Not saying, not doing: Convergences, contingencies and causal mechanisms of state reform and decentralisation in Hollande’s France

Are States in contemporary Europe subject to new forms of convergence under the impact of economic crisis, enhanced European steering and international monitoring? Or is the evolution of governance (national and sub-national) driven fundamentally by diverging, mainly domestic pressures? Drawing on extensive new data, the article combines analysis of the State Modernisation and Decentralisation reform programmes of the Hollande–Ayrault administration, drawing comparisons where appropriate with the previous Sarkozy regime. The limits of President Hollande’s anti-Sarkozy method were demonstrated in the first 2 years; framing state reform and decentralisation in negative terms prevented the emergence of a coherent legitimising discourse. The empirical data is interpreted with reference to a comparative ‘States of Convergence’ framework, which is conceptualised as a heuristic device for analysing variation between places, countries and policy fields. The article concludes that the forces of hard convergence are gaining ground, as economic, epistemic and European pressures continually challenge the forces of institutional inertia.

thereby taking at face value the claim for a more joined-up style of government. Both programmes were presented as forming part of a broader effort to restore the public finances in a delicate European and world context, as linked weapons in the target to meet the zero structural deficit promise by 2017.
As our article involves only one case, interpretation of the data is facilitated by reference to an analytical framework developed as part of a broader comparative study. 1 A 'States of Convergence' typology is proposed as the starting point for understanding the interplay between convergence and divergence pressures on the one hand, and processes of national and territorial adaptation and translation on the other. The typology identifies four distinct forms of policy convergence and divergence. 'Hard' convergence is 'top-down' in nature. It operates on the basis of clear institutional and policy templates and commitment to 'download' these; it involves specific criteria, intrusive monitoring and sometimes sanctions for noncompliance. A soft convergence perspective emphasises the exchange of policy relevant information, benchmarking 'best practice' or at least cognisance of other models. In our typology, soft divergence focuses on a specific form of lessondrawing or policy learning whereby organisations and institutions construct themselves against perceived negative models. Our final position is labelled hard divergence, which is understood in terms of the persistence of deep underlying structures that pre-shape institutional pathways. It is caused by the path-dependent development of institutions (historical institutionalism, Pierson, 2000;Steinmo, 2008), economies (represented in distinct ways by the varieties of capitalism, Hall and Soskice, 2001 and worlds of welfare, Esping-Anderson, 2013 literatures) or bounded governance concepts (see below). The typology is outlined in Table 1. Table 1 summarises the theoretical framework. 2 Results are the outcomes described by the concepts of convergence and divergence. The concepts of hard and soft refer to mechanisms. Mechanisms are dynamics that can be driven or constrained by processes emerging and evolving at different dimensions of collective action: for example, structures, ideas, institutions and interaction. For instance, new ideas can support soft convergence, while economic or political constraints cause hard divergence; institutionalised modes of multilevel governance might lead to hard or soft convergence, whereas institutional dynamics inside states and regions cause persisting divergence. Variation and contingency are the norm; regularities can emerge from sustained empirical investigation and comparison. In his work on causal mechanisms, Tilly (2001) stresses the likely combination of what he labels as environmental, relational and cognitive variables, but warns that 'we cannot decide in general or in advance how the elements interact ' (p. 38). The article draws inspiration from Tilly in particular, and proposes the adaptation of the States of Convergence to the linked fields of state reform and decentralisation in Hollande's France. State modernisation and decentralisation can be analysed as a single or linked phenomenon, but the case study makes most sense if framed in a broader (environmental, cognitive and relational) and at least implicitly comparative context. This argument is developed most fully in the concluding section, but is presented briefly here.
Processes of hard and soft convergence and divergence affect the Politics of Austerity in Hollande's France in different ways depending on whether formal outputs, informal outcomes, political discourses or institutional innovations are considered. Hard convergence might describe the ambition of the European Central Bank (ECB) or the European Commission to enact zero deficit budgetary rules within the EU, for example, or to engage in enhanced monitoring of welfare or local government expenditure. The increased propensity of Eurostat to challenge national statistics is a powerful driver of convergence, as is the obligation to respect new European Commission fiscal governance norms. 3 Converging pressures do not, of course, necessarily produce converging outcomes; the endogenous capture of policy processes might occur in fields, such as local government reform, that are high on the political agenda and under close observation from the EU and other international organisations. Processes of hard divergence might prove more powerful than injunctions to conform to international standards. They are sustained by deeply ingrained state traditions, party structures or widely understood collective action repertoires that tie-in domestic institutional developments in a path-dependent manner. Mechanisms are dynamics that can be driven or constrained by processes emerging and evolving across different dimensions of collective action; the value of the States of Convergence typology is that it accommodates these tensions andat bestoffers partial causal insights.
Thus defined, the States of Convergence framework adds value as it draws on different types of explanation and levels of analysis: material macro-level variables (economic indicators, institutional pathways or legal orders); middle-level explanations of specific reforms and institutional configurations (bureau shaping strategies, interest mobilisation, institutional path dependencies); individual or group-level discursive dynamics (via forms of emulation, 'othering' and discursive legitimisation). The States of Convergence framework engages fully with questions of structure and agency. Hard convergence and hard divergence are best conceived as structural mechanisms that allow limited room for human agency. Soft convergence and soft divergence are better conceptualised as processes that actively bridge the two dimensions of structure and agency. They each involve forms of positive discursive legitimisation of choices adopted, as well as representations of negative models (Schmidt, 2011). They imply strategic choices being made to embrace the logic of convergence or, on the contrary, to assert the appropriateness of existing institutions and interactions. The ensuing discussion will demonstrate that causal mechanisms might follow these levels of analysis.
How does this framework add value to the existing literature? A generation of scholars has reinvigorated debates in the field of state reform and decentralisation in France. It lies beyond the limits of this article to undertake a longer dissertation on the nature of valid explanations, or the status of the meta-theoretical debates underpinning the work of individual scholars; this would be of limited interest to the reader. A number of approaches have the most direct relevance for understanding state reform and decentralisation under President Hollande, the object addressed in this article and special issue. Four main dimensions might be identified from the existing literature, none of which contains the valid explanation that might pretend to Covering Law status (Tilly, 2001), but all of which contribute to meaningful reflection and are encompassed in some form by the States of Convergence typology. The dimensions are time and temporal dynamics, scale and rescaling, institutions and actor configurations and policy dynamics.
What is the temporal dimension? And how is time conceptualised? Is time understood as a series of bounded episodes? Or as precise choices linked to the sequence of reform? Le Lidec (2011b), for example, emphasises the importance of sequencing, as well as the tie-in of foundational trajectories in his major contribution to French decentralisation. 4 Is time best captured as an iterative processes, as with Pasquier's (2004Pasquier's ( , 2012 concept of territorial political capacity, whereby territorial elites and social groups produce a vision of the world that allows them to structure relations among themselves and to define the 'interests' that they are pursuing collectively. Both approaches are important; they provide material or constructed explanations that are valuable in their own terms of reference. This article considers the case of state reform and decentralisation under Hollande. For the sake of analytical clarity, it adopts a narrow time-frame (the first 2 years of the Hollande presidency, allowing comparison with the Sarkozy period).
What are the important spatial dimensions, and how important are debates about re-scaling (Négrier, 2011)? There are potentially competing literatures about space and scale that are relevant to this article. Our position of hard convergence is influenced by the literature that emphasises the converging dynamics of the international political economy (Dyson, 2014), new forms of central steering (Epstein, 2006, Bezes, 2009; public management (Pollitt and Bouckaert, 2004;Eymeri-Douzans and Pierre, 2011) and Europeanisation (Graziano and Vink, 2007;Ladrech, 2010) that is developed more fully in the next section. This literature generally underplays specific places. A similarly dense literature focuses on the diverging outcomes produced by multi-level governance, new regionalism and territorial capacity building (Keating, 1998;Keating et al, 2003;Piatonni, 2010;Pasquier et al, 2013), a literature that makes most sense in our final empirical section. In the narrower field of French decentralisation, Pasquier (2012), Barone (2011) and Dupuy and Pollard (2014) have developed the core expertise in terms of regional governance; Pinson (2009), Le Galès (2001, Payre (2010) and others have demonstrated the specific logics of cities and urban spaces. Insofar as it integrates the territorial dimension, the article draws on interviews in central government, one region (Brittany) and two cities (Rennes and Brest) since 2010. 5 Institutions, instruments and actor configurations form a third dimension (especially Lascoumes and Le Galès, 2004;Bezes, 2007;Epstein, 2013). Debates about policy instruments contribute indirectly to our reflection, especially in terms of the potentially converging outputs produced by the Europeanised instruments (such as the Stability and Growth Pact (SGP)) considered below. The analysis of specific policy instruments sometimes lends support to a hard convergence approach, or at least identifies common trends through detailed investigation of specific cases. 6 Processes of soft convergence rely in part on the concept of isomorphism, drawn from early new institutionalist analysis: this concept refers to the tendency of institutions to observe each other and to copy models of successful practice (Powell and DiMaggio, 1991). Other forms of new institutionalist analysis are required to make sense of the capacity of specific institutions to achieve discursive credibility, which is important in our framework for understanding softer forms of convergence and divergence. Schmidt's (2010) discursive institutionalism provides a useful constructivist counter-weight to process tracing accounts. Discursive institutionalism invites attention to the goodness of fit between specific discursive registers and policy outcomes, of key relevance when considering the Hollande presidency. Explicit process tracing accounts, which emphasise the role of old institutions and instruments in shaping legislative and policy outcomes over the long term, are also of key contextual importance. Le Lidec's (2007) work in particular brings to the fore the role of the French second chamber, the Senate and the local government associations as powerful domestic veto players shaping the pathways of reforms. Finally, 'policy sector' is, at least, a useful metaphor to differentiate between different types of policymaking activities and the range of political and professional interests associated with them (Dupuy and Pollard, 2014).
The States of Convergence framework provides a heuristic device for analysing variation between places, countries and policy fields. It sits naturally within an approach labelled as contingent governance (developed in the final section), whereby context, conceptualised as an intervening variable, is understood as a filter of change. The States of Convergence framework (and the contingent governance approach that underpins it) adds value, insofar as it attaches importance to theorisation on the basis of a mix of material and constructed data across distinct levels of analysis. The framework of analysis is thus appropriate to the research question in hand. The States of Convergence framework here is applied to the specific cases of state modernisation and decentralisation, but is designed also to have a broader utility (insofar as it points to varying types of causal mechanism or narratives). The article now considers convergences, taking as its starting point the widely cited definition of convergence is as 'the tendency of societies to grow more alike, to develop similarities in structures, processes, and performances' (Kerr, 1983, p. 3).

Convergences
Our first strong hypothesis is that of convergence (Bennett, 1991;Pollitt, 2001;Knill, 2005;Holzinger et al, 2007). Contemporary European States are subject to powerful and potentially converging transnational economic, ideational and institutional pressures, which arguably spillover into their management of inter-governmental relations. Different empirical referents of hard convergence might encompass fiscal coordination (Dyson, 2014), technical norms (Borraz, 2007), policy conditionality (Page, 2000), policy instruments Le Galès, 2004, 2012), respect for formal democratic and market norms by the European Union (EU), the International Monetary Fund (IMF) and the World Bank, or the external imposition of tough performance indicators by these international organisations (Eymeri-Douzans and Pierre, 2011). But what do we mean by convergence? Convergence of what? Of outputs, processes or norms? There are two versions of a convergence hypothesis: hard and soft. The relationship between economic crisis and territorial governance is interpreted in this article as a contender for hard convergence (especially in the field of public finances), as is a certain form of Europeanisation. Soft convergence, on the other hand, is better captured either in terms of institutional isomorphism or looser pressures to join up state modernisation and decentralisation reforms.
If the hard convergence hypothesis is robust then broader exogenous fiscal, economic and political pressures drive domestic policy change. In extreme cases, such as that of Greece, where the Troika (the European Commission, the IMF and the ECB) has intervened directly in domestic institutional arrangements, there would appear to be a clear association between externally imposed efforts at state retrenchment and domestic policy change (Zahariadis, 2012). More routinely, convergence presupposes an increasing oversight by the EU into internal budgetary affairs, including local government and welfare expenditures. Across Europe, devolved or regional governments are facing a new reality of enhanced international economic monitoring (embodied by the ratings agencies), European budgetary supervision (the 'Six Pack', the Treaty on Stability, Growth and Governance (TSGG)) and a reinvigorated EU-level control over member-states and their sub-state governments. In Germany, for instance, demands for greater Länder autonomy from more powerful regions have run against countervailing trends of fiscal equalisation, tighter federal controls over expenditure and borrowing, central definitions of minimal standards in fields such as education and administration and a model of European integration that has challenged traditional Not saying, not doing Land competencies (Benz, 2007;Vetter, 2010). In Spain, the economic crisis has produced new budgetary and regulatory controls over the autonomous communities (Colino, 2013). 7 Even in highly decentralised Belgium, the economic crisis has empowered the Belgian central bank in its dealing with the regions and communities. 8 A hard convergence hypothesis would posit a direct causal relationship between budgetary (debt and deficit) imbalances and (ultimately successful) reforms to rein-in public expenditure. The context of the Hollande presidency since 2012 is one of growing external attention on structural imbalances, leading to critical assessments by the IMF (2013) and the European Commission (2014). The scale of the public debt challenge faced by the Valls government is immense. The overall public debt (defined in terms of the Maastricht criteria) had been 64.2 per cent at the start of the Sarkozy period in 2007, but by 2013 it was approaching 90 per cent and was predicted by the IMF (2013) to peak at 95 per cent in 2015. The current account deficit had been brought down to 4.3 per cent in 2013 (from 7.5 per cent in 2009), but was still well above the official target of 3.7 per cent. The Stability Pact target of 3 per cent has now slipped to 2015 (Council of the European Union, 2013). The public sector, broadly defined, had accounted for 52 per cent of GDP at the start of the Sarkozy period in 2007, but by 2013 it was approaching 57 per cent, just behind Denmark (Marois, 2013). This constrained financial setting fed through into new pressures on the public sector broadly conceived, including local government, welfare organisations and agencies. A critical Court of Accounts report published in early 2014 identified further retrenchment in the field of local and regional authorities as a top priority for expenditure reductions (Cour des comptes, 2014).
The European Commission finally granted the French government a delayto 2015 rather than 2013for meeting the deficit criteria of the EU Stability Pact (deficit at 3 per cent and debt at 60 per cent of GDP). The counter-part was a commitment that structural reforms be undertaken, a euphemism for cutting public expenditure by tackling pensions, reforming the labour market and overhauling public administration. The Ayrault-Hollande administration (May 2012-March 2014) gave out mixed messages in relation to structural reform. The Competitiveness Pact (CICE) of November 2012, the labour market agreement of 18 January 2013 and pension reform all suggested some recognition of the problems of economic competitiveness and public debt. 9 Evidence from the first 22 months also strongly pointed to taxation increases (rather than public expenditure cuts) as being the main policy response of the Ayrault government to the problem of public debt, a fiscal choice criticised by the European Commission (2013). It is too early to provide a response to the question of whether a harder form of convergence will affect the policy outcomes of the Socialist-led government; but the reduction of business taxes (the figure of 50 billion euros was announced in Hollande's pacte de responsabilité of January 2014), along with some tightening of control over local and regional government finances (and the much more consequential social security budget), goes in the direction of the international competitiveness agenda.
The EU is sometimes presented as the symbol of hard convergence, especially for euro-sceptics. Has Europeanisation gone too far? The analytical value of Europeanisation has been challenged in some recent articles (Graziano and Vink, 2007;Bolleyer and Radaelli, 2009;Cole and Pasquier, 2012). For Graziano and Vink (2007), for example, Europeanisation must produce policy change if it is to be validated as a concept. Policy transformation involves a change in objectives, principles, procedures and financial instruments; policy recalibration requires two of the above, while policy inertia signifies no variation. This contingent approach is similar to that adopted by Cole and Drake (2000), for whom Europeanisation is best understood as an independent variable, or a causal mechanism of policy transformation, where the direction of change and causality runs clearly from the EU (and its multiple institutions) to member-states and regions and where EU decisions bind successor governments. In 2012, François Hollande campaigned against the ratification of the EU Fiscal Compact Treaty, engineered by his predecessor Sarkozy and Chancellor Merkel, which introduces the golden rule of virtually balanced budgets. After a few months in office, Hollande announced a volte face, and announced that France would ratify the TSGG, following the Constitutional Council's decision that the new treaty could be approved by means of an organic law (rather than a constitutional amendment). The TSGG, agreed in December 2011 and signed in 2012 by 25 of the (then) 27 EU member states, strengthened the automatic penalties to be paid by states who are unable to control their debt, or to bring into line their budgets to zero deficits by 2015. Reversed Qualified Majority Voting (RQMV) will make it much more difficult for states to avoid automatic penalties should they prove unable to fulfil the strict Stability Pact criteria. 10 The TSGG came after a significant fiscal and budgetary tightening in the form of the Six Pack and the Two Pack, allowing the European Commission, through the European semester process, a much more intrusive oversight into national budgets (including commenting on national budgets before they have passed the parliamentary stage). 11 The European Semester process produces annual reports on the strengths and, more usually, structural weaknesses of all EU states (including those, such as the United Kingdom and the Czech Republic, not having signed the TSGG). The details of these reports filter down into fields such as the housing market, wage indexation, pension ages: the core of traditional economic sovereignty.
The direct consequences of the TSGG on local and regional authorities appear limited in the French case. As representatives of local government associations pointed out in interviews, local authorities are legally bound to present balanced budgets. 12 Numerous competencies dealt with by local and regional authorities are potentially concerned by stricter budgetary rules, however, especially in those areas of infrastructure and investment such as road building, urban transport or schools that require long-term capital investment. The EU dimension also plays itself out in terms of local public services. In part under pressures from EU competition policy, the French state has gradually moved away from being a competitive provider of technical assistance to communes. Local authorities have had great difficulties in complying with EU public services legislation. They have had to devise expensive means for tendering out public service delivery and complying with the rules of public procurement. The specific rules for the provision of inter-communal services, whereby cross subsidies are considered as state aids, has created great practical difficulties for the smallest communes in providing basic public services (Duperon, 2011).
Irrespective of whether direct EU influence can be identified, the external context (EU and international organisations) weighs increasingly heavily on domestic management. Controlling local government finance represents one central feature of the state's close involvement in localities (Hertzog, 2010;Le Lidec, 2011a;Steckel, 2012). In 2012, French local government represented just over 20 per cent of the overall total public expenditure and 71 per cent of public investment (Laignel and Guené, 2013). From the perspective of an economy-minded central government, local government expenditure has increased well beyond its sustainable level. There was a substantial increase in transfers to local authorities in the period from 2003 to 2011. One source calculates that State transfers to local authorities rose by 31 per cent over the period, excluding the transfers of competencies from central government that were accompanied with specific additional funds (Botella, 2012). The average rise in local government expenditure was 4.8 per cent per annum over the period, set against an inflation rate of 1.6 per cent (Botella, 2012). There has also been a significant growth of local government employment (Biland, 2012;Bouquet, 2013). According to one source, in 2012 the local government sector represented 34 per cent of public sector workers in France and 7 per cent of total employment (Nunes, 2013). The communal bloc is by far the main provider of local government employment, provoking criticisms, such as those voiced by PS deputy Dosière or by a critical Court of Accounts report in 2013, that communes continue to employ staff even when competencies have been transferred upwards to the inter-communal level (Dosière, 2012;Cour des comptes, 2013). Local and regional authorities and their associations argue, on the other hand, that central government transferred new service delivery responsibilities in the 2004 decentralisation Act without adequate financial compensation. 13 The transfer of ancillary and technical staff working in high schools from central to regional control, in particular, greatly increased the numbers of staff working in the departments and regions. 14 The key decision to limit further local financial autonomy was taken during the Sarkozy presidency, responding to the political imperative to lower business taxes. 15 In a move reminiscent of Thatcher in the United Kingdom in the 1980s, the Fillon government announced in 2010 the abolition of the local collection and setting of business rates (taxe professionnelle), and its replacement by a more centralised formula-based method of tax collection, considerably limiting the fiscal autonomy of local authorities, especially the departments and regions (Hertzog, 2010;Steckel, 2012). The effects of this radical move were immediate; from 2010 to 2012, the proportion of local government finances raised by local taxes declined from 50 per cent to 20 per cent for the departments and from 38 per cent to 10 per cent for the regions (DGCL, 2013). 16 This nationalisation of the local tax base was criticised by the Socialists while in opposition, but the Hollande-Ayrault administration did not rescind the measure. The regions now depend on central government transfers or formulae for 90 per cent of their resource. The only local tax that the 101 departmental councils can vary is that of the built land tax (taxe sur le foncier bati) one of the four historic taxes. 17 The inter-communal public corporations (EPCI) have fared relatively better, gaining a share of dynamic household taxes in the new formulae used by central government to distribute grants to local and regional authorities (Guengant, 2012). 18 Future perspectives for financing local government are much tighter than in the golden decade from 2000 to 2010. The main transferthe dotation globale de fonctionnement, DGFhas been frozen since 2011 and is scheduled to decline by 3 billion euros in 2014 and 2015 (Le Monde, 2013).
These various factors strengthen arguments based on hard convergence: at the very least, the external context (the EU and other international organisations) weighs increasingly heavily on domestic public management and choices. The direction of change is broadly similar across euro-zone countries. Convergence pressures are never simply implemented according to a trans-national template, however; they require strategic and discursive choices that lie more squarely within the realm of human agency. We now consider soft convergence and soft divergence as actor-focused processes, involving a mix of strategic choices and discursive legitimisations.

Soft convergence
Soft convergence, in our framework, involves a more or less conscious effort to import institutions, instruments or ideas that have demonstrated their worth elsewhere. The concept draws on Rose's (1991, p. 21) analysis of lesson-drawing, which he defines as 'a program for action based on a program or programs undertaken in another city, state or nation, or by the same organisation in the past'. Within the relevant literature, the effectiveness of this process is contingent on a range of factors, such as the availability of information, comparability of cases and a capacity for institutional isomorphism (Evans and Davies, 1999;Dolowitz and Marsh, 2000;Radaelli, 2005;Payre and Pollet, 2013). A soft convergence perspective emphasises the regular exchange of policy-relevant information, benchmarking 'best practice' or at least cognisance of other models. The proxies identified as potential indicators of soft convergence are new public management (NPM) and the importing of policy programmes from elsewhere.
Do governments look to lessons from abroad? There were several similarities between the state reform programmes under Sarkozy (2007)(2008)(2009)(2010)(2011)(2012) and Hollande (since 2012), the RGPP and the MAP, both claiming their inspiration from the example of Canada in the 1990s. But this example is also routinely evoked by other state reformers (for example, the British Conservatives in 2010, with the implementation of the Comprehensive Spending Review, far more radical in scope than state reform programmes under Sarkozy and Hollande). To conclude that a mainly instrumental use is made of specific foreign examples (such as Canada for state reform, or Finland for education) would appear pertinent. The broader question, beyond specific cases, is framed by Eymeri-Douzans (2013) in terms of whether France is 'within the world movement' in terms of administrative reform. Is there a French-style NPM and, if so, does this make a difference in terms of territorial governance? The rise and trans-national diffusion, first within OECD countries and then worldwide, of NPM has been remarkable since the early 1980s (Hood, 1998). The content of this neo-managerialist repertoire includes well-known features such as agencification, benchmarking, performance indicators, accreditation procedures, citizens' charters, management by objectives, programme budgeting, one-stop-shops, public-private partnerships and delegation of public missions to private entities. These flagship recipes and techniques have repeatedly been broadcasted as 'best practices' in international and European forums (Cole and Eymeri-Douzans, 2010).
Interviewees in the Finance ministry in 2010 willingly described Sarkozy's General Policy Review (Révision générale des politiques publiques -RGPP) as a 'French-style pubic management'. 19 There was greater openness, under Sarkozy, to diffusion of international models of best practice such as NPM, specifically the emphasis on outputs, attainment of results and the involvement of private sector consultants and operators. 20 There was some evidence of synoptic reform under Sarkozy, especially during the early period: the RGPP (see below), the merging of corps (Gervais, 2010), the fusion of bureaux (especially those responsible for calculating and collecting taxes, the DGFP (direction générale des finances publiques) and DGI (direction générale des impôts)), the emphasis on performance indicators, along with some mild reforms to the system of competitive examinations 21 and timid measures of performance-related pay. At the very least, there was an embrace of the language of NPM and reference to broader international trends and best practices. 22 Efforts were undertaken to reduce administrative duplication within the state (especially via the reform of the territorial state (the RéATE) and the General Policy Review (RGPP)) and to tackle problems of central-local relations (the law of 2 December 2010) (Bezes, 2010;Bezes and Le Lidec, 2010;Cole, 2011Cole, , 2012Péneau, et al., 2012). Local government, however, was not subject to the two principal NPM approaches adopted in France in recent years: the budgetary law (LOLF) in operation since 2001, or the General Policy Review (RGPP) that held sway under President Sarkozy (2007)(2008)(2009)(2010)(2011)(2012). 23 The above discussion elucidates the limits of direct learning from foreign administrations. There remains a specific intellectual puzzle involving debates around governance; how are broad transnational trends, such as NPM, domesticated or obviated in different contexts (Kuhlmann, 2010)? Discursive strategies based on soft convergence provide one response. Sarkozy was much more ready than Hollande, for example, to embrace the 'world movement' as a legitimising device to justify domestic reform. President Hollande preferred to adapt a strategy of constructed divergence to provide distance from the record of former President Sarkozy, and, indirectly, from the models embraced by his predecessor.

Divergences
Two versions of a divergence hypothesis were presented in the introductory section: the constructed (soft) and the material (hard). In our typology, 'constructed divergence' focuses on a specific type of cognitive or interactional mechanism whereby organisations and institutions construct themselves against perceived negative models. Constructed divergence represents an 'other' as an adversary, identified with a particular political ideology, policy or approach to service delivery. Hard divergence is understood in terms of the persistence of deep underlying structures that pre-shape institutional pathways. It is caused by the path-dependent development of institutions, economies, legal orders or bounded governance concepts. In the ensuing section, constructed divergence is elucidated by the discursive frames and policy choices adopted by the Hollande-Ayrault administration in the field of state reform.
Francois Hollande's 2012 election campaign was based on his strategic political positioning as being a 'normal' candidate and potential president, a style deliberately adopted to be the counterpart of the flamboyant Sarkozy. Anti-Sarkozyism was a fundamental dynamic of the early Hollande period, a negative framing exemplified, in rather different ways, in the fields of state reform and decentralisation. To understand the sense of drift that characterised the Hollande-Ayrault government (May 2012-March 2014), one might profitably borrow insights from the concept of discursive institutionalism, as developed most fully by Schmidt (2008Schmidt ( , 2010Schmidt ( , 2011. Discourse is defined by Schmidt (2011, p. 107) as 'the interactive process of conveying ideas'. In one article, Schmidt (2007) emphasised the importance for policymakers of developing a legitimising discourse, especially in simple, majoritarian democracies, such as France, that require persuasion and justification in the public sphere. 24 When Schmidt (2007) describes politicians being 'trapped in their discourse', she refers to a phenomenon close to Bourdieu's (1980) concept of hysteresis, which refers to the dependence of a social phenomenon on its past, as well as its current environment. The case of Hollande exemplifies this poor fit between an existing repertoire and a changing field.
In an earlier article, I argued that processes of state reform in France since the late 1980s might be observed through the lens of two rather different forms of justification that are labelled as public service and state productivity (Cole, 2010). The public service narrative emphasised the need to ensure an efficient state and an effective public sector in order to defend a French social model; its proponents looked to forms of contractualisation between public service providers, but did not advocate a radical overhaul of existing processes, professional public sector norms or established bureaucratic hierarchies (Gaudin, 1999;Clark, 2000;Chevallier, 2003). The state productivity narrative emphasised NPM, an emphasis on outputs, the attainment of results, the private delivery of (some) public goods, the involvement of private sector consultants and operators and a clear preference for budgetary (over general administrative) actors (Bezes, 2009;Eymeri-Douzans and Pierre, 2011;Epstein, 2013). Framing state reform in terms of public services corresponded best to the Socialist periods in office under Rocard (1988Rocard ( -1991 and Jospin (1997Jospin ( -2002, while state productivity was embraced more openly from 2002 onwards with the return of the Right, especially under Sarkozy. One of the paradoxes under Sarkozy was that the state productivity narrative had great difficulty justifying the reforms adopted, as deficits and public debt escalated out of control as a consequence of the general downturn (Raymond, 2012).
In terms of this framework, the Hollande/Ayrault government presented a clear puzzle. There was no natural ideological sympathy with the prevalent themes of the state productivity narrative as articulated by his predecessor Sarkozy. Ideologically close to social democratic incrementalism, coordination and compromise, the natural clients of the Socialist-led government were to be found among the public sector workers and officials whose support was vital for Hollande's election. The Socialist candidate mobilised the core symbols of public service: promising 60 000 new recruits in the education sector, and undertaking symbolic side-payments that reveal a natural sympathy to the interests of public servants (such as restoring sick pay from Day 1 of an illness-related work absence) (Sawicki, 2013). Hollande was elected on a programme of preserving the education, interior and justice budgets, but also on the basis of a commitment to eradicate France's structural deficit by 2017 (Clift, 2012). The field of state reform is symptomatic of the tensions between Socialist traditions of public service on the one hand and the competitiveness agenda and powerful state retrenchment incentives induced by economic austerity on the other.
The process of constructed divergence involves the identification of an adversary and a process of 'othering'. State reform under Hollande started from the premise that the previous programme under Sarkozythe Révision générale des politiques publiques (RGPP)was a failure, as, indeed, was the whole Sarkozy enterprise. Yet the Ayrault government could not afford to call into question many of the achievements of the RGPP. One of the Ayrault government's first actions was to commission a report on the RGPP by the Financial, Social Affairs and General Affairs Inspectorates, a report that highlighted a number of positive aspects: the reform of the territorial state (RéATE), the re-organisation of local courts and tribunals (carte judiciare), the creation of the Regional Health Agencies and the merging of the tax collection and assessment bureaux into a unified service (Bondaz et al, 2012). The previous government had set out to economise €15 billion with its state modernisation measures and achieved almost €12 billion economies. The number of state employees had declined by 5.4 per cent, of which 3 per cent were imputable to the RGPP. All in all, 503 measures were implemented from 2007 to 2012 with the label RGPP. But the report was also critical about the RGPP, especially in relation to the lack of consultation and the almost exclusive budgetary concerns. The report recognised that the RGPP had become associated with the process of not replacing '1 in 2' staff taking retirement, with a financial, rather than an evaluation logic. And while the spin around the RGPP had been all about joining up reform of the State and decentralisation, in practice the dual processes of state territorial reform had been largely separate. 25 The Hollande version of the state reform programme was entitled the Public Policy Modernisation Programme (Modernisation de l'Action publique -MAP). There were important symbolic (and some substantive) differences between the two processes of state reform, the RGPP under Sarkozy and the MAP under Hollande. The Hollande administration was, in some respects, more joined up with its approach to state reform than its predecessor. Unlike the RGPP, the MAP encompasses not only state ministries, but also local government, the welfare system and public agencies. The MAP involves policy evaluations, rather than fixed targets to be applied indiscriminately of sector (Doriac, 2013;. Policy audits typically concern crossministerial fields of policy (the fields of housing policy and handicapped people were selected for the first audits in December 2012) and take around 6 months. The audits associated practitioners, local authorities and public service managers, as well as state officials. Unlike the NPM approach, targets would be determined at the end of a process of negotiation, rather than determined in accordance with a budgetary template. To succeed in its broad objectives, the MAP will need to be much harsher than the RGPP. Given the political commitment to increase employment in education, the police and the courts, the government will need to cut much deeper into public service employment elsewhere. The new government demanded cuts of 2.5 per cent in staff numbers in the non-priority areas of public policy: all except Education, the Police and Justice. One source estimated that there would need to be 8000 annual job losses in the Defence ministry, 2500 in Finance ministry and 3000 in the Ecology ministry (Paquier, 2013a).
One year after its creation, however, the MAP was barely visible as a cross-cutting public policy process, and its pretension to provide a holistic approach to policymaking was met with general scepticism in interviews. 26 In relation to the territorial dimension of the state's action, the legacy 1 year was far from insignificant. The second State Modernisation Interministerial committee (CIMAP), held on 17 July 2013, announced two potentially far-reaching reforms: one granting the regional prefectures control for the first time of budgets for all field services under their control; the second defining stricter conditions under which central ministries could issue circulars to their field services at the local, departmental and regional levels (CIMAP, 2013;Fargues, 2013). 27 But these rather technical 'modernisation' reforms appeared marginal to traditional defenders of public services who equated the MAP with state retrenchment. 28 The Association of Rural Mayors spokesperson complained, for example, that cut backs in state personnel working for one state agency, the Atesat, would deprive communes of technical assistance in their areas of traditional concern such as planning roundabouts or public squares (Berberian, 2013). The local public sector was earmarked for 20 per cent of the €10 billion that needed to be cut as part of the competitiveness pact (Paquier, 2013b); its contribution to the pacte de responsabilité was likely to be even higher.
Comparing Hollande and Sarkozy suggests that robust political leadership is required to drive reform that challenges institutional inertia; its chances of success appear correlated to timescale and intensity (Elgie, 1993;Burnham, 2009;Bezes, 2010). Presidential power is most effective in France when there is active presidential involvement in steering, as in the early RGPP under Sarkozy. 29 By comparison, such leadership was lacking in the case of Hollande, who remained one step removed from both processes of state modernisation and decentralisation. Hollande was faithful to his campaign commitment to introduce a new decentralisation reform (his Dijon speech, February 2012), but was not strongly personally implicated in either process.
State reform and decentralisation under Hollande has been marked by institutional confusion and the absence of a legitimising discourse for change. The nomination of Maryse Lebranchu as the Minister for State Reform, Decentralisation and the Civil Service was intended to present a more holistic approach to state modernisation than under the previous Sarkozy-Fillon administration, by removing this field from the exclusive prerogative of the Budget ministry and its associated General Directorate for State Modernisation (Direction générale de la modernisation de l'Etat -DGME). The creation of the General Secretariat for the Modernisation of Public Policy (Secrétariat Général de la Modernisation de l'Action Publique -SGMAP) was presented as a political sign that the Prime Minister's office had taken back control of the State Reform agenda from the budget division 30 This institutional tinkering illustrated the need never to lose from sight the bureau-shaping motivations of state reform. One consequence of the lessening centrality of the budget division, however, was to blur responsibility for state modernisation and to lessen its political saliency: from being a synoptic reform under Sarkozy, the MAP drifted into insignificance. 'Act 3' of decentralisation was even more dissipated across rival bureaux, as well as being shaped by the rigorous lobbying activities of the local government associations and the activism of PS deputies and senators. 31 These examples suggested an overall lack of practical efficiency, more than an exemplar of joined-up government. The process of constructed divergence was limited in time and ineffective in its outcomes. The blurring of discourse made it much more difficult for Hollande to retain credibility for related initiatives such as the administrative simplification (choc de simplification) announced in July 2013.

Hard divergence and institutional resilience: The case of decentralisation?
Does divergence best describe the evolution of the trajectories of (national and territorial) governance in Western Europe? There has been a general trend across Europe towards decentralisation and new modes of multilevel governance (Keating, 1998;Loughlin et al, 2010;Piatonni, 2010). Various studies have focused on divergence within the United Kingdom facilitated by the introduction of an asymmetrical form of devolution (especially Greer, 2004). Diverging dynamics have also been described, inter alia, in Belgium (de Visscher and Laborderie, 2013), Spain (Colino, 2013) and France (Pasquier, 2012). Rather than just constituting another variety of convergence (towards a new regionalism, for example), such asymmetry can support a divergence argument, insofar as the pathways of decentralisation and state reform are shaped by contextually specific institutional, economic or ideational variables and practices. Insofar as they exemplify path-dependent forms of governance, state and decentralisation reforms might reinforce arguments based on national distinctiveness, rather than describing a general trend. Our (necessarily partial) overview of French decentralisation in 2014 reaffirms the importance of three types of domestic constraint: binding normative and constitutional principles; the sub-optimal institutional outcomes of three decades of layering; and the partisan logics of shared decision making.
Domestic normative constraints refer to the existing constitutional settlement, and, more generally, the norms affirmed in successive Acts and implementing decrees since 1982 (Marcou, 2011;Le Lidec, 2012b). The original sin of French decentralisation relates to the tensions between two fundamental principles. The first principle, central to the 1982 reforms, is that decision-making responsibilities should be attributed to specific 'levels' of sub-national authority, the 'blocs de compétences' provision. The division of competencies between levels (communes and intercommunal syndicates, departments, regions) might be likened to a quasi-federal arrangement with a clear underpinning of the EU subsidiarity doctrine (Schondorf-Haubold, 2007). The second principle, however, that of the 'free administration of local authorities', cuts across the apparent clarity of the first (see especially Marcou, 2011). All authorities should enjoy the freedom of initiative to make policies in areas they deem important for their constituents, the general competency clause. The constitution was amended in 2003 to prevent any formal hierarchy of one local authority over another, the so-called absence of tutelle provision, added to the constitution in order to provide a safeguard for the communes and the communal bloc who feared being subsumed by more powerful departments and regions. The 'no hierarchy' norm provides an incentive to agree contractualised forms of relationship between local authorities and the State. Moreover, it recalls, in the Jacobin tradition, that only the central State can have precedence over specific local authorities. As a governance principle, the 'no-hierarchy' clause is a recipe for stasis; in practice, the various sub-national authorities have overlapping territorial jurisdictions and loosely defined spheres of competence. No single authority can impose its will on any other, or prevent a rival authority from adopting policies in competition with its own (Chavrier, 2013).
French sub-national governance operates within the system of the millefeuille institutionel, the colourful metaphor used to describe the many layers of local public administration: the communes, inter-communal public corporations, departments, regions, city-wide authorities and special statute authorities that were given constitutional recognition in the 2003 constitutional reform and confirmed in the law of 16 December 2010. A steady process of institutional accretion has taken place. The French constitution now recognises four levels of local authority within the constitution: the commune (and by extension the communal bloc, including the EPCI), the department, the region and those with a 'special statute' (Cole, 2006). Recent decentralisation reformsof 2003Recent decentralisation reformsof -2004Recent decentralisation reformsof , 2010Recent decentralisation reformsof and 2013Recent decentralisation reformsof -2014 have not altered the basic, highly fragmented structure of French local government. 'Who does what' might not be the most interesting question. French sub-national governance rests upon a complex actor system, whereby policy is managed by plural actors with overlapping responsibilities at several levels. Complex actor systems produce interdependent relationships, rather than clear-cut transfers of responsibilities. Network analyses are almost certainly required to capture the complexity of policy processes (Cole and John, 2001). And yet what has been analysed elsewhere as the system of competitive interdependency involves inflationary institutional layering, whereby institutions maximise staff and financial resources, and seek redress by insisting on a favourable interpretation of legal norms.
Partisan and institutional logics complete the endogenous triptych. After its victories in the 2008 municipal, 2010 regional and 2011 cantonal elections, the PS ran a majority of large cities, almost two-thirds of départements, and all mainland regional councils except Alsace. Building on this electoral success, the left finally conquered the Senate in 2011, which immediately invested substantial resources in and reflection upon local government reform (Brouard et al, 2013). Until Hollande's 2012 election, this asymmetry complicated inter-governmental relations, as central government was wary of transferring more competencies to the local or regional level. Partisan competition had a clear impact on the outcomes of the 2004 reform; why strengthen the regions if they were to be controlled by the opposition parties? A similar calculation explained the refusal to engage in a far-reaching overhaul of local government competencies in the 2010 Act (Cole, 2012). The partisan paradox revealed its pertinence once again in 2013. The Ayrault government made the calculation that the PS would retain control of most of the large cities governed by the Socialists (at least Paris, Lyons, Lille and premier Ayrault's own city of Nantes), hence the decision to focus on strengthening large cities in the first instance. 32 These domestic-level variables were at least as important as exogenous arguments in explaining the precise contours of the 2014 Law on the Modernisation of Territorial Policy and the Cities (loi de Modernisation de l'action publique territoriale et d'affirmation des métropoleshenceforth MAPAM), the first of three promised as part of the Decentralisation 'Act 3' package. 33 The MAPAM law was published on 28 January 2014. It contained a number of innovations (République Française, 2014). Starting from the precept that not all categories of local or regional authorities face the same substantive policy problems, the original bill contained provision for an asymmetrical distribution of powers. The bill, published in April 2013, referred to the right to transfer competencies between local and regional authorities and to allow a process of experimentation, whereby not all regions would exercise the same responsibilities (Assemblée nationale, 2013). The precise distribution of policy competencies would be determined on a region-by-region basis by negotiation in the Regional Territorial Conferences (chaired by the regional Presidents); they would be approved by a national-level High Council for the Territories (Haut conseil des territoires). These regional conferences would have influence over the norms to be implemented by local authorities, though the latter would not be forced to cooperate. 34 In the final law, the Regional Territorial Conferences were retained in a watered-down form, but the High Council for the Territories was suppressed. 35 The second key principle of the MAPAM law, corresponding to the logic of the blocs de compétences, was that of the lead authority (chef de file). In his speech at the Sorbonne on 4 October 2012, President Hollande announced that: 'each field of public policy ought to be led by one local or regional authority, with powers to fix the rules and procedures under which this policy might be delegated to other local authorities'. 36 The idea of a lead authority addressed the constitutional barrier to allowing any one authority hierarchy over any other, the so-called absence of tutelle reinforced in the 2010 Act; the other local authorities would continue to exist (Chavrier, 2013). The Regions were confirmed in their role as strategic councils. They were recognised as lead authority in training, transport, economic development, innovation, biodiversity, sustainable development, climate change, air quality, energy and support for higher education and research. They were expected to perform lead roles in the new Public Investment Bank and were granted control over European structural funds. The departments were strengthened in their role as service providers for the local welfare state, for example, obtaining control of social care centres for handicapped people, personal autonomy, social development and energy insecurity. Communes (and the EPCI) were recognised as lead authorities in relation to mobility, the organisation of local public services, physical environment and local development, the ecological transition and sustainable mobility. The Territorial Governance Pacts would attempt to harmonise the division of competencies between layers of local authority, to be negotiated in each region (Gerbeau and Paquier, 2013). 37 The main innovation of the 2014 MAPAM Act concerned the métropoles 38 (République Française, 2014;Gerbeau, 2014). The law conferred a special statute for Paris, Lyons and Marseilles (rather as an earlier law (known as PLM) had in the 1982 round of decentralisation laws). 39 The MAPAM law granted a special statute to the three largest cities (Paris, Lyons and Marseilles) and granted the status of métropoles to nine other cities (Bordeaux, Grenoble, Lille, Nantes, Nice, Rennes, Rouen, Strasbourg and Toulouse) with a further two (Brest and Montpellier) able to bid for the coveted status at a later date. The existence of the métropoles enhances further the principle of territorial differentiation by recognising specific metropolitan challenges. But there were numerous points of inconsistency: the separate provisions for Paris, 40 Lyons 41 and Marseilles; 42 the numerical threshold for obtaining the status of a métropole (lowered from 450 000 to 250 000 in the case of the Breton city of Brest, for example, supported by the [Breton] minister Lebranchu); the precise policy responsibilities of the métropoles, which vary according to place. 43 Decentralisation Act 3 bears a number of similarities with Decentralisation Act 2 of 2003-2004. Both processes reasoned at an early stage in terms of territorial differentiation and experimentation, but neither pushed this analysis to its logical conclusion. Both played with the idea of territorial leadership by recognising lead authorities, but stopped short of following the quasi-federal logic of allowing the regions a general authority over the other layers of local government. Both sets of reforms revealed the influence of the second chamber, the Senate, on the detailed legislation that eventually emerged from the parliamentary process (see Le Lidec, 2009, for the fullest discussion of [2003][2004]. 44 At the time of writing, numerous questions remained unanswered: would there be a move to vest genuine regulatory powers on the regional level, as hinted by President Hollande in his 14 January press conference? If so, why reintroduce the General Competency clause for the regions and the departments in the 2014 MAPAM law? Would there be a reduction in the number of regions, from 22 to 15 or fewer, as advocated in the 2009 Balladur report, and evoked again in early 2014? Would the departments be suppressed? These questions have been recurrent ones in the recent history of French decentralisation. One expert interviewed in 2013 calculated that there were now 15 distinct regimes for governing relations between central government and local and regional authorities, depending on their size, location, level or degree of stand-alone fiscal capacity. 45 The 2014 MAPAM reform provided no answers and, indeed, by creating the new métropoles, increased the inflationary institutional layering that was identified in the Attali (2008) and Gallois (2012) report as one of France's structural weaknesses.

France's Contingent Governance
The reconfiguration of European states is a core tenet in the abundant academic literature on governance (see recently Loughlin, 2009;Piatonni, 2010;Jessop, 2013;Le Galès, 2013;Pasquier et al, 2013). French politics has often required conceptualising a coherent centre, a global référentiel, a sense of direction, of scalar hierarchy, of equal treatment, of inalienable rights guaranteed by state acting in the general interest (Cole, 2008). But the French polity has also mutated under the combined impact of internal and external pressures for change. Complex legal orders and interdependent relationships lay bare traditional beliefs about the supremacy of the state (Keating, 1998). Multi-level dynamics and the requirements of multi-actor coordination create new challenges for actors vested with public authority (Wollmann et al, 2010). The operation of the international political economy has produced metaphors of a hollow state that go to the very heart of the French statist model. These material realities have been extensively discussed and debated (especially in Cole, 2008). This discussion is necessary because neither convergence nor divergence make sense in a purely national frame. Governance needs to be qualified, however. There are dangers with attributing mechanical effects to movements such as Europeanisation, globalisation, or even decentralisation; these constructions are interpreted and mediated in different ways in specific contexts, whether that context be organisational, sectoral or the form of the state. Policymaking takes place within a realm of discourse, a system of ideas and representations that is comprehensible to the actors involved (Bevir and Rhodes, 2002;Muller, 2005;Schmidt, 2008) and that informs ways of knowing that vary according to places, countries, policy fields and historical periods. Attempts to describe state reform or decentralisation in terms of overarching covering laws or meta-narratives without reference to context are unlikely to be very convincing.
The context is thus key to unlocking specific ways of knowing (Moses and Knutsen, 2012). But how best to conceptualise context? The core choice to operate in relation to France is an adjectival one: Ought governance to be qualified by the adjective of contingency or boundedness? Bounded governance represents one alternative; context is understood as an independent variable, the variable lourde that translates external and internal pressures into outcomes that are consistent with established institutions, interests and ideational frameworks. Bounded governance borrows from Simon's bounded rationality: agents act rationally, but their knowledge is limited and they prefer to act upon what they know and understand, even if this means accepting sub-optimal outcomes (Cole, 2007). As originally formulated by the author, the central argument was that governance in France is bounded by institutional structures and path dependencies, by collective action repertoires and domestic interest structures, by the dynamics of policy sectors, above all by the overarching ideational and discursive contexts within which sense-making occurs (Cole, 2007). The metaphor of bounded governance emphasises above all the pertinence of hard divergence modes of reasoning.
While national contexts matter, they must not be confused with a static version of national purpose, however, as the metaphor of boundedness might imply. Arguments based mainly on national, institutional or cultural distinctiveness cannot easily cope with political and policy change. French responses to change are embedded in precise contexts, but they are not literally pre-shaped. The French polity has adapted and will continue to adapt to processes of domestic and external change and slowly transform itself in the process. Hence the preference for the concept of contingent governance, where context is held to be important, but which operates as an intervening variable that acknowledges human agency. Our preferred approach acknowledges the contribution of interpretative and discursive approaches (how actors frame and justify change), as well as new institutional perspectives (especially how change is filtered by institutions, and how decisions can tie-in future trajectories) and praxeology (how practice and practical interactions affect outcomes) (Tilly, 2001;Bevir and Rhodes, 2002;Radaelli, 2005;Schmidt, 2010;Bueger, 2013). Patterns of multi-level, national and sub-national governance are subject to sets of conditions that are contingent upon variable environmental, cognitive and interactional contexts.
The preferred adjectival choice to qualify governance is that of contingency. A contingent governance approach captures best our understanding of the Politics of Austerity in Hollande's France. Processes of hard and soft convergence and divergence operate in distinctive ways depending on whether formal outputs, informal outcomes, political discourses or institutional innovations are considered. The four positions identified in our typology are more (soft) or less (hard) receptive to human agency. Hard convergence and divergence as results are best explained by material macro-level variables (economic indicators, institutional pathways and legal orders), which create tensions, sometimes evolve in opposing directions and require individual or collective-level interventions. 'Soft' convergence and divergence as mechanisms allow more room for human agency, specifically strategic choices and forms of discursive legitimisation, though they are ultimately bounded by harder material constraints. The following conclusions demonstrate our preferred explanations for the cases covered in this article, but, as Tilly (2001, p. 38) argues, law-like generalisations extrapolated from even complex contexts are likely to be of limited utility ( Table 2).
The article has provided strong arguments to support a hard convergence thesis. The financial crisis has provided evidence of some recentralisation of decentralisation, whereby the latest phase of EU integration has forced central governments in most instances to exercise a tighter supervision over local and regional government expenditures. Membership of the Eurozone creates a permanent pressure not to diverge that is likely to intensify as the zone becomes increasingly integrated. Europeanisation has produced a lessening of divergence in legal systems and in the provenance of much public policy (Ladrech, 2010). In the French case, it is too early to judge the impact of these pressures in the field of our investigation: state modernisation and decentralisation. In comparison with counterparts elsewhere in Europe, there has been no major calling into question of territorial equilibriums as a result of downgrading of local or regional authorities, as has occurred notably in Spain (Colino, 2013). The forces driving hard convergence have a differential impact depending, somewhat, on the size and resources of an individual member-state: France is in a stronger position, for example, than Spain (on an account of the interest rate spreads) or Belgium (a small divided country) to resist unwelcome influences from European policy. But this is a matter of degree. France is closely monitored by European authorities, international organisations and the markets and cannot determine its policy in a vacuum. Hard convergence arguments are much less convincing in terms of institutions; even common trends such as NPM appear shaped more by domestic traditions than by a genuine cross-national template. On balance, however, the forces identified as those of hard convergence are gaining ground.
To reason in terms of hard divergence makes sense mainly insofar as it refers to the capacity of existing institutional configurations to define the parameters and limits of change. This conclusion does not imply that the French model of territorial relations is radically different from others or that the direction of change is necessarily divergent. But it does lead us to conclude that the pressures producing hard or soft convergence are mediated by domestic party structures, interests and existing institutional configurations. Heavy domestic variables might be interpreted in part as the incremental adjustments of past bargains that are given legal-constitutional status (for example, in the constitutional reform of 2003 which prevented hierarchy of local authorities). Hard divergence is confirmed in key senses by the persistence of national administrative pathways (John and Cole, 2000). Domestic state and institutional structures, party systems and the political rules of the game still make sense nationally, leading Schmidt (2006) to diagnose a dangerous gap between (national) political competition and (European) policy formulation. As discussed above, the 2014 decentralisation reform demonstrates clearly that domestic variables remain a formidable source of resistance to change. The persistence of embedded institutional interests tends to produce satisficing forms of arrangement. The work undertaken by Le Lidec (2007Lidec ( , 2012a in particular allows us to make sense of the weight of established institutions, interests and sequences in determining the pathways of territorial reform. On balance, however, the forces of hard divergence are losing ground, as economic, epistemic and European pressures continually challenge the forces of institutional inertia. Hard mechanisms (whether of convergence or divergence) mainly involve structural-based explanations: they are less amenable to human agency than either soft convergence or divergence. Our investigation demonstrated limited evidence of soft convergence, at least during the first 24 months of the Hollande presidency. In practice, there was not much evidence of voluntary benchmarking. The dynamic of soft convergence appeared stronger under Sarkozy than Hollande, in the form of a more explicit espousal of the language of NPM and a closer connection between discourse and practice. What was labelled as soft or constructed divergence was pursued far more vigorously by the Hollande administration, defined above all in terms of reacting against the Sarkozy period (the abandoning of the RGPP, the repeal of core aspects of the December 2010 territorial reform). The first 2 years demonstrated the limits of the anti-Sarkozy method; framing state reform in negative terms had the major disadvantage, in Schmidt's terminology, of preventing the emergence of a coherent legitimising discourse. The Hollande-Ayrault administration was undermined by its absence of a legitimising discourse for change, by its lack of internal cohesion and by the weight of rival institutional lobbies. Hollande's discursive shifts were the equivalent of the couacs 46 that delighted critical commentators. They were interpreted as an inability to pursue a consistent line and to persuade the electorate of its necessity. Schmidt (2008, p. 305) urges us to taking seriously 'what actors say and do'; by extension, due consideration ought to be given to what they do not say and do not do. In the field of State Modernisation, the message was inaudible; in the case of decentralisation, the original coherent proposals were salami sliced. There was an eminently practical consequence of the lack of a consistent discourse: the government deprived itself of a legitimising device to tie in actors to shared visions. The limitations of discursive approaches must be acknowledged in some respects, but this discussion demonstrates their power in terms of defining the conditions for the exercise of discursive institutionalism. Discursive institutionalism is not a simple description of institutionalisation, but a challenge in relation to how to embed a paradigm or interpretive frame within an institution. Sarkozy came closer to achieving this mix than Hollande.

Notes
1 The research project (Territorial Governance in Western Europe: Between Convergence and Capacity) was supported by the Leverhulme Trust's International Network programme (IN-2012(IN- -109, 2012(IN- -2015. I am grateful to the Trust for its support. I am also grateful to the Collegium de Lyon, which provided additional support, and to Cardiff University for its University Research Fellowship award in 2013-2014. Twenty-nine interviews were carried out in Paris and in the French region of Brittany from July to September 2013. The article also draws on earlier interviews (1994-1996, 2001-2003, 2004-2005, 2006-2007, 2010) carried out as part of projects funded by the ESRC, the AHRC, the British Academy and the Fondation de France. I thank all of these agencies for their support. 2 I am indebted to Arthur Benz who provided valuable commentary in relation to this section. 3 The role of Eurostat was emphasised in several interviews, especially in Belgium in early 2014. 4 Le Lidec (2011a) emphasises the importance of sequential timing. In his interpretation, the sequencefinancial, political, administrativecan explain the character of French decentralisation reforms, initially interpreted in terms of the consolidation of the grands élus. Such sequencing is compatible with the contingent governance approach presented here. An alternative explanation might emphasise the differential logics present even in 1982-1986 (removing obstacles in the way of the grands élus, but also defence of the départements, and support for forms of regional capacity and identity building) and might stress the iterative nature of change, not locked into one definitive mode of regulation. 5 Interviews were carried out mainly in central government (in 2005 and 2010) and in the French region of Brittany (in the summer of 2013). As Bretons performed a key role in the Hollande administration, especially in this sphere of policy action, Brittany was the perfect place to be for 3 months just before the 'Red Beret' movement erupted. Interviews are also scheduled in Lyon in May-June 2014. 6 The case, for instance, for the 'low-profile instruments' analysed by Bezes (2007) to limit the growth of civil service wages in France, or the use of non-indexation to control salaries in the UK public sector under Cameron. 7 Twenty-five interviews were carried out in the Spanish region of Andalucía in October-November 2013 as part of the Leverhulme Trust project. 8 This conclusion was shared in interviews in the Wallon Regional Council and the Belgian central bank in January-February 2014, as part of the 25 interviews carried out in Belgium as part of the Leverhulme Trust project. 9 This point was acknowledged in interviews with the MEDEF, August 2013. 10 Qualified majority voting (QMV) is the basic principle for determining European Union policy decisions, except in those limited areas where the rule of unanimity still prevails. Votes for each country are weighted according to population. RQMV implies that a recommendation or a proposal of the Commission is considered adopted in the Council unless a qualified majority of Member States votes against it. Under the reverse QMV procedure, a qualified majority will need to be mobilised to prevent automatic penalties in the case of not respecting the revised Stability Pact criteria. Larger countries are better placed than smaller ones. 11 The 'Six Pack' refers to five regulations and one directive adopted as part of the reform of the EU's fiscal governance and the SGP in 2011. The many measures adopted included: country-specific medium-term objectives for deficit and debt reduction; the requirement that general government deficits and public debt must not exceed 3 per cent and 60 per cent of GDP, respectively, or at least diminish sufficiently towards the 60 per cent threshold.  (2010); ACuF (2013); ADF (2013). 14 Interviews in the Brittany regional council, July and August 2013. 15 Interviews AcuF (2013); Rennes métropole (2013). 16 Several interviews were carried out with public finance specialists and actors. All agreed that formal fiscal autonomy had been greatly reduced for the regions and departments. There was less agreement about how well local and regional authorities would adapt. One senior public finance official working for an EPCI was of the opinion that past financial largesse would easily allow the councils to absorb this cut. Officials working for the departments and the regions were far less sanguine. 17 Interview, ADF, September 2013. 18 The EPCI were given a new share of the two important taxeson residents and landwhile the regions and departments lost their share of these dynamic taxes. 19 The expression was used repeatedly in interviews in the State Reform directorate (DGME) in 2010. 20 'Aujourd'hui à Bercy c'est presque un dogme, dès qu'un poste se libère on fait appel à un cabinet de consultants, un chasseur de tête, on ne cherche pas dans la structure'. Interview, Finance ministry, 2010. 21 President Sarkozy made the commitment to abolish the Classement de sortie de l'ENA, but this was abandoned in one of the last acts of the Fillon government. 22 Interviewees in the DGME in 2010 celebrated a 'French-style new public management'. In an earlier interview in the Finance ministry in 2005, one interlocutor clearly identified the new budgetary policy in terms of opportunities: 'La gestion publique est dominée en France par la gestion de la régularité. L'émergence et le développement d'une culture de la performance au travers des lois organiques et des démarches d'objectif devraient faire basculer l'approche de la gestion publique plus vers la question des performances et des résultats que dans la question de la régularité …. Il faut absolument faire dominer l'approche des performances sur l'approche des régularités'. 23 Interview evidence (in the Financial Division) from the Brittany region in 2013 suggested that the regional council had adopted the methodology of the 2001 budgetary law (LOLF) in its own internal budgetary processes. 24 Schmidt distinguishes between simple (majoritarian) and compound (consensual) polities. 25 Evidence to this effect is provided in Cole (2012). 26 Local and regional government actors interviewed were, in general, sceptical about the link between the MAP and the local government reforms. One PS deputy, rapporteur of the first decentralisation bill in the National Assembly, argued that Minister Lebranchu had been unable to link the two processes. 27 State actors interviewed in September 2013 confirmed that the regional State would be given a further boost by these measures, which demonstrated continuity with territorial state reform under Sarkozy.
But doubts were expressed within one regional prefecture about the extent of the regional prefect's control over the budgets of the other state services: C'est un peu décevant. La décision numéro 30 du CIMAP, dit que c'est le préfet de région qui est RBop pour tous les services classés sous son autorité. C'est sous son autorité et c'est lui qui décide. C'est l'évolution logique de la remonte vers le plan régional. Mais la décision 31 a un contenu très décevant. J'ai demandé la fongibilité, encadrée et contrôlée, de certains crédits et de certaines masses de postes pour que le préfet de région puisse vraiment, là où il a besoin, affecter des personnels, affecter des crédits. Ça a été refusé catégoriquement. 28 A theme that emerged in several interviews in Brittany in summer 2013. 29 'Il y a eu un changement de président de la république, il a voulu affirmer de manière forte ses orientations sur l'administration: on ne remplace pas un emploi sur 2. Tout d'un coup dans ce ministère il y a eu un changement brutal, les français ont élu de manière forte un nouveau président de la République qui veut imposer un nouveau rythme, ça s'est fait immédiatement. Si il n'y avait pas eu un acte politique majeur, ca n'aurait peut être pas eu lieu. Ça c'est l'aspect positif'. Interview French Finance Ministry 2011. 30 The greater involvement of Matignon and distance of the Elysée was confirmed in several interviews. 31 According to one source very close to the discussions, the decision to split the original proposals into three came about as the result of 'a discussion between Ayrault, Lebranchu and the chair of the legislation committee in the Senate, who insisted that the text be cut into three to allow time for proper scrutiny'. 32 This interpretation was provided in several interviews, especially with the rapporteur of the MAPAM law. 33 The original plan had been to introduce one overarching reform that would have addressed the problems of multiple structures, overlapping competencies and the control of public finances. Faced with powerful, if contradictory, forces, however, Minister Lebranchu announced that the government would introduce three separate bills: the first one on city governments; the second on the clarification of competencies between the 22 regions and 101 departments; and finally in relation to the 'communal bloc'. 34 Local authorities refusing to cooperate would face financial penalties of up to 30 per cent. 35 More than anything else, this reflected the continuing influence of the Senate, which had steadfastly refused the High Council in the name of defending the principle of free administration. Interviews with UMP and PS politicians converged on this point (July-September 2013). PS Senator J.P.Sueur confirmed this interpretation in his opening remarks to the GRALE colloquium on 30 ans de la fonction publique territoriale, French Senate, 31 January 2014. 36 'A chaque grande politique correspondrait une seule autorité qui fixerait les modalités et l'action qui peut être déléguée à d' autres collectivités'. Cited in 'Ce qu'attendent les territoires' 15 January 2013, ActeursPublic.com, accessed 8 April 2013. 37 The issue area of tourism was pulled between the regions and the departments. In an interview, one proregionalist deputy claimed to have lobbied hard for tourism to remain a general competency, rather than granting the responsibility to the departments, the provision that initially emerged from the Senate's first reading. 38 I have left métropoles in the French. 39 For one sceptic, the first law was 'not a decentralisation law but a law on cities'. 40 One of the key aspects of the MAPAM law was the creation of the Greater Paris Council, an Intercommunal Public Corporation (Établissement public de coopération intercommunale -EPCI) with its own taxation powers. It will bring together the city of Paris, all communes in the three departments of the small Paris belt (Hauts-de-Seine, Seine-Saint-Denis, Val-de-Marne), along with some communes in the larger Ile-de-France region. It will exercise the classic functions of an urban community: planning, land use, social housing, economic, social and cultural development, along with some environmental and energy responsibilities. It is expected to come into existence on 1 January 2016. The Greater Paris council will take over certain functions from existing EPCI. Unlike in Lyons, the existing departments will continue to exist. 41 In Lyon, the proposed métropole is by far the most integrated, with the city of Lyon taking over social policy responsibilities from the Rhône-Alpes department.
42 In Marseilles, there is very serious opposition from nearly all local players to the proposed métropole, which will merge Marseilles city with a number of outlying urban centres. 43 The métropoles will be strengthened with respect to the classic intercommunal corporations. They will be the lead players in the structure plans (schéma de cohérence territoriale) within their boundaries. They are given stronger powers in relation to land use around train stations. Their powers in relation to housing are greater than a non-metropolitan EPCI, notably in the field of social housing. The métropoles might take over some responsibilities exercised by the departments if contractual agreements are concluded between the two (for example, in the field of assistance for deprived young people and housing grants). They might also take over some services currently delivered by the regions. The métropole is recognised as the lead authority in relation to local and regional government competencies that are delivered within the city boundaries 44 Interviews undertaken by the author in 2013 confirmed the strength of the Senate's opposition to initial proposals emerging from the National Assembly. 45 Interview with Ressources consultants, with an expert in local public finance. 46 Literally, the false notes in an orchestra.