Official Statistics

DSIT Public Engagement Survey 2025/2026: technical report

Published 16 July 2026

1. Introduction

1.1 Background

The Department for Science, Innovation and Technology (DSIT), focuses on improving people’s lives by maximising the potential of science and technology. DSIT priorities are to:

1. Accelerate innovation, investment and productivity through world-class science, research and development. 

2. Use technology for good by ensuring that new and existing technologies are safely developed and deployed across the UK, with the benefits more widely shared. 

3. Drive forward a modern digital government which gives citizens a more satisfying experience and their time back.

DSIT’s responsibilities include: 

• positioning the UK at the forefront of global scientific and technological advancement 

• driving innovations that change lives and sustain economic growth 

• delivering talent programmes, physical and digital infrastructure and regulation to support the economy, security and public services 

• R&D funding 

DSIT commissioned Verian to develop and conduct a new nationally representative ‘push-to-web’ survey to assess adult (16+) engagement with DSIT policy areas across the UK. This technical report documents the development and implementation of the first year of the survey (2025/2026 financial year), known as the Public Engagement Survey (PES). A second year of the survey will be conducted by Verian in 2026/2027.

1.2 Survey Overview

The scope of the PES is to deliver a UK representative sample of adults (aged 16 years and over) to assess engagement with DSIT policy areas. 

DIST had undertaken an initial scoping exercise (prior to the involvement of Verian) to identify the following broad topic areas for inclusion in the survey: 

• Technology adoption (including barriers and enables) 

• Trust and acceptability in technology 

• Digital inclusion 

• Emerging technologies, including Artificial Intelligence (AI) 

• Digital skills 

• Disinformation and misinformation. 

The data collection model for the PES is based on ABOS (Address Based Online Sampling) a type of ‘push-to-web’ survey method. Respondents take part either online or by completing a paper questionnaire. The target number of interviews for Year 1 of the survey was 31,500 and to be large enough to allow for International Territorial Level 2 (ITL2) geographical representation.

Code of Practice for Statistics

The DSIT Public Engagement Survey results have been used to produce official statistics. The data collected has been processed, stored and published in compliance with DSIT’s policies for compliance with official statistics. This is to ensure that the statistics serve the public good and provide insight to inform understanding and shape action. The departmental policies are underpinned by the Code of Practice for Statistics.

1.3 Survey design

The PES was designed to be conducted via an online and paper questionnaire using Address Based Online Surveying (ABOS). This approach enables broad population coverage, using the Royal Mail’s Postcode Address File (PAF), provides comprehensive coverage of the population, with more than 98% coverage of the UK resident population. The mixed mode implementation (web and paper modes) enables participation among those who are not online. In combination, this approach provides a scalable framework for population estimates, while remaining more cost-efficient than face-to-face methods. ABOS is also sometimes referred to as ’push-to-web’ surveying. 

The ABOS design is relatively simple: a stratified random sample of addresses is drawn from the Royal Mail’s postcode address file (PAF) and an invitation letter is sent to each one, containing username(s) and password(s) plus the URL of the survey website. Sampled individuals can log on using this information and complete the survey as they might any other web survey. Once the questionnaire is complete, the specific username and password cannot be used again, ensuring data confidentiality from others with access to this information.  

PES followed this approach, and after invitation letters were sent, a first reminder letter was sent to households where either no responses had been received or where responses had been received from at least one household member but not all household members. In line with standard ABOS design, the paper questionnaire was only offered on request in the invitation and first reminder mailing. It was then given more prominence in the second reminder mailing, being included for a targeted set of households (these were also randomly distributed to the remaining subset of households). 

Paper questionnaires provide coverage of the offline population and are especially effective with sub-populations that respond to online surveys at lower-than-average levels. However, paper questionnaires have measurement limitations that constrain the design of the questionnaire and also add considerably to overall cost.

2. Sampling

2.1 Sampling approach

Sample design: addresses

The address sample design is intrinsically linked to the data collection design (see Section 2.2) and was designed to yield a respondent sample that is representative with respect to neighbourhood deprivation level, and age group within each of the 46 ITL2 areas across the UK. This approach was intended to improve the representativeness of the achieved sample across key geographic and demographic groups, reducing the extent to which weighting is needed to correct for differential response. By building these features into the sample design, the approach helped to support more precise estimates than would be expected from a less targeted design, subject to the effects of weighting and differential responses. 

The design sought a minimum respondent sample of 650 for each ITL2 area, equivalent to an overall respondent sample of 30,000.

Stratification and master sample selection

As a first step, a stratified master sample of 255,560 addresses in England, Wales, Scotland and Northern Ireland was drawn from the Postcode Address File (PAF) ‘small users’ subframe, which primarily comprises of residential delivery points. The sample used this frame as provided and no additional bespoke filtering was applied at the sampling stage (e.g. to exclude specific address types). 

Before sampling, the PAF was disproportionately stratified by ITL2 area (46 strata) and, within area, sorted by (1) lower tier local authority, (2) neighbourhood deprivation level (five groups, each of a similar scale at the national level), (3) super output area, and finally (4) by postcode. This ensured that the master sample of addresses was geodemographically representative within each stratum.

Use of auxiliary data

This master sample of addresses was then augmented by data supplier CACI. For each address in the master sample, the data provided an estimate of the expected number of resident adults in each ten-year age band. Although this auxiliary data will have been imperfect, investigations by Verian have shown that it is highly effective at identifying households that are mostly young or mostly old. Once this data was attached, the master sample was additionally stratified by expected household age structure based on the CACI data^{[footnote 1]}: 

• all aged 35 or younger (15% of the total) 

• all aged 65 or older (21% of the total) 

• all other addresses (64% of the total) 

The conditional sampling probability in each stratum was varied to compensate for (expected) residual variation in response rate that could not be ‘designed out’, given the constraints of budget and timescale. The underlying assumptions for this procedure were derived from empirical evidence obtained from the 2023/24 Department for Culture, Media and Sport (DCMS) Participation Survey.

Sample selection

Verian drew a stratified random sample of 106,612 addresses from the master sample of 255,560. This selected sample was then divided into 20 equal-sized ‘replicates’, designed to have the same profile and scale (5,330 to 5,331 addresses). The use of replicates allowed for the sample to be issued in stages: the first 15 replicates would be issued (that is, 79,962 addresses), with the remaining 5 kept back in reserve and could be released if additional sample was needed. 

Sample was issued in three batches across the year. Sample productivity was reviewed towards the end of batch one to inform the sample issue for batch two, and again towards the end of batch two to inform the composition of batch three. These reviews were carried out at an ITL2 area level, leading to some differences between what was planned at the start of the year and what was issued in practice. 

In total, 83,694 addresses were issued: 3,732 more than planned (+5%). These were distributed as follows: 26,656 were issued in batch one, 33,315 in batch two, and 23,723 in batch three. 

Table 2.1 shows the combined issued sample structure with respect to the major ‘design’ strata: neighbourhood deprivation level and expected household age structure.

Table 2.1: Number of addresses issued by area deprivation quintile group and expected household age structure

Expected household age structure	Most deprived	2nd	3rd	4th	Least deprived
All household members aged 35 or below	4,223	4,246	3,237	2,624	1,822
All household members aged 65 or above	3,355	3,780	4,954	4,756	4,004
Other	15,840	14,692	12,982	11,437	9,270

Sample design: individuals within sampled addresses

All resident adults aged 16+ were invited to complete the survey. In this way, the DSIT Public Engagement Survey avoided the complexity and risk of selection error associated with remote random sampling within households. 

However, for practical reasons, the number of logins provided in the invitation letter was limited. The number of logins was varied between two and four, with this total adjusted in reminder letters to reflect household data provided by prior respondent(s). Addresses that CACI data predicted contained only one adult were allocated two logins; addresses predicted to contain two adults were allocated three logins; and other addresses were allocated four logins. The mean number of logins per address was 2.7. A maximum of four logins could be provided per address, no additional logins were issued beyond this cap. Paper questionnaires were available to those who are offline, not confident online, or unwilling to complete the survey this way.

2.2 Details of the data collection model

Table 2.2 summarises the data collection design within each principal stratum, showing the number of mailings and type of each mailing: push-to-web (W) or mailing with paper questionnaires (P). For example, ‘WWP’ means two push-to-web mailings and a third mailing with paper questionnaires included alongside the web survey login information. In general, there was a two-week gap between mailings.  

Table 2.2: Initial data collection design by principal stratum

Expected household age structure	Most deprived	2nd	3rd	4th	Least deprived
All household members aged 35 or below	WWPW	WWWW	WWWW	WWW	WWW
All household members aged 65 or above	WWPW	WWPW	WWP	WWP	WWP
Other	WWPW	WWW	WWW	WWW	WWW

As part of the Year 1 design, paper provision was randomly allocated across a subset of strata. Paper was randomly allocated in two thirds of the addresses in the most deprived quintile and in older addresses (all aged 65 or above), and one third of all other addresses. The table below shows the allocation of paper across strata. 

Table 2.3:  Paper allocation by principal stratum

Expected household age structure	Most deprived	2nd	3rd	4th	Least deprived
All household members aged 35 or below	2/3	1/3	1/3	1/3	1/3
All household members aged 65 or above	2/3	2/3	2/3	2/3	2/3
Other	2/3	1/3	1/3	1/3	1/3

3. Questionnaire development

The PES questionnaire was developed by Verian in consultation with DSIT. This chapter details the initial development of the questionnaire and the cognitive testing phases.  

The questionnaire was developed over several stages over the period June 2025 to November 2025: 

• review of scoping materials to help define the initial scope (June 2025) 

• stakeholder engagement sessions to further refine the scope (June 2025) 

• two rounds of cognitive testing (August/September 2025) 

• further modifications to the questionnaire following analysis of the pilot data (September 2025) 

• development of the paper version of the questionnaire (October/November 2025) 

Due to timetable constraints, the pilot fieldwork (9th September 2025 to 20th October 2025, see Chapter 4) ran concurrently with the cognitive testing.

3.1 Content scoping and stakeholder engagement

As the survey was to form the baseline wave of a new survey series it was important to develop the survey content with a view to longer-term policy objectives. 

Prior to the commencement of the 2025/2026 survey contract, DSIT had commissioned Ipsos to conduct a stakeholder engagement exercise involving interviews with 24 internal and external stakeholders. This included a review of evidence needs for the new survey, drawing on content from the DCMS Participation Survey and DSIT Public Attitudes to Data and AI tracker survey (which had previously been vehicles for tracking some of these topics) as well as scoping new topic areas for development.  

Verian reviewed this previous work alongside further scoping documents from DSIT and agreed with DSIT a broad list of topic areas for potential inclusion in the survey.  

The next stage was to conduct a series of workshops with different teams within DSIT to discuss and clarify the research objectives in more detail.  For each stakeholder workshop, Verian created an agenda and topic guide which included discussion topics such as the policy context, priority information needs, and how to ensure questions can be future-proofed as far as possible. 

Workshops with the following teams were conducted in June 2025:  

• GO-Science  

• Government Digital Services  

• Digital Economy  

• AI and AI regulation  

Additional teams that were also interested in contributing the survey were engaged with by other means at this stage. 

Following the workshops, Verian compiled a map of all topics identified as potentially in scope and stakeholders from each team indicated which areas were higher or lower priority and provided commentary against each topic of interest.  

Following this, Verian compiled a set of agreed initial measurement objectives which provided the starting point for question development.

3.2 Drafting of questions (pre-testing)

After confirming the objectives of the questionnaire, Verian developed a draft set of questions based on the agreed outline in consultation with DSIT and wider stakeholders. To aid this process, Verian conducted a rapid desk review of other relevant surveys to source existing questions that could potentially be used, adapted or updated. For new and emerging policy topics, Verian drafted questions from scratch. The draft was built iteratively, working closely with DSIT and stakeholders to refine and streamline the content.  

Prior to commencing any pre-testing, a draft of the questionnaire was circulated to the stakeholder groups to incorporate any final feedback. The questionnaire was then signed off for cognitive testing.

3.3 Cognitive testing

Cognitive testing describes a process of testing survey questions to ensure they work as intended before they are included in the main stage questionnaire. Interviews are carried out by members of the research team trained in cognitive testing who ask participants the new questions and then spend some time discussing their answers. Based on an agreed probe guide, the researcher probes to check how easily the participant can understand each question and explores how they decided on their answer. This provides valuable insight into how participants understand, process and respond to survey questions, and can help refine question wording and response lists.   

Two rounds of cognitive testing were conducted in August and September 2025 focussed primarily on the newly developed questions. In each round, 12 interviews were conducted (24 in total) according to quotas set on age, gender, region, level of education and digital literacy. Most interviews were conducted remotely via Zoom interface, although some interviews with the more digitally excluded (who were typically older in profile) were conducted face-to-face as not all of this group would have been able to join an interview on Zoom, and the views of the digitally excluded were especially important in this context.  

All participants were recruited by Acumen Fieldwork, a supplier of fieldwork recruitment services working on behalf of Verian. Participants received an appropriate payment in appreciation of their time.  

To manage the volume of questions to test, at each round the content was split into two halves which means that all questions were tested by at least 12 respondents over the two rounds. 

The detailed findings and recommendations from the cognitive testing were reported back to DSIT. In the most part, the testing led to recommendations for minor modifications in wording, although some more substantive issues were identified.  In summary, the main issues focussed around the following areas of questioning: 

• some questions on digital skills and digital access needed to be amended or simplified to account for lower levels of comprehension of some terminology among the more digitally excluded 

• there were challenges around accurately capturing how people accessed home internet given confusion associated with broadband packages involving other services and mobile broadband. These questions were restructured and simplified as a result 

• some questions on attitudes to scientific topics were found to be too challenging in terms of the concepts they covered, and were either amended or removed following cognitive testing 

• some questions on attitudes to AI were challenging: 

  • there was a wide range of knowledge about AI and some misperceptions about what ‘counts’ as AI with some respondents including non-AI products such as non-AI chatbots  

  • there was also found to be a need to better differentiate between ‘active’ (i.e. deliberate) use of AI and more passive forms of AI such as AI-generated web search summaries, AI chatbots and digital assistants such as Alexa.  As a result, questions were further clarified and questions on how people use AI were narrowed to those who actively used generative AI such as LLMs, picture/video editing AI, and agentic AI  

  • some questions seeking people’s attitudes towards more complex areas of AI regulation and AI use in society were also found to be problematic and were either amended or removed following testing 

After each round, findings were discussed, and changes were implemented. Changes from the first round were re-tested at the second round of testing.

3.4 Final questionnaire content

Following completion of the pilot survey (see Section 4), some further minor changes were made to the questionnaire following inspection of the pilot data.   

Interim analysis of the pilot data included a review of all multi-code response codes, and any that were selected by less than 2% of respondents were flagged for removal.  Some questions were identified as attracting very high rates of ‘don’t know’ responses. These were generally questions that had also been detected as problematic in cognitive testing, and these findings in combination led to a recommendation to remove these questions from the survey.   

Following the implementation of changes from round 2 of cognitive testing and analysis of pilot data, the online version of the questionnaire was finalised and signed off by DSIT in September 2025.  

A summary of final topics covered in the questionnaire is included below. As explained further in Section 3.5, the paper version of the questionnaire contained a reduced set of AI questions for practical reasons, and this is also indicated below. 

Topics asked about in the digital inclusion and skills section included: 

• frequency of internet use 

• where internet is used/use of Wi-Fi outside home 

• basic digital skills 

• reasons for not using the internet or not using it more often 

• participation in digital or online upskilling 

• perceived benefits and importance of improving digital skills 

Topics asked about in the internet connections and devices section included: 

• methods of connecting to the internet at home 

• cost, speed and quality of broadband/internet connections 

• ownership of devices that are able to access the internet 

• mobile data usage and cost 

• awareness of Fifth Generation (5G) mobile technology 

Topics asked about in the government digital services section included: 

• use of government digital services in the last 12 months 

• methods of accessing services 

• satisfaction with services and problems experienced while using services 

Topics asked about in the attitudes towards science, technology and data section included: 

• engagement with science 

• attitudes towards data use in society 

• trust in the safe handling of personal data by different organisations 

• trust in the accuracy of information about science and technology across different organisations  

• attitudes towards scientists and their role in public policy 

• preferred sources of information about science and health 

Topics asked about in the AI section included: 

• awareness of AI 

• extent of use of different forms of AI 

• contexts in which AI is used, mains reasons for use, self-rating of AI skills capability (online questionnaire only) 

• barriers to AI use (online questionnaire only) 

• perceived benefits and risks to of AI to society 

• perceived importance of government regulation of different aspects of AI  

• concerns about AI-driven misinformation 

In addition to these topics, a number of demographic questions were asked in the survey, including: 

• age 

• gender 

• number of children in household 

• tenure 

• educational qualifications 

• ethnicity 

• working status and occupation 

• disability  

• household income

3.5 Paper questionnaire development

Creation of the paper questionnaire

Once the content was signed off for the main online survey, the content and layout for the 24-page paper survey was agreed.  In the most part the paper survey matched the online survey, but some pragmatic modifications were made to suit the more static printed format and to aid navigation for paper survey users. 

• manual ‘Go to QX’ type routing arrows were added to direct the routing  

• a small number of questions were moved to a different order such that questions with the same routing could be grouped together.  Any changes in order were carefully reviewed to ensure that this would not create any ordering or context effects compared with online  

• complex filters such as ‘‘IF X AND ALSO Y GO TO QUESTION Z’ were avoided on paper. Where there was a complex filter online, the filter was widened on paper with a ‘not applicable’ type code added if appropriate to allow people to easily route themselves out of a block which is not relevant to them.  Where routing differed from online, data editing was used to ensure consistency in data structure across measures 

• the income question was simplified to a single annual measure to reduce the amount of space   

• some of the more detailed questions on AI use were removed on paper.  This was to avoid complex routing as well as to save space. Given the expected skew in paper-based completions to a more digitally excluded audience it was considered that these questions would be less relevant to this audience and many would have skipped them anyway  

Once the paper questionnaire was signed off, a small-scale usability test of the paper document was conducted (see below).

Usability testing of the paper questionnaire

A small-scale test of the paper questionnaire was conducted among six respondents who were selected based on being older in profile and having low digital confidence. This was to reflect the likely nature of people who would choose to do the survey on paper during fieldwork proper. Quotas were set on age, level of digital confidence and gender. All interviews were conducted face-to-face at Verian’s offices so researchers conducting the testing could directly observe participants completing the survey.   

The core aim of the usability testing was to explore how easily participants can navigate the paper questionnaire. Participants were observed completing a draft version of the paper questionnaire and their feedback was collected using a probe guide to ensure consistent coverage of key issues to observe and test. This insight helped to refine and improve the layout of the paper questionnaire to ensure it was fully accessible.  

As a result of the testing, some minor changes were made to the instructions cover page, and to the layout and format of questions.  

The interviews were also used as an opportunity to collect feedback on the draft invitation letters. At the outset of each interview, participants were shown different versions of an invite letter and asked for their feedback. Sharing the letters with the participants also helped to provide context to the survey and helped to mimic the ‘real life’ survey experience. As a result, it was determined which letter version worked best and made some minor adjustments to its wording.

4. Pilot survey

4.1 Overview

The aim of the pilot survey was to test and pilot the survey approach. In particular to provide a robust test of the questionnaire, provide a dry run of the mailing process and provide an indication of response rate. 

To do this, Verian conducted a pilot survey selecting 1,020 addresses using a simplified version of the main stage sampling approach. These were then issued using the same contact protocol as the main stage. However, due to time constraints in the timetable: 

• it was not possible to include paper questionnaires in this pilot stage  

• it was necessary to compress timings for the reminder stages (allowing c.10 days between mailings rather than two weeks)

4.2 Sample design

The pilot survey sample was drawn using a simplified version of the main stage sampling approach. Verian drew a stratified random sample of 1,020 addresses from a larger master sample drawn from the residential subset of the Postcode Address File (PAF) in England, Northern Ireland, and Scotland. The pilot sample was stratified by:

• age segment (based on CACI address-level data): 3 levels 

• deprivation level (based on Index of Multiple Deprivation 2019 LSOA-level data): 5 levels 

Within these 15 strata, PAF addresses were sorted by region and local authority within the region, to maximise the sample’s geographical dispersion. Sampling probabilities within each stratum were varied based on expected response rates as estimated by statistical models developed using response data from the DCMS Participation Survey. For the main stage, the sample would be further stratified by ITL2 area.

4.3 Methodology

The pilot methodology was as follows: 

• 1,020 households were sent a survey invitation inviting up to four members of their household aged 16 and above to take part in the online survey  

• this invitation explained the purpose of the survey and how to take part, provided log in details, detailed the survey participant website where they could go to for more information, and provided a survey email address and freephone number which they could use if they had any questions about the survey or wanted to opt out of the survey  

• up to three reminder letters were sent to households who had not taken part 

• once participants had taken part, they were eligible to claim a £10 incentive as a thank you. They could choose to claim this electronically online or request a paper voucher

4.4 Fieldwork

Pilot fieldwork took place from 9 September 2025 (when the survey invitation letters were despatched) to 20 October 2025.  

After the initial survey mailing, up to three reminders were sent to households, and the posting dates and quantities for these letters are shown in Table 4.1. Due to main stage fieldwork timetable requirements (with main stage fieldwork needing to be completed by the end of March 2026), the pilot fieldwork took place over a shorter period that the main stage fieldwork, with approximately ten days between each survey mailing rather than two weeks. 

Reminders 1 and 2 were sent to all households where not all expected household members had completed the survey (i.e. were not ‘fully’ complete). There were two versions of each reminder letter, one for households where no responses have been received, and one for households where responses had been received from at least one household member but not all household members. As detailed in Section 2.1, in letters to households where at least one response had been received, the predicted number of household members still to take part was updated to reflect the number who had already completed.  

Reminder 3 was a targeted reminder, sent to households who were not ‘fully’ complete, but also known to respond at a lower rate (those in deprived areas or with a younger household composition).

Table 4.1: Posting dates of pilot survey invitation letters and reminders 

Mailing type	Date	Households sent to
Survey invitation	9 September	1,020
Reminder 1	18 September	982
Reminder 2	29 September	949
Reminder 3 (targeted)	9 October	347

Due to time restraints in the development timetable, the pilot did not include the provision of paper questionnaires.

4.5 Response

From the 1,020 addresses issued for the pilot study, a total of 290 individual responses were received from 180 households.

4.6 Interview length

In the pilot, the median survey length was 21 minutes and 24 seconds, and the mean survey length was 22 minutes and 55 seconds.

5. Fieldwork processes

5.1 Contact procedures

All selected addresses were sent an initial invitation letter about the PES. This contained the following information:  

• a brief description of the survey 

• the URL of survey website (used to access the online survey, and to find more information about the survey) 

• a QR code that could be scanned to access the online survey 

• log-in details to the survey for the required number of household members 

• an explanation that participants would receive a £10 shopping voucher on completion of the survey as an incentive for participating

• information about how to contact Verian in case of any queries 

The reverse of the letter featured responses to a series of Frequently Asked Questions, including assurances about the confidentiality of the survey.  

All non-responding addresses were sent two reminder letters, at the end of the second and fourth weeks of fieldwork respectively. A pre-selected subset of non-responding addresses (see table 2.2, in Section 2.2) was sent a third reminder letter at the end of the sixth week of fieldwork. The information contained in the reminder letters was similar to the invitation letters, with slightly modified messaging to reflect each reminder stage.  

As well as the online survey, respondents were given the option to complete a paper questionnaire, which consisted of a reduced version of the online survey. Each letter informed respondents that they could request a paper questionnaire by contacting Verian using the email address or freephone telephone number provided, and a cut-off date for paper questionnaire requests was also included on the letters. In addition, some addresses received two paper questionnaires with the second reminder letter. This targeted approach was developed based on historical data Verian has collected through other studies, which suggests that proactive provision of paper questionnaires to all addresses can displace online responses in some strata. As discussed in Section 2.2, paper questionnaires were randomly allocated in two thirds of the addresses in the most deprived quintile and in older addresses (all aged 65 or above) and one third of all other addresses (as identified by the CACI auxiliary data).

5.2 Survey website and helpline

A survey website was set up for respondents (www.pesurvey.co.uk). Through this website respondents could access the online survey but could also find out more information about the survey. This included frequently asked questions, information around incentives, sources of further advice and information, details about how to content Verian, and the survey privacy notice.  

If respondents had any queries, they could contact Verian either via email, or through calling the survey freephone helpline.

5.3 Incentive management

All respondents who took part in the survey received a £10 voucher as a thank you for taking part.  

Those who completed the online survey were provided with details of how to claim their voucher at the end of the survey, and they were directed to the voucher website (administered by Merit, a specialist online incentives provider). On the website, they could choose from a range of different vouchers.  

Those who completed the paper questionnaire, were provided with a £10 gift voucher by post by Verian. The voucher could be used in a variety of different stores.

5.4 Fieldwork dates

Fieldwork took place between November 2025 and March 2026. The sample was issued in three batches, with the specific fieldwork dates for each batch shown in table 5.1 below. 

Table 5.1: Fieldwork and mailing dates 

Mailing types	Batch 1	Batch 2	Batch 3
Survey invitation	18 November 2025	6 January 2026	3 February 2026
Reminder 1	2 December 2025	20 January 2026	17 February 2026
Reminder 2	16 December 2025	3 February 2026	3 March 2026
Reminder 3 (targeted)[footnote 2]	29 December 2025	17 February 2026	17 March 2026
Fieldwork close	31 March 2026	31 March 2026	31 March 2026

It should be noted that fieldwork for batch 1 and batch 2 samples remained open until the 31 March for any late responders. However, there were no further mailings to households within these batches after the reminder 3 mailing and the vast majority were received no more than two weeks after the date of the final reminder.

6. Fieldwork response

As detailed in Section 5.4, fieldwork for the 2025/2026 PES survey took place between November 2025 and March 2026.

6.1 Fieldwork response rates

In this section, response rates are referred to in two different ways: 

• household response rate – this is the percentage of households contacted as part of the survey in which at least one questionnaire was completed

• individual response rate – this is the estimated response rate amongst all eligible adults that were contacted to complete the survey

Response rates were calculated via the standard ABOS method. An estimated 8% of ‘small user’ PAF addresses in the UK are assumed to be non-residential (derived from interviewer administered surveys)^{[footnote 3]}. The average number of adults aged 16+ per residential household, based on the Office for National Statistics’ (ONS) Labour Force Survey, is 1.89. Thus, the response rate formula:  

• household response rate = number of responding households / (number of issued addresses X 0.92) 

• individual response rate = number of responses / (number of issued addresses X 0.92 X 1.89)

Across the three batches 83,694 addresses were sampled, and in total 32,175 respondents completed the survey - 28,575 via completing the online survey and 3,600 from returning a paper questionnaire. Following data quality checks (see Section 7), 1,477 respondents were removed (1,222 web and 255 paper), leaving 30,698 respondents in the final dataset. 

This constitutes a 25% household-level response rate, and an individual-level response rate of 21%. 

The full breakdown of the fieldwork figures and response rates by batch are shown in table 6.1.

Table 6.1: Combined online and paper fieldwork by batch 

	Number of sampled addresses	Complete responses (online and paper)	Number of households completed	Household response rate	Individual response rate
Batch 1	26,656	9,312	5,899	24%	20%
Batch 2	33,315	13,099	8,344	27%	23%
Batch 3	23,723	8,287	5,309	24%	20%[footnote 4]
Total	83,694	30,698	19,552	25%	21%

The target number of interviews within each ITL2 area was 650. Out of the 46 areas, 650 interviews or more were achieved in 28 areas. Of the remaining 18 areas, at least 600 interviews were achieved in 16 areas, with only two areas where fewer than 600 interviews were achieved (Northern Ireland: 595 interviews, Outer London - West and North West: 554 interviews).

6.2 Interview length

For the online survey, the median completion time was 16 minutes and 37 seconds, and the average completion time was 17 minutes and 28 seconds^{[footnote 5]}.

7. Data processing

7.1 Data management

Due to the different structures of the online and paper questionnaires, data management was handled separately for each mode. The online questionnaire data was collected via the web script and, as such, was much more easily accessible. By contrast, paper questionnaires were scanned and converted into an accessible format. 

For the final outputs, both sets of interview data were converted into csv, with the online questionnaire structure as a base. The paper questionnaire data was converted to the same structure as the online data so that data from both sources could be combined into a single csv file. 

7.2 Quality checking 

Initial checks were carried out to ensure that paper questionnaire data had been correctly scanned and converted to the online questionnaire data structure. For questions common to both questionnaires, the data was compared to check for any notable differences in distribution and data setup. 

Once any structural issues had been corrected, further quality checks were carried out to identify and remove any invalid interviews. The specific checks were as follows: 

1. selecting complete interviews: Any test serials in the dataset (used by researchers prior to survey launch) were removed. Cases were also removed if the respondent reached but did not answer the quality declaration statement (online: QFraud; paper: Q72). 

2. duplicate serials check: If any individual serial had been returned in the data multiple times, responses were examined to determine whether this was due to the same person completing multiple times or due to a processing error. If they were found to be valid interviews, a new unique serial number was created, and the data was included in the data file. If the interview was deemed to be a ‘true’ duplicate, the more complete or earlier interview was retained. 

3. duplicate emails check: If multiple interviews used the same contact email address, responses were examined to determine if they were the same person or multiple people using the same email. If the interviews were found to be from the same person, only the most recent interview was retained. In these cases, online completes were prioritised over paper completes due to the higher data quality. 

4. interview quality check: A set of checks on the data were undertaken to check that the questionnaire was completed in good faith and to a reasonable quality. Several parameters were used: 

• interview length (online check only) 

• number of people in household reported in interview(s) versus number of total interviews from household. 

• whether key questions have valid answers 

• whether respondents have habitually selected the same response to all items in a grid question (commonly known as ‘flatlining’) where selecting the same responses would not make sense 

• how many multi-response questions were answered with only one option ticked 

Following the removal of invalid cases, 30,698 valid cases were left in the final dataset.

7.3 Data checks and edits

Upon completion of the general quality checks described above, more detailed data checks were carried out to ensure that the right questions had been answered according to questionnaire routing. This is generally all correct for all online completes, as routing is programmed into the scripting software, but for paper completes, data edits were required. 

There were two main types of data edits, both affecting the paper questionnaire data: 

1. single-response question edits: If a paper questionnaire respondent had mistakenly answered a question that they weren’t supposed to, their response in the data was changed to “-1: Item Not Applicable”. If a paper questionnaire respondent had neglected to answer a question that they should have, they were assigned a response in the data of “-2: Not answered but should have (paper)”. If a paper questionnaire respondent had ticked more than one box for a single response question, they were assigned a response in the data of “-3: Multi-selected for single response (paper)”. 

2. multiple response question edits: If a paper questionnaire respondent had mistakenly answered a question that they weren’t supposed to, their response was set to “-1: Item Not Applicable”. If a paper questionnaire respondent had neglected to answer a question that they should have, they were assigned a response in the data of “-2: Not answered but should have (paper)”. Where the respondent had selected both valid answers and an exclusive code such as “None of these”, any valid codes were retained, and the exclusive code response was set to “0”. 

In addition, some specific back edits of the data were conducted to prioritise the certain answers for routing. These edits primarily affected the paper data but did also occur in the web data too: 

1. if DIGRATE had a value of 5 then INTFREQ was set to a value 8 

2. if INTWHERE2_00X had a value of 1 then the corresponding variable of INTWHERE_00X was set to a value of 1 

3. if BBCONNECT_997 had a value of 1 and INTWHERE_001 had a value of 1, then INTWHERE_001 was revised to a value of 0 

4. if BBCONNECTMAIN had a value between 1 to 4 then the corresponding variable of BBCONNECT_00X was set to a value of 1

5. if BBSPEED had a value of 4 then BBCONNECT_001 was set to a value of 0 (paper only)

6. if BBSPEED had a value that wasn’t 4 or BBPAY or BBSPEED had been answered, then BBCONNECT_001 was set to a value of 1

7. if BBSPEED2 had a value of 1 or 2 then BBSPEED was set to a value of 1

8. if MOBPACK/MOBPAY/MOBDATA had been answered, then SMARTDEV_001 was set to a value of 1

9. if AIWHERE_997 had a value of 1 then AITYPEAE_A, AITYPEAE_C, AITYPEAE_E that had values of 1 or 2 were set to 998. And GENAIUSER would also be set to 2 (web only)

10. if DISAB3_00X had been answered, then DISAB2 should also have been answered and DISAB1 was set to a value of 1

7.4 Coding 

Post-interview coding was undertaken by members of the Verian coding department.  

The coding department reviewed verbatim responses, recorded for ‘other specify’ questions. If the verbatim answer code could be back coded to an existing answer response code, data edits were made to move the response from “Other” to the correct response code.  

7.5 Data outputs 

Once the checks were complete, a final csv file was created that only contained valid interviews and edited data.  

From this dataset, a set of data tables were produced. The tables were produced in R using the survey package to calculate the mean and confidence interval figures. The confidence interval estimates for each proportion were produced to the default 95% level with an upper and lower range, using a logit method to prevent out-of-bounds limits. This was done through applying weighting to the questions whether they were mixed mode or web only questions.

8. Weighting

8.1 Overview of weighting approach

A three-step weighting process was used to compensate for differences in both sampling probability and response probability.  

1. an address design weight was created equal to one divided by the sampling probability; this also served as the individual-level design weight because all resident adults can respond 

2. the expected number of responses per address was modelled as a function of data available at the neighbourhood and address levels. The step two weight was equal to one divided by the predicted number of responses 

3. the product of the first two steps was used as the input for the final step to calibrate the sample. The responding sample is calibrated using Random Iterative Method (RIM) weighting to the latest available ONS Labour Force Survey (LFS)^{[footnote 6]} and ONS Mid-Year-Population-Estimates^{[footnote 7]} with respect to (1) gender by age, (2) educational level by age, (3) ethnic group, (4) housing tenure, (5) ITL2 area by age and gender, (6) employment status by age, (7) household size, and (8) internet use by age within country^{[footnote 8]} 

RIM weighting, also known as raking, was used for the calibration step because the benchmarks were specified as a series of marginal distributions rather than a full cross-classified population matrix. The method iteratively adjusts the input weights so that the weighted sample matches each benchmark margin in turn, repeating until the target margins are met within tolerance.

Sourcing internet usage benchmarks

Internet usage benchmarks were derived using the most recent available estimates from large-scale, high-quality probability surveys covering the UK population. The following sources were selected due to their robust sampling designs and large achieved sample sizes: 

• England and Wales: Crime Survey for England and Wales (CSEW) 2023-24 

• Scotland: Scottish Household Survey (SHS) 2023 

• Northern Ireland: Continuous Household Survey (CHS) 2022-23 

Because the benchmark sources pre-dated PES fieldwork, the age bands used for the internet-use targets were adjusted to better reflect the approximate age composition of the population at the time of PES fieldwork. In practice, this involved rolling forward the older age groups used in the benchmark surveys by around one to two years, depending on the source period. This was a pragmatic adjustment intended to improve temporal alignment between the benchmark estimates and the PES fieldwork period; it does not fully account for time trends in internet use within age groups but assumes that once a person has been an internet user, they continue to be an internet user. 

Detailed benchmark distributions are provided in Annex 1.

Produced weights

The sum of these ‘grossing’ weights equals the UK population aged 16+. An additional standardised weight was produced that was the same but scaled so the weights sum to the respondent sample size. 

Equivalent weights were also produced for the (majority) subset of respondents who completed the survey by web. This weight was needed because some questionnaire items were included in the web questionnaire but not the paper questionnaire.

The final weight variables are: 

• ‘finalweight’ – to be used when analysing data available from both the web and paper questionnaires 

• ‘finalweightweb’ – to be used when analysing data available only from the web questionnaire

Limitations of weighting

It should be noted that the weighting only corrects for observed bias (for the set of variables included in the weighting matrix) and there is a risk of unobserved bias. Furthermore, the raking algorithm used for the weighting only ensures that the sample margins match the population margins. There is no guarantee that the weights will correct for bias in the relationships between the variables.

8.2 Design effects

The design effect is the ratio of the variance of a survey statistic under the complex design to the variance of the survey statistic under a simple random sample. It is a measure of how much the standard errors of estimates are affected by complex design features, such as weighting, clustering and stratification.

The estimated design effect due to the weighting is 1.67^{[footnote 9]}. This was calculated from the ‘finalweight’ using the standard approximation based on the coefficient of variation of the weights.  

Design effects varied across the ITL2 areas, reflecting differences in achieved sample sizes, response rates and the variability of weights. Across areas, the estimated design effect due to weighting ranged from approximately 1.29 to 1.47. It is expected behaviour that the individual ITL2 design effects are smaller than the overall design effect, as the overall design effect also reflects variation across areas.

The table of estimated design effect for each ITL2 area is in Annex 2.

Annex 1: weighting benchmarks

The tables below show the population benchmarks used in weighting.

Table A1: Weighting benchmark Age and Gender, ONS Mid-Year Population Estimates 2024 (UK 16+)

Gender by age	%
Male 16-24	6.7%
Male 25-34	8.1%
Male 35-44	7.9%
Male 45-54	7.4%
Male 55-64	7.7%
Male 65-74	5.7%
Male 75+	5.0%
Female 16-24	6.4%
Female 25-34	8.4%
Female 35-44	8.5%
Female 45-54	7.7%
Female 55-64	8.0%
Female 65-74	6.1%
Female 75+	6.5%

Table A2: Weighting benchmark education level by age, Labour Force Survey July-September 2025

Education level by age	%
Degree 16-24	2.9%
Degree 25-34	13.9%
Degree 35-44	13.5%
Degree 45-54	9.5%
Degree 55-64	7.3%
No degree 16-24	10.2%
No degree 25-34	2.6%
No degree 35-44	2.9%
No degree 45-54	5.5%
No degree 55-64	8.4%
65+	23.2%

Table A3: Weighting benchmark employment status by age, Labour Force Survey July-September 2025

Employment status by age	%
Working 16-24	6.8%
Working 25-34	13.8%
Working 35-44	13.9%
Working 45-54	12.6%
Working 55-64	10.4%
Working 65+	2.9%
Not working 16-24	6.3%
Not working 25-34	2.7%
Not working 35-44	2.5%
Not working 45-54	2.5%
Not working 55-64	5.3%
Not working 65+	20.3%

Table A4: Weighting benchmark housing tenure, Labour Force Survey July-September 2025

Housing tenure	%
Own outright	31.5%
Own with mortgage	33.5%
Other, mainly renting	35.0%

Table A5: Weighting benchmark ethnicity, Labour Force Survey July-September 2025

Ethnic group	%
White	84.9%
Indian	3.5%
Pakistani/Bangladeshi	2.5%
Black	4.7%
Other	4.4%

Table A6: Weighting benchmark number of adults in household, Labour Force Survey July-September 2025

Number of adults in household	%
One adult	17.7%
Two adults	52.3%
Three or more adults	30.0%

Table A7: Weighting benchmark internet usage by age within country 

Internet usage by age	%
England & Wales 16-66[footnote 10]	68.7%
England & Wales 67-76 internet users	9.6%
England & Wales 67-76 non-users	1.0%
England & Wales 77+ internet users	6.7%
England & Wales 77+ non-users	2.8%
Scotland 16-66	6.4%
Scotland 67-76 internet users	0.9%
Scotland 67-76 non-users	0.2%
Scotland 77+ internet users	0.5%
Scotland 77+ non-users	0.4%
Northern Ireland 16-67	2.2%
Northern Ireland 68-77 internet users	0.2%
Northern Ireland 68-77 non-users	0.1%
Northern Ireland 78+ internet users	0.1%
Northern Ireland 78+ non-users	0.1%

Annex 2: ITL2 design effects

Table A8: Estimated design effects within ITL2 areas

ITL2 Area	Estimated design effect
Bedfordshire and Hertfordshire	1.43
Berkshire, Buckinghamshire and Oxfordshire	1.36
Cambridgeshire and Peterborough	1.37
Cheshire	1.34
Cornwall and Isles of Scilly	1.47
Cumbria	1.37
Derbyshire and Nottinghamshire	1.45
Devon	1.47
East Central Scotland	1.33
East Yorkshire and Northern Lincolnshire	1.44
Eastern Scotland	1.42
Essex	1.39
Gloucestershire and Wiltshire	1.36
Greater Manchester	1.35
Hampshire and Isle of Wight	1.38
Herefordshire, Worcestershire and Warwickshire	1.40
Highlands and Islands	1.47
Inner London - East	1.34
Inner London - West	1.39
Kent	1.40
Lancashire	1.39
Leicestershire, Rutland and Northamptonshire	1.36
Lincolnshire	1.47
Merseyside	1.39
Mid and South West Wales	1.40
Norfolk	1.39
North Eastern Scotland	1.42
North Somerset, Somerset and Dorset	1.33
North Wales	1.43
North Yorkshire	1.33
Northern Ireland	1.31
Northumberland, Durham and Tyne & Wear	1.43
Outer London - East and North East	1.43
Outer London - South	1.41
Outer London - West and North West	1.35
Shropshire and Staffordshire	1.38
South East Wales	1.35
South Yorkshire	1.38
Southern Scotland	1.40
Suffolk	1.47
Surrey, East and West Sussex	1.35
Tees Valley	1.39
West Central Scotland	1.41
West Midlands	1.46
West of England	1.29
West Yorkshire	1.38

1. Due to rounding, the percentage figures per segment may sum to slightly more than 100%. ↩

2. Reminder 3 was targeted at all batches towards addresses in deprived areas or with a younger household structure. In addition, for batch 3, due to a strong response in this batch, this was further targeted by ITL2 area, only being also sent those in the 14 ITL2 areas who had less than 95% of their target number of interviews completed. ↩

3. No additional exclusions were applied to the response rate denominator beyond the standard ABOS adjustment for estimated non-residential addresses. In interviewer-administered surveys, addresses confirmed as vacant, derelict, second homes or otherwise ineligible may be removed from the denominator. For PES, this information was not collected systematically, so the denominator was not adjusted for these cases. Respondents were eligible to take part if the sampled address was either their main or secondary residence. ↩

4. It should be noted that (as detailed in Section 5.4) as well as being targeted to addresses in deprived areas or with a younger household structure in line, the third reminder for batch 3 was further targeted by ITL2 area. ↩

5. Interview lengths under 5 minutes are removed. If interviews are in the bottom 5 % of times, they are flagged in the system for the research team to evaluate; if they are flagged for multiple other quality checks, then those interviews are removed. ↩

6. Labour Force Survey July-September 2025 ↩

7. ONS Mid-Year Population Estimates 2024 ↩

8. Crime Survey of England and Wales 2023-2024, Scottish Household Survey 2023 and Northern Ireland Continuous Household Survey 2022-23 ↩

9. The design effect was calculated form the final analysis weight, as (1+ cv(w)2) where cv(W) is the coefficient of variation of the weight (cv= standard deviation by mean). ↩

10.  There is no split for internet usage for the under 67 age group as internet use is near universal within this age group. ↩