ONS’s approach to seasonal adjustment

Context

Key economic statistics produced by ONS that are seasonally adjusted include GDP and its components, and various labour market statistics. ONS is investigating the potential for seasonally adjusting the Consumer Price Index (the House Price Index is already seasonally adjusted).

Seasonal adjustment of GDP is undertaken through a “bottom up” approach, meaning that many component series are seasonally adjusted individually. Monthly or quarterly seasonal adjustment is carried out automatically within processing platforms using previously selected models. Headline aggregates are therefore adjusted indirectly, as they are the result of combining seasonally adjusted components. After this, the aggregates are checked for any residual seasonality, and the models for most outputs undergo an annual manual review.

Each year there is a comprehensive update of the National Accounts datasets, including monthly and quarterly GDP, as part of Blue Book and Pink Book changes. During this update, datasets are revised historically to ensure comparisons can be made over time, with the consequence that seasonal adjustment parameters are also reviewed. There is then a separate analysis of revisions to help inform users of revision properties for GDP aggregates.

In September 2025, ONS published an article on the effectiveness of its seasonal adjustment of quarterly GDP. The analysis reported in the article confirmed that, using ONS’s standard statistical tests, no statistical evidence was found of residual seasonality in the revised datasets for quarterly (or monthly) aggregate GDP.

The article was an update of an original article published in spring 2025, which was produced partly in response to external user claims about the potential presence of residual seasonality in quarterly GDP over recent years. Its publication also reflected ONS’s desire to become more transparent about how GDP is seasonally adjusted and to improve its communications with users.

ONS plans to provide a regular update each year as the National Accounts are revised.

For many years, ONS has also published a separate detailed revisions analysis of GDP estimates. This provides a comprehensive statistical assessment of the differences between first and revised GDP estimates.

Organisational factors

ONS has a centralised methodology team for time series analysis, including seasonal adjustment, rather than a devolved structure (although centralisation is not complete, with some expertise located within the teams compiling the component series). ONS considers that this structure promotes a degree of independence and challenge.

The central team is responsible for reviewing and applying updates to time series parameters. It has oversight of around 10,000 time series across ONS. Effective operation depends on good and collaborative working relations with compilers, who have sector-specific expertise. There is significant ad hoc demand for reviews, often driven by the economic statistics production schedule, where series with the largest impact on the published statistics are identified and are required to be assessed by time series experts in a short time frame.

Seasonal adjustment is a complex topic requiring deep expertise, often developed over many years. ONS is fortunate to have internationally recognised experts in this topic.  While those with the highest level of expertise do not currently work directly in the ONS time series area, they have provided detailed technical support and are in regular contact with the ONS time series team to help both in quality assurance and provide advice when needed. There is also a time series user group across ONS which brings together wider experience and meets on regular basis to share knowledge.

ONS has faced challenges in meeting the level of need, in transferring skills, and in building and retaining capability. The latter issues partly reflect more general concerns about the adequacy of incentives for recruiting and retaining specialist expertise, both within ONS and across government.

Over the last few years, there has been a loss of expertise within the central time series analysis team, but it is now being re-built and expanded. This recent expansion of the team reflects the allocation of additional resources by methodology senior management after this issue was raised by National Accounts teams.

The team is currently engaged in a complete review of the service provided. For example, the team is considering whether it may be possible to reduce pressures by changing the timetable for some reviews, so that the work is spread more evenly across the year. A new plan is being prepared, and the team is also considering ways to promote continuous improvement.

As part of its plan, the central team intends to revert to a model of annual reviews of seasonal adjustment, following the suspension of this approach in some areas due to resource constraints. The proposed model would be more flexible, extending existing capacity for reviews to take place more frequently if issues arise in high-risk and complex areas.

ONS has told us that recent specific steps to strengthen the central team’s capacity, capability and ways of working, have included:

• Major expansion of the team, although some vacancies remain. We were told that the team has expanded from 5 in April 2025 to 12 by February 2026, with a further 4 currently being onboarded.
• Measures to deepen the technical capability of the team (through increased training and development) and to improve guidance and documentation.
• Improvements to processes, including the more systematic assessment of risk, and increased automation (for example the processes that check for residual seasonality).
• The review of the service, which has started to identify areas for improvement, with the re-establishment of an expert group reflecting an early achievement.

In particular, since the initiation of this review ONS has taken additional steps to address emerging concerns about residual seasonality in GDP. These have included:

• Enhanced engagement with key stakeholders and external experts on this issue.
• An increased pace at which components series identified as requiring further investigation are reviewed.
• The initiation of a review of potential supplementary statistical tests for residual seasonality, with the results to be published (with a target date of autumn 2026).

Following this engagement and a recommendation from OSR, ONS commissioned an externally led review of its approach to residual seasonality.

Methodology and associated challenges

The software currently used to seasonally adjust (X-13-ARIMA-SEATS), based on moving averages, is robust and to an internationally endorsed standard. ONS has confirmed that the main alternative, a modelling-based approach (TRAMO-SEATS), gives broadly similar results and is also widely endorsed.

The current approach is well-documented, and ONS has high confidence in the methodology. However, as described above, ONS has faced challenges in resourcing the required parameter setting and updating processes at the component series level.

Detecting emerging problems with residual seasonality in real time is challenging – statistical methods need at least three years, and usually five, to detect effects. Hence, there is an important role for judgement in setting parameters and updating processes. This underscores the need for effective engagement between methodologists and the compilers, who have relevant contextual knowledge and expertise.

Subject to the completion of ongoing work to improve and increase resources in these areas, and the conclusions of the externally led review, as described above, ONS judges that the approach to seasonally adjusting low-level series is reasonably robust.

However, the approach does require closely monitoring the higher-level aggregates each month and quarter, to complement the annual review process as part of Blue Book and Pink Book changes.

Residual seasonality in higher-level aggregates

ONS states that its attention to residual seasonality in top-level aggregates has increased in recent years. GDP and some other prominent aggregates are now reviewed on a regular basis as part of the internal quality assurance process.

In the context of the regular monthly and quarterly production rounds, ONS conducts “curiosity” (or “data challenge”) sessions. These sessions bring together domain topic leads with statistics and economic experts. As part of these data confrontation sessions, additional analysis can be requested before the data are finalised. For example, movements for a series may suggest either emerging seasonality or require treatment for emerging one-off impacts. The seasonal adjustment parameters are then updated by experts in the time series methodology team, systems are rerun by production teams, and any updates are applied.

However, ONS acknowledges that there may be scope to do more, including, for example, introducing reviews of residual seasonality in additional series, such as within expenditure and income components. ONS also acknowledges that it may be possible to make further improvements through learning from other national statistical institutes and is planning to do so as part of its service review.

Current tests for residual seasonality are applied to the latest vintage of the data, which incorporates revisions to earlier periods. This approach has a limitation, because key decisions often rely on the first-release (unrevised) estimates for the most recent period. Those initial estimates can draw on different source data and collection frequencies (for example, higher reliance on modelled or survey-based early returns and short-term indicators) than later “mature” vintages and may therefore exhibit seasonal properties that differ from the revised series. In light of this, ONS’s view is that assessment of residual seasonality in first-release data is best addressed within the revisions review process, alongside analysis of how seasonal features evolve as estimates are updated.