New institutionalism and the problem of change

According to path-dependence theory (Pierson Reference Pierson2004; Thelen Reference Thelen1999), policies develop on historically embedded paths, confined by rules in institutional structures that limit the discretion of policymakers. The resistance to change intensifies over time and, as a result, the costs of change increase as well. Especially in advanced political economies, complementarities between different institutions are commonly assumed to constitute high barriers to institutional change, thereby reinforcing path-dependent development (Hall and Soskice Reference Hall and Soskice2001). ‘Orthodox’ conceptions of path dependence rely on the model of ‘punctuated equilibriums’ and ‘critical junctures’ to grasp institutional genesis and change. The model of punctuated equilibrium takes an evolutionary approach towards institutional change ‘in which long periods of stasis are broken by short, in geologic terms, episodes of rapid speciation’ (Krasner Reference Krasner1988: 77). Institutions develop following their path-dependent logic until this logic collapses and a new trajectory emerges. At these critical junctures, comprehensive change takes place ‘that shapes politics for years to come’ (Collier and Collier Reference Collier and Collier1991: 27), as a distinct new trajectory develops that prescribes path-dependent change by disclosing certain policy options.

The ‘problem with this model [of punctuated equilibrium] is that institutions explain everything until they explain nothing’ (Thelen and Steinmo Reference Thelen and Steinmo1992: 15); developments that comply with the image of sticky institutions can be explained, but change that goes beyond this pattern cannot be captured. In this understanding, change that takes place in ‘normal’ periods of path-dependent development is downplayed in significance, neglected or even ignored. The lack of interest reflects a predominant understanding of change as ‘path stabilization: marginal adaptation to environmental changes without changing core principles’ (Ebbinghaus Reference Ebbinghaus2005: 17; see Pierson Reference Pierson2004: 133). At the other extreme, we can find ‘path cessation or switching: intervention that ends the self-reinforcement of an established institution and may give way to a new institution’ (Ebbinghaus Reference Ebbinghaus2005: 17). However, the breakdown of a long-standing institution because of some exogenous shock which is then ‘displaced’ by a new institution is a rare empirical phenomenon. Hence, the image of minor but only reproductive and adaptive change, on the one hand, and major and therefore transformative change, on the other, is highly misleading – if not damaging – as this conception of path dependence tends to obscure and downplay a common pattern of policy development in advanced political economies; that is, incrementalism (Lindblom Reference Lindblom1959; Mahoney and Thelen Reference Mahoney and Thelen2010; Streeck and Thelen Reference Streeck and Thelen2005).

Building on the idea of incremental politics, Streeck and Thelen (Reference Streeck and Thelen2005) argue that minor institutional change does not necessarily reinforce the logic of an established path and therefore results in institutional reproduction. Instead, such processes of change can also deviate from this logic and facilitate path destabilization. Many subtle incremental changes can eventually add up to major discontinuity and result in a gradual transformation of an institution and path: ‘institutional discontinuity caused by incremental, “creeping” change’ (Streeck and Thelen Reference Streeck and Thelen2005: 9). For this reason, the persistence of some formal institution should not mistakenly be interpreted as institutional continuity. This mode of incremental but paradigmatic change between ‘path stabilization’ and ‘path switching’ can be described as a ‘path departure: gradual adaptation through partial renewal of institutional arrangements and limited redirection of core principles’ (Ebbinghaus Reference Ebbinghaus2005: 17). In brief, the core of the argument is that incremental changes can have a transformative effect without major disruption. But what are, in addition to exogenous shocks (such as comprehensive changes of the socioeconomic and political environment), the sources of institutional dynamism that destabilize patterns of reproduction?

Ideas as A source of institutional change

In the literature, ascribing a more prominent role to ideas in order to address the stability bias in new institutionalism is increasingly being called for, as ‘it is difficult to conceive of political and institutional change that is not propelled by some set of ideas about political outcomes and the arrangements to reach those goals’ (Peters et al. Reference Peters, Pierre and King2005: 1296). With a similar impetus, Béland and Cox (Reference Béland and Cox2011: 11) assert that ‘new perceptions and ways of thinking’ are at the core of paradigmatic policy change (see also Béland Reference Béland2009; Blyth Reference Blyth2001; Hay Reference Hay2006; Schmidt Reference Schmidt2008). However, ideas are ambiguous in that they can reinforce path-dependent development as much as they can facilitate paradigmatic change. Ideas, informally institutionalized in discourses and collective identities (Berman Reference Berman2001: 238), can work as ‘cognitive locks’ in the sense that actors are not prepared to consider certain policies because of their preconceptions (Blyth Reference Blyth2001: 4), and hence ideas contribute further to pursuing a given path. Apart from their significance in discourses, ideas may instead prove to aggravate change when they have been embedded in the formal institutional setting, reflecting ‘a set of dominant ideas translated through legal mechanisms into formal government organizations’ (Goldstein Reference Goldstein1988: 181). In other words, crystallized in formal institutions as well as institutionalized informally in discourses and identities, ideas create policy legacies that are typically difficult to change.

On the other hand, ideas help actors to redirect policies in the event of perceived policy failure and crisis; without ideas informing agency, political action is hardly imaginable. Providing guidance in situations of both overwhelming complexity and unclear preferences, ideas as institutional blueprints ‘reduce uncertainty, give content to interests, and make institutional construction possible’ (Blyth Reference Blyth2001: 3; see also Metha Reference Metha2011). At both the normative and cognitive levels, ideas therefore instruct agency about what is ‘desirable’ and about what is ‘feasible’ in processes of policy change, so that an independent significance can be ascribed to them. However, although ideas as blueprints possess the capacity to facilitate path-deviating change, policy legacies remain powerful. Regardless of crisis and failure discrediting previous policies, the emergence of new ideas and the challenging of ideas becoming predominant cannot be reduced to the inappropriateness of previous ideas. New ideas do not enter an institutional and political vacuum (Sikkink Reference Sikkink1991: 25–6). Synthesizing ideational and institutional analysis, Vivien Schmidt (Reference Schmidt2011) highlights the importance of the institutional context for policy deliberation and the communication of ideas. Institutions are argued to both enable and constrain the discursive practice of agency. With a focus on the latter, Weir (Reference Weir1992: 189) coined the notion of ‘bounded innovation’ to underline the narrowing impact of the prevailing institutional context for the emergence of new ideas.

With their emphasis on ordered patterns and regularity, ideational and institutional approaches have similar biases towards stability (Lieberman Reference Lieberman2002: 299). With reference to the agenda-setting literature (Kingdon Reference Kingdon1984), recent contributions in the ideational literature underline the importance of entrepreneurial agency and the instrumental use of ideas in the construction of reform imperatives (Béland Reference Béland2009; Schmidt Reference Schmidt2008; Stiller Reference Stiller2010; Stryker and Wald Reference Stryker and Wald2009). Although this sheds some important light on when new ideas can be expected to develop significance, it remains unclear why actors adopt new ideas that challenge established discourses. In other words, what is triggering ideational change? The argument pursued here is that policy learning constitutes a potentially potent mechanism that prompts ideational and corresponding institutional change that can contribute to overcoming the stability bias of the path-dependence theory associated with new institutionalism. This is based on the understanding that path dependence conceptually contains a dimension of material resources and incentives that feed a self-reinforcing and reactive sequence as well as an interpretative dimension that concerns the ideational factors fuelling path-dependent processes (Hay Reference Hay2006). Thus, learning is perceived as a mechanism that might destabilize the ideational foundations of institutional settings and their reproduction, thereby opening up room for path-departing institutional change.

Towards an institutional approach to learning

Policy learning can be differentiated between ‘instrumental learning’, which is ‘about the viability of policy instruments or implementation designs’ (May Reference May1992: 331), and ‘social learning’, which deals with the ‘social construction of policy problems, the scope of the policy, or policy goals’ (May 1992: 331). Learning is initiated by policy failure, which ‘serves as a trigger for considering policy redesign and as a potential occasion for policy learning’ (May 1992: 341). However, failure is considered a necessary but not sufficient variable for policy learning, since policy failure also needs to be perceived as a problem. This will open a ‘window of opportunity’ (Kingdon Reference Kingdon1984) that can become – under conditions of uncertainty – a ‘window of innovation’ in which policymakers search for policy solutions and are prepared to invest resources in learning. In such situations, policy learning can serve as a mechanism that decreases the uncertainty associated with decision making in complex environments by creating blueprints for institutional reform. Hence, policy making cannot be reduced to power struggles but needs to be complemented with some knowledge-based dimension in order to provide comprehensive explanations for political behaviour and paradigmatic policy change, which typically involves a great degree of uncertainty (Hall Reference Hall1993; Heclo Reference Heclo1974). Weyland (Reference Weyland2008: 290) asserts that policymakers are particularly ‘receptive to external inputs and eager to learn from foreign models and experiences’, as readily available policy solutions reduce the costs of policy learning. Yet Marier (Reference Marier2009) assesses the practical utility of foreign knowledge rather critically because of implementation problems related to institutional differences between countries.

Predominant approaches to policy transfers and learning provide important insights but are of limited utility as they essentially remain instrumentalist, with learning depicted as a deliberate exercise of which agency is in full control (Dolowitz and Marsh, Reference Dolowitz and Marsh2000; Hall Reference Hall1993; Heclo Reference Heclo1974; Knoepfel and Kissling-Näff Reference Knoepfel and Kissling-Näf1998; Sabatier Reference Sabatier1988). The literature lacks a comprehensive understanding of ‘embedded’ agency; that is, agency in its institutional context. Addressing the instrumentalist bias in the literature, the concept of policy learning is substantiated with insights from new institutionalism in order to improve the understanding of contextual factors and agency. Entrepreneurial agency is considered decisive for an endogenous account of institutional change that goes beyond the orthodoxy of punctuated equilibrium models (Scheingate Reference Scheingate2003). Such an approach to policy learning allows us to conceptualize learning as a mechanism of change that can be integrated sensibly into new institutionalism and allows improving our understanding of the embeddedness of path-deviating policy learning. The point of departure is the assertion that learning is ‘institutionally nested’ (Hemerijck and van Kersbergen Reference Hemerijck and van Kersbergen1999). Given the institutional dispersion of decision-making authority, even powerful actors cannot usually impose new policies. Consequently, learning by single actors may be regarded as necessary but nonetheless insufficient to cause learning-induced policy change. For collective learning, a ‘mediating institution’ is commonly required, which Knoepfel and Kissling-Näf (Reference Knoepfel and Kissling-Näf1998) consider as the policy network, where communication between relevant actors facilitates collective learning, thereby generating ‘consensual knowledge’. For policy change to take place, however, consensual knowledge does not need to be shared by all actors in the particular policy network but by all actors needed for the introduction of policy change – underlining the significance of veto players and formal rules in decision making (Tsebelis Reference Tsebelis1995).

Two additional points for an institutional approach to learning need to be considered. First, puzzlement and subsequent policy learning is institutionally nested in so far as actors’ learning activities are contingent upon the institutional structure of the policy, which includes not only the formal institutional framework (involving, for example, competencies and sunk costs) but also ideas guiding political behaviour (Jachtenfuchs Reference Jachtenfuchs1996: 23–5). Thus, the inheritance of the past strongly influences, not to say constrains, the scope of policy learning, although social learning – in contrast to instrumental learning – possesses the capacity to broaden the potential scope of political manoeuvre that is considered appropriate. For this reason, when analysing the capacity of policy learning as a mechanism of institutional change, it is imperative that path dependence is taken into consideration. Secondly, policy learning and its efficiency in fostering institutional change are facilitated by the availability of ‘learning mechanisms’ that can be utilized by actors for acquiring policy-relevant knowledge (Popper and Lipshitz Reference Popper and Lipshitz1998: 171–2). Expert commissions are prominently discussed as an institutionalized means of knowledge generation to inform policy making (Fleckenstein Reference Fleckenstein2004; Marier Reference Marier2009). Yet, whether actors actually use such tools effectively depends on their attitude towards them – that is the ‘learning culture’ (Popper and Lipshitz Reference Popper and Lipshitz1998: 172–5).

Hence, the contention that policy learning is ‘institutionally nested’ has two main dimensions. On the one hand, the scope of policy learning is considered to be bound to the formal and informal institutional setting, underlining the path-dependent nature of institutional development. On the other hand, the capacity and effectiveness of policy learning is thought to be contingent upon an appropriate institutional setting for learning (that is, mediating institutions, learning mechanisms and learning culture). Put briefly, the institutional structure of a particular policy must be perceived as providing both constraints and opportunities for policy learning.

To summarize the argument so far, new institutionalism has been presented as a powerful approach in public policy analysis, despite problems in grasping and explaining policy change. Addressing the stability bias requires an understanding of path dependence that ‘involves elements of both continuity and (structured) change’ (Thelen Reference Thelen1999: 104). Such an understanding includes self-reinforcing and reactive sequences, which allows the grasping of disruptive changes at critical junctures as well as gradual transformations without disruption. Equally important, ideas need to be taken seriously in new institutionalist analyses of change. Dealing with the interpretative dimension of path dependence, policy learning has been discussed as a mechanism of knowledge-induced institutional change. An institutional account of learning enables the conceptual integration of this mechanism of change into new institutionalism. In the following sections, the case of German labour market policy is scrutinized with regard to the significance of learning in policy formulation and actual institutional change. This examination of labour market policy making draws on 33 semi-structured interviews with policymakers and experts who were involved in designing labour market legislation, complemented with documentary analyses of governmental and parliamentary documents as well as secondary literature.

Labour market policy making in Germany

Germany has typically been depicted in terms of stability and inertia rather than change. However, despite unfavourable institutional conditions for major reforms, the German welfare state has departed from its conservative welfare trajectory. Cautious reforms geared towards workfare activation started in the mid-1990s under the last centre-right Kohl government (1994–8) and were first heavily opposed by the SPD. The political landscape changed with the Red–Green government (1998–2005). The Social Democrats did not adhere to their social policy programme in opposition but continued on the reform trajectory of the Kohl government, which further destabilized the previous policy path and eventually heralded the departure from the conservative path of welfare in labour market policy, as both main parties had renounced their support for the old paradigm.

In the revision of their programme, SPD policymakers were driven by the perception that Germany was increasingly ‘falling behind’ in comparison to most other OECD countries, where labour market performances had improved or stabilized at high levels. In Germany, by contrast, the picture first worsened and then remained fairly unfavourable by international standards. A widespread consensus emerged that the labour market policy was failing to deal with mass unemployment. Addressing this perceived underperformance was made central to the agenda of the Red–Green government, which strongly criticized the previous centre-right government for its failure to reduce unemployment. Because of the harsh criticism it had put forward in opposition, the new Red–Green government could hardly afford to ignore the continuing poor labour market performance, making it a political problem and underlining the politics in the reform of the labour market.

Different options for the reform of labour market policy were viewed in the discursive framework of globalization and competitiveness; that was the informal institutional setting that structured policy deliberation. Although there are good reasons to question the deterministic impact of globalization on welfare state retrenchment (Brady et al. Reference Brady, Beckfield and Seeleib-Kaiser2005; Swank Reference Swank2002), policymakers increasingly linked the crisis of the welfare state and – here more specifically – the underperformance of the German labour market to the impact of economic globalization on the nation state. Since the mid-1990s, German political discourse on economic and social policy has been dominated by the debate on Germany as a business location (the so-called Standortdebatte). The centre-right government of the 1990s vehemently fostered this discourse, calling for a reduction in (non-wage) labour costs and corporate tax cuts. The opposition at that time, the SPD and the Green Party, initially publicly rejected these claims by the government, while admitting the existence of cost pressure in principle in less ‘visible’ programmatic intraparty debates. Accordingly, the Standortdebatte became the predominant discourse in economic and social policy at the end of the 1990s. It established a powerful reform imperative that increasingly put the welfare state under immense political pressure for restructuring and retrenchment (Schmidt Reference Schmidt2000: 278–81; Seeleib-Kaiser Reference Seeleib-Kaiser2001: 107–23; Swank Reference Swank2002: 182–6).

The challenges of economic globalization and the claim that international competitiveness needed to be improved can be found throughout the reasoning in the legislative proposals of the Red–Green government, especially the Hartz Legislation (2003–5) (Bundesregierung 2003: 217; Deutscher Bundestag 2001: 24, 2002: 24, 2003: 1 and 2). Apparently the premises of the Standortdebatte were publicly adopted by the Schröder government, which explicitly took the option of increasing public expenditure for measures reducing unemployment off the table, so that an extensive strategy of job-creation and training programmes was ruled out because of the costs associated with it. Instead, labour market policy was expected to contribute to the reduction of social expenditure and budget consolidation (Deutscher Bundestag 2001: 226). This particularly posed a problem for SPD policymakers, who had long called for costly job-creation and training programmes. For this reason, Social Democrats in government saw themselves confronted with some uncertainty regarding their concrete policy preferences. Labour market policy learning took place in this new paradigm of the political economy, setting the broad parameters for policy deliberation. The imperatives of improving international competitiveness and consolidating the budget constituted major constraints on activation strategies biasing presumably cost-saving workfare activation. Accordingly, the labour market legislation of the second Red–Green term involved major efforts to reduce public spending on unemployed people (Bundesregierung 2003: 5; Deutscher Bundestag 2001: 3, 2002: 4–5).

Importantly, the perceived functional infeasibility of policy solutions of ‘old’ Social Democracy created an incentive structure to engage in learning as a mechanism for reducing uncertainty. The SPD did not have readily to hand labour market policy proposals that were compatible with the assumed imperatives of international competitiveness and budget consolidation, as reflected in the rather cautious labour market reform of the first Schröder government (1998–2002) and the inconsistent development towards activation. The labour market reform of the first Schröder government (the so-called Job-AQTIV Law) was restricted to measures of positive activation, of which many were transferred from European neighbours (in particular, the UK, the Netherlands, Denmark and Austria). Workfare policies were not pursued; these were, with support from trade unions, vetoed by the traditionalists within the SPD. The modernizers in government, although having some sympathy with UK workfare measures, were not in a position to prevent the traditionalists from exercising a partisan veto. In this context, it is important to acknowledge that organized business, which has long been viewed as providing support for Bismarckian unemployment protection (Mares Reference Mares2003), pushed vehemently for retrenchment (Schmidt Reference Schmidt2000), which suggests that there was a disintegration of the previous cross-class alliance for the German model of welfare. And unsurprisingly, the corporatist support sought by the first Red–Green government through the ‘Alliance for Jobs’ failed to promote consensus in social and labour market policy. On the contrary, it revealed deep divisions between employers and trade unions, which each forcefully pursued their interests, while the Schröder government lacked the political direction that might have allowed it to provide leadership for the Alliance (Streeck Reference Streeck2003). It is worth noting here that the so-called Benchmarking Group of high-profile academics and experts representing the social partners (established for the purpose of facilitating consensus building and policy learning in the Alliance) developed consensual policy proposals, but these were rejected by organized business and labour.

The political environment for labour market policy changed dramatically with the ‘placement scandal’ in the public employment service (PES) at the beginning of 2002. It emerged that the placement statistics of the PES were manipulated. As this was in the run-up for the general election in autumn 2002, Chancellor Schröder installed an expert commission, the Hartz Commission, to demonstrate that the problem of unemployment was taken seriously by the government. Crucially, agenda setting was deliberately handed over to the Hartz Commission. In his election campaign, the chancellor committed himself to implementing fully the policy recommendations of the Hartz Commission if he was re-elected (Dyson Reference Dyson2005). After the election, the second Red–Green government (2002–5) indeed started implementing the policy recommendations of the Hartz Commission. Specifically, the Hartz Legislation deregulated the labour market, strengthened workfare, cut benefits and reorganized the PES. With the Hartz IV Law, the most comprehensive piece of this legislation, social security for the long-term unemployed was restructured by integrating the previous unemployment and social assistance schemes at the level of social assistance. This new benefit, the unemployment benefit II (UB II), is in principle administered by the PES (although with some involvement of municipalities). Tightening the criteria of reasonable work, every legal job is considered acceptable; noncompliance of the benefit claimant is severely sanctioned. Additional legislation shortened the time frame for the earnings-related unemployment benefit I (UB I) to the regular maximum duration of 12 months (with the exception of unemployed persons of at least 55 years of age, who could receive UB I for up to 18 months; previously, UB I had had a maximum duration of 32 months). Complementing the reform of the benefit system for the unemployed, local employment agencies were transformed into one-stop agencies for jobseekers. Intensified efforts for job placement by Job Centres were expected to improve reintegration into the labour market. With the Hartz Legislation, Germany developed ‘a significant distance towards an Anglo-American model of largely means-tested, flat rate jobless benefits’ (Hassel and Williamson Reference Hassel and Williamson2005: 13).

Reviewing developments since the mid-1990s, we find that Germany has departed incrementally but with accelerating speed from its previous policy path. Workfare activation started in the last centre-right government, but it could not push for more comprehensive reforms because of opposition by the SPD in the Federal Council. In government, the Social Democrats continued with the move to activation. Whereas the first Red–Green government pursued (somewhat cautiously) reforms along the idea of enabling activation, the second Red–Green government pushed workfare activation on a scale previously unseen. The ideational foundations of the conservative welfare state (especially the objective of preservation of status) have been gradually undermined and replaced by the activation paradigm as the institutional blueprint (Bleses and Seeleib-Kaiser Reference Bleses and Seeleib-Kaiser2004). The development largely complies with the notion of ‘transformation without disruption’ (to a great extent through ‘functional conversion’; Streeck and Thelen Reference Streeck and Thelen2005), although we find some deviation from this pattern of incrementalism with the integration of social and unemployment assistance schemes, which rather qualifies as ‘displacement’. Both conversion and displacement are typically associated with an environment of weak veto possibilities (Mahoney and Thelen Reference Mahoney and Thelen2010). While this applied to most gradual changes in German labour market policy (where the Federal Council did not typically have veto rights), the Hartz IV Law was subject to approval in the second chamber, where it was confronted with a fierce opposition. In the next section, the empirical focus is turned to the path-departing Hartz IV Law as the most comprehensive means of re-commodification in the Hartz Legislation to illustrate how policy learning fostered the break with the conservative welfare trajectory.

Reforming social protection for the long-term unemployed

Looking at the integration of unemployment and social assistance, we find that the Hartz IV Law did not start with the Hartz Commission. Instead, the Red–Green alliance visibly embarked on the preparation of the reform of these schemes in the first government by setting up the so-called ‘Mozart’ scheme in 2000. Allowing closer cooperation between the PES and municipalities on an experimental basis, the Ministry of Labour intended to postpone this welfare reform to the next legislative period. Although the legislation could have been decided on in the second half of the first Schröder government from a lawmaking perspective, any comprehensive reform of the benefit system was considered politically infeasible before the general election in 2002 because an institutional veto was expected from the Federal Council, where the conservative opposition held a majority. If legislation had been proposed at that stage, the integration of the two benefit systems would not have taken place. Rather, a cautious approach of improving cooperation between the employment service and municipalities and reconciling the two welfare schemes for long-term unemployed people would have been preferred by both ministerial and parliamentary experts.

Before setting up the Mozart scheme, the Ministry of Labour asked – virtually unnoticed by the public – the Bertelsmann Foundation to establish an expert forum on the reform of unemployment and social assistance schemes. Experts of all relevant organizations, including ministerial bureaucrats, were involved in this forum, which ran from autumn 1999 until April 2003. When the forum started, the programmatic development was absolutely open, although most participants were fairly sceptical about the prospect of a comprehensive reform. However, in particular, the first insights from the Mozart projects and subsequent debates resulted in growing doubt about the functional viability of the two-tier system, so that the policy option of integrating the two schemes became increasingly popular and eventually the consensus among the group. Experts and policymakers considered the previous institutional setting inadequate for the reintegration of long-term unemployed people into the labour market. That the new benefit should refer to the social rather than the unemployment assistance scheme was hardly disputed – except by the trade unions – in the light of the scarcity of resources. Obviously, the imperative of budget consolidation was accepted among the experts in the Bertelsmann Forum.

At the end of 2001 (before the placement scandal of the PES and before installing the Hartz Commission), an internal project group was established in the Ministry of Labour in order to conceptually prepare legislation reforming unemployment and social assistance. Then, parallel to the Bertelsmann Forum, the ministerial policy experts, aware of the controversial nature of such a comprehensive reform project, worked on the issue in – even intradepartmental – secrecy. They referred extensively to the deliberations in the Bertelsmann Forum in their policy formulation. Departing from the social insurance idea, the creation of a means-tested basic security for the long-term unemployed was called for. The social assistance scheme became the main point of reference with respect to generosity and eligibility criteria, as the objective of covering only basic needs was now directing policymakers. The very idea of creating such a scheme for long-term unemployed people, however, was strongly inspired by corresponding universal systems in Anglo-Saxon and Scandinavian welfare states. The UK was particularly instructive, as implied by the generosity level, the strong workfare component and the emphasis on the ‘making work pay’ philosophy in combination with the reduction in the duration of unemployment benefit. In their search for reference models for a new administrative system for unemployed people, the ministerial project group paid considerable attention to one-stop agencies in Denmark, the Netherlands and, especially, the UK with its Job Centre Plus. The ministry was successful in placing its proposed idea of creating a new benefit system run by the PES in the report of the Hartz Commission.

Against this empirical evidence, it is argued that the Bertelsmann Forum and eventually the ministerial project group were the loci of policy learning. In particular, the former pushed the debate towards integrating both systems instead of sticking with the previously virtually unchallenged approach of cooperation between the assistance schemes. The expertise of both participating and external experts was utilized in the deliberations. For this reason, the Bertelsmann Forum can be described not only as a learning mechanism (as the ministerial project group) but also as a mediating institution that allowed the building up of consensual knowledge (i.e. collective learning). As the most relevant actors and experts were involved, the consensual knowledge generated could become significant for the subsequent policy process.

This is not to argue that the Hartz Commission was insignificant for the reform. The placement scandal of the PES and the Hartz Commission opened a window of opportunity, making key Red–Green policymakers believe that a comprehensive reform of the benefit system had become politically feasible. Resistance in the Social Democratic parliamentary group collapsed. The leadership of Chancellor Schröder (see Stiller Reference Stiller2010) put modernizers in the SPD and the ministerial bureaucracy in a position of extraordinary authority in policy making, highlighting the political dynamics in the reform and policy learning process. The chancellor essentially broke the partisan veto that traditionalists had previously been able to exercise. Hence, Schröder can be depicted as an ‘institutional opportunist’ (Offe Reference Offe2001: 368) who was able to utilize politically the ad hoc opportunity for change that arose without having a grand plan, and this can plausibly explain some inconsistencies in the Hartz Legislation. Rather than ‘binding the hands of government’ (Dyson Reference Dyson2005), the Hartz Commission enhanced the political room for manoeuvre of the Chancellery and modernizers in government, for which the exact content of the Hartz proposals was secondary. Crucially, Schröder as an entrepreneurial actor who shifted the balance of power within the coalition government was imperative for effective policy learning in so far as his political leadership lowered the demands for collective learning; the need of collective learning in policy formulation within the government was effectively confined to modernizers in the Red–Green coalition. Yet, despite its importance, the placement scandal should not be misinterpreted as an exogenous shock or contingent event in terms of the punctuated equilibrium model triggering a new chain of events out of the blue (Mahoney Reference Mahoney2000), as the previous institutional setting had not only been destabilized much earlier but policy deliberation among key experts and policymakers had also started prior to the scandal. The crisis of the public employment service, opening a window of opportunity, accelerated the path departure rather than causing it.

In the Federal Council, the Schröder government was faced with much harsher resistance, drawing attention to the institutional constraints imposed by the German polity – i.e. the formal institutional setting. Although the conservative opposition in the second chamber basically agreed with this reform, no consensus between the Red–Green government and the opposition was reached on the issues of the responsibility for the new benefit system and cost sharing, as the conservatives insisted on ascribing the responsibility for UB II to the municipalities. For this reason, the Hartz IV Law as first proposed by the government was vetoed in the Federal Council; this move was widely viewed as being driven by party competition. Further to a political compromise on the administrative responsibility for the new benefit scheme, the criteria of reasonable work were retightened (as initially planned by the government but dropped because of fierce resistance by some traditionalists). Obviously, the institutional veto player, the Federal Council, can be held responsible for changes in the legislation, but the second chamber did not stop the departure from the conservative path of welfare, as much of the institutionalist literature would have expected. Instead, it can rather be argued that this veto player increased the speed towards a more liberal model of welfare.

The pace of reform did not persist in the Grand Coalition of Christian (CDU/CSU) and Social Democracy (2005–9), though the parties in government announced in their coalition agreement further reforms along the activation paradigm as well as further efforts to reduce unemployment insurance contributions to improve the competitiveness of German businesses (CDU, CSU and SPD 2005). The latter was delivered by the Grand Coalition (in the context of falling unemployment), but the reform of employment promotion remained rather modest (see BMAS 2009). In particular, the coalition could not agree on a reform of the administrative system of the Hartz IV scheme, which had become necessary with a decision by the Federal Constitutional Court. And rather violating the activation idea, the coalition parties, in response to widespread public discontent with the Hartz Legislation, improved the maximum duration of UB I from 18 to 24 months for the unemployed of 58 years and older, with which a key component of the Red–Green reforms was partly reverted. This partial policy U-turn came as some surprise in light of the continued commitment to activation in the coalition agreement.

Looking at the political dynamics of this legislative change, one finds prominent Christian Democrats who wanted to channel the discontent with the SPD and Hartz IV into support for Christian Democracy, using the framing of a ‘social justice gap’. Social Democrats responded to this campaign by conceding the need to ‘correct’ the Hartz Legislation, in order to prevent the Christian Democrats from gaining more sway in the electorally sensitive issue of social justice. Hence, this policy initiative might be best viewed as an example of party competition: it did not only prevent the coalition partners from further substantively restructuring the labour market but was conducive to welfare state ‘expansion’. Apparently, the Christian and Social Democrats were competing for the pro-welfare vote in the upcoming election, which created a reform gridlock, despite the hegemonic activation paradigm. In this light, one might be tempted to argue that the Grand Coalition saw two partisan veto players (opposing each other's reform proposals), which reduced the speed of the departure from the conservative welfare trajectory.