CONCEPTS AND METHODS


SCOPE AND POPULATION

The experimental estimates in this publication are based on the 2006 edition of the Australian and New Zealand Standard Industrial Classification (ANZSIC) (cat. no. 1292.0) and by institutional sector, in accordance with the Standard Institutional Sector Classification of Australia (SISCA), which is detailed in Standard Economic Sector Classifications of Australia (SESCA) (cat.no. 1218.0).

The scope of the experimental estimates in this publication is based on the scope used for Australian Industry (cat. no. 8155.0). It includes all business entities in the Australian economy which are classified on the ABS Business Register (ABSBR), to ANZSIC Division C Manufacturing and excludes any entities classified to SISCA Sector 3 General Government. Note that government-owned or controlled Public Non-Financial Corporations are included.


STATISTICAL UNITS DEFINED ON THE ABS BUSINESS REGISTER

The experimental estimates in this publication have been created from businesses recorded on the ABSBR. The economic statistics units model used by the ABS allocates businesses on the ABSBR to one of two sub-populations. The vast majority of businesses are in what is called the ATO maintained population (ATOMP), while the remaining businesses are in the ABS maintained population (ABSMP).

ATOMP

Most businesses and organisations in Australia need to obtain an Australian Business Number (ABN). They are then included on the whole-of-government register of businesses, the Australian Business Register (ABR), which is maintained by the ATO.

The ATOMP is composed of those businesses on the ABR with simple structures (i.e. primarily comprised of a single ABN), and the ABN unit is used as the statistical unit for all ABS economic collections (in this case, the ABS has aligned its statistical units structure with the ABN unit).


ABSMP

For the population of businesses where the ABN unit is not suitable for ABS statistical requirements, the ABS maintains its own units structure through direct contact with the business, which constitutes the ABSMP. This population consists typically of large, complex and diverse businesses.


METHODOLOGY

The experimental estimates in this release were produced using a combination of Economic Activity Survey (EAS) data collected directly by the ABS and BAS data obtained from the ATO.


EAS Collection Design

In order to decrease the statistical reporting load placed on providers, the collection strategy for the EAS is to use directly collected data from a sample of businesses, in combination with BAS data sourced from the ATO.

Businesses in the ABSMP which have employment greater than 300, or are deemed to be 'significant', are completely enumerated via directly collected survey data.

Other businesses are available for random sample selection only if their business is identified as being an employing business (based on ATO records) or their turnover exceeds a threshold level. Turnover thresholds are set for each ANZSIC class so that the contribution of surveyed businesses accounts for 97.5% of total industry class turnover as determined by BAS data. Data for businesses selected from this part of the sample are obtained via direct collection.

Businesses which meet neither of these criteria are referred to as 'micro non-employing businesses' and are not eligible for selection in the EAS sample. For these units, BAS data are obtained and added to the directly collected estimates (with no modelling applied).

More detailed information about the EAS collection design can be found in Australian Industry (cat. no. 8155.0), Technical Note 1.


The Experimental Manufacturing Class Estimates Model

The estimation method used to create the experimental estimates makes use of observed linear relationships between data collected from businesses in the EAS and auxiliary information available from BAS data. Where the auxiliary information is strongly correlated with data items collected in the EAS, this information has been used to create predicted values for ATOMP businesses that were not selected in the survey. The auxiliary variables used to create predicted values were:

• BAS Turnover (to model sales and service income)
• BAS Wages and Salaries (to model wages and salaries and industry value added).

The ANZSIC class experimental estimates for 2006-07and 2007-08 were created subject to the constraint of being additive to ANZSIC subdivision estimates for 2006-07 and 2007-08, as published in the 2007-08 issue of Australian Industry (cat. no. 8155.0). In spite of this requirement, however, the experimental estimates presented at subdivision level in this publication do not exactly match those in Australian Industry owing to two main factors. The first is post-release editing of the directly collected data at a finer level of ANZSIC detail than required for Australian Industry. The second factor is a correction of the definition of IVA. The discrepancy between subdivision level estimates is explained in further detail in the section "Comparison with other ABS data"

For the purpose of compiling experimental ANZSIC class estimates for Division C Manufacturing in this publication, data for businesses are contributed via one of three categories (or 'streams') in accordance with significance and collection-related characteristics. The following diagram illustrates the ways in which the data streams contribute to the experimental estimates for the manufacturing industry.


SUMMARY OF DATA STREAMS


The Survey Stream

The survey stream consists of units with directly collected EAS data. Note that there was a significant reduction in the number of businesses in the survey stream between 2006-07 and 2007-08 due to changes in the ABS Economic survey program.


The Modelled Stream

The modelled stream includes employing businesses, not in the survey stream, whose turnover was higher than the threshold set for their ANZSIC class (where the threshold was set so that the contribution of surveyed businesses would account for 97.5% of total industry class turnover). Within this stream, data were modelled from auxiliary information available from BAS data. The BAS data were found to have a high correlation with corresponding data from the EAS. Modelling was used on the BAS data rather than directly substituting it as the BAS data items did not map directly to their corresponding EAS data item definitions.

Modelled data were created through the use of a robust, trimmed regression estimator, which uses survey data regressed against BAS data. The regression factor was obtained by utilising the sampled units in the ATOMP and comparing their reported survey data with their matching BAS data. The regression factor was created at the ANZSIC subdivision level. Sales and service income was modelled using BAS Turnover as the auxiliary variable; whilst wages and salaries and IVA were modelled using BAS Wages and Salaries. For most businesses in the modelled stream, predicted values for the three variables were created by multiplying their BAS data by the calculated regression factor.

A different methodology was used for ABSMP units which were not directly surveyed and had employment of 20 or more. These units were modelled from data from surveyed ABSMP units.


Micro Non-Employing Stream

This stream includes units in the ATOMP, which are non-employing businesses and whose turnover for each ANZSIC class was below the turnover threshold. For these units BAS data were obtained and directly added into the experimental estimates (with no modelling applied).


PRODUCING THE MANUFACTURING EXPERIMENTAL ESTIMATES

Initial experimental estimates for the manufacturing industry were produced by aggregating the contributing data streams, which were then adjusted so that the classes summed to the national subdivision estimates published in the 2007-08 issue of Australian Industry (cat. no. 8155.0). This adjustment removed some of the non-sampling error introduced through the regression modelling (see chapter 4 for discussion of modelling bias). This adjustment was obtained by first calculating the difference between the national subdivision estimates and the initial experimental subdivision estimates and then prorating the difference across the classes within the subdivision. The level of proration for each class was determined by the size of the modelled stream. Thus, proration has a stronger impact on those classes with a larger modelled stream.


ASSUMPTIONS IN THE MODEL

The modelling methodology used to create the experimental estimates presented in this publication is based on the following assumptions:

• the published ANZSIC subdivision estimates were of sufficient quality to warrant disaggregation at ANZSIC class level;
• it was valid to distribute the difference between Australian Industry subdivision estimates and the experimental subdivision estimates across the industry classes based on the size of the modelled stream;
• the relationship between the EAS data items and the BAS data items is meaningful and consistent. Analysis supports this assumption, with the correlation being of consistent quality to produce reliable estimates;
• the auxiliary (BAS) data was of high quality; and
• the industry coding was accurate on both the ATO maintained ABR and the ABSBR.

COMPARISON WITH OTHER ABS DATA

Following is a comparison of the manufacturing experimental estimates with manufacturing data presented in other ABS publications:


Australian Industry, 2007-08 (cat. no. 8155.0)

The manufacturing subdivision estimates presented in this publication do not exactly match those in the 2007-08 issue of Australian Industry (cat. no. 8155.0). This discrepancy is due to two factors.

Firstly, the process of producing the experimental ANZSIC class estimates resulted in editing the EAS data at finer levels than was required for the ANZSIC sub-division estimates in Australian Industry.

The second factor relates to a correction of the definition of IVA. In Australian Industry, 2007-08 payroll tax and land tax were included in intermediate input expenses, a component of IVA. However, for this information paper and future issues of Australian Industry, payroll tax and land tax is removed from the intermediate input expenses component of IVA. This definition change resulted in the experimental IVA estimates for manufacturing being approximately 2% greater than the IVA figures published in Australian Industry.


Manufacturing Industry, Australia, 2006-07, (cat. no. 8221.0)

Manufacturing Industry, Australia, 2006-07 published ANZSIC class estimates for the manufacturing industry. The experimental class estimates in this information paper will differ from those published under Manufacturing Industry primarily due to differences in methodology (e.g. the use of modelling for the modelled stream). It is important to note that estimates produced in this information paper are considered experimental and the ANZSIC class estimates for 2006-07 contained in Manufacturing Industry are the official ABS statistics.

Additionally, the ANZSIC subdivision estimates for Manufacturing Industry and Australian Industry differ slightly. The 2006-07 issue of Manufacturing Industry was released before the 2006-07 issue of Australian Industry. The discrepancy was due to additional survey information being received after publication of Manufacturing Industry and the resulting additional editing applied to Australian Industry prior to its publication.


Manufacturing Census, 2006-07

Data from the 2006-07 manufacturing census was published in the data cubes associated with the Manufacturing Industry, Australia, 2006-07 (cat no. 8221.0). The manufacturing census provides finer geographical data at a state and statistical division level. Aggregates in the census will differ from these other releases, as well as the experimental estimates in this publication, due to sampling error, methodological changes and definitional differences. For further information on the methodology used in the manufacturing census see the data cube (i-note) under Manufacturing Industry, 2006-07, Explanatory Notes.