Benefits conditionality for disabled people : Stylized facts from a review of international evidence and practice

While behavioural conditionality for disability benefit claimants has been increasing, there is little evidence as to its implementation or impacts. This paper therefore summarises existing studies, alongside an international review based on 140 documents and 38 expert interviews, into four ‘stylized facts’: (i) requirements for disability benefit claimants are common, but sanctioning is rare; (ii) assessment and support are critical for implementing conditionality for disabled people; (iii) the limited existing evidence suggests that sanctioning may have zero or even negative impacts on workrelated outcomes for disabled people; and (iv) there is little evidence on wider outcomes. It concludes by stressing the need for further research.


Introduction
Over the past 25 years, there has been an increasing turn towards behavioural conditionality in the beneits system (Clasen and Clegg, 2011: 338), in which claimants are required to carry out work-related activities under the threat of cuts to their beneits ('sanctions').While this initially focused on unemployed people and single parents, there has been a more recent move to extend conditionality to disability beneit claimants, not just in Anglo-Saxon countries, but also in Scandinavia and beyond.This has been actively encouraged by the inluential international think tank the Organisation for Economic Co-operation and Development (OECD), which has argued that 'helping jobseekers with mental health conditions should be achieved Benefits conditionality for disabled people 109 How common is conditionality for disabled people?

Disability benefits
Where conditionality for disabled people presently exists, there is always one group of disabled people who are required to take steps towards work, alongside another group of disabled people who are exempt.Those subject to conditionality are either separated into a particular category within the disability beneit (the UK), separated into a sickness beneit (the Scandinavian countries) or put on a wider beneit that also includes people without disabilities (Australia).Whichever structure is used, the onpaper requirements placed on disabled people are often based on the 'rehabilitationbefore-beneit principle' (OECD, 2010: 107).For example: • In Australia, all but the most severely disabled people are required to show 'active participation in a program of support' for 18 months before being eligible for the disability beneit (Department of Social Services, 2016).During this time, they claim Newstart Allowance alongside single parents and unemployed people, with conditionality requirements being lessened for 19% and temporarily suspended for 17% of claimants at any one time (Australian Bureau of Statistics, 2016).Young (under-40) disability beneit claimants are also required to attend workfocused interviews.• In Norway, when people move on to the long-term sick leave beneit after one year, they are required to participate in work-related activity based around an individual action plan (OECD, 2013a: 161).This is similar to the system in Sweden, where many people on sick leave are required to attend a face-to-face meeting ('Sassam') for which non-participation can result in sanctions (Engström et al, forthcoming).• In Denmark, people are referred for a multidisciplinary rehabilitation team assessment after an initial period of sickness absence.The assessment is not meant to award people with a disability pension until all eforts at rehabilitation have been exhausted; instead, the assessment sends them on a rehabilitation programme that they are required to participate in (see later).While individuals can refuse medical treatments, they are required to accept non-medical health-related treatments (Beskaeftigelsesministeriet, 2016).
An alternative to the rehabilitation-before-beneit principle can be found in the UK, where disabled people are sorted into those subject to conditionality (known as the 'Work-Related Activity Group' [WRAG]) on the basis of a functional capacity test.I go into more detail about these systems later in the section on implementation.
While requirements are common, actual sanctioning of disabled people seems rare.For example, Germany has a conditional unemployment beneit with high levels of sanctioning: over a million sanctions were applied in 2015 (Bundesagentur für Arbeit, 2016).However, only 8,700 sanctions were for failures to report for the medical or psychological service.Examining this is harder elsewhere as most countries do not publish data on sanctions for disabled claimants.Nevertheless, there are anecdotal reports from the Netherlands (NVVG, 2010), Germany (Aurich-Beerheide and Brussig, 2017), Denmark (Mehlsen et al, 2015), Norway and Sweden (Hultqvist and Nørup, 2017) that sanctions are uncommon for disabled people, for reasons we explore Ben Baumberg Geiger 110 more fully later.Claimants may still be worried about the possibility of sanctions (eg in Norway, see Breimo, 2016: 71), but the sanctions seem to be rarely carried through.
The exceptions to this are Australia and the UK, both of which publish unusually transparent data.In Australia, nearly 150,000 sanctions were applied in the irst quarter of 2016 to claimants who are likely to have a disability (claimants in jobactive stream C or with Disability Employment Services), about one quarter of all sanctions (Department of Employment, 2016).(Figures on the sanction rate are unfortunately not available.)Theseare predominantly temporary suspensions until the claimant starts complying, but a substantial minority are longer-lasting.Moreover, over 10,000 job-seekers in this period were sanctioned after giving a medical reason for nonattendance that was not accepted as reasonable.
The sanctioning of disabled people is also prevalent in the UK.At its peak in August 2014, 125,000 disability beneit (WRAG -see earlier) sanction referrals were being made each year, with almost 50,000 initial sanctions (where beneits are stopped pending potential challenges), and resulting in 35,000 conirmed sanctions (for sources and notes, see Web Appendix 1).Sanctions during 2009-10 were primarily for not attending interviews, whereas sanctions during 2013-15 were primarily for 'not carrying out activity', which is primarily failing to take part in the contractor-led 'Work Programme'.While sanctioning is still rare among those subject to conditionality (3% of WRAG claimants were sanctioned in 2014/15; see Webster, 2016b), it is clearly happening at a non-negligible scale, with over 400,000 conirmed sanctions being applied to disability beneit claimants since conditionality was introduced in 2008.

Non-disability benefits
Some disabled people are also subject to conditionality within non-disability beneits after being assessed as fully capable of work, especially as many countries have tightened eligibility for disability beneits (Clasen and Clegg, 2011: 338).Conditionality on these beneits is sometimes more demanding (and incurs greater penalties) than on disability beneits.However, there are often few data available, and no comparable data using consistent disability deinitions across countries.Nevertheless, there are data showing that, for example, 25% of disabled people in the US claim a non-disability beneit such as food stamps (the Supplementary Nutrition Assistance Program [SNAP]), Temporary Assistance for Needy Families (TANF) or unemployment insurance (Houtenville and Brucker, 2014).
The only country for which more detailed data are available is the UK, which publishes transparent igures on sanctioning among unemployment beneit claimants who self-declare a disability -though it is important to note the limitations of this self-declared disability data.The igures show that throughout 2000-15, 20-30% of unemployment beneit sanctions have been applied to disabled people.In total, across disability and non-disability beneits, 170,000 sanction referrals are made about disabled people in the UK every year (down from a peak of 620,000, as shown in Figure 1), of which 90,000 (250,000 at the peak) were subsequently conirmed as a sanction.While failure to attend an adviser interview was the main reason for sanctions in the early part of the period, failure to participate in work-related activity (the aforementioned introduction of the contractor-led 'Work Programme') Benefits conditionality for disabled people 111 and particularly work-related failures (such as not looking for work) became more important on the unemployment beneit from 2011.
Not only are such sanctions not uncommon in the UK, but it even appears that disabled unemployment beneit claimants are more likely to be sanctioned than nondisabled claimants.New estimates presented for the irst time in this article (based on oicial statistics, and detailed in Web Appendix 1) show that the sanctioning rate for disabled people on unemployment beneit is higher than for non-disabled peoplemuch higher in 2010 (disabled people being 50-53% more likely to be sanctioned), and still noticeably higher for the latest available statistics in 2014 (26-28% more likely).This is consistent with other estimates that, for example, homeless unemployment beneit claimants are much more likely to be sanctioned than others (Batty et al, 2015: 10).
It is not possible to look at the share of disabled unemployment beneit claimants in the UK who are sanctioned each month.However, we know that the proportion of all unemployment beneit claimants who were sanctioned over a 12-month period was 12.9% in 2014/15, down from a peak of 18.4% in 2013/14, and compared to 10-12% during 2007-10 (Webster, 2016b)).Over a longer period of six years, 24% of unemployment beneit claimants are sanctioned (21% after challenges) -and noticeably more if we focus on those claiming for more than a year (Webster, 2016a).While sanctions are not the majority experience for unemployment beneit claimants in the UK, neither are they a marginal experience.Combined with the sanctions on the disability beneit (see earlier), more than 2.7 million sanction referrals have been made for disabled people in the UK since 2008, including 350,000 initial sanctions that were later overturned and 1.2 million conirmed sanctions.Delivered by Ingenta to: Guest User IP : 129.12.60.151On: Tue, 18 Jul 2017 10:31:13 Copyright The Policy Press Ben Baumberg Geiger

112
How is conditionality for disabled people implemented?
The international literature has often distinguished two ways in which activation can be implemented: one that invests heavily in human capital; and one that relies on conditionality/sanctioning (Dingeldey, 2007).However, this may be misleading because (as Dingeldey argues) states can become both more enabling and move towards workfare -for example, there could be increased investment in employment support alongside increased levels of sanctioning.Moreover, when we focus on disability, the critical issue is investment in assessment (in order to assess what people are capable of) and employment support (in order to have suitable activities to require them to do), rather than human capital in general.
In this section, I therefore consider a typology of how conditionality is implemented in diferent countries, which combines two dimensions: (1) the level of conditionality itself; and (2) the extent to which conditionality is linked to rehabilitation (see Figure 2).

Ideal type 1: demanding systems (high conditionality, strong rehabilitation link)
The irst type of conditionality for disabled beneit claimants is a 'demanding' system: it can be broadly characterised as providing intensive assessment and rehabilitation to disabled beneit claimants, which they are then obliged to take up.Closest to this type is Denmark, where claimants are only eligible for a disability pension if rehabilitation has previously been attempted (or would not be expected to have an efect), and almost no under-40s are initially eligible (Hultqvist and Nørup, 2017).Instead, the assessment usually decides on appropriate, mandatory rehabilitation activities.Sanctioning practices seem to vary by municipality; some municipalities do not even threaten sanctions as they feel that this is counterproductive, whereas other municipalities use threats for those who are not motivated to participate (Mehlsen et al, 2015).Even in the latter case, however, sanctions themselves are rare.
Investment in expert assessment is central to the Danish process.Caseworkers are irst obliged to complete a preparatory rehabilitation plan with the claimant, based on a one-hour interview (OECD, 2013b: 102).This is then fed into a multidisciplinary rehabilitation team, which, in practice, includes four to six people representing multiple sectors, such as employment, health, social services and education (Deloitte, 2015).The other crucial element is employment support, and the multidisciplinary rehabilitation team have three options open to them.First, in the common situation that an individual's work capacity is unclear, they can be sent on a work trial/work

Conditionality STRONGLY linked to rehabilitation
Supportive systems Demanding systems test ('arbejdsprøvet/arbejdsprøvning'), after which the clariied understanding of the individual's work capacity is fed back into a future meeting.Second, individuals can be sent on Resource Activation ('Ressourceforløb'), a rehabilitation programme lasting for one to ive years.Third, individuals with some work capacity but who would be unable to get a job in the open labour market can be put on the 'Flex-Job' scheme.This provides considerable subsidies to employers to employ disabled people with as little as 1 hour a week work capacity.Here, activation is closely linked to conditionality (Hultqvist and Nørup, 2017); indeed, even before the latest round of reforms, jobcentre executives in 2008 agreed that one of their most important jobs was 'to use activation as a work-test' (Bredgaard and Larsen, 2008).
Other systems that are closest to the 'demanding' type include the WeCARE programme for welfare (TANF) recipients in New York City, where, unlike in the majority of US states, disabled claimants are required to participate in TANF.Again, there is investment in assessment (a bio-psychosocial assessment with a certiied physician), and employment support (condition management supported by a case manager; see Collins, 2015).It seems that conditionality is not implemented harshly -people can reschedule appointments easily -although the provider also meets their contractual obligations to closely monitor the time that claimants spend on workrelated activity.Another example comes from the Netherlands, where claimants can again be sanctioned if they refuse to participate.However, oicial guidelines make clear that four steps are necessary before the conclusion that participation behaviour is inadequate, with each step encouraging the participant to resume participation (NVVG, 2010).It is only when all of these steps have failed that sanctions can be applied, and, as a result, both claimants and insurance physicians report that sanctions are rare (NVVG, 2010).
It is diicult to know if the WeCARE, Dutch or Danish systems are successful.There seems to be no current public criticism of WeCARE, which is striking given that the programme was subject to a successful class action disability discrimination lawsuit in the 2000s (for requiring people to go to oices far away from their homes).There has been public criticism of the Danish system, but not because of sanctioning (although see Hultqvist and Nørup, 2017); instead, public concerns have focused on the appropriateness of forcing people with very low work capacity to go through a practical work evaluation or Resource Activation -in one widely reported case, a capacity to work of half an hour, twice a week (Lauth, 2015).
The efectiveness of these systems is therefore uncertain, and, furthermore, it does not seem that sanctioning is widespread in any of these systems (see earlier).They are therefore not as conditional as the compliance-based systems discussed later, despite being classiied as 'high conditionality' here.Still, from an implementation perspective, the Danish system and other examples such as New York and the Netherlands exemplify a form of conditionality that seems to require people to do activities that they are capable of by investing in assessment and employment support.

Ideal type 2: passive systems (low conditionality, weak rehabilitation link)
In a 'passive' system of disability conditionality, there are formal conditionality requirements for disability beneit claimants, but this is only weakly linked to the support and opportunities available for them to move towards work.It therefore becomes more diicult to require disabled people to undertaken a given activity -Ben Baumberg Geiger 114 partly because it is less clear to front-line staf if the person in question is capable of the activity, and partly because suitable activities to refer them to may not exist.Frontline staf may respond by simply not implementing conditionality in the way that formal policies propose, as seems to occur to some extent in Norway and Germany.
On paper, the Norwegian Work Assessment Allowance (WAA) is one of the most activating disability beneits worldwide.As the OECD (2013a: 161) have approvingly noted, WAA is a conditional disability beneit based around a mandatory activity plan, and there is suicient investment in rehabilitation services that front-line staf can refer WAA claimants to suitable work-related activities.(There are only a few exceptional circumstances where people can claim the beneit without having a plan, primarily during waiting periods; see NAV, 2015).Many of the letters from the social insurance agency repeat that complying with the activity plan is a condition of receiving the beneit, and it does appear that disability beneit claimants do feel that they need to comply with their individual action (Breimo, 2016: 71).After a maximum of four years, claimants are either found capable of work, found eligible for the permanent disability pension (with no activation requirements) or their work capacity is not considered to be clariied, in which case, they often move on to social assistance.
In practice, however, even the OECD (2013a: 154) concede that 'implementation of the procedure has been a challenge'.The assessment of work capacity relies on front-line workers who are often unsure of a person's work capacity (Gjersøe, 2016).Moreover, even where front-line workers are convinced that a person is unemployable, they do not want to refer someone to a disability pension if this might be denied by the back oice.This is partly because they do not want an unemployable claimant to be in a inancially precarious position, and partly because denied disability pension applications must leave WAA in order to reapply, which only serves to generate more work for front-line workers (Gjersøe, 2016: 138).As a result, the activation of disabled people can become a way to let claimants 'loat' (as one advisor put it in Gjersøe, 2016), and is used to build evidence of unemployability for a disability pension application, rather than helping claimants move towards employment.Unsurprisingly, a government-appointed committee has recently argued that assessment and activation need to be reformed (Vågeng Expert Committee, cited by Gjersøe, 2015: 12).
Similar issues can be seen in Germany, particularly in the article in the present special issue (Aurich-Beerheide and Brussig, 2017).While another oft-cited example of rehabilitation-before-beneits (OECD, 2010: 108), in practice, both conditionality and rehabilitation are limited.Partly, this is due to problems with the assessment of work capacity, which is done by a pension fund doctor who crudely reports the number of hours per day that a person is capable of working.This gives relatively little guidance about activation to front-line workers, who themselves have limited expertise in health-related issues, have limited time to build up a rapport with the claimant and struggle to efectively use the specialist medical services available.Moreover, it seems that people with health problems are generally protected by job centres, who have considerable discretion and whose incentives motivate them to only use sanctions where they feel that this will get people into work -something that they seem to feel often does not apply to those with more limited working capacity.Finally, the rehabilitation that is supposed to occur before disability pensions is a casualty of complex inter-organisational governance issues and incentive structures (see also Rauch and Dornette, 2010).As Aurich-Beerheide and Brussig conclude, disabled people fall into either incapacity pensions with little work focus or support, Benefits conditionality for disabled people 115 or the public employment service, which has a work focus but concentrates this on non-disabled people who are felt to be closer to the labour market.

Ideal type 3: compliance-based systems (high conditionality, weak rehabilitation link)
The previous section suggested that a lack of investment in assessment and rehabilitation makes it diicult to implement conditionality in practice.However, in 'compliance-based' systems, not only are requirements placed on disabled claimants despite problems with assessment or rehabilitation, but non-negligible numbers of claimants are sanctioned due to their failures to comply with these formal requirements.This comes at a cost: it seems less likely to help disability beneit claimants towards employment (as we discuss later), and may place requirements on claimants that some are simply unable to meet.
The UK seems to exemplify this.The decision about whether claimants should be subject to conditionality at all (whether on unemployment beneits or the 'WRAG' group of disability beneit claimants) 1 is governed by a functional capacity assessment that has been beset by problems, and that simply does not directly assess people's capacity to undertake work-related activity (Baumberg et al, 2015).Once claimants are allocated to one of these groups, decisions about conditionality depend on front-line staf (and government 'decision-makers' who conirm sanctioning decisions).The UK is not unique in depending on front-line staf, but this has caused particular problems: • Front-line staf are often low-skilled and have little health-speciic training, particularly for the unemployment benefit.While more skilled disability employment advisors do exist, they are inconsistently available and inconsistently used.• Front-line staf do not always have suicient discretion.Until March 2015, unemployment beneit claimants were not allowed to have more than two periods of two weeks' sickness without becoming ineligible for the beneit, and front-line staf did not realise that they could tailor conditionality for those with health conditions/impairments (Work and Pensions Committee, 2014: 37).For disability beneit and unemployment beneit claimants on the contractor-run 'Work Programme', contractors were required to refer claimants for possible sanction even if the contractor believed that there was good reason for non-participation.• Front-line staf have sometimes been monitored against performance standards that included sanctioning rates (PCS, 2014).While not formally a 'target', this nevertheless seems to have strongly inluenced the behaviour of front-line staf to sanction claimants even if they did not judge that this would help move claimants towards work.
Alongside these challenges around assessment, there have been challenges in providing appropriate employment support for disabled people.Some intensive support exists, but most disabled claimants have been placed on the contractor-run 'Work Programme', where there have been complaints about the quality and appropriateness of support (see also McNeill et al, 2017), and where employment outcomes for disabled people have been consistently low (see, eg, Public Accounts Committee, 2014).

116
While sanctioning is only applied to a minority of claimants (see earlier), the nonnegligible sanctioning rate has resulted in a number of cases where conditionality seems to have been inappropriately applied to disabled people, particularly for those on the unemployment beneit.For example, a recent qualitative academic research project interviewed 56 disabled unemployment and disability beneit claimants, of which 21 reported being sanctioned: Those who had been sanctioned reported being unable to attend appointments due to ill health, or being unaware of appointments.The latter could be due to poor communication (including letters going missing in the post) or not fully understanding the behavioural requirements attached to beneit receipt.Accordingly, they were unanimous in stating that their loss of beneit was inappropriate.(Dwyer et al, 2016: 8) The case studies in this special issue -taken from the same project -provide more detailed illustrations (McNeill et al, 2017).Among many other similar pieces of evidence, a convenience survey of 500 WRAG claimants found that many felt that the work-related activity was not tailored to their capacities (Hale, 2014), and a parliamentary select committee found that it was not uncommon for those with mental health conditions and learning diiculties to be set inappropriate requirements that resulted in sanctions (Work and Pensions Committee, 2015).While these case studies and convenience samples are not necessarily representative, and many only consider the claimant perspective, such practices would not be surprising given the design of the system, and we have already seen earlier that the sanctioning rate for disabled people on the unemployment beneit is noticeably higher than for non-disabled people.These concerns do not seem to exist on this scale in countries closer to the 'demanding' ideal type of disability conditionality, where sanctioning seems to be an emergency backstop for wholescale refusal to engage with rehabilitation (following expert assessment), rather than a non-negligible response by less well-trained staf to minor non-compliance.
In Australia, there seems to be greater level of investment in Disability Employment Services for those disabled claimants subject to conditionality, and it is perhaps not as clear an example of a compliance-based system as the UK.Nevertheless, in a recent consultation, there was 'widespread, almost universal concern' about the beneit eligibility assessment (Department of Social Services, 2015: 3).Many stakeholders reported that assessments were done on a tick-box basis, sometimes not face-to-face, nor by appropriately expert assessors, and based on a 'confusing' categorisation of claimants based on the hours a week that they are seen to be capable of -with the result that claimants were being referred to 'inappropriate services' (Department of Social Services, 2015: 4).
It is unclear whether this currently leads to inappropriate sanctioning in Australia.Historically, there have been complaints that this sanctioning has not been consistently fair for those with mental illness or at crisis periods (Jones et al, 2007), and the OECD have noted that some welfare-to-work providers described sanctioning practices as, at times, 'deeply lawed, punitive, damaging, and causing signiicant detriment and harm' (OECD, 2012: 157-63) -particularly for those with learning diiculties or mental health problems that are not necessarily disclosed (Disney et al, 2010).Yet, at other times, welfare-to-work providers have complained that conditionality is too easily circumvented by citing illness (Disney et al, 2010;OECD, 2012: 157-63), and it seems that the extensive use of individual exemptions may partly mitigate problems with the assessment (McClure et al, 2015: 133).Overall, it is clear that there are nonnegligible rates of sanctioning of disabled claimants, just as in the UK (see earlier), but there is too little recent research to know if implementation diiculties also exist.

Ideal type 4: supportive systems (low conditionality, strong rehabilitation link)
Within the typology of disability conditionality outlined earlier, a inal theoretical possibility is for conditionality to exist on paper, allied to a strong rehabilitation focus, but alongside low levels of conditionality in practice.Given that the focus of the international review was on countries that try to implement higher levels of conditionality (even if these are unsuccessfully implemented), there are no examples of 'supportive' systems within the countries selected here -though it is perhaps close to the autonomy-focused conditionality of the Swedish system, as described elsewhere in this special issue (Hultqvist and Nørup, 2017).

What impact does conditionality have on disabled people?
In examining the evidence base about the efect of conditionality on disabled people, there are four key questions: 'Does conditionality increase employment for nondisabled people, for which there is a relatively extensive evidence base?'; 'How far would we expect this wider evidence on conditionality to apply to disabled people?'; 'Is there any direct evidence that conditionality speciically for disabled people improves employment outcomes?';and 'What are the wider impacts of conditionality?'.

Does conditionality increase employment among non-disabled people?
Over the past decade, a number of studies have looked at the impact of sanctioning on employment outcomes among unemployment or social assistance beneit claimants.There are several methods that can be used.One is to compare the outcomes of claimants who are sanctioned compared to claimants who are not, but even when this is done well (eg Wu et al, 2014), it is diicult to be conident that we are picking up the efect of sanctions rather than diferences in the sort of people who are sanctioned.Instead, more robust studies look at the impact of the timing of sanctions on the beneit trajectories of sanctioned claimants, or look at natural experiments such as randomised policy pilots (or, failing that, quasi-experiments such as local variation in sanction rates), which have the added advantage of enabling us to look at 'threat' efects on those who are not sanctioned.None of these methods gives us absolute certainty, but these give us the strongest basis we can have for understanding the impacts of sanctioning (Griggs and Evans, 2010;McVicar, 2014).
This evidence shows that sanctioning increases short-term job entries.This is the conclusion of the few reviews that exist (McVicar, 2014;National Audit Oice, 2016), with one describing the evidence as 'compelling' (Griggs and Evans, 2010) -though it is worth noting that none of these are formal systematic reviews, and the individual studies difer widely in the estimated size of this efect.Moreover, there are two crucial caveats to this general conclusion: • While sanctions cause some people to leave beneits for work, they cause other people to leave beneits without having a job.This has been shown in several studies (eg Arni et al, 2013;Busk, 2016;National Audit Oice, 2016), and may be particularly strong for (though not restricted to) more advantaged claimants (Busk, 2016).• While sanctions increase employment, they cause people to be in worse (lower-paying, more unstable, more part-time) jobs.While Arni et al's (2013) study is often cited here, there are several other studies that come to similar results, for example, showing that increased days in employment translated to weak zero gains in earnings (National Audit Oice, 2016), or that sanctioning led to considerable increases in part-time rather than full-time work (Van den Berg and Vikström, 2014).While one recent article argues that these efects are simply because sanctions move people with historically lower earnings into employment (Lachowska et al, 2016), recent reviews share a consensus that short-term positive efects of sanctioning on employment need to be balanced against longer-term negative impacts on earnings (Griggs and Evans, 2010;McVicar, 2014;National Audit Oice, 2016).
While I have concentrated on the evidence on sanctioning here, the smaller evidence base on job-search monitoring shows similar (if slightly less consistent) results (McVicar, 2014).

Would we expect conditionality to increase employment for disabled people?
At its simplest, conditionality is based on three assumptions: that it will make claimants less selective about the jobs that they are willing to take, try harder to ind work (Griggs and Evans, 2010;Arni et al, 2013) and take fuller advantage of the support available to them.However, this has to be balanced against other mechanisms at play.Being less selective about jobs may make people move into work more quickly, but it may also lead them into poorer job matches, taking worse jobs that waste some of their human capital (Arni et al, 2013).There may be a 'lock-in efect' of work-related activity that diverts job-seekers from looking for suitable work (Molander and Torsvik, 2015).It may also undermine their relationship with their caseworker, which evidence suggests is one of the most important elements of employment support services (Hasluck and Green, 2007), and more generally lead to more supericial compliance with (or even disconnection from) government-provided support.The balance of these efects for non-disabled people was summarised in the previous section.The critical question here, however, is whether this balance will be the same for disabled people.It has been argued that there are many people with partial work capacity who do not voluntarily take up support when it is ofered; hence, conditionality could have positive efects for disabled people (Martin, 2015).However, other a priori considerations suggest a more negative efect.
Partly, this is because it is harder to apply conditionality fairly for disabled people -and unfair conditionality is unlikely to have positive impacts.This is even accepted by those in favour of conditionality; for example, a UK pro-conditionality think-tank conceded that 'it is undoubtedly the case that misapplied, conditionality runs the risk of worsening the position of the most vulnerable claimants' (Pickles et al, 2016: 43).Moreover, people with certain sorts of disabilities (eg cognitive impairments) may not be consistently aware of the requirements being placed on them, as has been reported in the UK (Dwyer et al, 2016) and Australia (Jones et al, 2007;Disney et al, 2010).In Benefits conditionality for disabled people 119 the light of the review of implementation, we would expect more positive impacts from 'demanding' systems (with considerable investment in assessing claimants fully, and providing rehabilitation that they are required to do) than 'compliance-based' systems (where less support is provided claimants to move towards work, and where unreasonable requirements are more likely).
More importantly, though, many welfare-to-work professionals, disability charities and disabled people themselves think that conditionality is not suited to the employment issues facing disabled people (Mehlsen et al, 2015;Work and Pensions Committee, 2016: #66).Conditionality implies that disabled people are partly out of because of a lack of motivation, which has been criticised for ignoring the real barriers that they face (eg Patrick et al, 2011;Garthwaite, 2014).However, even where claimants have some work capacity and may be able to get a job, they are often uncertain about their capacity to work and require considerable support.Claimants' relationship with their caseworker becomes critical, as does their willingness to experiment with a job in a safe context in which failure is not punished (arguably a key component of the success of the positively evaluated Individual Placement and Support model; see Nygren et al, 2016).Conditionality can undermine both the relationship with the caseworker (Work and Pensions Committee, 2016: #16) and this willingness to experiment.Rather than encouraging claimants to move towards work, it can lead to fear and anxiety (see later), prompting people to 'hunker down' (in the words of Paul Gregg, one of the architects of the UK's conditionality system 2 ).Again, using the typology from earlier in the article, this is likely to be an even greater problem in 'compliance-based' systems of disability conditionality than 'demanding' systems.

Direct evidence on conditionality for disabled people
This makes it crucial to have direct evidence on the impact of conditionality on disabled people, and while such evidence is rare, a small number of studies have recently been published. 3 Four studies look at the introduction of mandatory activities for sick and disabled people, in Sweden (Engström et al, forthcoming), Denmark (Rehwald et al, 2015), Norway (Markussen et al, 2015) and Australia (Broadway et al, 2014).The target group for the reform was either those near the start of sickness absence (all sick leave claims in Denmark, vague sickness diagnoses in Sweden), those after six months of sickness absence (Norway) or young disability beneit claimants (Australia).For Norway, Sweden and Australia, the policy being evaluated was a compulsory rehabilitationfocused meeting, while in Denmark, the policy was much more intensive, involving weekly meetings and the requirement to be participating in a return-to-work programme to continue claiming the beneit.Some discretion was usually allowed; hence, not all claimants were required to undergo conditionality. 4The studies were varied in their methodology, being based on either variation across geography and/ or time (Norway), comparing similar groups not afected by the reform (Australia), or randomised policy trials (Sweden and Denmark).
In contrast to the studies on non-disabled people mentioned earlier, some of these studies show negative impacts of conditionality on disabled people.This includes the intervention closest to the 'demanding' ideal type of conditionality, which looked at the efects of mandatory work-related activities rather than just rehabilitation-Ben Baumberg Geiger 120 focused meetings (the Danish study of Rehwald et al, 2015).Such conditionality led to sick claimants spending one week less in regular employment in the following year (5% less than the control group), with similar efects after two years (though the efect had efectively worn of the third year). 5The Swedish trial (Engström et al, forthcoming) also found that mandatory activation interviews led to higher levels of later disability pension receipt, though they only mention employment outcomes in passing.The other two studies show slightly diferent results.The introduction of mandatory interviews for young disability beneit claimants in Australia led to increased referrals to disability employment services, but had no signiicant impact on employment or earnings (Broadway et al, 2014).This leaves the Norwegian study (Markussen et al, 2015) as the single study that found positive impacts on return to work from sickness absence, as well as a small reduction in disability beneit claims in the longer term -though given that the Norwegian system has generally struggled to implement conditionality (see earlier), the extent of conditionality and sanctioning that was applied is unclear.
Beyond these studies of mandatory activities, there are also two studies that look at the impact of sanctioning on disabled people, both from the UK.Reeves's (2017) article looks at the impact on economic activity of the sanctioning of unemployment beneit claimants with a self-declared disability, based on spatial variation in sanctioning rates.He inds a link between sanctioning and greater levels of disability among the economically inactive (which is robust to multiple sensitivity tests), though with little consistent efect on levels of disability among employed people.While the design is partially limited by data availability and the possibility of confounding in nonexperimental cross-sectional data, it provides suggestive evidence that the sanctioning of disabled people on unemployment beneit may drive them into economic inactivity.
The other study is conducted by the UK National Audit Oice (2016), and uses a strong design, exploiting the random assignment of disability beneit claimants to employment service providers (with diferent sanctioning referral rates) within an area.They found negative impacts of sanctioning on employment outcomes among disability beneit claimants, which seem to increase over time.Sanctioning referral rates may relect other factors that inluence employment rates (eg quality of employment support), though the study does control for both diferences in claimant characteristics and diferences in the provider's overall performance.Still, given that the National Audit Oice study also shows positive impacts of sanctioning on employment for non-disabled people, the study strongly suggests that the employment consequences of sanctioning -at least within what I have described as the 'compliance-based' ideal type of disability conditionality -may be much worse for disabled than non-disabled people.
Overall, the limited but methodologically strong evidence that exists suggests that the employment efects of conditionality are much less positive for disabled than non-disabled people.Two studies even suggest that the stronger forms of disability conditionality are counterproductive, whether they are implemented with strong rehabilitation links (Denmark) or instead with a focus on compliance (the UK), with both studies inding a reduction in employment among disabled people.The four other studies show a mixture of negative, positive and null results; the one unambiguously positive impact (in Norway) may suggest that less onerous conditionality in a supportive context may have positive efects, or may simply relect a weaker methodological approach.

Benefits conditionality for disabled people 121
What are the wider effects of conditionality on disabled people?
The inal element to consider is whether conditionality has any wider impacts, particularly on income/destitution and health.If conditionality makes disabled people more likely to enter suitable work, then any increased employment could potentially improve both of these outcomes.However, not only are there concerns about whether conditionality drives even non-disabled people towards poorer job matches, but the evidence also ofers little reason to believe that disabled people's employment chances will be increased by conditionality -and they may even be harmed (see the previous section).Yet, even without inluencing employment, there are reasons to believe that conditionality and sanctioning may inluence poverty and ill-health.
Given that sanctioning involves withdrawing money from people who lack jobs (and also raising the numbers of people without work or beneits), there is a clear mechanism by which sanctioning afects income.This will partly be mitigated by other sources of support, but it seems unlikely that such support is universal.It is therefore unsurprising that, for example, almost one third of destitute people in the UK in Fitzpatrick et al's (2016: 29) study said that they had been sanctioned, while a UK area-based study found that the increased sanctioning of unemployment beneit claimants led to greater food bank use (Loopstra et al, 2016).For disabled people, the issues may be even more acute given the greater costs of disability, the greater challenges that many disabled people have in the labour market and the added challenges of responding to sanctions among those with learning disabilities and mental ill-health (Inclusion London, 2016).Both case studies from disabled people's organisations (eg Inclusion London, 2016) and qualitative research (Dwyer et al, 2016) in the UK have found that disabled beneit claimants who have been sanctioned struggled inancially, sometimes with very damaging consequences.While the sanctioning of disabled people is also non-negligible in Australia, no evidence on its impacts is currently available.
Conditionality may also negatively afect disabled people's health.Partly, this may stem from inancial issues following sanctions (as in one prominent UK case 6 ), but it may also come from the stress of conditionality itself -particularly in 'compliancebased' systems such as the UK, where unreasonable requirements may sometimes be set, and where minor lapses in compliance may lead to signiicant inancial penalties (see earlier).The rollout of a new disability assessment in the UK has been linked to suicides and mental ill-health (Barr et al, 2016), and while much of this efect is likely to be around inancial worries around beneit eligibility (Garthwaite, 2014), there also seems to be considerable anxiety around the conditionality regime that accompanied it.For example: • Disabled unemployment and disability beneit claimants in a large-scale piece of qualitative research regularly referred to the anxiety-inducing efects of conditionality (Dwyer et al, 2016), which can also be seen in the case studies in the present special issue taken from the same project (McNeill et al, 2017).• A recent inquiry into beneit sanctions by a UK parliamentary select committee mentioned the possibility that sanctions 'can have negative impacts on mental health', based on evidence from disabled people's organisations and disability charities on the harmful consequences experienced by their members and clients (Public Accounts Committee, 2016).
• In one non-representative survey, 85% of WRAG claimants reported feeling anxious about sanctions (Hale, 2014).• The British Psychological Society (2016) and four other associations involved in mental health treatment recently signed a statement saying that 'the sanctions process may be detrimental to people's mental health and wellbeing'.• There are even several anecdotal reports in the UK of claimants dying from suicide or heart attacks while subject to conditionality, 7 though the extent to which conditionality can be blamed for these individual cases is contested.
It is unclear how common these issues are, and robust quantitative studies of the average health efects of UK conditionality on disabled people should be undertaken.However, the volume of case-study evidence from claimants, campaigners, health professionals and qualitative researchers strongly suggests that the UK's compliancebased system of conditionality can have negative efects on health.Further research from outside the UK is needed to understand whether all forms of conditionality for disabled people potentially have similar efects.
Other efects of conditionality and sanctioning have been suggested in the literature.For example, Lohman et al (2004) show that children in sanctioned families in the US had worse developmental outcomes, and Paxson and Waldfogel (2003) ind that US states with tougher sanctions policies had greater levels of child maltreatment.However, in both cases, the methodology is relatively weak -there are other diferences between sanctioned families and sanctioning states that are diicult to account for -and this evidence does not justify strong conclusions (as accepted by Griggs and Evans, 2010: 27).Similarly, while Machin and Marie (2006) ind that areas afected by a tougher beneit regime in the UK had greater rises in crime, Fallesen et al (2014) ind that workfare in Denmark reduced crime levels.In both cases, plausible mechanisms can be outlined, but irmer conclusions await further research.

Conclusion
While bodies such as the OECD recommend introducing conditionality for disabled people, they provide little evidence as to either its implementation or its impacts.In this article, I therefore reviewed available evidence in the academic and grey literature, alongside a new international review of disability assessment in 10 countries based on over 140 documents and 38 expert interviews.The indings can be summarised as the following 'stylised facts' (an approach taken from Martin, 2015): 1.Many states formally require disabled people to participate in steps towards work.
However, sanctioning itself is rare, with most countries using sanctioning as an emergency backstop for repeated non-compliance with rehabilitation.Only in Australia and the UK is sanctioning less rare, seemingly being used as a nonnegligible response to instances of perceived non-compliance.2. There are several models of implementing conditionality for disabled people.
Demanding systems (high conditionality, strong rehabilitation links) invest heavily in assessment and support, which claimants are then required to participate in.
Passive systems (low conditionality, weak rehabilitation links) have on-paper conditionality requirements but do not sanction claimants in practice because of a lack of assessment/support.Compliance-based systems sanction around procedural Benefits conditionality for disabled people 123 requirements despite poor links to assessment and support -potentially sometimes implementing sanctions unfairly.3.In terms of employment outcomes: -The evidence, on balance, suggests that sanctioning increases job entry among non-disabled people -but often to low-quality jobs, and it also increases the numbers of people not claiming beneits who are not in work.-However, a priori, the impact of conditionality on disabled people is likely to be more negative than its impact on non-disabled people, and is likely to vary according to the way conditionality is implemented.-The limited but robust existing evidence focusing on disabled people suggests that sanctioning may have zero or even negative impacts on work-related outcomes.4. In terms of wider impacts, individual case studies in 'compliance-based' systems suggest that sanctioning in the absence of other support can lead to destitution, and that conditionality can negatively inluence mental health.
It is worth ending by emphasising that conditionality for disabled people is simply diferent to conditionality for non-disabled people -its implementation is more challenging, and its impacts may well be more negative.While the present article has reviewed the available evidence and conducted a new international review of implementation, the published evidence is relatively scarce, and multi-country reviews necessarily sacriice some depth in order to achieve international breadth.
Rather than simply assuming that the existing evidence base for unemployment and social assistance beneits will suice, it is therefore crucial to add to the limited existing evidence base on disability and conditionality in all respects: its extent, its implementation, its impacts on employment and its wider impacts.

Notes
1 .Requirements are more demanding on unemployment beneits (where claimants can be required to take particular jobs), and the sanctions more severe (on unemployment beneits, claimants lose their entire beneit for four to 52 weeks; on disability beneits, they can lose about two thirds of their beneit [soon to rise to the full beneit] for one to four weeks).Around 40% of Job Seekers Allowance (JSA) sanctionees and 10-20% of WRAG sanctionees receive hardship payments (Webster, 2015: 8-11).
2 .See: https://inequalitiesblog.wordpress.com/2012/09/27/paul-gregg-disabilityemployment-speech/ 3 .Evaluations of reforms that combine conditionality and support (such as 'Pathways to Work' in the UK) have been cited elsewhere as evidence for the efectiveness of conditionality (OECD, 2010: 108;Pickles et al, 2016: 41).However, I do not consider this evidence here as the contribution of conditionality itself is unclear (as admitted by the OECD, 2010: 108). 4. Exemptions were 13-30% in Norwegian municipalities (after six months sickness absence) and 17% in Denmark (at the start of sickness claims). 5. Rehwald et al (2015) further focus on the diferences between diferent types of mandatory activities (inding that some are harmful and others beneicial), but these results can no longer exploit the randomised nature of the intervention and are therefore subject to greater possible biases.

Figure 1 :
Figure 1: Sanctions given to disabled benefit claimants in the UK, by benefit and reason (last 12 months, rolling)

Figure
Figure 2: Typology of implementation of conditionality LOW conditionality HIGH conditionality