Research and analysis

Precarious Work and Economically Inactive Survey: technical report

Published 12 August 2026

Introduction

Background

The Precarious Work and Economically Inactive Survey was commissioned by the Ministry of Housing, Communities and Local Government (MHCLG) to generate robust, up-to-date evidence on the experiences of individuals in precarious employment, as well as those currently out of work but seeking to return.

For the purpose of this study, MHCLG defined precarious workers as individuals aged 16 to 64 living in the United Kingdom who earned less than £30,500 per year (after tax and deductions) and met at least one of the following criteria:

• self-employed individuals earning less than £19,700 per year
• individuals on zero-hours contracts
• individuals in non-permanent work (casual, seasonal, fixed term, or agency contracts) who reported being unable to find a permanent job
• under employed individuals, defined as those working fewer than 16 hours per week while expressing a desire to work more
• individuals with volatile pay, reporting monthly fluctuations of more than 10%

In addition to precarious workers, the survey also targeted economically inactive individuals who expressed a desire to return to work. This group included:

• people who were unemployed and actively seeking employment
• people not currently working due to a long-term health condition or disability but interested in re-entering the workforce
• people looking after home or family who wished to return to work at a later stage

Summary of approach

The Precarious Work and Economically Inactive Survey was designed and delivered through a structured, multi-phase process to ensure methodological rigour and national representativeness. The project began with an initial meeting between MHCLG and Verian during which the study design, timeline, and deliverables were agreed. The questionnaire was co-developed by MHCLG and Verian, and included 26 closed questions on employment status, contract type, income volatility, working hours, workplace rights, financial impacts, and barriers to employment.

The primary sample was drawn from Verian’s nationally representative Public Voice^{[footnote 1]} panel using random probability methods. This was used to help calculate the incidence rates of precarious workers across the UK. To enhance the precision of survey estimates and enable more detailed subgroup analysis, the sample was then supplemented with the CINT panel network^{[footnote 2]}.

Fieldwork was conducted in two phases. The initial phase, running from 21 August to 9 September 2025, drew on the Public Voice panel and used a mixed-mode protocol combining online and telephone interviews to maximise accessibility and response rates. A soft launch pilot was carried out to test the questionnaire flow, script accuracy, and survey length. Eligible Public Voice participants received a £10 gift voucher, while those screened out received a £1 charity donation. In total, 1,370 eligible precarious workers and 590 economically inactive individuals wishing to return to work were interviewed in the initial phase. This also enabled us to calculate the national incidence rate of precarious workers which was 18.7%.

The supplementary phase used the CINT panel network, with fieldwork conducted online with 1,883 precarious workers completing the survey. This survey was conducted from 18 September until 9 October. Overall, combining both the Public Voice and CINT panel data in total 3,253 eligible precarious workers were interviewed and 594 economically inactive respondents.

All collected data underwent quality assurance checks, including validation of responses, and removal of incomplete or low-quality data. As a result of these checks, 23 respondents categorised as precarious workers were removed, and 56 respondents categorised economically inactive but wanting to return to work were removed. As such, the final dataset comprised 3,230 precarious workers and 538 economically inactive individuals wishing to return to work.

The dataset was sequentially weighted. See section 6 (Weighting) for details on this process. Verian then prepared detailed data tables, and a technical report documenting the survey methodology, sampling, fieldwork procedures, and weighting approach.

Data tables have been simplified for publication. In instances where the level of response in an individual cell amounts to fewer than 5, demographic breakdowns have been aggregated to avoid the risk of disclosure; including secondary suppression.

Ethics and data protection

Ethical review

The Precarious Work and Economically Inactive Survey underwent a formal ethics review by Verian’s internal expert panel at the start of the project. The overall ethical risk was rated low. Key risks identified included the potential for distress due to raising sensitive topics such as financial hardship or insecure work. To mitigate these, all questions were optional, and participants were given contact details for questions and any support. Interviewers for telephone surveys were experienced and briefed on topics and were aware of Verian’s safeguarding procedures. During fieldwork no safeguarding issues were flagged.

Data protection

All survey data was collected and managed by Varian, the contractor for this work. The survey data was securely stored in accordance with Verian’s data protection procedures. Personal data was stored in password-protected folders, with access restricted to authorised members of the project team. All data handling complied with Verian’s data protection policies.

A privacy policy was shared with all respondents. Verian will retain the data for 12 months following project completion, after which it will be permanently deleted. Participants were able to request deletion of their personal data prior to anonymisation or sharing; once anonymised, deletion is no longer possible. These measures ensured full compliance with GDPR and industry standards for data security and confidentiality.

Sample and fieldwork design

The Public Voice panel

At the time of this survey (August-September 2025), the Public Voice panel comprised 25,571 nationally representative members from across the UK. Most of these panel members were recruited via the ‘ABOS’ method in which (probabilistically) sampled individuals complete a 20-minute recruitment questionnaire either by web or on paper ensuring coverage of those who rarely or never use the internet. Targeted refreshment samples are conducted annually, reflecting Verian’s ongoing commitment to maintaining the quality and representativeness of the Public Voice panel.

CINT panel

The CINT panel network was used to supplement the Public Voice sample and increase the number of interviews with precarious workers. CINT provides access to around five million UK respondents from a wide range of panel sources, enabling efficient recruitment of low-incidence groups. Following completion of the Public Voice survey, the weighted sample distribution was used to generate specified quotas for the CINT survey of precarious workers. The inclusion of the CINT panel significantly enhanced the precision of survey estimates and enabled more detailed subgroup analysis.

Fieldwork design

Fieldwork for this survey was implemented in two phases: the first using the Public Voice panel, and the second using the CINT panel network.

Phase One: Public Voice panel

The first phase involved all members of the Public Voice panel between the ages of 16 and 64, employing a mixed mode methodology that combined online and telephone interviews to enhance accessibility and response rates. Given this age restriction the total Public Voice sample frame for this survey consisted of 19,197 panel members.

To begin, a soft launch pilot was conducted online randomly sampling 500 panel members. This was used to test the questionnaire flow, accuracy, and survey length, allowing for refinements before the main fieldwork commenced.

The main launch involved the remaining 18,697 eligible panel members to participate. Each contacted panel member had previously indicated their preferred mode of participation – either via a web link sent directly to their email or through a telephone interview. In this case, 134 panel members preferred to be contacted via telephone.

Emails sent to online participants contained personalised survey hyperlinks, along with unique survey logins and passwords. Additional verification was carried out using the panel member’s birthdate (including year), ensuring respondent authenticity. Three reminders were sent to increase participation rates. Where an email address was unavailable, SMS text messages were used only as supporting communication, and this was sent to panel members who had not opened the email 24 hours after it had been sent. A reminder letter (including QR code for survey access) was also sent to online non-responders alongside the third reminder email. It also informed respondents that a telephone interviewer may contact them via telephone to arrange an interview.

Telephone interviews were conducted to maximise response rates. Each respondent was called at least four times, with at least one attempt made over the weekend to accommodate varying schedules and increase participation. All respondents – regardless of mode – had access to a dedicated helpline via email or telephone, where they could request data deletion or ask questions about the research.

Phase Two: CINT Panel Network

The second phase of fieldwork utilised the CINT panel network, with all data collection conducted online. This phase targeted solely precarious workers. Demographic quotas were set by Verian’s methods team to enhance the precision of survey estimates and enabled more detailed subgroup analysis. 

CINT panellists were incentivised through the panel’s points system, which allows respondents to accumulate points for each completed survey and redeem them as vouchers. This system helped maintain engagement and encouraged participation across a diverse respondent pool.

Due to the low incidence rate of eligible respondents – those considered precarious workers or economically inactive but wishing to return to work – a large volume of invitations was required in both phases to achieve the target sample size, as outlined in the fieldwork outcomes section below.

Data Collection and Management

All survey data was collected via a secure platform. This ensured that only Verian had access to the survey responses, enabling secure data management and real-time tracking of progress.

Table 1: Fieldwork process

Panel	Activity	Date
Public Voice	Soft launch	21 August 2025
Public Voice	Main Launch	26 August 2025
Public Voice	Reminder 1	29 August 2025
Public Voice	Reminder 2	04 September 2025
Public Voice	Reminder 3	09 September 2025
CINT	Main Launch	23 September 2025

Questionnaire design

Questionnaire development

Prior to the project kick-off meeting, MHCLG provided an initial draft of the questionnaire and the range of questions which they wanted the study to explore and discussed the overall objectives and the survey content. Verian then facilitated a review process, working closely with MHCLG to finalise the questionnaire.

In this study, precarious work was defined as meeting at least one of the following criteria:

• self-employed individuals earning less than £19,700 per year
• individuals on zero-hours contracts
• individuals in non-permanent work (casual, seasonal, fixed term, or agency contracts) who reported being unable to find a permanent job
• under employed individuals, defined as those working fewer than 16 hours per week while expressing a desire to work more
• individuals with volatile pay, reporting monthly fluctuations of more than 10%

To ensure the survey did not capture higher earners who met these criteria, an income threshold was introduced. As such respondents earning more than £30,500 per year were screened out. This was to avoid including individuals earning above the median salary, such as high-paid consultants who may otherwise be seen as in precarious employment. A question was added to identify respondents earning less than £19,700 per year, allowing for separate analysis of this group. Verian also incorporated demographic questions.

Eligibility questions were developed to identify individuals who were economically inactive but expressed a desire to return to work in the future. These included people who were unemployed and actively seeking employment; those not currently working due to a long-term health condition or disability but interested in re-entering the workforce; and individuals looking after home or family who wished to return to work at a later stage.

The questionnaire underwent several rounds of review over a three-week period. The review process focused on best practices in survey design, including the use of plain English, comprehensive and clear response options, logical question ordering, and effective routing.

It should be noted that one of the key objectives of the survey was to quantify the financial impacts associated with precarious work. To address this, respondents were asked to estimate how much money they had lost or had to spend in the past four weeks due to problems with their work situation, across several categories (e.g. lost earnings from cancelled shifts or reduced hours, additional childcare or care costs, borrowing costs). However, given the high cognitive burden involved in recalling and monetising such events, as well as providing a sufficiently comprehensive range of response options, the resulting data should be interpreted with some caution. Self-reported financial losses are susceptible to recall error and variation in interpretation. While qualitative methods such as case studies or income diaries can capture these financial effects with greater accuracy and contextual depth, they do not offer the same level of generalisability as a survey. The inclusion of this question represented a pragmatic attempt to capture, in broad terms, one of the survey’s key objectives.

The Government Statistical Service guidance currently states that there is no finalised harmonised standard for collecting gender identity data, therefore no agreed best practice to collecting gender identity. Indeed, ONS is undertaking further research to better understand the limitations of the census 2021 gender question. Therefore, the question on gender is self-reported, rather than sex, and we do not treat it as directly comparable with, for example, ONS 2021 Census sex or gender identity statistics.

Questionnaire structure

The topics covered in the questionnaire are listed below. For the full survey questionnaire see the Annex. 

1. Demographics
2. Employment status and sector
3. Income and pay volatility
4. Employment contract type and working hours
5. Workplace benefits and rights
6. Job satisfaction and quality
7. Workplace issues and insecurity
8. Financial impacts of precarious work
9. Barriers and concerns for returning to work
10. Protections and support needed to return to work

Table 2: Breakdown of fieldwork outcomes (prior to quality checks)

	Target	Achieved online	Achieved telephone	Achieved total	% of target achieved
Total Completes	3,750	3,809	38	3,847	102.59%
Eligible Precarious Workers	3,250	3,231	22	3,253	100.09%
Eligible Economically inactive	500	578	16	594	118.80%
Screened Out - Age (under 16 or over 64)	-	949	4	953	-
Screened Out - Not working and not looking to return to work in the future	-	3,556	18	3,574	-
Screened Out - Earning £30,500 or more	-	8,713	57	8,770	-

Weighting

Public Voice respondents

The Public Voice respondent sample including screen outs were weighted in three stages:

1. Survey base weight

For each respondent, a base weight was calculated by dividing their panel weight by their probability of being sampled for the survey (sampling probabilities varied considerably across panel members). This was used as weight (1).

2. Propensity score weight

A propensity score weight was estimated for each respondent using a set of recruitment survey variables. Technically, this weight represents the estimated odds of a respondent appearing in the weighted panel dataset rather than in the survey-base-weighted respondent dataset, when the latter is added to the former (i.e. respondents are present in both datasets). To avoid over-reliance on the model, the propensity score weight was constrained to fall within the inter-95th percentile range. This was used as weight (2).

3. Calibration to population benchmarks

The product of weights (1) and (2) was used as the starting point for calibration, with values trimmed to lie between the 1st and 99th percentiles. The respondent sample was then calibrated to the latest available weighted ONS Annual Population Survey (2024), with respect to sex by age group and region. The calibration was performed using the classic raking algorithm, which is transparent, avoids negative weights, and preserves much of the sample’s covariance structure.

To assess the representativeness of the weighted respondent dataset, we compared it to the weighted panel dataset across 345 category-level proportions derived from 92 recruitment survey variables (both demographic and non-demographic). The median difference between the two datasets was 0.4 percentage points, with 99% of differences falling within three percentage points.

The overall weighting efficiency was 50%, corresponding to a design effect of 2.00 and an effective sample size of 4,255 (calculated as 8,535 ×50%). Respondents who did not pass the data quality checks, described in the section below were excluded from the weighting process.

The table below presents the calibration matrix used for the survey, based on the ONS Annual Population Survey (2024).

Table 3: ONS Annual Population Survey population estimates, 2024, UK adults aged 16-64

Variable	Category (all)	% of population (100.0)
Sex/age group	Male 16-24	8.4
Sex/age group	Male 25-34	10.9
Sex/age group	Male 35-44	10.2
Sex/age group	Male 45-54	10.3
Sex/age group	Male 55-64	10.0
Sex/age group	Female 16-24	8.1
Sex/age group	Female 25-34	10.7
Sex/age group	Female 35-44	10.4
Sex/age group	Female 45-54	10.7
Sex/age group	Female 55-64	10.4
Region	NE England	3.9
Region	NW England	10.8
Region	Yorkshire & The Humber	8.1
Region	E Midlands	7.1
Region	W Midlands	8.7
Region	E England	9.2
Region	London	14.8
Region	SE England	13.5
Region	SW England	8.1
Region	Northern Ireland	2.8
Region	Wales	4.6
Region	Scotland	8.3

After weighting was complete, screen outs were removed from the dataset, leaving just those who completed the full questionnaire. The weights for these cases were rescaled to have a mean value of 1. For target group (i) – those in precarious employment – the effective sample size was 694 (actual sample size = 1,350; weighting efficiency = 51%); for target group (ii) – those economically inactive individuals wishing to return to work – the effective sample size was 274 (actual sample size = 538; weighting efficiency = 50%).

CINT respondents

Once the Public Voice survey was complete, the weighted distribution for target group (1) was used to form quotas for a supplementary sample from the CINT network. An additional quality assured respondent sample of 1,880 from CINT was added to the Public Voice sample of 1,350. The CINT sample was calibrated to the weighted Public Voice sample with respect to sex/age group, region, working status, and highest education level.

The weighting efficiency of the CINT sample was 71%, corresponding to a design effect of 1.40 and an effective sample size of 1,342 (calculated as 1,880×71%).

The two samples were then combined, giving equal weight to both sources. This meant multiplying the Public Voice weights by 1.194 and the CINT weights by 0.860.

The combined sample weighting efficiency was 57%, corresponding to a design effect of 1.77 and an effective sample size of 1,829 (calculated as (1,350+1,880)×57%).

Data quality checks

The first stage of data cleaning and processing involved conducting quality checks separately on the Public Voice and CINT raw datasets. Initial checks focused on identifying duplicate responses, of which none were found.

The second step involved detecting speeders (completing the survey in less that one third of the median completion time) and straight-liners (those providing the same response across multiple questions). As a result of these checks, 23 respondents categorised as precarious workers were removed, and 56 respondents categorised economically inactive were removed.

Further cleaning was undertaken for the financial impact questions (e.g. reported lost earnings due to precarious work). To reduce the influence of outliers, responses falling within the bottom 1% and top 1% of reported values were re-coded as “Prefer not to say”. This approach minimised the effect of anomalous entries, such as extremely small or inflated figures (e.g. additional zeros entered by mistake). The remaining responses were then reviewed individually to identify any residual anomalies, and a small number were subsequently recoded as “Prefer not to say.”

These steps ensured that the final dataset was robust, internally consistent, and ready for analysis. As all questions were closed and completed in English, no additional coding or cleaning for open-ended responses or translation was required.

It should be noted a total of 770 respondents reported having a permanent contract in their main job yet still met the definition of a precarious worker. This occurred because eligibility was based on multiple indicators beyond contract type. For example, 496 reported volatile pay exceeding 10% month to month, making financial planning difficult. In addition, 212 of these respondents were working fewer than 16 hours per week despite wanting to work more, classifying them as under-employed. A further 76 respondents were classified as low paid self-employed, having identified as self-employed in a previous question and earning less than £19,500 per year. These figures sum to more than 770 due to some respondents meeting multiple criteria^{[footnote 3]}.

Table 4: Breakdown of fieldwork outcomes (post quality checks)

	Target	Achieved online	Achieved telephone	Achieved total	% of target achieved
Total Completes	3,750	3,735	33	3,768	100.48%
Precarious Workers	3,250	3,208	22	3,230	99.40%
Economically inactive	500	527	11	538	107.60%

Table 5: Breakdown by precarious worker criteria (post quality checks)

	Total	% of total precarious workers
Earn under £30,500 take home pay per year	3,230	100.00%
Self-employed respondents who earn under £19,700 take home pay per year.	663	20.52%
Respondents on zero-hours contracts.	587	18.17%
Respondents in non-permanent work (casual, seasonal jobs, fixed term and agency) who are unable to find permanent work.	1,292	40.00%
Respondents who report working less than 16 hours a week despite wanting to work more (i.e., under-employed).	639	19.78%
Respondents who self-report volatile pay while being below median income (changes in pay of more than 10%).	973	30.12%

1. Public Voice is Verian’s own random probability panel. Most of these panel members were recruited via the ‘ABOS’ method in which (probabilistically) sampled individuals complete a 20-minute recruitment questionnaire either by web or on paper. Recruitment surveys were carried out yearly  and the respondent samples have been linked together via a weighting protocol to form a single panel. A full technical report for Public Voice is available separately. ↩

2. CINT’s Panel Exchange provides access to approximately five million UK respondents drawn from a range of panel sources. It uses multi-channel recruitment, calibrated incentives, and machine learning validation to ensure demographic diversity, respondent engagement, and data quality. ↩

3. See table 5 for a breakdown of the criteria for a respondent to be categorised as a precarious worker. All eligible respondents must earn under £30,500 take home pay per year, and at least one of the other criteria listed in the table. Please note respondents could meet more than one of the criteria. ↩