Official Statistics

UK Public Survey of Risk Perception, Resilience and Preparedness 2026: Technical Report (HTML)

Published 19 August 2026

Introduction

This report describes the key technical features of the UK Public Survey of Risk Perception, Resilience and Preparedness 2026. It covers the following elements of technical survey delivery:

• This ‘Introduction’ covers the purpose of the project and the supplier
• ‘Questionnaire design’ covers the process of changing the questionnaire
• ‘Sample design’ describes the approach taken to design the sample and conduct fieldwork
• ‘Data production’ describes the data processing methods used to assure the quality of the dataset
• ‘Weighting’ describes the weighting method employed, and the variables used

Savanta was commissioned to deliver the survey and obtained respondents from online panels. 11,727 respondents were surveyed online in a single wave, which took place between 27 February and 11 May 2026. This included 10,791 surveyed as part of the core survey, and 936 surveyed for the LRF Boost sample described below. During and after fieldwork, the data underwent quality assurance checks conducted by Savanta. Following fieldwork, final quality control checks were applied by UK Health Security Agency (UKHSA) analysts, prior to the data’s use and publication. All work was carried out in line with the ethical standards of the Market Research Society’s Code of Conduct and the Standards for Official Statistics. 

Questionnaire design

The findings from the first wave of the UK Public Survey of Risk Perception, Resilience and Preparedness were published on 23 July 2025. The survey then underwent development between August and November 2025. During this period the Cabinet Office conducted a short user feedback survey (between July and October 2025) and held discussion sessions with the survey’s primary users, national and local officials responsible for communicating with the public about risks and preparedness.

Following this period of consultation, the questionnaire was reviewed by the Cabinet Office in collaboration with Savanta, UKHSA and the Department for Environment, Food and Rural Affairs. The first draft of the revised questionnaire was shared with Savanta in December 2025. The questionnaire was then revised, updated, and finalised in February 2026.

The median time taken by respondents to complete the updated survey was 15 ½ minutes. 

Changes to main survey questions in 2026

A list of changes made to the 2026 questionnaire is provided below. Changes include the addition of questions on new topics, and changes to existing questions. Changes were made in response to feedback, and to the areas of improvement identified in 2025. In addition to these listed changes, the demographic questions have been reordered, and some questions have been removed. The changes to the main questionnaire have been grouped below into two types: questions that covered new topics, and main survey questions revised as part of developing the survey. Changes to demographics made to improve quotas and weights will be addressed further down in this report.

Questions covering new topics

MEDICATION was added as a new question to discover whether respondents were taking essential medication.

Three questions – Q37, Q38 and Q39 – were added in order to measure respondents’ self-assessed confidence in their first aid skills and knowledge.

Q40 was added to find out how aware self-employed respondents were of advice on how to prepare their businesses for emergencies or disruption.

Q41 was added to learn if respondents had a place they could stay for a few days, if they had to evacuate their homes due to an emergency.

Q42 was added to learn about the access respondents had to health items at home. The list of items it presented respondents with combined two items (hand sanitiser and wet wipes) asked about in Q26 in the 2025 survey, with two new items (supplies of essential medicine and first aid kits).

Q43 was added to learn about how many days’ worth of non-perishable food people had stored in their homes.

Adapted main survey questions

Following completion of the 2025 survey, a small number of very large outliers were discovered in the results for Q27, which asked how much bottled/mineral water respondents had at home. These outliers were deemed plausible, given the possibility that respondents were either describing water stored in potable water tanks or they could have been using millilitres instead of litres. To prevent misinterpretation of this question, in 2026 the question was revised:

• The question (now using the ID Q27_2026) was reworded to ask how many litres of bottled drinking water respondents had. An image of common water bottle sizes with accompanying numbers listing how many litres they each contain was added to the question to further clarify the subject and aid respondents in answering.
• A threshold for a large amount of water storage was set at 21 litres per person in the respondent’s household. Respondents who responded with a number of litres of water higher than this threshold were asked a follow-up question – Q27b – to clarify whether this water was stored in bottles, or in a tank. If they confirmed they were not referring to bottled water, it was decided that the respondent should be excluded on the basis they had not paid sufficient attention to the revised Q27_2026. Following Q27b, the respondent was then given an opportunity to change their answer to a revised number of litres, at Q27c.

In the 2025 survey, 11% of respondents selected ‘none of these’ at Q30 when asked about barriers to preparedness, raising the possibility that the list of potential barriers did not include all significant options. To test this, an open-end response – ‘Other (please specify)’ – was added to the list of options, allowing respondents to submit additional barriers if the existing list was not sufficiently exhaustive. In addition to this, response option 10 ‘I feel I have taken all the steps I need to’ was replaced with option 13 ‘I have already taken a lot of action to prepare for emergencies’, as this wording was deemed easier to understand.

Following the 2025 survey, there was some concern from user/s of the statistics that respondents might have found it difficult to answer Q29, which asked how many days’ worth of baby formula respondents had at home. To make it easier to answer, the question was changed from an open text question, where the respondent typed in the number of days’ worth of baby formula they had, to a single code question, that presented respondents with a list of numbers they could select from, ranging from 1 to 7+. To reflect this change, the question ID was changed to Q29_2026.

Other changes made to the main survey questions due to feedback from statistics users included:

At Q2 and Q3, the options ‘Poor air quality incident’ and ‘Landslide’ were removed from the list of emergencies tested at each question. ‘Radiation release’ and ‘disruption to food supply’ were added as new options. In addition, respondents were asked about water outages/power cuts that lasted for 3 days or more, replacing last year’s options which asked about outages/cuts that lasted for over 12 hours.

At Q9, the options ‘Elected mayors’, ‘Emergency services’ and ‘Utility providers’ were removed from the list of organisations tested. In addition, the options ‘Voluntary organisations/charities’ and ‘Community groups’, which were kept separate in the 2025 survey, were in 2026 merged into one option: ‘Voluntary organisations / charities / community groups’.

At Q12, the options ‘From a media personality, celebrity or influencer’, ‘From a UK politician’, ‘From a member of the UK emergency services’, and ‘Other (please specify)’ were added to the list of tested information sources. As at Q9 ‘Voluntary organisations/charities’ and ‘Community groups’ were merged into one option: ‘Voluntary organisations / charities / community groups’. ‘On social media’ was split into two separate options: ‘On social media, from a UK government account’ and ‘On social media, not from a UK Government account.’

At Q13, the options ‘War’, ‘Food supply disruption’ and ‘Other (please specify)’ were added to the list of tested information topics.

At Q25, the option ‘Power bank to charge mobile phone’ was added to the list of tested items.

At Q26, the options ‘Bottled drinking water for 3 days (at least 9 litres per person in your household’, ‘Non-refrigerated drinks (other than water or alcohol) that would last you 3 days’, or ‘Water purification tablets’ were added to the list of tested items. Options asking about stocks of perishable food or baby formula that would last 7 days were removed, replaced with Q43 and Q29_2026 respectively.

Changes to demographic questions

Qualifications

In the 2025 survey, Savanta recruited fewer people who reported having no qualifications than was expected given census data. Increasing the number recruited was a key Development Plan goal for the 2026 survey.  One potential explanation for this was a difference in question wording between the 2025 survey, and the ONS, NISRA and Scotland censuses. The 2025 survey asked respondents, ‘What is the highest educational level that you have achieved to date?’ and supplied a list of options that mixed types of school (e.g. ‘Primary school’, ‘Secondary school’) with types of qualification (e.g. ‘University degree’). The censuses by contrast only ask respondents what qualifications they have obtained. To more closely align the Resilience survey with the censuses, the education question was revised to ask only about qualifications.

Specifically, the single qualification question was replaced with three separate qualification questions, listed below. This design, and the wording of the questions, was taken from the 2021 ONS Census:

• QUALIFICATION_A. Have you completed an apprenticeship?
• QUALIFICATION_B. Have you achieved a qualification at degree level or above?
• QUALIFICATION_C. Have you achieved any other qualifications for which you received a certificate? 

Using these three questions, respondents were assigned a ‘highest qualification’ (degree level or above, below degree level and no qualifications).

The change in question wording appears to have improved the alignment of people’s educational attainment, as captured in the survey, with educational attainment captured in the censuses. Whereas in 2025, 1% of the sample reported having no qualifications, compared to 18% in the UK as a whole, in 2026 11% of the sample indicated that they had no qualifications. Further progress against these criteria is uncertain, as the length and the complexity of the survey, and the tendency of online survey panel respondents to have a higher level of educational attainment, are all barriers to the recruitment of respondents without formal qualifications.

NS-SeC

When selecting variables on which to weight a dataset, it is important that those variables meet the following three criteria:

• The variable predicts the probability of a respondent answering the survey
• The variable predicts how a person will answer key questions in the survey
• We know what proportion of the population falls into each variable category

For example, we would weight on age because:

• People in different age groups are likely to over/under-respond to an internet panel survey
• People in different age groups are likely to respond differently to questions on how prepared they are for emergencies
• We know how many people in the UK, and each of the devolved nations, are aged 18-24, 25-34, etc.

Since completing the 2025 Resilience survey, it has become clear that approximated social economic grade (SEG) no longer met these criteria, as it is not known what proportion of the 18+ population are truly AB, C1, C2 or DE. The 2021 ONS Census, 2021 NISRA Census and 2022 Scottish Census all determined that, because they did not have sufficiently comprehensive questions about respondents’ economic status, they could not reliably assign retired people to a social grade. This meant that people in households where the Household Reference Person was retired were not assigned a social grade by any UK census. As roughly 20% of people across the UK live in such households, existing official data on SEG cannot be said to give us population-level information.

Given these concerns surrounding SEG we instead based our socioeconomic quotas and weights on National Statistics Socio-economic Classification (NS-SeC) categories. Much like SEG, NS-SeC determines a respondent’s socio-economic position based on their current or most recent occupation. Unlike SEG however, everyone who answered the 2021/22 Censuses were assigned to an NS-SeC category, making it a true population-level measure, and therefore more appropriate as a quota and weighting variable. The categories used for quotas/weights were also similar to the AB, C1, C2 and DE categories used in the previous survey. These categories are as follows:

• Managerial, administrative and professional occupations (similar to AB)
• Intermediate occupations (similar to C1 and C2)
• Routine and manual occupations (similar to D)
• Never worked, long-term unemployed or full-time students (similar to E).

There is however one key difference between the two variables, beyond the greater reliability of NS-SeC. An individual’s SEG classification is based on the classification of their Household Reference Person (HRP). The HRP serves as a reference point to characterise a whole household, and in households with multiple occupants, is selected primarily on the basis of economic activity, with age used as a tie-breaker where economic activity is similar. When calculating SEG, all occupants of a household are given the HRP’s grade. For example, if a household’s HRP is employed as a doctor, everyone in their household would share the HRP’s AB SEG classification, regardless of their personal occupation or economic activity. 

By contrast, a person’s NS-SeC is based on their individual economic activity. To take the example above, if the doctor HRP lived with a long-term unemployed partner, and a child in a routine occupation, SEG would classify all three people as AB, whereas NS-SeC would assign them three separate classes. 

To code NS-SeC, the following questions were added to the survey. Derived variables were not answered directly by respondents but instead calculated based on how respondents answered other questions. In addition to the questions below, the question EMP_STATUS was moved from the final section of the questionnaire, to the first, as it was key to calculating NS-SeC.

• WORKED
• NSSEC_A (derived)
• NSSEC_B
• NSSEC_C
• NSSEC_D
• NSSEC_E
• NSSEC_F
• NSSEC_STATUS (derived)
• NSSEC_CLASS (derived)
• NSSEC_QUOTA (derived)

Postcode

Following discussion sessions with key survey users, it was agreed that a priority for the Resilience survey would be to produce estimates of preparedness at the Local Resilience Forum (LRF) level. LRFs are multi-agency partnerships that plan for emergencies in England and Wales. The results of the 2025 survey had shown that the majority of survey respondents in England and Wales had not heard of LRFs before taking the survey (73%). Given these findings, asking respondents directly to identify their LRF would not have been practical. As such, we have instead taken two approaches.

For the majority of respondents in the data, Savanta determined their LRF by coding it from their postcode. As LRF areas are based on combining Local Authorities, it is possible to derive LRFs from postcode, via Local Authority. However, postcode is classed as Personally Identifiable Information (PII), and following MRS guidance, Savanta cannot exclude people from a survey on the basis that they refuse to provide PII. 1,161 (14.5%) of English respondents refused to provide their postcode, and could not be coded directly to their LRF. For this subgroup, we instead asked them to identify their Local Authority. If they did so, they were coded to an LRF based on their response. If they did not, they were excluded from the survey.

Sex

In last year’s survey, to discover whether respondents identified as male or female, we asked: ‘In which of the following ways do you identify?’, with the options ‘Female’, ‘Male’, ‘I identify in another way’ and ‘Prefer not to say’. The 2025 survey was launched prior to the publication of the Independent review of data, statistics and research on sex and gender in March 2025. Following the recommendations from this review , a new question was added to the survey that explicitly asked respondents for their sex at birth. As this question could be sensitive for some respondents, a ‘Prefer not to say’ option was provided, to allow respondents to opt out. This is in line with both MRS guidelines, and recommendations from the independent review.

Additional questions

In line with user feedback two additional demographic questions were added to the survey:

dURBANRURAL was added to capture whether people live in an urban or rural area. This information was either collected based on an individual’s postcode, or by asking the respondent directly, if they either refused to provide their postcode or lived in Northern Ireland. Northern Ireland does not currently have a rural/urban classification that can be derived from postcode.

INCOME was added to capture respondents’ sense of their own financial wellbeing.

Sample design and limitations

As in 2025, the UK Public Survey of Risk Perception, Resilience and Preparedness sought to provide robust data for each of the UK’s constituent nations. To ensure this was the case, respondents from Scotland, Wales and Northern Ireland were oversampled during fieldwork, to ensure enough participants were recruited in these areas for the analysis to be robust. The targets were to recruit 1,500 people living in Scotland, and 1,000 living in Wales and Northern Ireland respectively. These results have been provided in separate data tables. For the UK-level analysis weights have been used to adjust for the oversampling and ensure the results are representative of the UK adult population.

In addition to focusing on the devolved nations, the UK Public Survey of Risk Perception, Resilience and Preparedness also sought to estimate the opinion of individuals in each of England’s Local Resilience Forums (LRF) areas. To accomplish this, a target was set to recruit 150 people in each Local Resilience Forum area. For some LRF areas this target was achieved through natural fallout in the main survey. For many LRF areas however this was not the case. As a result, a boost sample was used to push numbers above the target threshold. As can be seen in Table 1.5, we were not able to recruit 150 in every LRF area, despite leaving the survey open for longer than intended. However, we were able to recruit above 100 people in almost all of the LRFs, the minimum number needed to estimate opinion, with the sole exception of Cumbria. The LRF Area results are presented in their own separate table. 

The method of data collection used in the survey was Computer Assisted Web Interviews (CAWI). Respondents were recruited using the Savanta Hub API from multiple MRS-member and/or ESOMAR-accredited online panel providers. Each panel has their own procedures for recruiting people, which can include direct recruitment from telephone research; face-to-face recruitment after completing on-street or in-home interviews; referrals from existing panellists; direct sign-ups on the panel website; targeting specific online communities; and recruiting via social media and mobile advertisements. 

To aid survey completion and engagement, the Savanta survey platform is designed to be highly accessible. Screen-reading and other accessibility features are built in to ensure a good respondent experience for common disabilities such as visual impairments, hearing impairments and motor impairments. 

Respondents who completed the survey were rewarded with incentives, as is normal practice in panel-based surveys. The value of these incentives varied over time, and according to whether or not the respondent was included in a quota boost sample plan. Given this variation in incentives paid, it is not possible to report payments at an individual level. The amount paid per respondent fell within the range of £0.25 - £7.00.

Using online access panels to recruit respondents has many advantages. It is a particularly efficient means of gathering survey data, allowing for the recruitment of large samples at speed and at low cost. It also reduces social desirability bias, as there is no interviewer present to shape responses. However, recruiting a sample from online panels does carry a risk of nonresponse bias. This is when estimates taken from a survey do not accurately reflect the true opinions of the population, due to the survey sample differing markedly in composition from the population (in this case people aged 18+ and living in the UK). Certain nonresponse bias is unavoidable. For example, people without access to the internet will not be part of an online access panel. However, according to the latest research from Ofcom, 98% of all residential properties in the UK currently have access to superfast broadband coverage.^{[footnote 1]} 

Steps were taken to address nonresponse bias in the sample, through the application of quotas. Using data from a range of official sources (listed below) the proportion of the population that lives in each constituent nation of the UK, is aged 18 and over, and falls into a set of different demographic categories, was calculated. This was done for each constituent nation to ensure the final sample was representative at constituent-country level.

These proportions were then applied to the planned sample size for each constituent nation, to calculate how many people in each category should be recruited. During fieldwork the number of people recruited in each of these groups was monitored, with the goal of having the final number of respondents fall between 90% and 120% of quota. Where that goal could not be achieved, weighting ensured that the sample remained representative. 

Results at LRF level should not be treated as representative of the local population in the same way as the UK or devolved nation samples. Due to the small size of many LRFs, it would not have been feasible to employ a full suite of restrictive quotas to shape their demographics. As such the results for the LRF Boost sample should be treated as indicative, rather than strictly representative.

Quotas were set for the following characteristics for the England, Wales, Scotland and Northern Ireland samples. Population statistics used for the quotas were sourced from the ONS Mid-Year Population Estimates and the England & Wales, Northern Ireland and Scotland Censuses. 

• Age and gender (interlocked)
• Region
• Ethnicity
• NS-SeC
• Religion or religion brought up in (Northern Ireland only)^{[footnote 2]}
• Highest qualification obtained (degree/no degree)

A comparison of quotas applied and the demographic balance of the dataset is provided below in the following four tables, showing the number of respondents in each quota recruited from each part of the United Kingdom.

Table 1.1. Target quotas and sample achieved in each demographic category in England

Category	Level	Quota Count	Respondents recruited	% of Quota Recruited
Gender / Age	Male x 18-24	380	400	105%
Gender / Age	Male x 25-34	591	558	94%
Gender / Age	Male x 35-44	577	572	99%
Gender / Age	Male x 45-54	533	518	97%
Gender / Age	Male x 55-64	548	539	98%
Gender / Age	Male x 65-74	400	419	105%
Gender / Age	Male x 75+	360	377	105%
Gender / Age	Female x 18-24	360	386	107%
Gender / Age	Female x 25-34	613	603	98%
Gender / Age	Female x 35-44	618	634	103%
Gender / Age	Female x 45-54	554	575	104%
Gender / Age	Female x 55-64	572	569	100%
Gender / Age	Female x 65-74	432	460	106%
Gender / Age	Female x 75+	463	456	98%
Region	North-West	921	960	104%
Region	North-East	334	343	103%
Region	Yorkshire & Humberside	677	720	106%
Region	West Midlands	729	755	104%
Region	East Midlands	608	628	103%
Region	South-West	720	700	97%
Region	South-East	1149	1077	94%
Region	Eastern	781	730	93%
Region	London	1081	1169	108%
Ethnicity	White	5830	5984	103%
Ethnicity	Mixed	136	144	106%
Ethnicity	Asian	623	605	97%
Ethnicity	Black	268	286	107%
Ethnicity	Other	143	35	24%
NS-SeC	Managerial, administrative and professional occupations	2389	2672	112%
NS-SeC	Intermediate occupations	1584	1412	89%
NS-SeC	Routine and manual occupations	2056	2059	100%
NS-SeC	Never worked, long-term unemployed or full-time students	972	939	97%
Highest qualification	Degree	2440	2717	111%
Highest qualification	Below degree	4560	4365	96%
Total	Total	7000	7082	101%

Table 1.2. Target quotas and number of respondents recruited in each demographic category in Wales

Category	Level	Quota Count	Respondents recruited	% of Quota Recruited
Gender / Age	Male x 18-24	54	54	101%
Gender / Age	Male x 25-34	76	81	106%
Gender / Age	Male x 35-44	75	81	108%
Gender / Age	Male x 45-54	71	74	104%
Gender / Age	Male x 55-64	83	83	100%
Gender / Age	Male x 65-74	66	82	124%
Gender / Age	Male x 75+	59	62	106%
Gender / Age	Female x 18-24	49	50	102%
Gender / Age	Female x 25-34	78	82	105%
Gender / Age	Female x 35-44	80	82	103%
Gender / Age	Female x 45-54	75	77	102%
Gender / Age	Female x 55-64	88	96	109%
Gender / Age	Female x 65-74	71	97	137%
Gender / Age	Female x 75+	74	48	65%
Region	Mid and West Wales	206	227	110%
Region	North Wales	181	174	96%
Region	South Wales Central	238	257	108%
Region	South Wales East	204	206	101%
Region	South Wales West	171	185	108%
Ethnicity	White	947	993	105%
Ethnicity	Ethnic minority (excluding white minorities)	53	56	107%
NS-SeC	Managerial, administrative and professional occupations	305	349	115%
NS-SeC	Intermediate occupations	222	221	99%
NS-SeC	Routine and manual occupations	333	341	102%
NS-SeC	Never worked, long-term unemployed or full-time students	140	138	99%
Highest qualification	Degree	324	371	115%
Highest qualification	Below degree	676	678	100%
Total	Total	1000	1049	105%

Table 1.3. Target quotas and number of respondents recruited in each demographic category in Scotland

Category	Level	Quota Count	Respondents recruited	% of Quota Recruited
Gender / Age	Male x 18-24	77	81	105%
Gender / Age	Male x 25-34	117	123	105%
Gender / Age	Male x 35-44	116	125	108%
Gender / Age	Male x 45-54	110	118	107%
Gender / Age	Male x 55-64	127	139	110%
Gender / Age	Male x 65-74	97	101	104%
Gender / Age	Male x 75+	75	77	102%
Gender / Age	Female x 18-24	78	80	102%
Gender / Age	Female x 25-34	122	135	110%
Gender / Age	Female x 35-44	122	136	111%
Gender / Age	Female x 45-54	117	123	105%
Gender / Age	Female x 55-64	135	141	104%
Gender / Age	Female x 65-74	106	110	104%
Gender / Age	Female x 75+	100	86	86%
Region	North Eastern Scotland	133	149	112%
Region	Highlands And Islands	134	142	106%
Region	Eastern Scotland	550	578	105%
Region	West Central Scotland	421	455	108%
Region	Southern Scotland	262	254	97%
Ethnicity	White	1408	1480	105%
Ethnicity	Ethnic minority (excluding white minorities)	92	94	102%
Approximated social grade (Scotland calculation)	Managerial, administrative and professional occupations	510	584	114%
Approximated social grade (Scotland calculation)	Intermediate occupations	303	312	103%
Approximated social grade (Scotland calculation)	Routine and manual occupations	485	488	101%
Approximated social grade (Scotland calculation)	Never worked, long-term unemployed or full-time students	202	194	96%
Highest qualification	Degree	712	814	114%
Highest qualification	Below degree	788	764	97%
Total	Total	1500	1578	105%

Table 1.4. Target quotas and number of respondents recruited in each demographic category in Northern Ireland

Category	Level	Quota Count	Respondents recruited	% of Quota Recruited
Gender / Age	Male x 18-24	52	59	113%
Gender / Age	Male x 25-34	80	65	81%
Gender / Age	Male x 35-44	84	82	98%
Gender / Age	Male x 45-54	80	99	124%
Gender / Age	Male x 55-64	82	104	126%
Gender / Age	Male x 65-74	61	83	137%
Gender / Age	Male x 75+	48	63	132%
Gender / Age	Female x 18-24	48	51	105%
Gender / Age	Female x 25-34	80	84	105%
Gender / Age	Female x 35-44	90	89	99%
Gender / Age	Female x 45-54	83	89	107%
Gender / Age	Female x 55-64	86	94	109%
Gender / Age	Female x 65-74	64	74	116%
Gender / Age	Female x 75+	62	42	68%
Region	East	505	614	121%
Region	South	205	217	106%
Region	North	153	135	88%
Region	West	136	116	85%
Ethnicity	White	970	1058	109%
Ethnicity	Ethnic minority (excluding white minorities)	30	21	70%
Approximated social grade	Managerial, administrative and professional occupations	298	388	130%
Approximated social grade	Intermediate occupations	221	251	114%
Approximated social grade	Routine and manual occupations	319	302	95%
Approximated social grade	Never worked, long-term unemployed or full-time students	162	141	87%
Religion or Religion brought up in	Catholic	448	552	123%
Religion or Religion brought up in	Protestant and Other Christian	465	452	97%
Religion or Religion brought up in	Other religion/None	87	78	90%
Highest qualification	Degree	331	413	125%
Highest qualification	Below degree	669	669	100%
Total	Total	1000	1082	108%

Table 1.5. Number of respondents recruited in each Local Resilience Forum

LRF	Respondents Recruited
Avon and Somerset	199
Bedfordshire	142
Cambridgeshire and Peterborough	147
Cheshire	147
Cleveland	114
Cumbria	67
DCIOS	231
Derbyshire	148
Dorset	144
Durham and Darlington	151
Essex	202
Gloucestershire	143
Greater Manchester	391
Hampshire and IOW	238
Hertfordshire	147
Humber	135
Kent	242
Lancashire	223
Leicestershire	145
Lincolnshire	148
London	1193
Merseyside	166
Norfolk	146
North Yorkshire	149
Northamptonshire	147
Northumbria	253
Nottinghamshire	150
South Yorkshire	194
Staffordshire	150
Suffolk	149
Surrey	150
Sussex	232
Thames Valley	224
Warwickshire	146
West Mercia	150
West Midlands	436
West Yorkshire	333
Wiltshire and Swindon	146

Data processing

To ensure the data collected as part of this project was always of the highest quality, all survey data went through several stages of quality assurance (QA) with automated cleaning processes used where appropriate.

Pre-processing checks

Using their in-house survey platform, Savanta automatically removed the following:

• Respondents who took 33% or less of the median time to answer a question, on 40% of questions.
• Respondents who took 33% or less of the median time to answer the survey as a whole
• Respondents who gave the same answer to 66% of the ‘scale’ questions used in the survey.
• Respondents flagged as potentially fraudulent by RelevantID. RelevantID is an industry-standard tool that uses a variety of respondent metadata (e.g. the respondent’s browser, operating system and location) to create a ‘machine fingerprint’ to determine the likelihood of a respondent taking the same survey on multiple occasions. It then provides both a ‘dupe’ (duplicate) score and a fraud profile score. Scores above particular levels cause the associated respondent to be shadow banned from taking part in future surveys.

Savanta then reviewed the data collected in Excel, and removed respondents according to the following criteria:

• Those respondents whose full postcode or postcode outward code was not a ‘live’ code in the ONS Postcode Directory
• Those who gave their full postcode, and their stated ITL1 Region did not match the Region coded from their postcode 
• Those who gave only the first half of their postcode, and whose stated ITL1 Region did not match at least one of the possible ITL1 Regions calculated using the first half (outward code) of their postcode
• Those that did not write the word ‘blue’ (or a close misspelling, such as ‘bluu, bluey’ etc.) in response to the question COLOUR_TRAP (“Please write ‘blue’ in the box below. This is a quality check to test attention”)
• Those whose stated age differed from their age as calculated using their stated year of birth, by 4 or more years
• Those respondents who were below the age of 65 but claimed they were in receipt of a state pension
• Those respondents who were below the age of 55 but claimed they were in receipt of a private pension
• Those respondents who said they had children aged 0-17 living in their household at CHILDREN_AGE, but reported having no children aged 0-17 living in their household at HOUSEHOLD
• Those respondents who report not owning a mobile phone at MOBILE, report being the only person who lives in their household, and who do report owning a mobile phone charger at Q25_11.
• Those respondents who said that they had talked to their children about what happens in an emergency at Q24b_3, and who reported only having children aged 0-1.
• Respondents whose responses to Q27b made no sense given the wording of the question, or, who revealed they had given the amount of water they own based on wells, tanks etc. and not bottled water, as asked for at Q27_2026, or, who did not give their responses at Q27c in litres despite multiple requests to do so.
• Respondents who reported either having enough baby formula to last 3 days at Q26_3, but only 1-2 days’ worth at Q29_2026, or, who reported not having enough baby formula to last 3 days at Q26_3 but 3 or more days’ worth at Q29_2026.
• Respondents who reported having enough non-perishable food (food that can be stored in a cupboard rather than a fridge) that doesn’t require cooking and would last their household approximately 3 days at Q26_1, but who reported having approximately 1-2 days’ worth of non-perishable food (that does or does not require cooking) at Q43. 

Northern Ireland

In order to collect a representative sample of around 1,000 people in Northern Ireland, Savanta both recruited respondents via the Savanta Hub API, and contracted a local specialist company LucidTalk to obtain additional respondents. The resulting combined sample was both too large, and demographically skewed to be included in the database directly. To adjust for this, Savanta extracted a more demographically representative sub-sample from this total in the following way.

Using the R programming language, a small random sample of respondents was selected from the combined Savanta/LucidTalk sample. Respondents were then selected to be included in the new sub-sample, according to what quotas needed to be filled. This process was automatically repeated until all quotas were filled as much as was possible, without overfilling any of them. Overfilling was defined as the demographic group’s proportion in the sample exceeding 120% of their proportion in the population. The R code used to select the sample can be shared on request.

Weighting

Weighting was applied to the final results to ensure the demographic profile of the sample matched both the adult (18+) UK population and, for country-level tables, the adult (18+) population of England, Scotland, Wales and Northern Ireland. 

Weights are values that cause individual respondents to count for more, or less, depending on whether their demographics are underrepresented or overrepresented in the sample relative to the target population. For example, if the survey sample is 50% men, but the UK population is 48% men, men in the survey sample would be underweighted so that they would comprise 48% of the weighted sample. 

Savanta used RIM weighting, also known as raking, to calculate the survey weights. This method iteratively adjusts the proportions in the sample for each variable in a list, until they match the target proportions. The list of variables used in the UK and country-level weighting schemes are shown in the table below, alongside the relevant proportions. The data sources used to construct these weights are listed in the annex of this report.

Weights used differed slightly from the quotas. As in 2025, the UK Public Survey of Risk Perception, Resilience and Preparedness was not weighted by highest qualification. In addition, while in the quotas the oldest age groups were 65-74 and 75+, we have weighted the sample using 65+ as the oldest age group. This was done to keep weights consistent with the 2025 survey.

As results at LRF level are indicative, rather than strictly representative, the LRF area tables are not weighted.

Table 2.1. The impact of the UK weighting scheme

Category	Level	Unweighted proportion	Weighted proportion
Sex / Age	Male x 18-24	5.5%	5.4%
Sex / Age	Male x 25-34	7.7%	8.5%
Sex / Age	Male x 35-44	8.0%	8.2%
Sex / Age	Male x 45-54	7.5%	7.6%
Sex / Age	Male x 55-64	8.0%	7.8%
Sex / Age	Male x 65+	11.7%	10.9%
Sex / Age	Female x 18-24	5.3%	5.1%
Sex / Age	Female x 25-34	8.4%	8.8%
Sex / Age	Female x 35-44	8.7%	8.8%
Sex / Age	Female x 45-54	8.0%	7.9%
Sex / Age	Female x 55-64	8.3%	8.2%
Sex / Age	Female x 65+	12.7%	12.7%
Region	Northern Ireland	10.0%	2.8%
Region	Scotland	14.6%	8.3%
Region	North-West	8.9%	11.2%
Region	North-East	3.2%	3.9%
Region	Yorkshire & Humberside	6.7%	8.1%
Region	Wales	9.7%	4.7%
Region	West Midlands	7.0%	8.9%
Region	East Midlands	5.8%	7.4%
Region	South-West	6.5%	8.6%
Region	South-East	10.0%	12.8%
Region	Eastern	6.8%	10.1%
Region	London	14.6%	13.3%
Ethnicity	White	88.5%	85.0%
Ethnicity	Mixed	1.7%	1.8%
Ethnicity	Asian	6.1%	8.0%
Ethnicity	Black	3.3%	3.4%
Ethnicity	Other	0.4%	1.8%
NS-SeC	Managerial, administrative and professional occupations	37.0%	33.8%
NS-SeC	Intermediate occupations	20.4%	22.4%
NS-SeC	Routine and manual occupations	29.6%	29.9%
NS-SeC	Never worked, long-term unemployed or full-time students	13.1%	13.9%

Table 2.2. The impact of the England weighting scheme

Category	Level	Unweighted proportion	Weighted proportion
Gender / Age	Male x 18-24	5.6%	5.4%
Gender / Age	Male x 25-34	7.9%	8.5%
Gender / Age	Male x 35-44	8.1%	8.2%
Gender / Age	Male x 45-54	7.3%	7.6%
Gender / Age	Male x 55-64	7.6%	7.8%
Gender / Age	Male x 65+	11.2%	10.9%
Gender / Age	Female x 18-24	5.5%	5.1%
Gender / Age	Female x 25-34	8.5%	8.8%
Gender / Age	Female x 35-44	9.0%	8.8%
Gender / Age	Female x 45-54	8.1%	7.9%
Gender / Age	Female x 55-64	8.0%	8.2%
Gender / Age	Female x 65+	12.9%	12.7%
Region	North-West	13.6%	13.2%
Region	North-East	4.8%	4.7%
Region	Yorkshire & Humberside	10.2%	9.5%
Region	West Midlands	10.7%	10.5%
Region	East Midlands	8.9%	8.8%
Region	South-West	9.9%	10.2%
Region	South-East	15.2%	15.3%
Region	East of England	10.3%	12.0%
Region	London	16.5%	15.8%
Ethnicity	White	84.8%	83.4%
Ethnicity	Mixed	2.0%	1.9%
Ethnicity	Asian	8.6%	8.9%
Ethnicity	Black	4.1%	3.8%
Ethnicity	Other	0.5%	2.0%
NS-SeC	Managerial, administrative and professional occupations	37.7%	34.1%
NS-SeC	Intermediate occupations	19.9%	22.6%
NS-SeC	Routine and manual occupations	29.1%	29.4%
NS-SeC	Never worked, long-term unemployed or full-time students	13.3%	13.9%

Table 2.3. The impact of the Wales weighting scheme

Category	Level	Unweighted proportion	Weighted proportion
Gender / Age	Male x 18-24	5.1%	5.4%
Gender / Age	Male x 25-34	7.7%	7.6%
Gender / Age	Male x 35-44	7.7%	7.5%
Gender / Age	Male x 45-54	7.1%	7.1%
Gender / Age	Male x 55-64	7.9%	8.3%
Gender / Age	Male x 65+	13.7%	12.5%
Gender / Age	Female x 18-24	4.8%	4.9%
Gender / Age	Female x 25-34	7.8%	7.8%
Gender / Age	Female x 35-44	7.8%	8.0%
Gender / Age	Female x 45-54	7.3%	7.5%
Gender / Age	Female x 55-64	9.2%	8.8%
Gender / Age	Female x 65+	13.8%	14.5%
Region	Mid and West Wales	21.6%	20.6%
Region	North Wales	16.6%	18.2%
Region	South Wales Central	24.5%	23.8%
Region	South Wales East	19.6%	20.4%
Region	South Wales West	17.6%	17%
Ethnicity	White	94.7%	94.7%
Ethnicity	Ethnic minority (excluding white minorities)	5.3%	5.3%
NS-SeC	Managerial, administrative and professional occupations	33.3%	30.5%
NS-SeC	Intermediate occupations	21.1%	22.2%
NS-SeC	Routine and manual occupations	32.5%	33.3%
NS-SeC	Never worked, long-term unemployed or full-time students	13.2%	14.0%

Table 2.4. The impact of the Scotland weighting scheme

Category	Level	Unweighted proportion	Weighted proportion
Gender / Age	Male x 18-24	5.1%	5.1%
Gender / Age	Male x 25-34	7.8%	7.8%
Gender / Age	Male x 35-44	7.9%	7.8%
Gender / Age	Male x 45-54	7.5%	7.3%
Gender / Age	Male x 55-64	8.8%	8.4%
Gender / Age	Male x 65+	11.3%	11.4%
Gender / Age	Female x 18-24	5.1%	5.2%
Gender / Age	Female x 25-34	8.6%	8.1%
Gender / Age	Female x 35-44	8.6%	8.1%
Gender / Age	Female x 45-54	7.8%	7.8%
Gender / Age	Female x 55-64	8.9%	9.0%
Gender / Age	Female x 65+	12.4%	13.6%
Region	North Eastern Scotland	9.4%	9.0%
Region	Highlands and Islands	9.0%	9.0%
Region	Eastern Scotland	36.6%	36.7%
Region	West Central Scotland	28.8%	27.9%
Region	Southern Scotland	16.1%	17.4%
Ethnicity	White	94.0%	93.8%
Ethnicity	Ethnic minority (excluding white minorities)	6.0%	6.2%
NS-SeC	Managerial, administrative and professional occupations	37.0%	34.0%
NS-SeC	Intermediate occupations	19.8%	20.2%
NS-SeC	Routine and manual occupations	30.9%	32.2%
NS-SeC	Never worked, long-term unemployed or full-time students	12.3%	13.6%

Table 2.5. The impact of the Northern Ireland weighting scheme

Category	Level	Unweighted proportion	Weighted proportion
Gender / Age	Male x 18-24	5.5%	5.2%
Gender / Age	Male x 25-34	6.0%	8.0%
Gender / Age	Male x 35-44	7.6%	8.3%
Gender / Age	Male x 45-54	9.1%	7.9%
Gender / Age	Male x 55-64	9.6%	8.2%
Gender / Age	Male x 65+	13.5%	11.0%
Gender / Age	Female x 18-24	4.7%	4.8%
Gender / Age	Female x 25-34	7.8%	8.0%
Gender / Age	Female x 35-44	8.2%	8.9%
Gender / Age	Female x 45-54	8.2%	8.3%
Gender / Age	Female x 55-64	8.7%	8.6%
Gender / Age	Female x 65+	10.7%	12.5%
Region	East	56.7%	50.5%
Region	South	20.1%	20.6%
Region	North	12.5%	15.2%
Region	West	10.7%	13.7%
Ethnicity	White	98.1%	97.0%
Ethnicity	Ethnic minority (excluding white minorities)	1.9%	3.0%
NS-SeC	Managerial, administrative and professional occupations	35.9%	29.8%
NS-SeC	Intermediate occupations	23.2%	22.0%
NS-SeC	Routine and manual occupations	27.9%	31.8%
NS-SeC	Never worked, long-term unemployed or full-time students	13.0%	16.4%
Religion or religion brought up in	Roman Catholic	41.8%	44.8%
Religion or religion brought up in	Protestant and Other Christian	51.0%	46.6%
Religion or religion brought up in	Other religion/None	7.2%	8.6%

Devolved Nation Regions

For UK and England weights, respondents were assigned to an ITL1 region based on their own self-reported residence (validated using their postcode). For Wales weights, respondents were assigned to Senedd Cymru Electoral Regions based on their Unitary Authorities. However, for Scotland and Northern Ireland, there are no similarly official regions that respondents can be easily assigned to. As such, in Northern Ireland, Savanta grouped Local Government Districts based on their location into four regions named North, South, East and West. In Scotland, Savanta grouped Council Areas into five regions based on old NUTS2 regions, named North Eastern Scotland, Highlands and Islands, Eastern Scotland, West Central Scotland, and Southern Scotland. A table showing how Local Authorities were matched to regions in Northern Ireland and Scotland is available on request.

Annex: Source Data

Unless otherwise specified, the data sources below were used to create both quotas and weights. UK ethnicity, social grade and education statistics were created by aggregating the results from the following three Censuses:

• 2021 ONS Census. This covers England and Wales and was conducted in 2021 by the Office for National Statistics.
• Scotland’s Census. This was conducted in 2022 by the National Records of Scotland.
• 2021 NISRA Census. This covers Northern Ireland and was conducted in 2021 by the Northern Ireland Statistics and Research Agency.

Age by gender

• UK: ONS Mid-Year Population Estimates
• England: ONS Mid-Year Population Estimates
• Wales: ONS Mid-Year Population Estimates
• Scotland: ONS Mid-Year Population Estimates
• Northern Ireland: ONS Mid-Year Population Estimates

Region

• UK: ONS Mid-Year Population Estimates
• England: ONS Mid-Year Population Estimates
• Wales: ONS Mid-Year Population Estimates, Ward to Unitary Authority to Senedd Cymru Electoral Region Lookup
• Scotland: ONS Mid-Year Population Estimates
• Northern Ireland: ONS Mid-Year Population Estimates

Ethnicity

• UK: Combined results from all three Censuses (weights only)
• England: 2021 ONS Census
• Wales: 2021 ONS Census
• Scotland: Scotland’s Census
• Northern Ireland: 2021 NISRA Census

National Statistics Socio-economic Classification (NS-SeC)

• UK: Combined results from all three Censuses (weights only)
• England: 2021 ONS Census
• Wales: 2021 ONS Census
• Scotland: Scotland’s Census
• Northern Ireland: 2021 NISRA Census

Education quotas

• UK: Combined results from all three Censuses
• England: 2021 ONS Census
• Wales: 2021 ONS Census
• Scotland: Scotland’s Census
• Northern Ireland: 2021 NISRA Census

1. Ofcom, Connected Nations 2024, p.12 ↩

2. The quotas used in this survey ensure that the sample recruited reflects the population of each constituent country, with a focus on those demographics which are most likely to have an association with peoples’ attitudes. Northern Ireland differs from the rest of the United Kingdom in that being Protestant or Catholic, whether by personal religiosity or by heritage, is strongly correlated with a number of other social and political attitudes, a result of the region’s distinctive history. ↩