Conditionality, activation and benefit sanctions

Entitlement to unemployment benefits has never been fully unconditional; initial access, for example, has always been regulated by rules that require claimants to be available for work (Adler, Reference Adler2016). In recent decades, however, a process of ‘benefit activation’ (Clasen and Clegg, Reference Clasen, Clegg, Clasen and Clegg2011: 9) has seen the re-configuration of international social security systems so as to increasingly demand the fulfilment of work-related behavioural requirements as a condition of ongoing benefit receipt. Consequently, claimants are now expected to meet various availability and suitable work criteria, as well as job-search requirements and participation in training programmes or even unpaid work placements (Immervoll and Knotz, Reference Immervoll and Knotz2018). These conditions are enforced through monitoring and the threat and imposition of financial penalties, known as benefit sanctions. Importantly, the particular work-related conditions, as well as the magnitude and length of sanctions, vary across countries, including the claimant groups to which they apply. In the UK, for example, conditions and sanctions now affect benefits designed to support unemployed individuals, lone parents, long-term sick and disabled people as well as people in low-paid employment (Dwyer and Wright, Reference Dwyer and Wright2014).

The shift towards work-related behavioural conditionality is an important constitutive element of a broader ‘activation turn’ (Bonoli, Reference Bonoli2010: 435) in international social security and labour market policy. Distinctive from earlier forms of active labour market policies (ALMPs), initiatives since the 1990s have combined work incentivisation and job-search assistance through a variety of ‘punitive and enabling mechanisms’ (Raffass, Reference Raffass2017: 350). With regards work incentives, activation has involved the development of minimum wages and in-work benefits alongside cuts in the generosity of benefits and sanctions (Immervoll and Scarpetta, Reference Immervoll and Scarpetta2012). In terms of employment assistance, furthermore, activation differs according to ‘human capital development’ and ‘work first’ approaches (Lindsay et al., Reference Lindsay, McQuaid and Dutton2007). Human capital development emphasises investment in skills, education and training, whereas work first approaches focus on job search and more basic skills training in order to get unemployed individuals into work as quickly as possible. Prior to 2010, the UK’s work first approach to activation combined both disciplinary and assistive measures, in the form of sanctions and a variety of employment-related support initiatives (Lindsay, Reference Lindsay, Pascual and Magnusson2007). These supportive aspects were then diminished, however, as a result of the Coalition government’s (2010–15) subsequent ‘punitive turn’ (Fletcher and Wright, Reference Fletcher and Wright2018: 324).

A variety of demanding social security reforms were implemented by the Coalition government; this article focuses specifically on sanctions imposed on unemployed individuals claiming Jobseeker’s Allowance (JSA), which saw two important developments. First, the frequency of sanctions was consistently high compared with previous rates observed since the introduction of JSA in 1996 (NAO, 2016a). Between 2010 and 2015, nearly a quarter (24%) of JSA claimants received at least one sanction; monthly sanctions rates varied dramatically, more than doubling from approximately 3% in May 2010 to a peak of over 7% in October 2013, before gradually returning to their pre-Coalition level by late 2015. Second, the severity of sanctions was increased by the Welfare Reform Act 2012. Following its implementation in October 2012, the minimum sanction period increased from one to four weeks and the maximum from 26 to 156 weeks, depending on the type and number of sanctionable actions incurred (DWP, 2013). Such sanctions represent full benefit withdrawal for the period in question, from an already low benefit level. Sanctioned claimants can apply for hardship payments amounting to 60% of JSA or 80% for those deemed ‘vulnerable’, which begin in the third week after a sanction for the former group and immediately for the latter group. Before the October 2012 changes, less than 10% of sanctions resulted in a hardship award, a figure that rose to over 40% by the end of 2014 (Webster, Reference Webster2015).

Sanctions policy generally, including these developments specifically, has generated significant debate regarding impacts on claimants, and it is to this literature that this article now turns.

The impacts of benefit sanctions

Work-related conditionality and benefit sanctions are explicitly intended to improve labour market outcomes for claimants, which has motivated an extensive literature on labour market effects (Griggs and Evans, 2010). Evidence from across different social security systems arguably provides some support to the claim that sanctions improve rates of employment re-entry in the short-term. In the longer-term, however, the same literature suggests that there are negative impacts on wages and job stability as well as no perceivable employment effect (Arni et al., Reference Arni, Lalive and Van Ours2013; van den Berg and Vikström, Reference van den Berg and Vikström2014; van den Berg et al., Reference van den Berg, Uhlendorff and Wolff2017; Taulbut et al., Reference Taulbut, Mackay and McCartney2018). The most methodologically robust UK study is based on long-term unemployed Work Programme participants (NAO, 2016b). It finds that sanctions increase the likelihood of employment up to a year after being sanctioned, though the impact on earnings is negative. Indeed, sanctions are also found to increase exits from JSA without employment, a result that is supported by additional UK analysis conducted at the local authority-level (Loopstra et al., Reference Loopstra, Reeves, McKee and Stuckler2015). Importantly, the results of the NAO (2016b) study differ by the claimant group in question; employment re-entry effects are in fact negative for those claiming Employment and Support Allowance (ESA) due to sickness or disability.

In light of the mixed evidence on labour market impacts, a growing area of research identifies a wider range of possible, consistently negative, sanction-related impacts. The risk that sanctions will lead to financial hardship for claimants is confirmed by several policy reviews carried out by separate UK government departments (Vincent, Reference Vincent1998; Saunders et al., Reference Saunders, Stone and Candy2001; Peters and Joyce, Reference Peters and Joyce2006; Dorsett et al., Reference Dorsett, Rolfe and George2011). In the largest of these studies, Peters and Joyce (Reference Peters and Joyce2006) found that over two-thirds of sanctioned claimants experienced financial hardship, whilst many were forced to borrow money from friends and family in order to survive. These individuals had difficulty paying utility bills, rent and managing debt; many had already been struggling to get by financially on JSA, a situation that was aggravated by the imposition of a sanction. These findings are supported by the academic literature, where sanctions have been linked to rises in the number of people being fed by food banks in both qualitative and quantitative research (Lambie-Mumford, Reference Lambie-Mumford2014; Loopstra et al., Reference Loopstra, Fledderjohann, Reeves and Stuckler2018). Sanctions have also been associated with negative impacts on third parties, including the friends, family and children of claimants (Watts et al., Reference Watts, Fitzpatrick, Bramley and Watkins2014).

It is increasingly recognised, furthermore, that sanctions are likely to affect the mental health and wellbeing of claimants. Evidence with regards the impact of JSA sanctions specifically is, however, scarce. In particular, Stewart and Wright (Reference Stewart and Wright2018) conduct longitudinal qualitative interviews with JSA claimants and find that sanctions are commonly associated with impacts such as stress, anxiety and depression, caused both by the fear of and actual imposition of benefit sanctions. Additional qualitative research identifies negative psychological impacts of sanctions on groups such as lone parents, disabled people and homeless people (Dwyer et al., Reference Dwyer, Jones, McNeill, Scullion and Stewart2018; Johnsen et al., Reference Johnsen, Watts and Fitzpatrick2018; Johnsen and Blenkinsopp, Reference Johnsen and Blenkinsopp2018). Quantitative research in this area has focused more broadly on the impact of work-related behavioural conditionality. A natural experiment study by Katikireddi and colleagues (Reference Katikireddi, Molaodi, Gibson, Dundas and Craig2018), for example, finds that conditionality negatively impacts the mental health of lone parents in the UK, which is supported by evidence from the US (Davis, Reference Davis2019). The research carried out here, in contrast, contributes to the emerging literature through consideration of the quantitative relationship between sanctions and mental health, focusing in particular on JSA sanctions and rates of antidepressant prescribing.

The determinants of antidepressant prescribing

Antidepressant medication is prescribed to treat individuals suffering from anxiety and depression, and therefore represents a salient – albeit imperfect – means of investigating these particular aspects of mental health. Not all individuals suffering from anxiety and depression will be prescribed antidepressant medication, since there are differences in the likelihood of recognising and reporting mental health problems, as well as in GP prescribing behaviour and the exploration of alternative treatments (Hyde et al., Reference Hyde, Evans, Sharp, Croudace, Harrison, Lewis and Araya2005). A high correlation, nevertheless, exists between the two (Barr et al., Reference Barr, Taylor-Robinson, Stuckler, Loopstra, Reeves and Whitehead2016). Existing quantitative research on the determinants of antidepressant prescribing in the UK considers the factors that explain variations in prescribing rates at the GP practice-level, and takes into account a combination of the characteristics of registered patients, the GP practice itself as well as area-level determinants. Spence and colleagues (Reference Spence, Roberts, Ariti and Bardsley2014), for example, find that antidepressant prescribing is higher in GP practices that have patients with higher proportions of older people, women and white people. Morrison and colleagues (Reference Morrison, Anderson, Sutton, Munoz-Arroyo, McDonald, Maxwell and Wilson2009), furthermore, find that greater proportions of GPs who are female, young or born in the UK are also associated with higher levels of antidepressant prescribing.

The quantitative analysis in this study is carried out solely at the local authority-level, which – due to data availability – is a feature of the UK quantitative literature on the impacts of sanctions more generally. Area-level factors associated with higher antidepressant prescribing at the GP-level include higher levels of deprivation as well as greater levels of urbanisation (Morrison et al., Reference Morrison, Anderson, Sutton, Munoz-Arroyo, McDonald, Maxwell and Wilson2009; Sreeharan et al., Reference Sreeharan, Madden, Lee, Millett and Majeed2013). Additional area-level research emphasises the influence of economic factors, with higher levels of antidepressant prescribing being associated with higher unemployment rates and lower rates of economic output (Barr et al., Reference Barr, Taylor-Robinson, Stuckler, Loopstra, Reeves and Whitehead2016). The study by Barr and colleagues (Reference Barr, Taylor-Robinson, Stuckler, Loopstra, Reeves and Whitehead2016) also highlights the role of the social security system more broadly, finding that rates of antidepressant prescribing are higher in local authorities with a greater cumulative proportion of Work Capability Assessments (WCAs) for claimants of the main out-of-work disability benefit. Sanctions represent an important additional determinant to the factors already discussed, not least as a result of the serious material implications that they hold; in terms of JSA withdrawal absent of a hardship payment, a four-week sanction amounts to the loss of over £230 for somebody aged 18–24 and over £290 for somebody aged 25 and over (NAO, 2016a).

Ecological analyses are affected by well-known limitations, whereby correlations that hold at the area-level do not necessarily apply at the individual-level, and vice versa. Using aggregate-level data is nevertheless informative given current data constraints, as it permits investigation of the relationship between sanctions and antidepressants as indicated by variations in local authority-level rates through time. In addition, this level of analysis is able to capture potential mental health impacts of sanctions on claimants themselves as well as any possible third-party impacts on friends and family. As previously indicated, this article focuses on the period of Coalition government (2010–15) which imposed a comparatively high number of benefit sanctions and increased their severity through the Welfare Reform Act 2012. These developments in sanctions policy inform the following research questions:

1. 1. Are benefit sanctions associated with higher rates of antidepressant prescribing at the local authority-level?

2. 2. Does the observed relationship strengthen following the implementation of the Welfare Reform Act 2012?

Analytic sample

NHS Digital publishes monthly antidepressant prescribing data for all practices in England, beginning in June 2010; this study carries out a quarterly analysis and thus begins at the third quarter of 2010, coinciding with the early months of the Coalition government and the initial rise in rates of JSA sanctions. February 2015 marked the start of the national expansion of Universal Credit (UC), the new benefit that replaces six existing means-tested benefits, including JSA (DWP, 2015). The rollout of UC systematically altered the composition of the remaining JSA claimant group by initially only being open to younger unemployed individuals without dependents (DWP, 2014). Due to data availability, however, UC sanctions could not be included in the analysis; in order to minimise the influence of compositional change, therefore, data are included up to and including the fourth quarter of 2014, prior to the national rollout. In the remaining pre-2015 sample, 31 local authorities were affected by the Pathfinder phase of UC that began in April 2013, totalling 78 local authority quarters. These local authority quarters are removed in the analysis presented here, though the results remain substantively unchanged with or without their inclusion; this similarity is unsurprising given the small number of individuals claiming UC by December 2014 (DWP, 2018). There are 326 local authority areas in England, though the City of London and the Isles of Scilly are excluded from the analysis due to their small population size.

Antidepressant prescribing data

Data on the number of antidepressant items prescribed by GP practices in England are accessed from NHS Digital (2018), measured using ‘Selective Serotonin Re-Uptake Inhibitors’ (SSRIs), the first-line medication for treating anxiety and depression (NICE, 2015). SSRIs are the most appropriate indicator to capture impacts on anxiety and depression since the broader total antidepressant measure includes items prescribed to treat non-psychiatric health conditions such as chronic pain (Spence et al., Reference Spence, Roberts, Ariti and Bardsley2014). Prescription items are single supplies of a medicine that generally refer to month-long prescriptions, though the length of prescription items will vary depending on the length of treatment or quantity of medicine prescribed (HSCIC, 2015). Quarterly rates of SSRI prescribing per 100,000 population are constructed for each local authority by aggregating monthly GP practice-level data and using mid-year population estimates that were updated in 2018 and available through Nomis (ONS, 2018). A key limitation is that the prescribing data do not contain any patient-related information, meaning that it is not possible to construct prescribing rates per working age population, which is the group who are at risk of sanctioning if claiming JSA. Consequently, all variables included in the analysis are expressed as quarterly rates per 100,000 total population. This is unlikely to unduly influence the results obtained, however, which control for age and estimate the effect of sanctions by exploiting variations in local authority-level rates through time.

Sanctioning data and additional explanatory variables

Data on the number of JSA sanctions are accessed from Stat-Xplore (DWP, 2018). This database is limited in that it records only the latest decision for each sanction case, meaning that it is not possible to calculate the total number of sanctions imposed in any one quarter. Sanctions that have gone through the review, reconsideration or appeal process, for example, will be recorded at a later point in time from the original sanctioning decision, even though claimants will have had their benefits stopped throughout. Consistent with other studies in the literature, the analysis here uses original adverse sanctions as its main sanctions variable (Loopstra et al., Reference Loopstra, Fledderjohann, Reeves and Stuckler2018; Taulbut et al., Reference Taulbut, Mackay and McCartney2018). This measure benefits from its specificity to the quarter in question, though it underestimates the true sanctions figure as it does not include sanctions that are challenged, reaching up to a fifth of sanctions during the period of analysis (Kennedy and Keen, Reference Kennedy and Keen2016). One additional limitation is that it is not possible to include ESA sanctions at the local authority-level. Similar to JSA sanctions, ESA sanctions saw a rise and fall in the frequency of their application around October 2013, whilst their severity increased from December 2012 (Webster, Reference Webster2016). These variations, nevertheless, were smaller than those for JSA sanctions and occurred from a lower base-level. Indeed, Figure A1 in the online appendix contrasts the JSA and ESA sanctions rates throughout the period using original adverse sanctions in England; the contrasting trends suggests that the omission of ESA sanctions does not represent a serious source of omitted variable bias on the results of the analysis.

Additional explanatory variables are included in the analysis, informed by the previous discussion on the determinants of antidepressant prescribing and sourced from Nomis, Stat-Xplore and additional UK government departments (see online appendix Table A1). These include data on the monthly number of JSA claimants, averaged over the quarter to provide an estimate of the quarterly claimant count, as well as GVA per head – a local authority-level equivalent of GDP – to capture the influence of economic trends. Unemployment measured according to the ILO definition is also investigated, as well as economic inactivity, using Annual Population Survey (APS) data for 12-month periods beginning every quarter. In addition to GVA per head, a residence-based measure of local authority-level income – GDHI per head – is investigated, which did not alter the substantive results presented here. Demographic characteristics such as annual proportions of separate age groups, gender and ethnicity are included, as well as quarterly rates of WCAs, the Index of Multiple Deprivation and rural-urban classification. Finally, quarterly rates of antibiotic prescribing are included as a proxy for the propensity of GPs to prescribe in general, reflective of discretionary prescribing behaviour (Spence et al., Reference Spence, Roberts, Ariti and Bardsley2014).

Statistical analysis

To investigate the relationship between sanctions and SSRI prescribing, the analysis estimates fixed effects models with a basic form described in Equation 1:

(1) $$SSR{I_{i,t}} = {\beta _0} + {\beta _1}Sanction{s_{i,t}} + {\beta _2}{X_{i,t}} + {\mu _i} + {\lambda _t} + {\varepsilon _{i,t}}$$

In Equation 1, i denotes the local authority and t denotes the quarter. SSRI is the SSRI prescribing rate per 100,000 population, Sanctions is the JSA sanctions rate per 100,000 population and X represents a vector of additional control variables. μ denotes local authority fixed effects, λ denotes time fixed effects and ϵ represents the error term. The inclusion of local authority fixed effects controls for time-invariant unobserved differences between areas, meaning that the analysis estimates the average association between sanctions and antidepressants within local authorities. The inclusion of time fixed effects controls for the influence of factors that are constant across local authorities but that vary over time, without the need to impose a functional form on the relationship between SSRI prescribing and time. All estimated models use Driscoll-Kraay standard errors, which are robust to heteroscedasticity, correlation through time within local authorities and general forms of cross-sectional dependence (Hoechle, Reference Hoechle2007).

As part of the initial regression modelling process, additional time-variant control variables were first included (see online appendix Table A2, Model A1). As a sensitivity check, this model was repeated with the inclusion of separate time trends by quintile of deprivation and rural-urban classification, which allows the trend in antidepressants to vary by level of baseline deprivation and rurality (Model A2). The coefficients for these interactions estimate how the effect of deprivation and rurality change over the period, with their main baseline effects absorbed into the local authority fixed effects as in Equation 1 (Allison, Reference Allison2009). Given the significance of these interactions and the increase in within-R ², this model was ultimately favoured and its results will be discussed in full in the next section. Diagnostic tests for this model are detailed in the online appendix. As a final check, the analysis was repeated through estimation of a random effects model, which adjusts for time-invariant factors by making the relatively stricter assumption that any omitted variables are uncorrelated with the included explanatory variables (Model A3). A Hausman test of this assumption indicates that a random effects framework should not be favoured here (p < 0.001).

The next stage of the analysis considers the second research question outlined above, and investigates the impact of the Welfare Reform Act 2012. To capture the effect of this reform, which increased the average length of sanctions, Equation 2 modifies the initial fixed effects analysis through inclusion of an interaction term with Sanctions and Reform:

(2) $$\eqalign{SSR{I_{i,t}} & = {\beta _0} + {\beta _1}Sanction{s_{i,t}} + {\beta _2}(Sanction{s_{i,t}}*Refor{m_t}) \cr & \quad + {\beta _3}{X_{i,t}} + {\mu _i} + {\lambda _t} + {\varepsilon _{i,t}}}$$

Reform is a dummy variable that marks the quarters before and after the implementation of the harsher sanctions regime brought about by the Welfare Reform Act 2012; it is coded 1 for quarters Q4 2012 onwards and 0 before that date.