Average Weekly Earnings, Australia methodology

This is not the latest release View the latest release
Reference period
November 2021
Released
24/02/2022

Introduction

The Survey of Average Weekly Earnings (AWE) is a survey of business units which has been conducted since 1981. Prior to August 1981, the AWE series was based principally on information from payroll tax returns. The survey is currently conducted biannually, in May and November, and was conducted on a quarterly basis until May 2012.

The purpose of the survey is to measure the level of average gross weekly earnings associated with employees. While AWE is not designed to produce movement in earnings data, the frequency of collection supports a time series of these level estimates.

In practice, AWE estimates are used in commercial contracts, more broadly by the private sector for economic and labour market analysis, and in Commonwealth, state and territory legislation for adjusting a variety of government payments, supporting minimum wage claims and monitoring wage equity, and as an indicator of change in underlying wage rates for economic or taxation policy analysis.

Earnings in AWE are broadly defined as current and regular payments in cash to employees for work done, excluding amounts salary sacrificed (except for the Average Weekly Cash Earnings series which are inclusive of salary sacrificed). Descriptions of the underlying concepts of Australia's AWE statistics (including the concept of earnings used in AWE), and the sources and methods used in compiling these estimates, are presented in Labour Statistics: Concepts, Sources and Methods.

Reference period

The reference period for the survey is the last pay period ending on or before the third Friday of the middle month of the reference quarter (i.e. May and November). Where a pay period is fortnightly or monthly, etc., the employer is requested to report only one week's proportion.

The data is collected in respect of a typical week and, therefore, may not reflect events such as Christmas trading. The data excludes irregular and infrequent payments, such as annual bonuses. For these reasons, caution is advised if using AWE to derive annualised average earnings.

Scope and coverage

The AWE survey includes all wage and salary earners who received pay for the reference period, except:

  • members of the Australian permanent defence forces;
  • employees of enterprises primarily engaged in agriculture, forestry and fishing;
  • employees of private households;
  • employees of overseas embassies, consulates, etc.;
  • employees based outside Australia; and
  • employees on workers' compensation who are not paid through the payroll.
     

Also excluded are:

  • casual employees who did not receive pay during the reference period;
  • employees on leave without pay who did not receive pay during the reference period;
  • employees on strike, or stood down, who did not receive pay during the reference period;
  • directors who are not paid a salary;
  • proprietors/partners of unincorporated businesses;
  • self-employed persons such as subcontractors, owner/drivers, consultants;
  • persons paid solely by commission without a retainer; and
  • employees paid under the Australian Government's Paid Parental Leave Scheme.

Collection method

Details of the total number of employees and earnings paid for the survey reference week are obtained on a biannual basis (in May and November) from selected businesses. Data are collected via online electronic collection.

Follow-up procedures are in place to obtain information from respondents who do not lodge a completed form by the due date. The target minimum response is 91% for the survey as a whole, and approximately 90% for each state and/or industry.

Sample design

The sample is selected from the ABS Business Register (ABSBR) which is a list of businesses and organisations operating in Australia primarily based on registrations to the Australian Taxation Office's (ATO) Pay-As-You-Go Withholding (PAYGW) scheme. The population is updated quarterly to take account of new businesses, businesses that have ceased employing, changes in industry and employment levels and other general business changes.

Sample redesigns are undertaken periodically to ensure the survey design continues to be optimal. The most recent sample redesign for the Survey of Average Weekly Earnings was implemented for November 2017.

A probability sample of statistical units (employing businesses) is drawn from the ABSBR. The statistical unit for the survey comprises all the activities of an employer in a particular state or territory based on the Australian Business Number (ABN) unit or Type of Activity Unit (TAU). Each statistical unit is classified to an industry which reflects the predominant activity of the business. Variables used to stratify the survey frame are:

  • public/private sector;
  • industry;
  • state/territory; and
  • employment size – the ranges used vary between states/territories and industries.

Statistical units with benchmark employment greater than a set cut off (this cut off will vary for different states/territories) are completely enumerated. Strata with a very small number of statistical units may also be completely enumerated, but such strata may become sampled strata if the number of units increases sufficiently. Within each stratum, statistical units are selected with equal probability.

Sample selection is constrained by ensuring that there is minimum overlap with other labour-related business surveys.

For details on the ABS Business Register and ABS economic units model, see the Methods used in ABS Business Surveys chapter in Labour Statistics: Concepts, Sources and Methods.

Sample size

A sample of approximately 5,500 employer units is selected from the ABS Business Register to ensure adequate state, industry and sector representation. This yields a live sample of approximately 5,130 units.

The sample is allocated optimally across sampled strata using a technique designed to minimise the variance of AWE estimates at both the national and state/territory level.

Sample rotation

The sample is updated each period to reflect changes in the ABS Business Register. Approximately 16% of the sample selected from the non-completely enumerated strata is replaced each period. Refer to the Methods Used in ABS Business Surveys chapter in Labour Statistics: Concepts, Sources and Methods for further information.

Sample rotation is implemented for the majority of sampled strata, but is not implemented where the population of a stratum is so small that units rotating out of the sample would be rotated back in after only a short interval.

Use of Australian Business Register (ABR) data

The results of these studies are based, in part, on ABR data supplied by the Registrar to the ABS under A New Tax System (Australian Business Number) Act 1999 which requires that such data is only used for the purpose of carrying out functions of the ABS. Any discussion of data limitations or weaknesses is in the context of using the data for statistical purposes, and is not related to the ability of the data to support the ABR’s core operational requirements. 

Legislative requirements to ensure privacy and secrecy of this data have been followed. In accordance with the Census and Statistics Act 1905, results have been confidentialised to ensure that they are not likely to enable identification of a particular person or organisation.

Estimation

Ratio estimation is used in all sampled strata, except in small sized strata where number raised estimation is used.

In both completely enumerated and sampled strata, an automatic imputation procedure is used for units not responding, by applying imputed growth rates to the most recently reported employment and earnings data for these units, provided that data have been reported in a previous period. This is referred to as Beta imputation. Otherwise, the Live Respondent Mean method is used to impute for missing data items.

For the quarters from May 2020 to May 2021, the AWE imputation process was further refined to incorporate JobKeeper recipient information in the imputation class structure. This improved the accuracy of imputation for the small number of businesses unable to respond in those quarters, by better controlling for COVID-19 impacts. From the November 2021 quarter, the JobKeeper recipient information was removed from the imputation class structure.  

The winsorisation methodology is used as the primary method to treat outliers in AWE. Winsorisation moderates the impact of an outlier business without the harsh impact of the surprise outliering approach. For more information, refer to the Overview of Survey Methods chapter in Labour Statistics: Concepts, Sources and Methods.

Survey estimates include an adjustment called Business Provisions to allow for births and resurrections of businesses that have occurred up to the end of the survey reference period but which are not reflected on the survey frame.

For further information on estimation methods used in ABS business surveys, refer to the Methods used in ABS Business Surveys chapter in Labour Statistics: Concepts, Sources and Methods.

Seasonal adjustment and trend estimation

Both seasonally adjusted and trend estimates are usually produced for key series from this survey. Trend estimates have been suspended during the COVID-19 period and historical trend series (up to November 2019) are available in previous releases.

During the COVID-19 period, the ABS will use forward seasonal factors to produce seasonally adjusted estimates for some AWE series. Forward factor adjustments are generally better suited to managing large movements at the end point of a series and ensure that large movements do not have a disproportionate influence on the seasonal factors.

Seasonal adjustment

Seasonal adjustment is a means of removing the estimated effects of normal seasonal variation from the series so that the effects of other influences can be more clearly recognised. Seasonal adjustment does not aim to remove the irregular or non-seasonal influences which may be present in any particular series. Influences that are volatile or unsystematic can still make it difficult to interpret the movement of the series even after adjustment for seasonal variation. If a time series has no identifiable seasonality it is not seasonally adjusted.

The AWE survey uses the concurrent seasonal adjustment technique to estimate seasonal factors, based on a synthesised quarterly original series. Linear interpolation is used to impute "missing" quarterly original observations based on the succeeding and preceding survey estimates. Under concurrent seasonal adjustment, the estimates of seasonal factors are improved as new or revised original estimates become available each period.

Trend estimates

Seasonally adjusted estimates can be smoothed to reduce the impact of irregular or non-seasonal influences. Smoothed seasonally adjusted series are called trend estimates.

The ABS considers that trend estimates provide a more reliable guide to the underlying direction of the original estimates and are more suitable than either the seasonally adjusted or original estimates for most business decisions and policy advice.

The trend estimates in the AWE survey are calculated using a centred 7-term Henderson moving average of the seasonally adjusted estimates of quarterly synthesised original data. Estimates for the three most recent periods cannot be calculated using this centred average method; instead an asymmetric average is used.

The changes to the moving average formulae can lead to revisions in the trend as data for subsequent periods becomes available. Revisions to the original data and re-estimation of seasonal adjustment factors also cause revisions to trend estimates. If a series is highly volatile then the trend estimates will be subject to greater revision for the latest few observations as new data become available. However, it is important to note that this does not make the trend series inferior to the seasonally adjusted or original series.

For further information, see A Guide to Interpreting Time Series - Monitoring Trends.

Survey output

Estimates are published biannually (for May and November) in Average Weekly Earnings, Australia, released approximately three months after the reference period.

Three main series are published:

  • Average weekly ordinary time earnings (commonly referred to as AWOTE) for full-time adult employees;
  • Average weekly total earnings for full-time adult employees (comprising weekly ordinary time earnings plus weekly overtime earnings); and
  • Average weekly total earnings for all employees.
     

Estimates of the annual percentage change for average earnings are published for each key series. Estimates are cross-classified by sector, state or territory, and by industry at the Australian level for males, females and persons.

Estimates are published on original, seasonally adjusted and trend basis, with seasonally adjusted and trend estimates available by Australia and state/territory for each of the three main series listed above.

For further information in understanding AWE statistics, see the feature article A Guide to Understanding Average Weekly Earnings Statistics.

Movements in average weekly earnings and comparability with the wage price index

The AWE survey is designed to provide estimates of the level of average earnings at a point in time and, while not designed for movements in earnings, the frequency of collection supports a time series of these level estimates. The standard errors for the period-to-period movements are much higher proportionally than for the level estimates, so should be interpreted with this in mind.

Changes in average earnings are affected not only by changes in the level of earnings of employees, but also by changes in hours worked and the overall composition of the workforce over time.

This compositional change (may apply differently within different states and territories, and over time) can include variations in:

  • the proportion of full-time, part-time, casual and junior employees;
  • occupational distribution within and across industries; and
  • the distribution of employment between industries.

     

Movements from the AWE survey therefore account for the changing workforce over time (i.e. the average earnings based on the composition of the workforce, and how they have changed from the earnings based on the composition of the workforce at an earlier time).

An alternative source for estimates of movements in earnings (wages growth) is the quarterly Wage Price Index (WPI).

Unlike AWE, the WPI is designed to measure the change over time in the price of wages and salaries (i.e. a pure price change, unaffected by changes in the quality or quantity of work performed or the composition of the workforce).

As a result, period-to-period movements for the AWE series are not comparable with those for the WPI.

The WPI is published in Wage Price Index, Australia. For further information on comparability between AWE and WPI, please refer to the feature article Average Weekly Earnings and Wage Price Index - What do they measure?.

Reliability of estimates

Estimates are subject to sampling and non-sampling errors. For information on the reliability of estimates see the Technical Note.

The Bootstrap approach is used to calculate estimates of variance for this survey. The Bootstrap is a variance estimation method which relies on the use of replicate samples, essentially sampling from within the main sample. Each of these replicate samples is then used to calculate a replicate estimate and the variation in these replicate estimates is used to calculate the variance of a particular estimate.

Rounding

Estimates of average weekly earnings are rounded to the nearest 10 cents, and discrepancies may occur between sums of the component items and totals. Estimates of percentage change have been calculated using unrounded estimates, and may be different from, but are more accurate than, movements obtained from calculating percentage change using the rounded estimates presented in this release.

History of changes

In order to provide a high degree of consistency and comparability over time, changes to survey methods, concepts, data item definitions, and frequency of collection are made as infrequently as possible. Significant changes have included:

    Show all

    Technical note - sampling error

    Reliability of estimates

    As the estimates in this release are based on information relating to a sample of employers, rather than a full enumeration, they are subject to sampling variability. That is, they may differ from the estimates that would have been produced if the information had been obtained from all employers. This difference, called sampling error, should not be confused with inaccuracy that may occur because of imperfections in reporting by respondents or in processing by the ABS. Such inaccuracy is referred to as non-sampling error and may occur in any enumeration whether it be a full count or a sample. Efforts have been made to reduce non-sampling error by careful design of questionnaires, detailed checking of returns and quality control of processing.

    The sampling error associated with any estimate can be estimated from the sample results. One measure of sampling error is given by the standard error which indicates the degree to which an estimate may vary from the value which would have been obtained from a full enumeration (the 'true value'). There are about two chances in three that a sample estimate differs from the true value by less than one standard error, and about nineteen chances in twenty that the difference will be less than two standard errors. Standard errors are provided in tables 3, 6, 9, 10 and 13 to 17.

    An example of the use of a standard error is as follows. If the estimated average earnings were $1,100.00 with a standard error of $7.00, then there would be about two chances in three that a full enumeration would have given an estimate in the range $1,093.00 to $1,107.00 and about nineteen chances in twenty that it would be in the range $1,086.00 to $1,114.00.

    Another measure of the sampling error is the relative standard error, which is obtained by expressing the standard error as a percentage of the estimate.

    Glossary

    Show all

    Quality declaration

    Institutional environment

    Relevance

    Timeliness

    Accuracy

    Coherence

    Interpretability

    Accessibility

    Abbreviations

    Show all

    Back to top of the page