Seasonal adjustment and changing seasonality in Labour Force statistics

An overview of the approach that the ABS takes to seasonal adjustment and accounting for changing seasonality in Labour Force statistics

Cite
Print
Released
16/05/2024
Source
Labour Force, Australia, April 2024

This article provides an overview of the approach that the ABS takes to seasonal adjustment and accounting for changing seasonality in Labour Force statistics. Its release coincides with the latest annual series review, undertaken between the March and April releases.

Seasonal adjustment is a statistical technique that seeks to measure and remove the effects of systematic calendar-related patterns (that is, things that happen at the same time every year, like lower employment and hours in January). 

The COVID-19 pandemic was a period of major disruption, with large month-to-month changes in the seasonally adjusted Labour Force series. Now, four years on from the start of the pandemic, we still periodically see larger-than-usual increases or decreases in seasonally adjusted data. 

This challenge is not unique to Australia, with countries around the world also faced with similar changes that, like the ABS, they are actively assessing, managing, and explaining.

In exploring seasonal adjustment and seasonality, this article highlights recent larger-than-usual monthly changes in seasonally adjusted Labour Force series, focusing on total hours worked and employment. It concludes with information on some recent changes in dynamics and behaviours that have contributed to these larger movements in the seasonally adjusted series.

Seasonal adjustment and accounting for changing seasonality in Labour Force statistics

Times series provide insights into changes in statistics of interest over time (such as employment, unemployment and hours worked in the labour market). The 'original' Labour Force series (that is, the unadjusted time series) reflects the sequence of data from each monthly Labour Force Survey and reflects three components - the underlying trend, seasonal elements, and irregular/short-term elements. 

Seasonal adjustment is a statistical technique that seeks to measure and remove the effects of systematic calendar-related patterns. Seasonally adjusting a time series is useful if you wish to understand the underlying patterns of change or movement in a series, without the impact of the seasonal elements (for example, reduced employment and hours worked during January). In contrast to the trend series, seasonal adjustment does not aim to remove the short-term or irregular influences in the original data. For a simple explanation of this, see the ABS guide to time series.

To seasonally adjust a series, the ABS uses a method called concurrent adjustment to derive ‘seasonal factors’, which are an estimate of the calendar and seasonal effects. This method uses data up to the current month (or quarter for quarterly data) to estimate the seasonal factors for the current month and re-estimate (and revise) them for previous months. Further information on this can be found in the Labour Force methodology summary, the Time Series Analysis Frequently Asked Questions and A Guide to Interpreting Time Series – Monitoring Trends. 

In addition to the monthly estimation, seasonal factors are also reviewed annually, at a more detailed level. The annual series review uses the context gained from an additional year of original data to assess the appropriateness of seasonal adjustment parameters and prior corrections. This process is typically done around March or April for Labour Force statistics and generally only results in minor revisions, as can be seen in the April 2024 Labour Force release.

By their very nature, seasonal patterns are only visible when looking across multiple years. Given this, there are limitations in how quickly a change in a pattern will be reflected in changes to the seasonal factors. 

When there is a discernible change in a seasonal pattern, it will generally be reflected in corresponding changes in the seasonal factors after three years, and the seasonal factors for a month will be reasonably stable after six years (after which revisions tend to be minimal). This is an important part of the methodology, as some changes may be part of a continuing change, while others may be more related to specific economic or social conditions (that is, stronger or weaker changes that then start to revert to a more long-term seasonal pattern).

The seasonal adjustment approach applied by the ABS is data driven. This is important since attempting to intervene through applying adjustments to seasonal factors would risk introducing bias or other errors into the Labour Force series, given it would rely on the ABS making assumptions around the extent to which the seasonal patterns are changing. This would be particularly challenging for something as dynamic and complex as the labour market, where the relatively small changes in aggregates usually reflect the net difference in very large underlying gross flows and changes, which are influenced by a broad range of contributing factors. 

Instead, when seasonal patterns in Labour Force statistics may look a little unusual, the ABS generally focuses on providing additional guidance and advice to users to help interpret the data. This involves sharing key insights from detailed analysis of historical seasonal patterns in underlying flows and changes in the composition of the Labour Force Survey sample. 

This commentary is routinely included in media releases and supporting articles, which provide a useful and transparent guide to how to make sense of unusual data movements. The ABS will also generally couple this advice with supporting analysis of trend series and how changes align with similar changes observed in other non-Labour Force Survey data (for example, changes in statistics based on Single Touch Payroll data or the Job Vacancies Survey). 

Some recent examples of articles that include analysis and guidance can be found in the January 2024 Labour Force release and the November 2023 Retail trade release).

Recent behaviour in the seasonally adjusted hours worked and employment series

Chart 1 shows the monthly change in seasonally adjusted total hours worked. Following the large changes during the pandemic-related disruption (from April 2020 to March 2022), some larger changes have continued to be seen in recent periods, particularly around the peak holiday periods of January, April, July and October. 

Source: Labour Force, Australia, Table 19

While less pronounced than changes in total hours worked, Chart 2 shows that there have also been some relatively large recent movements in seasonally adjusted employment, including a larger fall in December 2023 and a larger rise in February 2024.

Source: Labour Force, Australia, Table 1

The recent larger-than-usual increases and decreases in seasonally adjusted hours worked and employment at these times are not necessarily an indication of ongoing structural changes in seasonal patterns in the labour market. Some changes may only be temporary changes in the dynamics and behaviours of people and businesses, within a tight labour market characterised by relatively low unemployment, high participation and high levels of job vacancies. For example, more hours might be worked by existing employees who reduce or postpone their leave to cover a vacancy that cannot be filled, or job starters may have more flexibility and choose to start in February after the summer school holidays, rather than in January. 

Future data will be required to fully understand the extent to which recent changes in the labour market may reflect enduring changes in seasonal patterns or irregular/short-term elements. 

The power of trend data during periods of potentially changing seasonality

Trend data provides the most reliable measure of the labour market, as it specifically excludes seasonal elements (for example, lower employment and hours in January) and irregular/short-term elements (such as a one-off event, like a flood or a major sporting event). Trend also smooths out month-to-month changes related to sampling variability (that is, changes in who is in the Labour Force Survey sample, month to month), which also contributes to the irregular elements. 

Trend data therefore provides the best means of determining whether the labour market is strengthening or softening, and how the composition of the labour market is changing over time. While it is slower to move, as it draws heavily upon data across multiple adjacent months, it provides a more reliable indication of a sustained change.

Trend data is particularly useful in periods where seasonality may be changing more than usual and contributing to larger-than-usual movements in the seasonally adjusted data. During this time, the relative slowness of trend is likely to be less of an issue than the time it takes to get close to the final estimate of a seasonal factor.

Since the middle of 2022, trend data has been particularly valuable in informing analysis of changes in total hours worked in January, April, July and October, and changes in employment (and related changes in unemployment and the unemployment rate) between November and February. This can be seen in charts 3, 4 and 5.

Source: Labour Force, Australia, Table 19

Source: Labour Force, Australia, Table 1

Source: Labour Force, Australia, Table 1

The following two sections provide some examples of factors that have contributed to larger-than-usual month-to-month changes in seasonally adjusted hours worked and employment series. 

Understanding seasonality in hours worked and employment

Understanding seasonality in hours worked: looking at when people take leave

Understanding seasonality in employment: looking at when people start of return to jobs

How to contact us with questions or feedback

The ABS always seeks to provides guidance and advice to users to assist in interpreting changes in Labour Force data over time. 

If you have a question or have any feedback on additional explanatory that would be useful, please contact us at labour.statistics@abs.gov.au.