Chapter 4: Coherence of Public Services Productivity Published Data and the Demand for Faster Statistics

The Office for National Statistics (ONS) publishes a range of statistics on public service productivity (PSP). This has been developed incrementally to meet user needs in the years following the Atkinson Review.

The ONS publishes two estimates for PSP. Methodological differences between the annual and quarterly statistics are explained in Section 9 of Sources and methods for public service productivity estimates:

• Annual estimates which are badged as accredited National Statistics with data on a two-year time lag and are quality adjusted.
• Quarterly estimates which are badged as Official Statistics in development; these are timelier but experimental, and only published for the most recent time periods. Unlike the annual series, these hold quality adjustment constant from the last quality adjusted annual period available, accounting for published quality levels, but not recent growth.

One of the key issues identified by users is the time lags involved in sourcing some of the data required for producing the annual estimates, particularly the quality measures.

The second is the complexity in interpreting the published quarterly and annual estimates given methodological differences, and the degree to which these present a unified and coherent dataset.

4.1 Coherence of ONS public service productivity publications

Coherence is a cross-cutting theme of the Code of Practice for Statistics, which states that:

“producers must demonstrate that they do not simply publish a set of numbers, but that they explain how they relate to other data on the topic, and how they combine with other statistics to better explain the part of the world they describe.”

The Review has considered how coherence can be improved across the ONS suite of publications on PSP, and with other ONS productivity publications, given user interest in comparison with whole economy and market sector productivity.

Using the Office for Statistics Regulation’s framework for achieving coherence, a review of web analytics has provided insight into how the publications are currently used. A user engagement survey launched in January 2025 will identify changes needed to optimise the purpose of each publication and the relationship between them, to improve the clarity of the overall narrative.

4.2 Work programme to improve coherence and timeliness of ONS estimates

The Review found that the post-review steps needed to improve the coherence and timeliness of the ONS PSP estimates were:

• Development of ‘nowcasting’ methods aiming to bring changes in quality into the most recent quarterly time periods, and to address the two-year time lag in availability of annual estimates. This was a key priority for the Review, and nowcasts were initially incorporated from November 2023, then methods were reviewed in December 2024 and updated from February 2025. This is described in more detail in the section Nowcasting methodologies for quality adjustment.
• Taking faster account of quality adjustment data with shorter time lags (for example, education data are available after one year, but is implemented alongside other quality data with a two-year lag).

• Improvement of the quarterly experimental PSP publication to include disaggregation for individual services. The ONS introduced publication of quarterly experimental service estimates starting with the largest sector (Healthcare) from February 2025 with the aim of rolling out to other sectors.

• Replacement of the current ‘contribution to growth’ compilation method with that of ‘chain volume measures’ to better align quarterly and annual compilation methodology in line with UK National Accounts best practice. The differences in methods in the historic series make interpreting the published quarterly and annual estimates relatively complex. Addressing this will allow the ONS to present a more unified and coherent dataset. This work is underway.
• Reconciliation of the quarterly estimates with the annual accredited National Statistics PSP estimates. For the years covered by the annual publication the sum of the quarterly series will equal the annual figure (known as ‘benchmarking the quarterly series to the annual series’). Quarterly data after this point follows this new annual level. This is in line with the ONS’ UK National Accounts protocols and will simplify understanding for users in interpreting the growth shown in the quarterly and annual estimates.

4.3 Nowcasting methodologies for quality adjustments

The timeliness of outcome data, which are frequently related to longer-term changes for citizens means that annual estimates are currently produced with a two-year time lag. For example, the measure of re-offending used counts the number of offenders who commit further offences in the year following the end of their sentence. A further year is then given for the courts to complete the process of re-conviction. These are explained in Section 9 of Sources and methods for public service productivity estimates. The Review therefore prioritised the development of ‘nowcasts’ (estimates for the years annual data are not yet available).

Nevertheless, prior to the Review, the ONS already published quarterly estimates, holding quality constant at the last published date, alongside ‘annualised’ estimates (the sum of the relevant four quarters) at the end of each calendar year as a ‘nowcast’ to provide timelier annual estimates for users.

However, this approach has limitations. Although both the annual estimates and annualised quarterly estimates track changes in the quantity of services delivered, only the annual series includes an up-to-date picture of quality adjustment. Moreover, the quarterly indices (which are in themselves experimental) have lower coverage and are less granular than the accredited National Statistics annual estimates.

The inference from using an annualised approach of the quarterly time series is that the annual, quality adjusted productivity series will follow the same path as the non-quality adjusted series (derived from the quarterly estimates), implying that quality increases or falls over time with a mean of zero, whereas the Review knows it has a generally upward path.

With the aim of improving upon this the Review developed and implemented a ‘nowcast’ method to produce new experimental estimates of annual PSP, incorporating the ability to estimate the likely change in quality adjustment, and started publishing this from November 2023.

The Review has iteratively improved these techniques. The ONS initially produced an experimental modelled estimate using Dynamic Regression (DR). This is an extension of autoregressive integrated moving average (ARIMA) modelling which allows inclusion of predictor variables. Unlike forecasting, which relies heavily on projections and assumptions about the future economic situation, the DR approach uses data on annual quality adjusted productivity trends for previous years, and information on more recently published quarterly PSP, to provide timelier estimates of annual PSP.

To address the two-year time lag, these nowcasts are a product of the observed annualised quarterly series for the latest two years, and the relationship between the observed annual series (including quality adjustments) and annualised quarterly series preceding those years (excluding quality adjustment).

The publication of the 2021 accredited official annual PSP statistics allowed assessment of the accuracy of the initial DR model. It was found that the nowcast estimate was an over-estimate, particularly in some service areas. Given nowcasts are reliant on recent data, the pandemic and its impact on the UK economy made this nowcasting exercise particularly challenging. The DR model under-estimated the effect of the pandemic. Consequently, the Review analysed options to further improve this approach, testing different treatment of 2020 (the year most affected by the pandemic).

In December 2024 the Review published an article recommending a model utilising Quarterly cumulative Average Growth Rates (QAGR), which in general performed better in terms of accuracy than dynamic regression. Productivity nowcasts for total productivity, and the Healthcare and Public Order and Safety services saw improvements using the QAGR method. However, the new Education nowcast continued to over-estimate growth relative to published accredited annual statistics.

From February 2025, in quarterly publications, the ONS has published experimental annual nowcast estimates using the QAGR approach. The performance of the two most optimal methodologies in terms of accuracy (QAGR and DR) should be reviewed on an annual basis to ensure this conclusion remains valid as the analysis moves further away from 2020 (the biggest impact from the pandemic). Future developmental work could also include exploring other nowcast methodologies such as a hybrid approach or state-space models. For further information please see Developing nowcast methodologies for public service productivity, UK”.

4.4 Future work

The work programme developed by the Review is a substantial programme of methodological work that has been started and partially implemented by the Review as described in section 4.2. An associated plan to deliver the work programme has been developed for financial year ending 2026. Delivery will require resources to conduct this development work in parallel with production and publication of estimates using the existing methodologies until they can be replaced.