Fair Work Commission

Introduction

The AWRS data collection strategy leads to 2 units of analysis:

• employers, and
• employees.

The data sources used to inform the construction of weights for these 2 units of analysis are outlined below.

All enterprises that participated in the AWRS completed the Recruitment questionnaire and the Employee Relations questionnaire (3057 employers).

The remaining 3 employer questionnaires (the Structure and Operations CATI questionnaire, the Online Workforce Profile questionnaire and the Online Financial Information questionnaire) were not completed by all enterprises. In addition, a shortened CATI version of the Workforce Profile and Financial Information questionnaires were administered so that at least some key information from those questionnaires was collected. As a result, there are varying sample sizes according to the origin of the data items of interest, particularly where analysis sources data items from different questionnaires.

The construction of weights for the AWRS has enabled all data to be used in analysis for the First Findings report.

A total of 48 weights have been constructed that cover all of the analysis permutations according to the source of each data item and the amount of sample available.

The Commission will evaluate the complexity of the weighting strategy, including the documentation that explains how to select the correct or most appropriate weight for any analysis to be undertaken. Use of the survey weights requires a good understanding of the research design, response rates and origin of data items.

Care has been exercised in constructing employer and employee weights drawing on data from different sources as inconsistencies can arise. Checks have been carried out on the employer data and the employee data during the course of constructing the weights and following construction of weights through applying the weights to conduct analysis.

The information provided below is an overview of the process. Further information about the construction of weights will be made available when AWRS data is released.

Employer population weights

The employer population weights have been calculated as the ratio of best available employer counts to the number of responding employers at the ANZSIC 2006 division x employment size category (defined appropriately).

Note that the industry strata in the sample design were defined at the ANZSIC division (one-digit) level apart from Accommodation and Food Services division, which was defined at the industry subdivision level (two-digit). Apart from 4 industry groups there were also 2 employment size strata based on 5–19 employees and 20+ employees in the sample design.

The starting point for calculating the weights was deriving approximate population counts from the most recently available ABS business counts. These were for June 2013, obtained from publication 81650, Counts of Australian Businesses, including Entries and Exits, Jun 2009 to Jun 2013 (released 23 June 2014). This publication contained 2 relevant datasets:

• (1) Business counts by SA2 x ANZSIC Division x employment size (1–4, 5–19, 20–199, 200+), and
• (2) Business counts by Main state x ANZSIC Class (four-digit) x employment size (1–19, 20–199 and 200+).

A minimum requirement for constructing weights is to base the weighting cells on the sample design. It was decided, however, that weighting for employer survey data would be carried out post-stratifying by finer employment size classifications for those ANZSIC Divisions with sufficient population in the 200+ employees category . This was used to not only ensure that the selection probabilities were accounted for, but also to reduce the impact of differential response rates across size within the survey strata. Further information about the construction of survey weights will be provided as survey data becomes available for analysis.

The ANZSIC division data sourced from the ABS provided, for each industry division, business counts for employment sizes: 5–19, 20–199, and 200+ and these employment size categories were considered initially for weighting at this level for each ANZSIC division, provided there was sufficient sample size. The rule used in this weighting was to avoid collapsing across employment size unless the weighting cell sample sizes were less than 5 enterprises.

The table below shows the resulting employer counts for the relevant ANZSIC divisions and subdivisions by employment size 5–19, 20–199, 200+ employees. The 2 974 employers with Unknown ANZSIC were imputed to an ANZSIC division x employment size in proportion to the employers with known industry and size.

Business counts were also required for the non-profit enterprises.

These counts were obtained from the ABS as a special request for unpublished data. For reasons of preserving confidentiality, the ABS was not able to provide the business counts for all ANZSIC divisions classified by employment size 20–199 and 200+ employees. For these ANZSIC divisions the information was only provided for employment size 20+ employees.

The Public Administration and Safety division counts needed to be supplemented with counts for the public sector part of the division which was in scope of AWRS. These counts were compiled from various lists of organisations in the public sector that are in the national system of workplace relations.

Employee population weights

In addition to calculating employer weights for use in analysing the employer data, an employment/workforce weight was calculated at the employer level to be applied when analysing numeric employee count data collected from employers (e.g. number of employees that have their pay set by an award).

Weights for analysing data from the Employee questionnaire component have also been constructed.

ABS employment data was obtained from the ABS’ monthly Labour Force survey for the purpose of constructing survey weights.

The estimates for May 2013 were used as these were closest in time to the reference period of the business counts used for construction of the employer weights.

The table below shows the ABS employment estimates from publication 6291.0.55.003 Labour Force, Australia, Detailed, Quarterly for May 2013.

One limitation of the Labour Force survey catalogue is that data on employment size of the employees’ employer is not collected, and so these data would include employees who are employed by enterprises with fewer than 5 employees that would not be within the scope of the AWRS.

These deficiencies were addressed by computing the relative share of employment for employees employed by enterprises with 5+ employees for the ranges 5–19, 20–199 and 200+ from data generated from the Commission’s Award Reliance survey. This survey data was suitable for the purpose of constructing weights as it covered employers of all employment sizes.