Australian Bureau of Statistics
8168.0.55.001 - Microdata: Business Longitudinal Database, Australia, 2004-05 to 2009-10 Quality Declaration
Previous ISSUE Released at 11:30 AM (CANBERRA TIME) 05/12/2012 Reissue
|Page tools: Print Page Print All RSS Search this Product|
Together these two sub-populations (of ABN units and TAUs) make up the ABSBR population.
SCOPE AND COVERAGE
The scope of the BLD is restricted to the ATOMP, i.e. businesses with a simple structure. This facilitates the confidentialisation process as larger businesses may be more easily identified and avoids difficulties in dealing with matching ATO data for businesses with a complex structure. On this basis, the statistical unit for the BLD is the ABN unit. For the purposes of the BLD, the term business is interchangeable with the ABN unit.
The businesses that are included in the BLD are classified:
The scope of the BLD is actively trading businesses in the Australian economy. An actively trading business is one which is registered for an ABN and remits Goods and Services Tax (GST). The BLD includes both non-employing and employing businesses in the Australian economy except for:
For detailed information on ANZSIC 1993 classes included in the Food Industry component - see Appendix.
The sample design involves the use of panels that represent the Australian business population at the point in time that each panel is initiated into the BLD. Panel One is representative of the in-scope business population as at 30 June 2005. Panel Two is representative of the in-scope business population as at 30 June 2006. Each panel is directly surveyed once a year for a period of five years. A new panel is normally initiated each year.
The sample for each BLD panel is stratified by industry division and business size. Industry is based on ANZSIC 1993 division, and business size is based on a derived employment size indicator - Derived Size Benchmark (DSB). DSB is a derived item using ATO data which models employment and formed a part of stratification for all ABS business surveys at the time of the panels' selection. Once included in a panel, and irrespective of changes to business size or industry division, the selected business remains in the stratum for which it was originally selected. However, as employment is collected in the survey component, users will be able to observe businesses which have changed size ranges over time. As DSB is created at the time the business first appears on the ABSBR, elapsed time can lead to reported employment differing from DSB. Changes to industry division, however, are not recorded.
There are four business size ranges used:
State/territory is not included in stratification and no level of geography is available on the BLD CURF.
The intent of the BLD design is to maximise longitudinal, micro level comparisons across the five years during which each panel remains in the BLD. While the sample is stratified on the basis of business size and industry division, the sample is not allocated to enable the creation of population or cross-sectional estimates within any reasonable accuracy constraint. The major consideration in allocating the sample in the current BLD design is ensuring that enough sample is included to facilitate longitudinal analysis, i.e. that sufficient live sample remains in each stratum at the end of five years.
During the development of the BLD, consultation with users determined that the minimum number of businesses considered to be viable for longitudinal analysis was approximately 30 businesses per stratum still live at the end of five years. As part of BLD design, expected death rates for each panel over the five years were calculated for each individual stratum. To achieve the endpoint number of 30 live businesses per stratum, taking death rates into consideration, a starting point of 40 businesses per stratum, on average, was required. Therefore, the sample size for the entire BLD (excluding the Food Industry component), is approximately 2,000 businesses per panel.
Two factors have impacted on the intention of having approximately 30 live businesses per stratum in each panel at the end of five years, these are:
Food Industry Component
The Food Industry component is defined by ANZSIC 1993 classes predominantly associated with food for human consumption in the Agriculture, Forestry and Fishing; Manufacturing; and Wholesale Trade industry divisions. Consequently, there are "Food Industry" and "Non-Food Industry" flags for units in these three Divisions. The definition of the Food Industry component was determined through user consultation during the initial BLD development.
In addition to the core BLD sample, extra sample of approximately 1,000 businesses per panel has been included for the Food Industry (i.e. approximately a third of selections in each panel are Food Industry additional sample). The original Food Industry stratification was based on food industry class by employment size. As part of the confidentialising process, the original Food Industry stratification has been collapsed to food industry division in the BLD CURF. The Food Industry additional sample was allocated on a similar basis to that of the core sample, on average an equal number of selections per stratum. Unlike in the core BLD, resource limitations did not permit an initial allocation of approximately 40 businesses per stratum. Therefore, in the Food Industry component, the allocation was reduced to a maximum of 20 businesses in total for each class, i.e. for the combination of the four business employment size ranges.
POPULATION COUNTS AND PANEL SAMPLE SIZE
Panel One sample was selected from a survey frame created in June 2005 and includes 2,732 businesses. The Panel One frame contained 1,563,857 businesses. In this BLD CURF, Panel One contains its completed set of five reference periods of data (2004-05, 2005-06. 2006-07, 2007-08 and 2008-09).
Panel Two sample was selected from a survey frame created in June 2006 and includes 3,432 businesses. The Panel Two frame contained 1,336,515 businesses. In this BLD CURF, Panel Two also contains its completed set of five reference periods of data (2005-06. 2006-07, 2007-08, 2008-09 and 2009-10).
Design weights are provided for each panel separately.
In the BLD CURF, weights are based on the stratum (i.e. Industry division by business employment size) with the exception of units in Agriculture, forestry and fishing; Manufacturing; and Wholesale trade. For units in the Food Industry component of these three industries, weights are based on the original stratification (food industry class by business employment size). For the non-food industry units in these three industries, weight is based on "Non-food industry part of division by business employment size". Users should be aware that food industry class is not available on the BLD CURF as an indicative item.
Use of the BLD CURF to calculate population or cross-sectional estimates is not recommended as the BLD sample is not allocated to enable the creation of such estimates with reasonable accuracy, and as each BLD panel is selected independently (i.e. each is designed to represent the Australian business population at the time of initiation), combining panels is difficult.
Users are advised that combining panels for the purpose of linear modelling can be valid under certain circumstances. For example, it can be valid to perform linear modelling using data for businesses from both panels provided that users ensure:
The BLD is populated from administrative data provided to the ABS for statistical purposes by other government agencies and data directly collected via the ABS Business Characteristics Survey (BCS).
Administrative data included in the BLD are:
Business Characteristics Survey
The BCS is an annual ABS survey and is conducted via a mail-out/mail-back questionnaire. It is designed to collect characteristics data. It is intended that each year the survey will contain a consistent set of core questions to allow longitudinal analysis. There have been some changes to the core content of the survey in these early iterations as its purpose and scope has been refined.
In developing the BCS instrument, as with all ABS surveys, it has been necessary to achieve the appropriate balance between developing a survey which enables the collection of comprehensive, integrated data with the responsibility of managing the reporting load of businesses.
One of the key aims of the BLD is to enable analysis of business performance, including the capacity of businesses to undertake activities which lead to performance growth and the relative importance of these activities in driving that growth. To better meet this objective, the scope of the questions included in the 2004-05 BCS was refined in the 2005-06 survey. The 2005-06 survey reflected these changes through a greater focus on aspects of business characteristics or operations which may have an impact on business performance generally and, in particular, productivity. Therefore, some of the BCS data items available for the first year of Panel One (2004-05) were only collected once and are not available for the first year of Panel Two. Several replacement questions have been introduced in later years and relatively minor changes have been made to existing questions, resulting in differing time periods for these questions being available for analysis.
The BCS is also used as the survey vehicle for the collection of detailed ABS Business Use of Information Technology (BUIT) and Innovation outputs. Some of the BLD sample is also used in compiling these outputs. Detailed BUIT and Innovation data are collected in alternate years via additional questions relating to these topics included in the BCS. To meet this additional purpose, a second form type was introduced for the 2006-07 BCS.
Most of the items included in the BCS are categorical in nature (i.e. require a yes/no response).
The reference period for the data included in the BCS is the year ended 30 June or as at 30 June of the relevant year.
The volume of data included in the BLD is substantial and resource constraints mean it is not possible to quality assure each individual item. Users should take the following into consideration when using the BLD CURF.
Business Activity Statement
BAS data are supplied by the ATO to the ABS. Users should take into consideration that any discussion of the data limitations or weaknesses is in the context of using the data for statistical purposes (i.e. this BLD CURF), and is not related to the ability of the data to support the ATO's core operational requirements.
In some cases, BAS data may be partially missing for an individual business (for example, data may be available for three out of four quarters). No imputation is applied for missing data. The impact of this approach is somewhat mitigated by the timing of compiling these data. BAS data included in the BLD were compiled in July 2011, therefore, the time that has elapsed since the end of the latest included reference period (June 2010) has resulted in BAS records that are predominantly complete.
Any BAS data items which are either missing or have not been reported have been assigned the missing value code of 999999999 within the BLD CURF. Users should ensure that they do not inadvertently use the missing value code as a numeric value. This will ensure that these missing value codes do not incorrectly contribute to BLD CURF usage or outputs.
BLD CURF users should also be aware that there are specific reporting requirements for businesses with Pay As You Go withholding payments. For example, these requirements may impact on values for wages and salaries as shown in BAS data included in the BLD CURF. For this reason, missing values for this item should not necessarily be interpreted as nil payment of wages and salaries. The ABS is investigating how this can be rectified for future BLD CURF releases.
Business Characteristics Survey (BCS)
The approach to quality assurance for the BCS aims to make the best use of ABS resources to meet user prioritised requirements - both in terms of data quality and timing of release. The approach specifies the level and degree to which each data item is quality assured, noting that only some of the total number of items in the BCS are able to be quality assured to the highest standards. Most of the data contained in the BLD CURF are 'as reported' by businesses with limited editing and cross-validation being applied. The qualitative nature of most of the categorical information included in the BCS also contributes to the difficulty of quality assuring these items.
While every effort is made to aid reporting accuracy through form design and checking of a selection of the "more important" items (such as employment and innovation indicators), the BCS is subject to reporting error.
For example, there are a small number of items where businesses have been asked to 'tick one box only' yet multiple responses may have been provided. In these cases, contact would need to be made with the respondent to ascertain the correct response. However, it has not been possible to make contact with all businesses that may have reported in this way, in all instances. These responses have been appropriately flagged.
There are three reasons why data may be missing in the BLD CURF: question not completed; question not required to be completed (i.e. sequencing); or question not asked.
Question not completed
It is common in mail-out surveys for there to be some degree of missing information as a result of forms not being completed. Forms which are received without being completed at all (i.e. blank forms) are returned to respondents for completion. Forms which are received partly completed are not returned and may have some missing data in the BLD CURF. The ABS has not imputed for missing data and, where a response could be expected, the variable has been flagged as missing. Users should be aware that investigations by the ABS into non-response at particular questions indicates that, in the majority of cases, non-response equates to a negative response.
Question not required to be completed
There are a number of topics included in the BCS for which filter questions are used and a negative response usually results in the respondent being sequenced past other questions for that topic. Businesses for which this is a legitimate scenario will have the relevant variables flagged as not requiring a response.
Question not asked
The BCS serves the dual purpose of collecting information to populate the BLD and to replace previously separate collections related to Innovation and IT use in Australian business. To meet this dual purpose, and to minimise provider load and remain within available resources, different form types were introduced into the BCS from the 2006-07 reference period. A long form is used for the component of the BLD panel sample that also contributes to the alternating Innovation and IT use outputs; and a short form is used for businesses that are only included in the BLD panels. The difference between the long and short forms is the more detailed Innovation or IT related questions used to produce these outputs. Businesses for which this is a legitimate scenario will have the relevant variables flagged as not being asked.
All units in their fourth and fifth years receive the short form, therefore a number of data items introduced in 2007-08 on the long form were not asked of any units in Panel One, as this was the fourth year of that panel.
Data related to imports and exports of goods (i.e. trade items)
Data included in the BLD CURF related to imports and exports of goods are available from three sources:
The reporting requirements for data collected by the ATO and Customs are detailed and there are differences in requirements between the two. As the information from each of the three sources is collected on a different basis, it may result in some inconsistencies between the trade data items included in the BLD CURF. For example, a business may have self-identified as an exporter on the BCS yet no corresponding BAS export sales and/or value of exports from merchandise trade data are available. The differences between the three sources relate to:
The ABS has not contacted businesses in the sample to clarify specific reasons for the differences. BLD CURF users are advised to use the source best suited to their specific analytical needs.
Other quality considerations
Some additional information related to quality of reporting for selected BCS data items is included in Using the BLD CURF.
The business population in Australia is not static, changing constantly as a result of new businesses being created (often referred to as births), businesses which cease operation (deaths) and businesses which undergo structural change. As each panel in the BLD represents the business population at the time of selection, it is these latter two changes (deaths and structural change) that need to be dealt with in the BLD.
No action is taken to adjust for births as this would be inconsistent with the principle that each panel represents the business population at the time of initiation and not the population as it may be after five years (i.e. when the panel lapses).
The statistical unit for the BLD is the ABN and it is the operations of the business represented by that ABN that are followed for the life of the panel. ABNs (i.e. businesses) which cease to operate during the life of the panel are considered to be deaths. Businesses may cease to operate for a variety of reasons, for example, they may be wound up or they may be sold to or taken over by another businesses entity. For the purposes of the BLD, businesses (as represented by ABNs) which cease to operate or are wound up during the life of the panel remain in the sample and are appropriately flagged. Any data for that business prior to it ceasing to operate is included in the BLD. As noted above in SAMPLE DESIGN, each BLD panel includes an allowance for expected deaths.
There are a number of reasons why a selected business will change in structure over the life of the panel. A business may merge with another business; be wholly or partly taken over; split into multiple new businesses; take over part or all of another business; or, any combination of these. For the purposes of the BLD, if the original ABN continues to operate, the business will remain in the sample without having any flags incorporated to indicate change. It is recognised that this may create a difficulty for the analyst as the business has undergone a change and should be treated as such, but such change is not easily identifiable. Businesses which undergo structural change and cease using the original ABN are treated as deaths.
Businesses which are out of scope
It is possible that businesses which have been selected in a BLD panel may have incorrect structural information (industry and/or employment size) on the ABSBR. Part of initialising each panel includes checking the structural information to ensure that the business is actually in scope of the BLD, for example, not a large business or part of a business with a complex structure. Businesses may be made out of scope during the life of the panel if they become part of a complex business structure. These will be flagged appropriately and no subsequent data will be present in the BLD CURF. Businesses which, at the time of initiation, were below the employment size cut-off but grow to have 200 or more employees remain in scope and continue to have data collected and included in the BLD CURF.
Businesses with a "live" ABN but no operations
There are businesses in the BLD panels which have become dormant and are flagged as nils. These are businesses which retain their ABN but no longer operate in the market. Nils include businesses which have temporarily ceased their operations.
While every effort has been made to achieve full response for businesses selected in each BLD Panel, there is a small level of non-response. In combining survey data with that from administrative sources, it is possible for a business which did not respond to the BCS to have ATO or Customs sourced data present. All non-responding businesses have been confirmed as live and operating and have been flagged in the CURF using the status fields.
CONFIDENTIALISING THE DATA
The BLD CURF is released under the provisions of the Census and Statistics Act 1905. The Act allows for the release of data in the form of unit records where the information is not likely to enable the identification of a particular person or organisation. Accordingly, there are no names or addresses of survey respondents on the BLD CURF, and other steps have been taken to protect the confidentiality of respondents. These include:
Users should be aware that the methodology applied to perturb financial data does not impact on the comparability between selected businesses within a time period or over time.
These documents will be presented in a new window.
This page last updated 17 June 2013