APCP-T(26)02: Variances update

Advisory Panel on Consumer Prices Technical Panel 17 April 2026 presentation slides

Members in attendance

• Mr Grant Fitzner (ONS, Chair)
• Professor Paul Smith
• Dr Jens Mehrhoff
• Mr Rupert de Vincent-Humphreys
• Dr Martin Weale
• Dr Jan de Haan
• Professor Huw Dixon

Presenters

• Kevin Sharp (ONS)
• Professor Paul Smith

Secretariat

• Ms Aimee North (ONS)
• Ms Rifah Abdullah (ONS)

ONS attendees

• Mr Chris Payne (ONS)
• Mr Alun Rees (ONS)

Apologies

• Ms Frances Krsinich
• Mr Peter Levell
• Dr Bernhard Goldhammer
• Mr Mike Hardie (ONS)

1. Introduction and apologies

1. Mr Fitzner opened the meeting as Chief Economist and passed on apologies from members unable to attend.

2. CPI Variances update

1. Professor Paul Smith provided an update on his ongoing work to calculate variance of consumer price index estimates. He hopes to bring a more detailed paper with estimates to the panel later this year.
2. Professor Smith explained the complexity of calculating variance on UK consumer prices index (CPI) 12-month rate as there are no explicit product weights at an elementary aggregate level nor sampling weights. He explained that, internationally, countries such as the US do not use explicit weights in their variance calculations. He suggested that there are two paradigms to explore for UK CPI variance, a design-based approach which considers the effect of selecting different representative items for the basket or a model-based approach which considers variance compared to an assumed model.
3. Professor Smith further explained that variance due to prices can be calculated using model-based estimators however this would not consider the effect of replication factors, where a weight of greater than one is assigned to multinational retailers with national pricing. A design-based model would consider the variability of weights however the weights used in the CPI from the Living Costs and Food survey are adjusted prior to use. He suggested that a combination of both approaches would be most effective however this is difficult in practice as they are on different frameworks. Additionally, Professor Smith asked the panel on how best to explain either approach to calculating variance to non-technical users.
4. A panel member asked for clarification on replication factors and their relations with the shop weight. They highlighted there is still some randomness in choosing the product to represent the item in a retailer with a replication factor greater than one.
5. Another panel member questioned how chain-linking and variance from base price to current price will be considered as the variance increases as time goes on. Prof Smith explained that they are developing a statistic where the chain-linking will address these every 12 months. The panel member then questioned whether variance estimates will be produced for other CPI measures outside of the 12-month rate. Professor Smith explained that the current focus is only the 12-month changes as each approach is very resource intensive however there is potential to explore other measures in the future.
6. A panel member agreed that model-based approach seemed sensible but questioned whether a superpopulation model is needed especially when considering the inclusion of scanner data. Another panel member highlighted that non-technical users would focus more on whether the variance is significant rather than the methodology used to calculate it, especially if the calculated difference between the two approaches is minimal. They suggested that it is more important to communicate the difference between uncertainty and unreliability. Professor Smith explained that they have not yet looked at scanner data as part of their research but plan to.

3. Proposals and options to produce and publish seasonally adjusted CPI and CPIH

1. Mr Kevin Sharp presented an update on ONS’ plans to produce a seasonally adjusted Consumer Prices Index (CPI) and Consumer Prices Index including owner occupiers’ housing costs (CPIH) series, following on from the work done by the National Institute of Economic and Social Research (NIESR). Mr Sharp outlined the stakeholder panel’s preference for the indirect aggregation method, which is also used in other areas of ONS, the suggested revision policy for a seasonally adjusted series and how policy impacted events would be approached. He requested feedback on the technical aspects of the proposal.
2. One panel member suggested following the direct approach and providing a note for users explaining that the lower level seasonally adjusted components do not add up to the seasonally adjusted headline figure. This would remove potential residual seasonality from aggregating the more volatile lower-level series.
3. A panel member questioned the use of SEATS software vs X11 and on how outlier calendar effects will be approached. They raised that there will be volatility when applying the indirect approach at a 4-digit COICOP level, however that at a 2-digit level it would be stable, although more inconsistent with aggregation in lower levels. The panel member also questioned whether there would be a concurrent revision policy. Mr Sharp replied that JDemetra+ software will apply some of the outlier analysis that may be missed using X-13 software. He agrees there is a trade-off and the current preferred revision window is referred to as partially concurrent. Ms North explained that House Price Index (HPI) estimates follow a similar process. Mr Rees explained that there is work to develop a custom specification with automation included for this such as calendar effects and outliers.
4. A panel member suggested treating policy events as outliers so that the impact is still seen in the index but not adjusted away. Another member questioned how the difference between a policy effect and seasonal effect would be determined if policy announcements are seasonal. A panel member countered that regular policy effects should be removed as they are seasonal and only irregular policy effects should be kept in. Mr Rees confirmed that ONS policy will account for regular changes.
5. When the panel were asked on using time series breaks, they stated it depends on the data therefore it may be prudent to conduct more analysis on the time series to assess where would be best for structural breaks in the series. The panel cautioned a continuous treatment of historic data especially as the time series goes back to 1988 and agreed a spilt would be best practice.

4. Any other business

1. Publication of papers
   • Papers to be published alongside the minutes and the slides.
2. Reporting lines for Advisory Panels
   • Mr Fitzner explained that as there is no National Statistician in post, the advisory panel will report to James Benford the Director General for Economic Statistics for ONS in the interim.
3. Name change for Advisory panels (Grant)
   • Mr Fitzner proposed changing the panel name to reflect the current wider prices remit of the panel and to become Advisory Panel on Price Statistics – Technical (APP-T). It was clarified that this would just be a change in name only, and not activities of the panel, and was accepted.
4. Travel guidance for panel members
   • Ms Abdullah outlined the travel reimbursement policy for panel members and is to circulate a copy for members to look through in their own time.
5. Suggestions for future panel meeting topics
   • Ms Abdullah requested the panel members to provide suggestions of future topics ONS could bring to meetings.

Action:

Secretariat to circulate panel travel reimbursement policy and link to information on how Living Costs and Foods survey data is used in National Accounts.