Weekly Payroll Jobs and Wages in Australia methodology

The Australian Taxation Office (ATO) receives payroll information from businesses with Single Touch Payroll (STP) enabled payroll and accounting software each time the business runs its payroll. The ATO provides selected business and job level data items from the STP system to the ABS for the production of statistics.

The scope and coverage of these estimates are largely defined and constrained by the characteristics of the data sources from which these estimates are produced. As such, users should note that not all jobs and wages in the Australian labour market are captured within these estimates.

Payroll jobs as reported to the ATO through STP are in scope of these estimates. All payroll job holders regardless of age or Australian residency status are included. Persons reported via STP must hold either a Tax File Number (TFN) or an Australian Business Number (ABN).

A payroll job is a relationship between an employee and their employing enterprise, where the employee is paid in the reference week through STP-enabled payroll or accounting software and reported to the ATO. Where an employee is paid other than weekly, the established payment pattern is used to include payroll jobs paid in weeks outside the reference week.

Payroll jobs reported via STP exclude owner managers of unincorporated enterprises.

Employers with 20 or more employees (large employers) commenced transition to STP reporting on 1 July 2018, with approximately 99% of large employers reporting through STP at the time of this release. Payroll reporting via STP is still relatively new and some employers have been granted concessions to enable a longer transition period to mandatory STP reporting.

Employers with less than 20 employees (small employers) began transitioning to STP on 1 July 2019. The ATO has made reporting concessions available for small employers where they:

• employ family members or other ‘closely held’ payees,
• are micro employers with one to four employees,
• employ intermittent or seasonal workers, or
• don’t have access to a reliable internet connection.

As such, at the beginning of September 2020 approximately 77% of small employers are reporting through STP. This figure is lower than reported previously, as the ATO has recently redefined and revised their underlying employer population from a 2017 basis to a 2020 basis. This resulted in a decrease in the proportion of small employers reporting through STP as a large number of small employers (with closely held employees) are exempt from STP reporting for this financial year.

In addition, payroll jobs reported in the Defence Industry (ANZSIC Class 7600) are excluded from these estimates by the ABS to better align with other Labour estimates.

The STP reported wages associated with each payroll job are in scope of these estimates. Wages are gross amounts, prior to taxation and deductions and include:

• salary payments and allowances,
• labour hire payments and foreign income,
• the value of payments in kind (where a fringe benefit amount is recorded),
• bonuses where they are reported in the same field as normal payments.

The total wages concept broadly aligns with the Australian System of National Accounts (ASNA) definition of wages and salaries, with the exception of payments to employee's superannuation and severance and termination payments which are excluded.

More specifically, the following STP reported income items are included in the production of wages estimates;

• gross income amount (including bonuses),
• allowance income,
• fringe benefit amount (reportable, taxable),
• fringe benefit amount (reportable, tax exempt),
• other income (not specified),
• foreign income amount including tax exempt income,
• Community Development Employment Project income.

The STP data are enhanced by combining with other administrative data held by the ABS (also sourced from the Australian taxation system).

With the exception of Age (which is calculated from date of birth in STP data), sex and geographic variables are primarily sourced from Client Register data (supplied by ATO to the ABS as part of the transfer of Personal Income Tax data). When this information is not available, these variables are sourced from STP data.

Industry of activity and employment size variables of the employing business are sourced from the ABS Business Register.

To produce estimates from STP data, a number of processes and treatments are applied.

The STP data are reported on a cash basis (the time when the payment was made) rather than an accrual basis (the time when the payment was earned). Production of real time estimates require the conversion of STP data from a cash basis to an accrual basis. This is done through a “calendarisation” method. This method breaks down all records to a common period (daily), which allows the data to be aggregated and analysed for any longer period (e.g. weekly).

The calendarisation method includes the following steps:

• calculation of the periodicity (payment frequency) using the start and end date of the payment period,
• calculation of a daily pay rate by dividing the total payments by the payment frequency (for example, weekly pay is divided by seven), and
• an adjustment to the periodicity for a job to exclude the days before commencement (or after termination), where the start or termination date for a job occurs within the payment period.

Towards the end of the financial year, unadjusted STP data includes higher than usual week-to-week changes in total wages paid. In particular, some employers report lumped fringe benefits tax (FBT) payment amounts for eligible employees at the end of the financial year. This is most evident in the Health care and social assistance industry. In order to reduce the reporting variability introduced by these payments, the ABS has determined and applied an adjustment factor to accrue reported FBT amounts across the relevant financial year wages series. It is applied to all records which include reportable FBT amounts, not just those in the Health care and social assistance industry.

This adjustment factor has been applied for the 2020-21 financial year, but will be revised at the conclusion of each financial year using business reported data to ensure it remains current. This treatment enhances the existing calendarisation methodology and is consistent with the definition of wages and salaries used in the Australian System of National Accounts. More information about employer reporting of FBT is available from the ATO website.

This adjustment methodology is not possible for other extraordinary payments (such as bonuses) where they are included with the wages data in the period they are paid. Bonuses are not currently readily distinguishable from the wages component of the STP job level dataset (unlike reportable FBT amounts) and can be paid at any time.

In addition to cash reporting, the STP payment data that are extracted for a specific week cycle will be incomplete due to different payment and reporting frequencies. For example, a business with a fortnightly payroll will only report payments in alternate weeks in the STP dataset. As at mid-January 2021, the data indicated that 32% of employees are paid weekly; 51% are paid fortnightly; 12% are paid monthly; and 5% are paid quarterly or infrequently.

To produce reliable weekly statistics, an imputation method is applied to account for different payment and reporting habits which includes the following considerations:

• Business reported data is included for all employees paid weekly due to the 17-day lag between the reference week and the release of estimates. Actual payments for most employees paid fortnightly and some employees paid monthly are also available. Analysis indicates that STP data for the most recent reference week at the time of initial publication is approximately 75-80% complete and can take several months to be fully complete (due to quarterly STP reporters).
• Imputation is not applied where no established pattern of payment exists to enable forward extrapolation. For example, a small proportion of employees who are paid on an ad hoc basis.
• If an employee has not yet had payment data reported and they have not been flagged for termination, it is assumed that their payment status is consistent with their previous reporting record. The previous calculated daily rate will be imputed for the current period.
• If an employee has no payment data reported for two consecutive pay periods, it is assumed that their employment has been terminated. No further imputation is applied.
• No imputation is applied for new employees without historical payment information, until a pattern can be determined. This means that there is an inherent and unavoidable lag before new payroll jobs appear in the data after their initial pay period. The lag is longer for new jobs with employers who have less frequent payment and reporting periods, which is accounted for via a coverage adjustment.

The imputation method has been updated in this release. For more information, see the Imputation update subsection of Methods review.

Once STP data are converted via the calendarisation method and imputation is applied, the data are aggregated for each week (ending Saturday) to produce:

• total payroll jobs, which is the average of the 7 days of payroll job counts,
• total wages, which is the sum of all daily wages for the week, except for employees who cease a job during the week where only the wages for the days worked are included, and
• average weekly wage per payroll job, which is calculated by dividing the total wages value by the total payroll jobs number.

To protect the confidentiality of individuals, data are subject to a suppression process. Data suppression is generally applied to finer level estimates, where a smaller number of individuals contribute to the estimate (such as sub-state indexes). The existing data suppression process has recently been reviewed and updated, and applied to estimates since the 20 October 2020 release. This has resulted in an improvement in the quality of the underlying data used to produce the sub-state datacube and interactive maps. Previously released national, state and territory and industry indexes were not changed by this updated process.

Estimates are supplied as indexes to provide an indication of movements (rather than level estimates) during the COVID-19 period. In order to compare changes over time, the week in which Australia recorded its 100th confirmed coronavirus case (i.e. the week ending 14 March 2020) is used as the reference period for constructing the indexes and given an index value of 100.0. These indexes differ from the ABS' suite of price indexes (including the Wage Price Index) which measure changes in price over time unaffected by quality or quantity and should not be directly compared.

Indexes allow comparison of data between two points in time, the points in time can be adjacent (this week and the previous week) or many weeks apart. Movements in the index from one period to another can be expressed as either points or percentage change and these are rounded to one decimal place. The following example illustrates the method of calculating changes in index points and percentage changes between any two periods:

The following example illustrates the method of calculating a recovery percentage change between any two periods:

STP is a dynamic administrative data source, hence these estimates may be subject to the following sources of error:

• Conceptual misalignment - The Australian tax system is purpose-built and complex, and in some cases it is difficult to determine how a particular STP item should be used to describe impact on payroll jobs and wages. While all care is taken, some income items are subject to this type of validity error. Coherence with other sources indicates that this has a low impact on the aggregate series.
• Reporting error - This is likely to be present in both person and business information used. Most reporting errors are unable to be determined or corrected; however, coherence with other similar statistics demonstrates that this has a low impact on the aggregate series.

For further information, please see Data limitations and revisions.

There are differences between these estimates and similar statistics produced by the ABS. When compared to other ABS sources, the change in payroll jobs, change in wages paid and change in average weekly wage per job in these estimates may differ due to differences in the concepts, scope and methodology used. For example, these estimates:

• contain a combination of administrative data collected for taxation purposes from businesses, whereas other ABS data sources are compiled for the explicit purpose of producing statistics,
• exclude unreported cash in hand payments which may be included in household and business surveys,
• may include information relating a reference week, rather than a particular point in time,
• are not currently able to be adjusted with respect to seasonality, unlike other Labour Force releases, and
• provide a view of payroll jobs, whereas the Labour Force survey presents a view of employed persons. The difference being those people who hold more than one job at a time (i.e. secondary jobs).

For further information, please see Differences to Labour Force employment statistics.

All estimates are presented for weeks ending on a Saturday. Core estimates are released fortnightly on Tuesdays during the COVID-19 period (i.e. a 17-day time lag after the reference week). The ABS has worked to optimise both the timeliness and quality of these data, however revisions in subsequent releases are necessary. Please see Data limitations and revisions for more information.

The following core estimates are produced for each release:

• payroll jobs and total wages, presented as indexes and percentage change movements, and
• average weekly wages per payroll job, for selected characteristics.

Other than indicative numbers on changes in jobs between March and the current period, levels for jobs and wages are not available for release. The payroll jobs index provides a measure of changes in jobs over time since the week ending 14 March 2020. Information on levels for jobs is best sourced from estimates of filled jobs from Labour Account Australia and estimates of employed persons from Labour Force, Australia. More information is included in Differences to Labour Force employment statistics.

Estimates are available at the national, state and territory and Australian and New Zealand Standard Industry Classification (ANZSIC) division by selected personal attributes, including sex and 10 year age group.

Australian Statistical Geography Standard sub-state regions (Statistical Area 4 and Statistical Area 3) and ANZSIC sub-division estimates are updated on alternate releases for payroll jobs only. These estimates are published the day after the main release.

The estimates are presented as an original series only. Seasonally adjusted and trend estimates are not yet available. A number of years of data will be required before seasonal patterns can be observed and adjusted for.

The calendarisation and imputation methodologies applied to the estimates account for calendar related variations, such as the number of days in a month, and different payment frequencies.

The data underlying these estimates are revised for each release and reflected in percentage change movements and indexes. Please see Data limitations and revisions for more detail.

Legislative requirements to ensure privacy and secrecy of this data have been adhered to. 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.

All personal information is handled in accordance with the Australian Privacy Principles contained in the Privacy Act 1988. For more information, see ABS Privacy.

For more information on this methodology please email labour.statistics@abs.gov.au

As part of the transition of the Weekly Payroll Jobs and Wages series from a temporary COVID-19 product to an ongoing ABS output, the ABS are reviewing the methods and refreshing the data sources supporting these estimates. The ABS is currently investigating methods to improve the quality of payroll job and job holder characteristics variables: Age; gender; industry; employment size; and geography. 

In the coming months, the ABS will be progressively implementing updated methods which may result in a refresh of existing characteristics values and the population of ‘unknown’ values.

A regular refresh of the data sources, which support the determination of characteristics variables in these estimates, will also be established to maintain the currency and relevance of these data. As updated methods are implemented, more information will be provided in the How data are processed and Data limitations and revisions sections. 

As part of the ongoing methods review, the imputation method has been reviewed and updated in this release. Imputation is used to estimate values for incomplete data in the most recent pay period, to account for different payment frequencies and business reporting patterns. For more information on how imputation is applied, see the Imputation section of How data are processed.

The imputation method uses a derived ‘final pay frequency’ (based on reported pay events in the STP data) to determine which pay events will be imputed. In the previous method, the final pay frequency was the mode frequency of the three largest pay events of a payroll job. Investigations determined that this method did not impute pay events where a mode could not be calculated. 

The updated method seeks to optimise the number of pay events imputed, while ensuring that imputation is not applied continuously. This is achieved by:

• determining final pay frequency based on the mode of the three most recent pay events, using the median pay frequency if the mode cannot be determined; and
• removing imputation when subsequent business reporting indicates that imputation is not required.

Across the time series, all indexes (including component indexes) are affected by the updated method. The updated imputation method has been applied across all historical data, resulting in revisions across the time series. Revisions may be more pronounced in finer component indexes, such as state by industry. The ABS recommends that analyses of previously published estimates be refreshed with the data from this release.

The updated method is designed to deliver a reduced magnitude of future revisions, particularly in the most recent week’s data. The following graphs present separate comparisons of payroll jobs and wages, using the updated and previous methods against data as released on 2 February 2021 for the week ending 16 January 2021.

In line with the Methods review outlined above, an investigation of existing methods highlighted an issue with the derivation of age (and consequently age group indexes). The issue related to how age was determined with respect to the current reference week. For example, at the time of the first release on 21 April 2020, there were approximately 509,000 payroll jobs reported to be held by 15-19-year olds in the week ending 14 March 2020.  As at the 19 January 2021 release, this cohort was estimated to contain 434,000 payroll jobs for the week ending 14 March 2020.  

Previously, a person’s age (and age group) was redefined at each reference week and updated to the start of the time series included in that release. This resulted in the age group population being measured in each reference week changing over time, and becoming increasingly different to the population of interest. This issue particularly impacted the time series of the youngest (aged 15-19 years) and oldest (aged 70 years and over) persons age groups. 

In the 2 February 2021 release, the derivation method was updated to: hold a person’s age (and consequently age group) constant until their birthday month; and ‘age’ a person at the start of their birthday month. That is, at the start of each calendar month, the age of all job occupants with a birthday in the coming month will be updated to that of their upcoming birthday. 

Privacy protection has been increased by using calendar month to update age, rather than the more precise day and month. However, it may introduce a small amount of volatility in the week on week change at the start of each calendar month. That is, a portion of the movement in age group at the start of the month (payroll jobs and wages) may relate to the ‘aging’ of persons as they transition up an age group. 

Indexes for national, state/territory, industry and employment size estimates are not affected by the updated method, as only those indexes with an age dimension have been impacted.

A comparison of payroll jobs is presented in the following graph for the age groups most affected by the change in method, using data from the 2 February 2021 release.  

Weekly Payroll Jobs and Wages in Australia estimates are derived from data collected via the STP system, which effectively supports employer reporting obligations and ATO operational requirements through enabled software. However, it is not primarily designed to support the production of statistics and therefore some inherent limitations of the data require specific treatment and result in data revisions between statistical publications. 

The weekly change estimates for the most recent weeks of data contain a higher degree of reporting variability and imputation (described further below). The ABS recommends that users exercise caution when focusing on change in the most recent weeks, as these estimates are subject to a greater levels of revision in subsequent releases.

As the compilation of these estimates evolves, reporting patterns sometimes arise which require detailed analysis and treatment. Outcomes from these analyses may result in changes to methodology or revisions to data which will be updated here, as required.

Weekly Payroll Jobs and Wages in Australia are experimental estimates, compiled in near real time and published fortnightly. This information provides a complementary insight to Labour Force statistics on employment, which provide a more comprehensive view of the Labour market. 

The differences in concepts, scope and methodology used to produce changes in employment (as reported in Labour Force statistics) and changes in payroll jobs can affect their interpretation as economic measures. The following key differences should be considered when comparing these statistics.

Recently released Labour Force analysis of employment versus payroll jobs up to August 2020 can be found in Strong employment growth for non-employees. More information on the coherence of Weekly Payroll Jobs and Wages estimates with similar ABS statistics are detailed in the Coherence section of How data are processed.