Introduction
‘Activation’ has become a policy paradigm extending far beyond labour market or ‘welfare to work’ policies. Where not only people but also the entire systems of social protection are being ‘activated’ (Barbier, Reference Barbier, Lind, Knudsen and Joergensen2004: 236), activation may cast its spell on health care systems (Sundmacher, Reference Sundmacher2006), on pensioners (Casey, Reference Casey2004), and in particular, on people hitherto considered as ‘incapacitated’ or ‘disabled’ (Carcillo and Grubb, Reference Carcillo and Grubb2006). A main goal of Jobcentre Plus in the United Kingdom was to get people on incapacity benefits back to work through work-focused interviews (Finn et al., Reference Finn, Knuth, Schweer and Somerville2005; Konle-Seidl and Lang, Reference Konle-Seidl and Lang2006; Stafford et al., Reference Stafford2007; Adam et al., Reference Adam, Bozio, Emmerson, Greenberg and Knight2008). In Norway, the ongoing merger of the National Employment Directorate and the National Social Insurance Directorate – plus compulsory cooperation and co-location of this new NAV bureaucracy with municipal social assistance offices – is aimed primarily at persuading recipients of disability benefits to go back to work (Overbye, Reference Overbye2007). Switzerland is reinforcing both preventive and rehabilitation schemes for (potential) recipients of disability insurance (Bonvin and Rosenstein, Reference Bonvin and Rosenstein2008). In Sweden, where people may remain – still formally employed – on sick leave for long periods, reactivation of this category of ‘inactive’ people has become an issue (Hetzler, Reference Hetzler2008).
With regard to the ascription of a social and benefit status to inactive people, the German situation is unique. Gatekeeping before disability pensions (Erwerbsminderungsrente) is very strict both in terms of medical definition and of the assessment process, the latter being entirely in the hands of institutional doctors, social courts, and medical experts commissioned by those courts. Consequently, percentages of the German population of working age receiving incapacity-, disability-, or sickness-related benefits are low by international comparison (Grubb and Miyamoto, Reference Grubb and Miyamoto2003). In contrast, it almost goes without saying that unemployment – and long-term unemployment, in particular – is high in Germany. There is ample evidence of a trade-off between the status ascriptions of ‘unemployed’ and ‘incapacitated/long-term ill’. The role of disability insurance benefits as an absorber of labour market shocks has long been described by many authors and for several countries (Parsons, Reference Parsons1980; Gruber, Kubik, Reference Gruber and Kubik1997; Autor and Duggan, Reference Autor and Duggan2003; Campolieti, Reference Campolieti2004; Beatty and Fothergill, Reference Beatty and Fothergill2005; Becker, Reference Becker2000). ‘Activation’ in one status category may actually lead to the crowding out of recipients into another category (Clasen et al., Reference Clasen, Davidson, Ganßmann and Mauer2006). Conversely, an analysis of European Social Survey data has demonstrated that individuals who are comparable in terms of a number of demographic characteristics and in their subjective health reporting have a much higher propensity to describe their status as ‘incapacitated’ in a number of countries compared with Germany (Erlinghagen and Knuth, Reference Erlinghagen and Knuth2008).
Thus, we have good reason to assume that the kind of persons targeted for ‘re-activation’ from a sickness- or disability-related scheme in a number of other European countries tend to be in a regime of social protection related to worklessness in Germany. This particular allocation of social risks may give rise to ambivalent appraisals: if one believes that relevant proportions of people with problems that, at least once, were severe enough to justify a health- or disability-related benefit can be reactivated for employment in other countries, then this should be all the easier to accomplish in a setting like the German one where most such people were never released from ‘adult worker’ obligations. However, if one tends to believe that poor physical or mental health will increasingly lead to exclusion from an employment environment that is becoming ever more demanding and stressful, then the inclusion of high proportions of customers with impaired health in an activation-oriented regime of labour market and welfare policy would seem a heavy deadweight, which might finally shipwreck the whole activation mission and which would distort a nation's self-perception with regard to its unemployment problem.
The recently created German benefit and activation regime of ‘basic income support for job seekers’ (Grundsicherung für Arbeitsuchende, vulgo: ‘Hartz IV'), with a benefit called Unemployment Benefit II (Arbeitslosengeld II) provides an ideal test case for these alternative hypotheses. Concentrating the workless population of working age not covered by contribution-based unemployment benefits – that is, the long-term unemployed, those only marginally and intermittently employed and thus unable to earn an entitlement, and spouses considered ‘inactive’ before the reform – the new regime should be replete with ‘bad risks’ with regard to health.
The paper is organised as follows: the following section very briefly explains the policy context of the new German regime and data used here. The next section describes and summarises the health condition of the relevant population and the activation received by ‘customers’ in different health conditions. The penultimate section explores the effects of both activation and health condition on two different employment outcomes. The final section summarises the findings and discusses their policy and research implications.
Policy and research context
Basic income support for job seekers: the new German benefit regime
As of 2005, Germany merged two benefits for workless people devoid of unemployment insurance entitlements into a new flat-rate benefit financed mainly from the federal budget. Federal unemployment assistance (Arbeitslosenhilfe) for those who had exhausted their eligibility for contribution-based unemployment benefit (roughly 2 million recipients at the end of 2004) and municipal social assistance (Sozialhilfe) for people of working age and considered capable of working (roughly 1.6 million) were replaced by the new benefit Unemployment Benefit II (Arbeitlosengeld II). The justification for this far-reaching and – for many concerned – ‘retrenching’ benefit reform was that only by merging the benefits, also employment and social services previously delivered by Agencies for Work and municipalities, respectively, could be merged. The aim was to form unified ‘job centres’ that would be better prepared than their forerunners to activate job seekers and to address their manifold restraints against reemployment in a holistic manner (see Knuth, Reference Knuth2007 for details).
‘Capability to work’ (Erwerbsfähigkeit) defines the watershed between the new benefit on the one side and disability pensions or – in the absence of a pension entitlement – residual social assistance on the other. This is institutionally defined as ‘being capable of working for at least three hours per day under normal conditions of the general labour market’ or being inhibited from doing so by health problems merely for a foreseeable period. Thus, both the temporarily ill and those currently not available for work because of caring responsibilities are considered ‘able to work’. It is obvious that under such a broad definition of working capability, many people with a poor health condition will be included in the new benefit regime called ‘for job seekers’, although individuals may be exempt from job search requirements for the time being.Footnote 1 Therefore, the question on the meaning and effects of ‘activation’ arises.
Data
The data used for the following analysis originate from a computer-aided telephone ‘customer survey’ of 25,000 respondents receiving UB II, part of which was organised as a panel. Sampling was restricted to 154 (of roughly 440) regional units in Germany, and was stratified in several dimensions to capture sufficient numbers of individuals in defined target groups. The analysis presented here is based on the panel, that is those 11,108 respondents who had been sampled from the caseload as it existed between 19 September and 18 October 2006, and who were actually interviewed twice. The first wave of interviews was conducted between January and April 2007, the second wave between November 2007 and March 2008. As far as possible, individual interviews were sequenced in such a way as to have roughly equal time spans between the first and the second wave. Percentages calculated in this paper have been adjusted for sample stratification and for differing response rates in different strata of the sample. However, it should be noted that, strictly speaking, these findings are representative only for the 154 regional units and not representative for the Federal Republic as a whole.Footnote 2 Descriptive values must therefore be taken ‘with a pinch of salt’, whereas the multivariate models should not be affected by this regional bias in the sample design.
The survey contains several indicators of self-reported working capability and health, descriptions of the kind of activation respondents experienced with job centres and subsequent employment outcomes between the two waves of the panel, which were conducted roughly 12 months apart. Because part of the benefitFootnote 3 continues to be paid ‘in work’ to ‘working poor’ unable to support their families, taking up work while continuing to draw benefits provides an additional indicator of integration into employment not yielding a living wage.Footnote 4
Health and activation
Health condition of the target population
The customer survey confirms the expectation that health would be a relevant problem among the target population. About 11.4 per cent of male respondents and 6.8 per cent of female respondents report an officially recognised disability, and more than half (6.5 per cent of male respondents and 3.9 per cent of female respondents) say that they fulfil the legal requirement of ‘severely disabled’ (schwerbehindert). Around one-third of those with a disability say that they would be very strongly restricted with regard to work.
Approximately 60 per cent (59.4 per cent males and 61.2 per cent females) describe their current health condition as ‘very good’ or ‘good’. Comparable values for the population at large (69.8 per cent for males and 68.6 per cent for females) can be derived from the EU SILC database (Statistisches Bundesamt, 2006). This confirms our expectation that people with weaker health would be concentrated among the recipients of UB II.
Asked how many hours of daily work they would be able to sustain for a longer period, 3.8 per cent of male respondents and 4.1 per cent of female respondents estimated their daily ability to work at less than three hours. If these self-assessments were to be confirmed by medical examination, these respondents would have to be considered as wrongly allocated to the benefit in question, given that the capability of working at least three hours per day is a constituent definition of the benefit. Another 8.1 per cent of male respondents but 19.3 per cent of female respondents said that they would be able to perform daily work of between three and less than six hours. These recipients also do not possess ‘full’, or ‘normal’, working ability. However, because gender differences on the other health indicators are much smaller, the high share of female respondents reporting restricted working capability should in part be attributed to a mingling of capability with restricted availability due to caring responsibilities and role expectations. This must be seen against the background of an institutionally still strong German male breadwinner model (Gustafsson et al., Reference Gustafsson, Wetzels, Vlasblom and Dex1996) and lack of childcare facilities for children under three in West Germany (Statistische Ämter des Bundes und der Länder, 2007). Additional descriptive indicators of health condition not discussed in the body of this paper are to be found in the appendix.
For the purposes of multivariate analyses, the two indicators – subjective health condition and daily working capability – will be combined into one index of health-related capacity (see Table 1). The combination of temporally unrestricted working capability (eight hours or more per day) and at least ‘good’ health is recoded as (1) very good health-related capacity (47.1 per cent of the respondents), whereas a health condition of only ‘satisfactory’ or a slightly restricted working capability from six to less than eight hours makes up index (2) good health-related capacity (28.8 per cent of the respondents). A health condition less than satisfactory combined with a daily working capability from three to less than six hours makes up for (4) poor health-related capacity (5.7 per cent), and the same categories of health condition combined with a daily working capability of less than three hours is defined as (5) very poor (3.1 per cent). The 15.1 per cent diagonally situated ‘in-betweens’ are lumped together as (3) fair health-related capacity.
Table 1 Index of health-related capacity (percentages of responses)
Source: UB II customer panel (stock sample only), own calculations.
The index of health-related capacity is consistent with other response items related to health: on average, disabled respondents score significantly lower on this index than respondents who did not report a disability. Among the disabled respondents, work restrictions become significant as the index of health-related capacity increases in value. The same holds true for the incidence of certain medical diagnoses as well as for the number of diagnoses mentioned. It appears, then, that the index of health-related capacity adequately summarises the various health aspects covered in the customer survey. Within the population of respondents, those aged 50 and older score less favourably on the index of health-related capacity than the average (see appendix).
Activation and health
Generally speaking, more than two years after their creation, the newly established job centres still fell far short of their official mission of comprehensive activation. Only slightly more than two-thirds (69.8 per cent) of the respondents had had at least one interview with their personal adviser during the six months prior to the survey. Only less than a half (47.7 per cent) had a currently valid personal action plan (Eingliederungsvereinbarung) and only slightly more than one quarter (27.9 per cent) had ever received an offer for a job or (in the case of young people) of an apprenticeship since entering the system or since being referred to it from the two preceding benefit systems as of 1 January 2005.
Large groups are practically exempt from activation, which explains the above findings to some degree. Mothers with children younger than three years do not have to be available for work because, as previously mentioned, childcare facilities for children of this age are almost nonexistent in the western part of Germany. Recipients aged 58 years or more, at the time of the survey, could still opt out of job search obligations (a provision which has since been abolished). Young people of working age but still attending school are not suitable targets for activation. Although the ‘working poor’ drawing in-work benefits (see p. 2 and footnote 4 for details) are in theory called to reduce their dependency by trying to earn more, it can be assumed that they are often left alone because they are regarded as ‘integrated’ into the labour market. These caveats notwithstanding, even among those recipients of the benefit officially registered as unemployed and thus considered available for employment (which includes those working up to 15 hours per week but not those working more), only 70.4 per cent had attended a job centre interview during the last six months, and only 50.5 per cent had a valid personal action plan. In other words, even among those whose need for activation is beyond doubt, considerable proportions are being neglected.
Against this background, it seems relevant how activation is related to the index of health-related capacity introduced above. Are those with health problems activated more because they need more support? Or are they activated less due to ‘creaming’ decisions in an environment with caseloads that are still much too high for personal advisers?
Whereas the incidence of interviews varies little with health, personal action plans as well as job offers are fewer in the groups with weaker health (Table 2). On the one hand, this seems logical and reflects the emphasis on work in the new regime of basic income support for job seekers. On the other hand, the fundamental justification for the reform that led to the merger of benefits and services from the two preceding benefit systems was to create more comprehensive services, including psychosocial and other concomitant services. Even where work is not an immediate option, a personal action plan might include steps toward improving a person's health status, or there could be job offers adjusted to the individual's health condition. However, low intervention rates were found with regard to social or psychological problems (ZEW et al., 2008). Furthermore, health has not been explicitly addressed in the reform discourse. Medical rehabilitation in cases of officially recognised disabilities has suffered from the institutional split between unemployment insurance and basic income support for job seekers (Dornette et al., Reference Dornette, Rauch, Schubert, Behrens, Höhne and Zimmermann2008). Concepts and measures suited to address the often multi-morbid or unspecific syndromes of psychosomatic ill-being among long-term unemployed are evolving, but slowly and sporadically (Büttner et al., Reference Büttner, Knuth, Neuffer, Neukirch, Schweer and Stegmann2007). This explains why the health score is inversely correlated with the activation score. It is worth noting that those with reduced health-related capacity are not simply left alone, which is reflected in the almost even distribution of job centre interviews.
Table 2 Activation and health-related capacity
Source: UB II customer panel (stock sample only), own calculations.
Health, activation, and employment outcome
Does activation matter?
The impact of different aspects of activation (three single items and one composite indicator) has been estimated for two different outcomes: (1) employment take-up and (2) quitting the benefit in conjunction with employment take-up. As previously explained, the outcomes labelled ‘quitting the benefit in conjunction with employment take-up’ are a subset of the total of employment take-ups.Footnote 5
In all the eight resulting probit models (Table 3), health-related capacity, gender and the condition of the regional labour market are used as control variables. Activation indicators do show effects on both types of employment outcome at least on a 5 per cent level of statistical significance (with the exception of quitting the benefit after only one item of activation, which is still significant at the 10 per cent level). Effects increase with the number of activation items that apply. Effects on mere employment take-up are mostly larger than that on quitting the benefit in conjunction with employment take-up, which is a more demanding criterion of success than just finding any, possibly low-paid, job.
Table 3 Effects of activation on employment outcomes
Notes:Also controlled for age and belonging to one or more target groups of labor market policy (parents with small children, lone parents, disabled people and migrant background). Coefficients, in parentheses standard errors. ±, **, *: significant on 1.5 and 10 per cent level, respectively.
Source: UB II customer panel (stock sample only), own calculations.
In all the eight models, the index of health-related capacity works in the expected direction and in a consistent pattern. ‘Poor’ and ‘very poor’ health results in negative coefficients, where ‘fair’ health is taken as the reference category. Except for ‘poor’ health, which does not seem to discriminate enough from ‘fair’, all health-related coefficients are highly significant. The effects of ‘very good’ and ‘good’ health are stronger with regard to the more demanding outcome of quitting the benefit than with regard to simply taking up any employment.
Female gender – compared with male gender – has a slightly negative but highly significant effect in these models in which health is being controlled for. The effect of regional labour market conditions is weak compared with many of the other variables, and notably ‘above average’ labour market conditions are not significant for three of the four models for the more demanding outcome of quitting the benefit. As was often found, ‘objective’ labour market indicators explain little when individual and treatment characteristics are taken into account (Büttner et al., Reference Büttner, Knuth, Muth, Beer and Schumann2001; Büttner et al., Reference Büttner, Knuth, Neuffer, Neukirch, Schweer and Stegmann2007).
How much does health matter?
So far, it has been explored whether activation has an impact on employment and benefit recipiency outcomes when health is controlled for. Although both variables showed the expected effects, these models do not sufficiently take into consideration that people with a weaker health-related capacity are activated less intensively (see Table 2) and perhaps in different ways. How do health and activation interact with regard to employment outcomes? Does activation work differently for groups with different health-related capacities?
To explore this question, in Table 4 another four probit models were estimated for the two employment outcomes known from Table 3, separated for two health categories: superior (‘very good’ + ‘good’) and inferior (‘fair’ + ‘poor’ + ‘very poor’).Footnote 6 Activation is represented here only by the number of applicable items, and the control variables are the same as in Table 3 – except, of course, for health, which has become constitutive of the models.
Table 4 Health-differentiated effects of activation on employment outcomes
Notes: Also controlled for age and belonging to one or more target group of labor market policy (parents with small children, lone parents, disabled people and migrant background). Coefficients, in parentheses standard errors. ±, **, *: significant on 1.5 and 10 per cent level, respectively.
Source: UB II customer panel (stock sample only), own calculations.
As Table 4 shows, activation does work for both subpopulations in the expected direction and in a consistent pattern. However, those with inferior health need more activation (at least two applicable items) to experience a statistically significant improvement in the more demanding employment outcome indicator; that is, quitting the benefit. Only where all three items of activation are applicable do differences between the coefficients for the two subpopulations become negligible. The significance of gender becomes weaker than in the previous models and is lacking for persons of inferior health with regard to taking up ‘just any’ employment. In general, the significance of regional labour market conditions becomes even weaker than in the previous models.
To sum up, activation does work even for people with reduced health-related capacities, but at a higher rate before effects become visible. At a sufficient level, the differential effect of activation as compared with no activation becomes equivalent for the two subgroups with ‘superior’ and ‘inferior’ health-related capacity. This does not alter the fact that those of inferior health are less likely to take up employment, be it with or without activation. Because activation of people with impaired health requires higher intensity to be effective at all while still producing lower rates of desired outcomes, it is more costly, but these expenses are not wasted.
Summary and discussion
Using a very recent two-wave panel database of roughly 11,000 recipients of the new German Unemployment Benefit II, the expectation inferred from institutional considerations, namely to find high proportions of people with an impaired health condition, could be confirmed. Self-reported daily working capability of approximately 4 per cent of the benefit population runs below the legal threshold that defines the benefit. Even allowing for misreporting by some female respondents, there is another group of at least 10 per cent of the recipient population, whose daily work capability is critical (at least three but less than six hours), adding up to something like 15 per cent who would not be able to sustain normal working days because of their health condition.
The level of activation produced by the new German benefit regime still falls short of its pretence with regard to all its so-called customers. Most notably, more than two-thirds of the respondents said they never received a job or apprenticeship offer. Two of three activation indicators used are positively correlated with health – activation is directed more at those whose health condition seems to promise positive results.
Controlling for health, gender and regional labour market conditions as well as age and belonging to defined policy target groups, positive effects of activation on taking up employment can be demonstrated. However, these effects are weaker for the more ambitiously defined outcome of quitting the benefit in conjunction with taking up employment than for taking up just any job. As would be expected, health-related capacity has significant and consistent effects on employment outcomes in the expected direction. It can be shown that activation has a positive employment effect even on the slight majority of the respondents who do not enjoy a ‘good’ or ‘very good’ health condition. However, to have this effect on people with inferior health, a higher intensity of activation is needed.
Referring to the two policy alternatives hinted in the introduction, it can be said that, generally speaking, it actually does make sense to keep unemployed and needy people with superior and inferior health within one uniform regime of employment-oriented activation rather than defining them as ‘out of the labour market’ and assigning them to a benefit without work requirement. Having said this, it still remains to be questioned whether smaller parts of the recipients are perhaps hopelessly misallocated in this benefit regime – namely those reporting daily working capabilities below or only slightly above the legal definition of eligibility for UB II. More differentiated analyses focusing on activation and employment outcomes for this particular group would have to assess the degree to which these respondents are simply misjudging or misreporting their general health condition, comparing it with the extent of institutional misallocation.
A policy implication of our findings would be that job centres need more adequate possibilities for intervention and support with regard to their customers’ health condition. As far as manifest diagnoses and chronic conditions are concerned, cooperation with the health insurance funds and among all the organisations responsible for rehabilitation – a very complex institutional mix in Germany – is urgently needed. With regard to less specific conditions of ill-being, integrating health aspects and health-related behaviour into approaches of employment-related coaching and empowerment seems promising. Such projects now only occasionally commissioned by the job centres to providers should be expanded.
Appendix: Health indicators for recipients of unemployment benefit II
