Policy background

IWC is a form of ‘conduct conditionality’ whereby entitlement to social benefits is dependent upon recipients’ fulfilment of behavioural requirements (Clasen and Clegg, Reference Clasen, Clegg, Clasen and Siegel2007: 172). Although conduct conditionality has become increasingly common across advanced welfare states following the activation turn of the 1990s, it is central to the UK’s ‘work-first’ activation model, in which unemployed individuals’ entitlement to social security is conditional on requirements to search for, apply for, and accept jobs, with relatively harsh penalties, known as sanctions, for non-compliance (Lindsay, Reference Lindsay, Serrano Pascual and Magnusson2007; Dwyer and Wright, Reference Dwyer and Wright2014; Langenbucher, Reference Langenbucher2015; Immervoll and Knotz, Reference Immervoll and Knotz2018). In contrast, up until the introduction of IWC within Universal Credit in 2012, low-income families combining cash benefits with paid work were not subject to conduct conditionality, with policies for these groups designed to incentivise rather than enforce certain behaviours (Dwyer and Wright, Reference Dwyer and Wright2014: 31).

Until recently, means-tested benefits for households categorised as inwork and those categorised as unemployed have followed distinctive reform paths (Clegg, Reference Clegg2015). The Conservative government’s Jobseekers Act of 1995 is considered a ‘watershed moment’ as it ‘intensified’ the monitoring of unemployment benefit recipients’ job search behaviour in light of concerns about welfare expenditure and claimants’ attitudes (Watts et al., Reference Watts, Fitzpatrick, Bramley and Watkins2014: 3). However, this approach was not applied to in-work benefits. Similarly, from 1997, New Labour sought to ‘take tough measures’ by intensifying conduct conditionality, including for disabled people and lone parents, while expanding support for working households to ‘extend opportunities’, promote social inclusion through work and reduce child poverty (Blair, Reference Blair1996; Walker and Wiseman, Reference Walker and Wiseman2003; Patrick, Reference Patrick2012; Fletcher, Reference Fletcher2014).

This separation has become less clear-cut. The Conservative-Liberal Democrat coalition government, elected in 2010, increased sanctions attached to out-of-work benefits to their highest levels, justified on the grounds that out-of-work claimants are workshy, lazy and scrounging from the state, and also retrenched in-work benefits (Patrick, Reference Patrick2012; Dwyer, Reference Dwyer, Horsfall and Hudson2017; Fletcher and Wright, Reference Fletcher and Wright2018: 329; Watts and Fitzpatrick, Reference Watts and Fitzpatrick2018: 9). The introduction of Universal Credit in 2012, which merges in- and out-of-work benefits and extends conduct conditionality to working households, blurs the distinction between these groups (Dwyer and Wright, Reference Dwyer and Wright2014: 31). Later Conservative governments also weakened conduct conditionality from previous record high levels, with the maximum sanctioning period reduced from three years to six months (DWP 2019b; Sage, Reference Sage2019), suggesting a softening of punitive mechanisms for out-of-work households.

More specifically, IWC means that some working recipients of UC with earnings below a defined threshold could be required to take steps to increase their earnings. The earnings threshold is approximately thirty-five hours per week at the level of the statutory minimum wage, or lower for claimants with limited work capacity (e.g. because of disabilities or caring responsibilities) (Economic Affairs Committee, 2020: 64–66). The requirements, which are mandated in a contract (the ‘Claimant Commitment’) drawn up with employment services, are not clearly defined by the government, but could include asking a current employer for a pay rise or more hours of work, applying for alternative jobs with better pay or hours, and/or applying for additional jobs (Jones et al., Reference Jones, Berry, Rouse and Whittle2019: 4; DWP, 2020a).

Early analysis argued that the sanctions bolstering IWC would be tough (Dwyer and Wright, Reference Dwyer and Wright2014: 30–31). However, according to the Department for Work and Pensions’ statistics (2020b), UC claimants categorised as ‘Working – with requirements’ were required to undertake work-related requirements with a conditionality regime that is ‘less frequent and lighter’ than for unemployed claimants and those with very low earnings (DWP, 2020b). From January 2019 – 2020, just under a sixth of all UC claimants were registered under the IWC regime (Table 1). This shows that UC still partly distinguishes recipients’ conditionality regime on the basis of work-status, even if the extension of conduct conditionality to working households nevertheless marks a significant shift (Dwyer and Wright, Reference Dwyer and Wright2014: 31).

Table 1 Universal Credit conditionality regimes

Notes: Authors’ calculations. Average claimants in each group based on monthly figures for Jan 2019 - Jan 2020. Source: DWP, ‘Number of people on Universal Credit by conditionality regime’. Accessed at: https://stat-xplore.dwp.gov.uk/.

The ambiguity surrounding IWC is partly because employment service staff members are afforded some discretion in setting requirements and applying penalties (DWP, 2020a). These details were also not confirmed in the original legislation and the government tested different applications of IWC after it became law, finding limited evidence that more intensive requirements lead to a minor increase in earnings (DWP, 2018; 2019a). This apparent caution in the application and communication of IWC indicates the government may not be convinced of the effectiveness of the policy – indeed, there is a limited evidence base for IWC – and/or it may be aware of legitimacy issues among its prospective voters. The next section therefore sets out our framework for analysing the relationship between citizen-voter attitudes towards IWC, and political parties’ vote-seeking and policy-seeking behaviour (Beramendi et al., Reference Beramendi, Häusermann, Kitschelt and Kriesi2015).

The politics of punitive (in-work) conditionality

Our framework focuses on the relationship between citizen-voter attitudes and political parties’ vote-seeking and policy-seeking behaviour, and integrates insights from traditional and new perspectives on the politics of social policy (Bonoli and Natali, Reference Bonoli, Natali, Bonoli and Natali2012; Häusermann et al., Reference Häusermann, Picot and Geering2013; Beramendi et al., Reference Beramendi, Häusermann, Kitschelt and Kriesi2015). Traditional perspectives map the preferences of voters and political parties on a single left-right dimension and assume alignment between parties and their voters on the basis of industrial social class (Häusermann et al., Reference Häusermann, Picot and Geering2013: 221). This perspective would suggest that left parties oppose IWC, as IWC restricts entitlement to benefits that their working-class voters are more likely to receive. Similarly, right-wing parties would support IWC because reducing entitlements to welfare appeals to voters further along the income distribution who may stand to gain from lower social expenditure and taxation. Indeed, both those that identify as left wing and those on a low income or unemployed are more likely to oppose conduct conditionality applied to unemployment benefits (Larsen, Reference Larsen2008; Achterberg et al., Reference Achterberg, van der Veen and Raven2014; Fossati, Reference Fossati2018)

However, several interrelated factors require us to go beyond the traditional view. First, IWC relates to areas of contestation that transcend or nuance conflict relating to more or less tax and expenditure. For example, the government frames IWC in UC as support for low-income working households to progress in the labour market (Millar and Bennett, Reference Millar and Bennett2017: 177). This goal conceivably aligns with a social investment agenda to foster inclusion of groups at risk in post-industrial societies and therefore could appeal to parties and/or voters, including those on the left, supportive of these ideals (Häusermann, Reference Häusermann2012: 114–115, 119). However, IWC within UC defines ‘progression’ in the labour market primarily with reference to earnings, which likely limits the scope of employment support provided to claimants (Jones et al., Reference Jones, Berry, Rouse and Whittle2019: 15). IWC may therefore be viewed as aligning with work-first ideals that emphasise supply-side barriers to progression (Häusermann, Reference Häusermann2012: 119).

Relatedly, IWC could also relate to preferences for punitive welfare, i.e. the use of penalties to punish undesirable behaviour. Studies suggest there are distinct preferences for punitive policies that do not necessarily map on to economic interests (Achterberg et al., Reference Achterberg, van der Veen and Raven2014: 217). Political parties and voters may be supportive of IWC if they believe working households in receipt of benefits should be penalised when their behaviour is socially undesirable e.g. when they are not working as hard as they can be and are claiming benefits to make up for their lack of earnings (Dwyer and Wright, Reference Dwyer and Wright2014: 31). This could also work in reverse – e.g. opponents may reject IWC on the basis that it is unnecessarily punitive.

However, the dimensions of conflict among parties and voters that are likely to be emphasised in the debate on IWC should also be considered in light of a changing context of welfare state politics. In post-industrial societies, socio-economic transformations have shifted the composition of the electorate, including in the declining proportion of industrial working-class voters, driving partisan dealignment and prompting efforts from political parties to reconcile multi-dimensional policy preferences of new coalitions of voters (Bonoli and Natali, Reference Bonoli, Natali, Bonoli and Natali2012; Beramendi et al., Reference Beramendi, Häusermann, Kitschelt and Kriesi2015). Importantly, the decline in the size of the industrial working class has prompted many left-wing parties to appeal to middle-income voters by expanding social investment and activation, such as in-work benefits, and fiscal constraint in other areas (Gingrich and Häusermann, Reference Gingrich and Häusermann2015). Thus, the precise framing and application of IWC in ways that are conducive to activation are likely to influence its support among middle-income voters.

Moreover, growing inequalities based on education are argued to have driven the emphasis on the ‘cultural’ dimension of welfare politics by political parties aiming to appeal across groups with possibly divergent economic interests (Bovens and Wille, Reference Bovens and Wille2017; Iversen and Soskice, Reference Iversen and Soskice2019). Studies show that lower levels of education are associated with greater support for sanctions in unemployment benefits (Achterberg et al., Reference Achterberg, van der Veen and Raven2014). As individuals with lower education are more likely to receive these benefits, some attribute these attitudes to the relationship between education and authoritarian values (Knotz, Reference Knotz2016) or notions of deservingness (Van Oorschot, Reference van Oorschot2000). Indeed, Labour and Conservative governments have utilised authoritarian frames, such as being ‘tough on welfare’, to justify conduct conditionality, while political parties trying to appeal to university graduates have critiqued punitive welfare in later years (Sage, Reference Sage2019). The conflict surrounding IWC could therefore lie on the cultural axis with patterns of public support mirroring educational divisions, depending on the emphasis on and application of its punitive elements.

However, cutting across the previous points is the argument that political parties’ and voters’ preferences are not formed in a vacuum (Bonoli and Natali, Reference Bonoli, Natali, Bonoli and Natali2012). IWC could find support if it is perceived to be the best option to address new challenges in a constrained policy context. In the UK, there are concerns that UC could disincentivise work because, as a merged in- and out-of-work benefit, it does not have a minimum working hours threshold for eligibility. In contrast, the outgoing Working Tax Credits are conditional on employment of at least sixteen hours a week (for those aged twenty-five to fifty-nine without children the minimum number of hours is thirty) (Finch, Reference Finch2016: 31). The rate at which UC is withdrawn is also expected to reduce work incentives for some households (Dwyer and Wright, Reference Dwyer and Wright2014: 30). IWC may be considered a necessary fix to issues of work incentives under UC, where alternatives, such as abolishing UC and returning to a previous system of in- and out-of-work benefits, are not considered feasible (e.g. because of sunk costs).

Finally, political parties’ understanding of the (dis)alignment between their policy-seeking and vote-seeking preferences are also likely to drive their position on, and framing of, IWC (Pierson, Reference Pierson1994; Bonoli, Reference Bonoli, Bonoli and Natali2012: 100). If parties expect their prospective voters to support IWC, they are likely to ‘claim credit’ for the reform by, for example, publicly expressing support and/or claiming ownership of the reform, its implementation and/or its ideas (in the case of parties that did not implement the reform), as was broadly the case for conduct conditionality in out-of-work benefits (Sage, Reference Sage2019: 100). Yet, if parties support or preside over IWC, and yet suspect their prospective voters could oppose the reform, they may adopt so-called ‘blame avoidance’ strategies, designed to minimise potential electoral harm of unpopular reforms (Pierson, Reference Pierson1994).

On the latter, if the government responsible for IWC implements a blame avoidance strategy, we might expect this to take one of three main forms (Vis, Reference Vis2016: 128-131, 137). The first is the manipulation of procedures by, for example, ‘agenda limitation’ (i.e. preventing decisions from being taken openly). The second is the manipulation of perceptions by, for example, ‘redefining the issue’ (i.e. mobilising new groups that may perceive the reform as in their interest) or ‘strategic re-framing/communication’ (e.g. downplaying less popular aspects of the reform in its communication). Finally, parties can manipulate payoffs through ‘exemption’ (i.e. exempting specific categories of opponents), ‘concentration’ (i.e. imposing losses on groups that are politically weakest) or ‘creative accounting and lies’ (i.e. redefining terms or measures).

Given the range of factors possibly influencing preferences for and positions on IWC, we do not set out precise expectations on the key question of whether politics parties and their voters are aligned on IWC and how or why. Instead, we examine the arguments parties use in support of or against IWC to gauge the coalitions of voters that parties are trying to mobilise and assess the kinds of blame avoidance or credit claiming strategies pursued. Individual-level factors associated with IWC preferences are also likely to indicate the dividing lines of the politics of IWC, and we examine if this resembles the income and educational divide found for other forms of welfare.

Data, variables and analysis

We examine the preferences of voters and the preferences, rationales and strategies of political parties and their elected representatives in Parliament and Government concerning IWC. The study period, 2011 to 2020, is informed by the introduction of IWC in the Welfare Reform Bill, first presented to the House of Commons in early 2011. During this time, the UK voted to leave the European Union, there were three general elections (2015, 2017 and 2019) and high profile changes in the leadership of the main political parties. Yet welfare reform debates remained salient because of the introduction and implementation of UC. Changes in party leadership, especially the election of Jeremy Corbyn as leader of the Labour Party, and the associated rise of the left wing of the party (Sage, Reference Sage2019), are nevertheless significant for our analysis because they coincided with a shift of position on welfare issues that subsume IWC, including punitive welfare and the question of whether to keep or abolish UC. Party positions on IWC were gauged primarily through analysis of MPs’ contributions in Parliament and secondarily through party manifestos for the 2015, 2017 and 2019 general elections (UC was not announced prior to the general election in 2010). Relevant contributions were identified through searches of Parliamentary records online, the records of the Public Bill Committee (PBC) of the Welfare Reform Bill, and in manifestos, using the keywords ‘in-work conditionality’ and ‘in-work progression’. The searches and sources included in the analysis are summarised in the Supplementary Material (Tables A-C). MPs’ contributions were coded systematically using NVivo 12 Software (NVivo, 2020), starting with broad preferences (e.g. support) and characteristic nodes (e.g. political party), followed by thematic coding of the associated arguments (e.g. reduce welfare expenditure). The manifestos searches yielded no relevant references (Conservative Party, 2015; 2017; 2019; Green Party of England and Wales, 2015; 2017; 2019; Labour Party, 2015; 2017; 2019; Liberal Democrats, 2015; 2017; 2019; Scottish National Party, 2015; 2017; 2019). Yet the manifestos form part of the analysis as they show parties’ positioning and strategic communication to voters on the overarching topic of conduct conditionality and welfare reform (Budge et al., 2001). Moreover, the only time IWC was debated before becoming law was in the Public Bill Committee (PBC). The policy debates in PBCs differ from those in the main chamber of the House of Commons because their purpose is to scrutinise the proposed legislation line by line in ‘an arena in which MPs are not vying for press coverage of their comments’ (Thompson, Reference Thompson2016: 44). PBC members are selected primarily by their party’s leadership (Russell et al., Reference Russell, Morris and Larkin2013: 11) and are proportional to party representation in Parliament. The success rate of non-government amendments to legislation is therefore very low (Thompson, Reference Thompson2016: 42). We therefore focus on what MPs say during PBCs, rather than what they do (e.g. table amendments), to gauge parties’ policy-seeking intentions on IWC.

We also analyse voter preferences using data from an online, cross-sectional survey data conducted by a professional polling company, IPSOS-Mori, in August 2017, on behalf of researchers from the Institute for Policy Research at the University of Bath. A representative sample of 1,111 adults aged eighteen to seventy-five in the UK was drawn from IPSOS-Mori’s (non-representative) omnibus panel and post-stratification weights constructed based on population statistics sourced from Eurostat. The use of an online survey raises concerns about the sampling procedure, e.g. the under-representation of certain groups, especially the digitally excluded, and data quality, e.g. respondents providing satisficing answers (Frippiat et al., Reference Frippiat, Marquis and Wiles-Portier2010). There is an over-representation of Labour voters in the sample relative to their 2017 election performance, even after weighting. Nevertheless, we found highly comparable results in a near identical question on unemployment benefit conditionality in the European Values Study conducted in the UK less than a year later between February and July 2018 (see Table E in the Appendix).

Attitudes to IWC were analysed using dependent variables comprising four survey items.

IWC was described in the survey with reference to tax credits for households undertaking paid work. Yet the description of IWC closely resembles that which it is suggested is in place in UC. UC was not explicitly referenced in the question because it is high profile and politicised as a flagship government policy (Dwyer and Wright, Reference Dwyer and Wright2014; Millar and Bennett, Reference Millar and Bennett2017). Tax credits for working households are less politically aligned (Clegg, Reference Clegg2015). Our aim was to isolate (to some degree) attitudes to IWC, rather than measure attitudes to UC as a whole.

The survey items differentiated between four types of in-work conditionality: two ‘passive’ and two ‘active’ (see items 1.1 to 1.4 in Table 2). Passive in-work conditionality reduces benefits if recipients turn down an offer of more hours of work or an offer of a better-paid job. The active forms of in-work conditionality reduce benefits if recipients did not actively look for a better-paid job, or opportunities to increase their hours of work. Respondents were asked whether each was ‘Acceptable’ or ‘Not acceptable’. A third ‘Don’t know’ response was included.

Table 2 Overview of in-work conditionality dependent variables

We converted the four IWC survey items into a binary variable to carry out logistic regression analysis. This involved four coding stages: (1) Each item was coded as 1 for pro-IWC responses (‘Acceptable’), -1 for anti-IWC responses (‘Unacceptable’) and 0 for ‘Don’t know’; (2) The items were summed giving a range of -4 to 4; (3) Values -4 to 0 were recoded as 0, while values 1-4 were recoded as 1; (4) Respondents who answered ‘Don’t know’ to all four items were coded as missing (N= 90, 8.1 per cent). Thus, the binary variable identifies those that were more supportive of IWC across the four items than they were opposed, while excluding those with no opinion at all.

Independent variables were included to explore both socioeconomic characteristics and political partisanship as determinants of attitudes to IWC. These included age (continuous), highest level of education (GCSE level, A level or Degree and above), gender, whether respondents had children in their home (dummy), household income (<£20k a year, £20k-£35k a year, £35k a year+ and Prefer not to say) and employment status (full-time, part-time, self-employed, unemployed, retired and not employed). Political party choice at the last general election (2017) was used to test for the relevance of political partisanship. The sample was too small for individual analysis of the parties other than Conservatives and Labour in the regressions, but these are included (as ‘other’ parties). Attitudes to conduct conditionality within unemployment benefit (see Table E in Appendix) were also included in a final model to test whether opposition to conditionality in general may explain the relationship between variables: ‘Don’t know’ responses (2.89 per cent) were coded as missing. Responses 1-5 were coded as support for conduct conditionality (1) and 6-10 were coded as opposition.

Findings

Voter preferences and in-work conditionality

Table 3 summarises the weighted responses for the four survey questions on in-work conditionality. The results indicate that the majority of voters support the principle of a passive form of IWC that would see benefits reduced for recipients turning down an offer of better-paid work (54 per cent support, 29 per cent oppose) or more hours of work (50 per cent support, 33 per cent oppose). However, active forms of IWC revealed a plurality of respondents opposed sanctions for not searching for better-paid work (36 per cent support, 45 per cent oppose) or not searching for more hours of work (36 per cent support, 42 per cent oppose).

Table 3 Voter preferences for four in-work conditionality types

While a high proportion of ‘Don’t know’ responses for all four IWC questions could suggest indifference or lack of understanding from nearly 20 per cent of respondents, only eight per cent of respondents answered ‘Don’t know’ to all four questions, suggesting that most had a formulated opinion on at least one of the requirements (see Table D in the Supplementary Material for the breakdown of ‘Don’t know’ responses by party).

Overall levels of support and opposition are helpful for gauging the social legitimacy of IWC, but parties are also concerned with the views of specific groups that they identify as potential voters. Figure 1 summarises the responses according to party preference for the 2017 general election. The findings indicate there is a stark partisan divide between voters of the two main parties in preferences for IWC. Support for IWC was much higher among Conservative voters than among Labour voters and voters of other parties. The pattern of higher support for passive IWC than for active IWC is consistent across all voters, despite different levels of support among each party.

Figure 1. Support for in-work conditionality by party support. The Y-axis shows the proportion of respondents agreeing it is ‘Acceptable to reduce payments’ when a recipient of in-work benefits (a) turned down an offer of more hours, (b) turned down an offer of a better paid job, (c) did not actively search for more hours, (d) did not actively search for a better paid job. The denominator includes ‘Don’t Know’ responses. The X-axis shows a respondent’s recalled vote in the 2017 general election. DNV = did not vote, N/A = respondent did not provide an answer or did not know.

Preferences for IWC were analysed using a binary variable distinguishing between those more in favour of IWC across the four survey items versus those that are not. Logistic regression models predicting attitudes to IWC are shown in Table 4. Three models were constructed to regress support (in favour of) for in-work conditionality. Model 1 includes education, income categories, a dependent children dummy (‘parents’), gender, and labour market status. Model 2 adds political party preferences and Model 3 adds attitudes to unemployment benefit conditionality. The last two columns in Table 4 show the proportion of respondents within that category in favour of IWC and the total number of respondents in the sample to aid the interpretation of the regression results.

Table 4 Multivariate logistic regression models predicting support for in-work conditionality

The results of the regression analyses further highlight the significance of the relationship between political partisanship and attitudes to IWC. Compared to the reference group (Conservative voters), voting for Labour and ‘other’ parties have a significant negative effect on support for IWC even with the inclusion of sociodemographic covariates. The effect size was reduced partly when controlling for unemployment benefit conditionality preferences, but nevertheless remains highly significant.

The results also indicate that economic interests are likely to be an important driver of attitudes to IWC. Belonging to the highest income group is associated with significantly higher support for IWC (compared to the low-income reference group) across all three models. Considering the multivariate nature of the model, the increase in the coefficients for income between Model 1 and Model 2 indicates that, even when controlling for political party preferences, there is a significant effect of income on support for IWC. This means, for example, that even among Labour voters, who are significantly more likely to oppose IWC compared to Conservative voters, there is an independent, positive effect of income on their support for IWC. The negative and significant effect of unemployment in Model 1 is also in line with the argument that economic interests would lead to opposition to IWC. The effect becomes insignificant with the inclusion of party preferences in Model 2, indicating unemployment and party preference are correlated, although it is also primarily due to the small sample of unemployed respondents. On the other hand, part-time employment and self-employment are not significantly related to attitudes to in-work conditionality, compared with full-time employment, in any model. The former is particularly surprising given part-time workers are more likely to be subject to in-work conditionality than full-time workers.

The consistently significant negative effect of having dependent children on support for in-work conditionality is also a striking finding. One explanation could be that individuals with additional caring responsibilities view IWC as at odds with their caring responsibilities, which aligns with the challenges to IWC made by some Labour and SNP MPs in Parliament.

Employment statuses furthest removed from the labour market – retired and not employed – have a significant negative effect on support for IWC, compared to full-time employment. The result only becomes significant for retired individuals in Models 2 and Models 3, indicating that this only holds when controlling for their greater propensity to vote Conservative and support unemployment benefit conditionality. The effect and significance of those not employed also increases in Model 2. Similar to the negative effect of parenthood, this may be because these groups value and/or rely on the unpaid work of those potentially affected by IWC, such as care work. Contrary to this assumption, gender is insignificant, with women, if anything, slightly more likely to be in favour of IWC.

Education, typically a strong predictor of authoritarian values, is also not significant in any of our models. This casts doubt on the idea that IWC is likely to cut across the economic left-right dimension of politics and tap into the educational cleavage across advanced democracies. Table F shows that higher levels of education are a significant predictor of opposition to unemployment benefit conditionality, as found in other studies, suggesting it is not simply the sample or the contemporary UK context behind the null result. While support for unemployment benefit conditionality (binary) was strongly and significantly related to support for IWC, even when controlling for party preference and the other socio-demographic factors, there are important distinctions between the types of voters split on these issues.

The importance of voter mobility in the theoretical framework, and the findings thus far, warrant further probing of variation of preferences among those who voted for the two main parties in 2017. Figure 2 further illuminates the limited differences between voters’ attitudes to IWC across levels of education, age and gender, as found in the regression analysis. The graphs indicate there is more division among Conservative voters than Labour voters. The most striking within-party conflict is among Conservative voters with and without dependent children, with over 20 per cent difference in the levels of support. There is also a larger divide between the oldest age bracket (55-75) and the middle-aged bracket (35-54) as well as those with A level and Degree qualifications among Conservative party voters compared to Labour voters. This points to potential problems for the Conservatives with certain groups of voters, despite a majority of their supporters broadly in favour of the policy.

Figure 2. Support for in-work conditionality by party vote and education/gender/age/parental status. The Y-axis shows the proportion of respondents that agreed it was ‘Acceptable to reduce payments’ more often than they agreed it was ‘Unacceptable to reduce payments’ across the four survey items that asked about a recipient of in-work benefits that (a) turned down an offer of more hours, (b) turned down an offer of a better paid job, (c) did not actively search for more hours, (d) did not actively search for a better paid job. The denominator excludes respondents that stated ‘Don’t know’ to all four items. The X-axis in the top-left quadrant shows a respondent’s recalled vote in the 2017 general election and their highest level of education (GCSE-level or lower, A-level or similar and undergraduate degree or higher). X-axis in the top-right quadrant shows a respondent’s recalled vote and their self-identified gender. The X-axis in the bottom-left quadrant shows a respondent’s recalled vote and their age group (18-34, 35-54, 55-75). The X-axis in the bottom-right quadrant shows a respondent’s recalled vote and whether they had children under the age of 18 living in their household. DNV = did not vote, N/A = respondent did not provide an answer or did not know.