planned, properly funded, and had better governance arrangements in place than the previous exercise. An ANZSIC Implementation Coordination Program Team, comprising representatives of all major stakeholder areas, was established to manage the implementation of ANZSIC 2006 and coordinate the activities of affected areas. Implementation plans, progress, risks and issues are reported regularly to a high level project board, comprising senior managers from the ABS Economic Statistics Group. A list of key milestones is shown in Attachment 2. In addition to these governance arrangements, an independent Quality Adviser was appointed to provide guidance on the program process and the management of the program.
6.As detailed later, staff in the Australian Tax Office (ATO) have a critical role in the implementation of ANZSIC 2006. The work of the ATO is being overseen by a high level ATO/ABS ANZSIC 2006 Implementation Steering Committee, comprising senior managers from both agencies. The ABS has a long and close relationship with the ATO, which has provided a solid basis for collaboration on all aspects of introducing ANZSIC 2006, including planning, funding, training, and implementation.
7. On the funding side, funding for both the ABS and ATO were supplemented as part of a New Policy Proposal which was accepted by the Australian Government. The ABS was provided with extra funding of approximately $AUD 11.4 million for ANZSIC 2006 implementation, of which $AUD 1.5 million was allocated to the Business Register Unit (BRU). As part of the same proposal, the ATO was provided with an extra $AUD 7.8 million which covered staff, IT, accommodation and other infrastructure costs for ANZSIC 2006 implementation.
Statistical Strategy
8.The statistical strategy for ANZSIC 2006 Implementation has been devised taking account of two critically important elements. Firstly, establishing the requirements (including timing aspects) of implementation within the National Accounts, and secondly, a determination not to release indicator series on a different conceptual basis than the National Accounts.
9.The ABS has determined that late 2009 is the optimal point for the National Accounts to be released on the revised ANZSIC 2006 basis. This timing permits all areas to fulfil current obligations such as the Tax Reform Stream 2 changes for National Accounts and BSIP-related changes, and ensures that all of the required changes for a successful transition have occurred.
10.The statistical strategy involves;
the ABS Business Register carrying two ANZSIC codes from Dec 2005 to June 2009,
full survey redesigns to ANZSIC 2006 basis from July month and September quarter 2009 reference periods for sub-annuals; 2006-2007 reference year for annuals. Publication on an ANZSIC 06 basis will only occur following the redesign activity.
annual collections producing 3 years of 'extended overlapping bridging' data on the new basis,
sub-annual collections will follow a similar strategy to annuals with regard to producing 'extended overlapping bridging' data on the new basis. In this case, however, the period of increased sample sizes will relate to the 5 quarters up to, and including, the June 2009 reference quarter.
for the period between July 2006 and June 2009, all irregular collections such as the range of service industries collections, will need to be dual coded. Publication relating to these collections will include statistics on both ANZSIC 1993 and ANZSIC 2006 bases.
Business Register Strategy
11.The ABS Business Register comprises two populations, referred to respectively as the ABS Maintained Population (ABSMP) and the ATO Maintained Population (ATOMP).
12.The ABSMP covers all businesses with complex business structures, the majority of government units, and the majority of businesses with significant employment in more than one State for which State data are required. The ABS conducts a range of ongoing maintenance activities for businesses in this population, including personal and mail-based profiling, to ensure that statistical structures and classificatory variables such as ANZSIC are kept up-to-date.
13.The ATOMP covers all other businesses. The primary source for maintaining information about these simple units is a monthly snapshot from the Australian Business Register (ABR) which is administered by the Australian Tax Office (ATO). Industry codes are determined by ATO staff for all new ABR registrants. The quality of this coding work is assessed each month by the ABS, with accuracy targets of 90% at the 4-digit ANZSIC class level for employing units and 80% at the 2-digit ANZSIC subdivision level for non-employing units.
14.To realise broader ANZSIC implementation plans, all units on the ABS Business Register are required to hold both 1993 ANZSIC and 2006 ANZSIC from December 2005 until June 2009. For about two thirds of industries, there is a direct 1:1 or many:1 correspondence between the 1993 and 2006 versions of ANZSIC. Coding of units in these industry classes is relatively straightforward, as 2006 ANZSIC codes can be derived automatically from their 1993 ANZSIC. Generally, remaining units will require some level of clerical intervention to obtain 2006 ANZSIC codes.
15.As there are separate maintenance strategies for the two component populations, the strategies for implementing ANZSIC 2006 are described separately.
Classifying units in the ABSMP
16.Within the ABSMP, the primary unit for statistical reporting is the Type of Activity Unit (TAU). The composition of the ABSMP in relation to 2006 ANZSIC coding is shown in Table 1.
1:1 or many:1 / 1:many (clerical recoding required) / Total
Non-profiled / 3,400 / 2,900 / 6,300 (36%
Profiled / 7,900 / 3,200 / 11,100 (64%
Total / 11,300 (65%) / 6,100 (35%) / 17,400
17.For TAUs with a 1:1 or many:1 relationship, the 2006 ANZSIC can be derived directly from their 1993 ANZSIC, so no clerical recoding is required. Table 1 shows that 11,300 (65%) TAUs in the ABSMP were in this category. The remaining 6,100 (35%) required clerical intervention to determine a correct 2006 ANZSIC. The primary means of maintaining units in the ABSMP is through profiling activity, either by mail or personal visits, however, not all units are profiled each year. Table 1 shows that of the 6,100 TAUs requiring clerical intervention, 3,200 were in enterprise groups to be profiled as part of regular maintenance activity. For the remaining 2,900, a units survey was conducted between February and June 2005, with the primary objective of determining a correct 2006 ANZSIC. The opportunity was also taken to update employment and enterprise group structures for these units.
18.As at mid September, 83% of all TAUs in the ABSMP had been coded to a 2006 ANZSIC. 86% of TAUs with a 1 (or many):1 correspondence had been recoded and 77% of TAUs with a 1:many correspondence had been recoded. Based on this progress, the BRU is well on track to meet the end-December deadline.
19.Following completion of recoding the stock of ABSMP units, profilers will maintain both 1993 and 2006 ANZSIC codes for all units as part of regular maintenance, for a period of approximately three years.
Classifying the ATOMP
20.The most significant component of work for the ATOMP is to determine 2006 ANZSIC codes for the stock of units on the ABR. This work, which is being undertaken by ATO staff, commenced in June 2005, and is scheduled for completion by December 2005. Table 2 shows the approximate number of units in the ATOMP which need to be recoded by December 2005. Units in this population are defined as Australian Business Number (ABN) registrants.
Table 2: Number of units in the ATOMP by 2006 ANZSIC coding category1:1 or many:1 / 1:many (recoding required) / Total
employing units / 618,000 / 522,000 / 1,140,000
non-employing units / 116,400 / 955,000 / 2,120,000
Total / 1,782,000 (55%) / 1,477,000 (45%) / 3,260,000
21.Table 2 shows that 2006 ANZSIC codes can be derived automatically for 1.78 million (55%) of the 3.26 million units on the ABR. A phased approach is being used to determine 2006 ANZSIC codes for the remaining 1.48 million units. An initial step is to utilise an auto-coding tool which was developed by the ABS to derive a 2006 ANZSIC code. This tool, which uses the description of activity provided by registrants on their original registration forms, is able to derive a 2006 ANZSIC codes for about 35% of units. Remaining units have to be clerically coded by ATO staff.
22.At this phase of industry coding, different procedures are applied for coding employing and non-employing units, mainly due to the differing levels of required accuracy for these two sub-populations (90% at the ANZSIC class level for employing units and 80% at the ANZSIC subdivision level for non-employing units). For employing units, trained ATO staff code units based on activity descriptions and other relevant information held by the ATO. Where necessary, registrants are contacted if there is insufficient information available to accurately derive a 2006 ANZSIC code. For non-employing units, ATO staff determine 2006 codes based on information available on the registration form. These non-employing units are not contacted during this phase.
23.In order to undertake this recoding work, the ATO recruited about 120 temporary staff for a period of seven months. An ABS officer is outposted to the ATO for the duration of this recoding work, to provide on-site training, support and advice for these ATO staff. As at the end of August 2005, 64% of all ATOMP units had been updated with a 2006 ANZSIC. 82% of ATOMP units subject to automated recoding (ie. 1:1 or many:1 correspondences) were recoded and 44% of units subject to clerical recoding (ie. 1:many correspondence) were recoded. For employing units, 74% had been recoded, with 96% of automated recodes completed and 48% of clerical recodes completed. For non-employing units, 61% had been recoded, with 77% of automated recodes completed and 43% of clerical recodes completed. Based on these results, it is pleasing to note that the ATO is ahead of schedule with recoding work. This means that
in addition to ongoing quality assurance and repair, the ATO will be able to undertake a global review and repair process before the end-December deadline.
24.The ABR is held on a computer system in the ATO which cannot hold both 1993 and 2006 ANZSIC codes. A 'snapshot' has been taken of the ABR and the information has been stored in an off-line environment by the ATO. Recoding work is being done in this off-line environment. Regular files are provided to the ABS, where the Business Register system (Inteframe) has been modified to store both 1993 and 2006 ANZSIC codes. To date, these files are being used solely for quality assurance purposes. Loading of 2006 ANZSIC codes to Inteframe will commence in November 2005.
25.Until the ATO operating system is converted to an ANZSIC 2006 basis, any new ABN registrations made after the snapshot was taken will continue to be classified on an ANZSIC 1993 basis. These new registrations will be copied across to the off-line environment and will be reclassified to a 2006 ANZSIC. Recoding of the off-line snapshot is expected to be completed by the end of 2005. During the first half of 2006, the ABR computer system will have been updated to cater for the new classification.
26.From early May 2006, all new registrations will be classified by the ATO using only ANZSIC 2006. To support the statistical strategies for economic collections outlined above, the ABS business register will need to carry both 1993 and 2006 ANZSIC codes from early 2006 until June 2009. This will require ABS staff to 'reverse' industry code new units to obtain 1993 ANZSIC codes. Where possible, these units will be coded using a series of auto-coding modules developed by the ABS. These modules are expected to provide 1993 ANZSIC codes for about 35% of new registrants. The remainder will be coded by temporary ABS staff, recruited specifically to undertake this work. This 'reverse' coding work will continue until June 2009.
Quality assurance strategy
27.A comprehensive quality assurance strategy was developed to measure the quality of both automated and manual processes. A summary of the various components of this strategy is shown in Attachment 3. Work to recode both the ABSMP and ATOMP, which is being undertaken by the ABS and ATO respectively is being separately quality assured by the ABS.
Concluding comments
28.Implementation of ANZSIC 2006 is one of the most significant statistical change programs undertaken by the Economic Statistics Group. Recoding the ABS Business Register is of critical importance to this work, as a major lesson learned during the previous industry classification change was that if the Business Register conversion at the front end was not successful then the downstream costs at the output end were multiplied.
29.As reported above, conversion of the ABS Business Register is well on track for completion by the end of December 2005. This will provide a solid platform as statistical collection areas implement their strategies for conversion, with results ultimately feeding into Australia's National Accounts. Significant factors in achieving this outcome have been the coordination and governance provided by the Economic Standards area in the ABS, the level of funding provided to undertake the work, active management of risks, and the cooperation of the ATO in planning and undertaking this work for the majority of units.
30.An unanticipated benefit of the ANZSIC 2006 implementation program is that it has strengthened the relationship between the ABS and the ATO. The ATO has expressed a high degree of satisfaction with the support being provided to them and this has generated a large amount of goodwill. The support activities themselves, and the strengthened relationship, have set us up well for progressing ABS strategic priorities regarding ATO's ABR. The degree of cooperation between the two agencies on this program has benefited other ABS/ATO work such as that relating to the politically sensitive issue of the definition of a business.
Glossary of Terms
ABNAustralian Business Number
ABRAustralian Business Register (maintained by ATO)
ABSAustralian Bureau of Statistics