Federalism, multi-level elections and social policy in Brazil and India

ABSTRACT Comparative studies often highlight the negative effects of federalism for welfare state expansion. We examine Brazil and India, which have both enhanced their welfare effort despite political fragmentation. We argue that federalism’s effects must be seen together with degrees of party system nationalisation. In Brazil, new social policies have reinforced a move towards greater party system nationalisation. Control over anti-poverty programmes has been recentralised leading to more even outcomes. In India, while the central government also introduced new social policies, expansion has been filtered by political regionalisation. The effectiveness of social provision relies on state governments, producing substantial territorial differentiation.

The dispersal of political power within complex multi-level structures of government can present challenges for the development of welfare states. A sizeable literature on federalism and welfare states in advanced industrialised economies suggests that the fragmentation that accompanies federalism is bad news for social policy. Cross-national comparative studies highlight features such as horizontal competition between federal sub-units, vertical inter-governmental competition and the multiplication of veto points within multi-level government which are argued to undermine coordination, empower special interest groups and make it harder to achieve improvements in social provisioning (Bonoli & Champion, 2015;Champion & Bonoli, 2011;Huber, Ragin, & Stephens, 1993;Obinger, Leibfried, & Castles, 2005;Pierson, 1995; see also Gerring, Thacker, & Moreno, 2005 for a related argument applied to a wider range of cases). Furthermore, the complex and often opaque relationships and divisions of responsibilities between levels of government in federal systems reduce the ability of voters to hold governments accountable for policy performance (Cutler, 2004). The very goals of federal constitutional settlements, to protect self-rule alongside shared rule, may work against efforts to extend equal social rights across a country because federalism, especially in multi-ethnic contexts, also depends upon the recognition of difference (Obinger et al., 2005, p. 2). 1 Yet much of the existing literature treats federalism as an undifferentiated variable which often serves simply as a proxy for political fragmentation or decentralisation without taking into account significant variation in structures of federalism across countries and the effects of intervening variables such as degrees of political, fiscal and administrative (de)centralisation or party system nationalisation. Cross-national studies also depend typically on national averages that obscure the subnational differences that result from the decentralisation of power over social policy in federal systems (Greer, Elliott, & Oliver, 2015). Lastly, most of the existing literature on the effects of federalism on social policy draws on case studies from advanced industrial countries with mature welfare states. The implications of federalism in low and middleincome countries, often in the process of attempting the simultaneous consolidation of both democracy and welfare provision, have been little studied. In the wake of several decades of decentralisation across the developing world, as well as the increasing attention to federalisation within middle-income countries as a response to conflict and debates about regional equity, the impact of federalism on welfare in such countries deserves greater attention.
In this paper we explore the cases of Brazil and India. Both countries saw the adoption of far-reaching new social policies in the 2000s despite the tendencies towards fragmentation and policy stasis supposedly inherent in their federal institutions. In earlier decades, such fragmentation had been seen as a reason for the lack of progress on 'equity enhancing change' (Weyland, 1996, p. 5). We argue here that one feature of federalism that fuelled and shaped social policy expansion in a period of growing government revenues was competition between different levels of government to claim credit for anti-poverty programmes in the eyes of voters. But we argue that this tendency must be interpreted in conjunction with patterns of party system nationalisation. Both countries were governed at the federal level by parties with social democratic commitments which predisposed some of their leaders to pursue welfare as a goal in its own right. In both places national parties saw the opportunity to use social policy to reach out directly to poor voters over the heads of subnational governments. Yet their strategies had different results.
In Brazil, conditional cash transfers (CCTs)the largest of new social policy commitmentshave been a vehicle for the nationalisation of credit claiming around social policy by political parties at the centre. 2 Whereas previously the delivery of anti-poverty programmes was more clientelistic, programmes such as Bolsa Família challenged local oligarchies and enhanced the prestige of the Workers' Party at a national level (at least before the anti-corruption investigations that began in 2014). This helped to ensure that the programme was implemented fairly evenly across Brazil (Fenwick, 2016). The centrality of social provision in the political legitimation strategies of successive presidential coalitions from the mid-1990s onwards reflected what Carlos Pereira and Frederico Bertholini in this special issue describe as a 'dominant belief' in social inclusion with fiscal responsibility that cut across parties (see also Alston, Melo, Mueller, & Pereira, 2016, p. 116). By contrast, in India, even in an era since 2004 when the federal government took on a larger role in the design and funding of new or enhanced national social policies such as the Mahatma Gandhi National Rural Employment Guarantee Scheme, or the National Food Security Act (NFSA), it is state governments that have been more successful in claiming credit for welfare measures. Although the Congress Party at the national level sought to derive electoral mileage from new social policies, state governments (including opposition-ruled states) were relatively successful in claiming a significant degree of credit for federal policies and many also ran their own programmes. Moreover, reforms to ensure the more effective delivery of anti-poverty programmes have depended on the initiative of individual states. There is considerable unevenness in the extent to which chief ministers have over-hauled local systems to insulate social programmes from clientelistic pressures and enabled more reliable implementation. 3 We show in this paper that inter-governmental competition had different consequences in Brazil and India because of significant differences between the two countries in degrees of party system nationalisation. In Brazil, the introduction of new federally financed social policies, especially CCTs, has reinforcedand been reinforced bya move towards greater party system nationalisation. These programmes allowed the national government to move into areas that had previously been the preserve of states, effectively recentralising control over anti-poverty programmes. As the programmes became popular they helped bolster multi-level electoral alignment and increased the incentives for subnational politicians to take on federal party labels or support federal coalitions (Melo, 2008, p. 184). In India, state governments successfully claim much of the credit for social policy in the eyes of voters, 4 even when these policies are designed and funded by the government in New Delhi. This scenario is supported by the regionalisation of India's party system. Unlike Brazil, state-and national-level elections take place according to different timetables in India. Elections to the subnational tier have a substantial degree of autonomy from the national level, and social welfare has become an increasingly prominent part of state-level election contests in recent years. The decentralisation of credit claiming for social policy supports, and is supported by, the relative regionalisation of India's party system.
In emphasising the interaction of federalism with degrees of party system nationalisation, we bring together two literatures that do not often refer to each other but may have important implications for understanding the trajectory of welfare state development beyond Europe and North America. We note that historically the development of welfare states in post-war Europe was located alongside the development of national party systems and the replacement of more regional or particularistic political identities by nationwide social cleavages framed around class (worker-owner) cleavages in the post-industrialisation context (Caramani, 2004;Lipset & Rokkan, 1967). The experience of war also had the effect of increasing national solidarity which many authors have linked to the growth of welfare states (Burke, 1985;Singh, 2016, pp. 244-248). There is some evidence to suggest that more nationalised party systems see higher levels of social spending because parties are incentivised to provide broad public goods to appeal to a wide electorate, rather than providing more localised, targeted goods to particular constituencies (Hicken, Kollman, & Simmons, n.d.;Castañeda-Angarita, 2013;Jurado, 2014).
In both India and Brazil, the development of social provision has been debated without the kind of national ideological political cleavages that were associated with welfare state emergence in earlier periods and other geographies. Although ideology and class matter in both countries, in neither are party systems organised around a clear left-right axis. Therefore, the extent of party system nationalisation per se may loom larger as an explanation for the design of social provision. While Brazil has a lower level of party nationalisation than other parts of Latin America because of the profound proliferation of parties encouraged by its electoral system, it has a considerably more nationalised party system than India where the institutionalisation of territorialised identities within its multi-ethnic federal system has underpinned the regionalisation of its party system. It has been argued recently that in India, it is the cohesiveness of subnational identities within individual states that is pivotal for understanding development outcomes (Singh, 2016).
In this article we first bring together a discussion of the existing literature on federalism's effects for welfare provision with that on multi-level electoral systems to establish the contrasting institutional and electoral patterns in Brazil and India. We then move on to compare structures of federalism and party systems in the two countries, before analysing how these structures produce different outcomes for the politics and implementation of social policy in both countries. The article focuses largely on the period of fiscal and economic expansion in both countries in the 2000s and early 2010s. In Brazil, growth fuelled by the commodity boom combined with an increase in tax revenue facilitated expanded redistribution. In India, rapid servicesector driven economic growth also produced new revenues that were invested in part in the social sector. 5 A brief epilogue will consider the period since 2014 in which new fiscal and political pressures have altered the environment in which social policy has been designed.

Federalism, multi-level elections and the welfare state
Key issues arising from the institutional structures of federalism that are highlighted in comparative literature include coordination problems between levels of government, competition between and within levels of government over policy jurisdiction and spending, and the multiplication of institutional veto points. Where institutional structures multiply veto points, the ability of the central government to oversee new policies or difficult reforms may be diminished (see, for instance, Ames, 2001;Gerring et al., 2005;Tsebelis, 2002), although this same feature can also make retrenchment policies more difficult (Obinger et al., 2005;Pierson, 1996). Core to the central tenets of self and shared rule under federalism is the dispersal of political authority between multiple jurisdictions. Not only are institutions themselves dispersed by design, civil society organisations are likely also to be fragmented because they have to achieve influence in multiple settings where the actors, interests and context of decision-making are likely to be different (Pierson, 1995).
Coordination between multiple centres of political authority can be a crucial problem for the development of federation-wide social rights. Social policy often becomes the focus of inter-governmental competition as both central and subnational governments seek to claim the political credit associated with social policies. In such instances, social policy debates become central to the act of state-and nation-building itself, as multiple tiers of government attempt to use them to bolster their legitimacy and popular allegiance to either national or regional communities (Banting, 2005, p. 90). Further issues arise where the initiation of new social policies in certain subnational units pre-empt initiatives in related fields at the national level, and make agreements over new national policies more contentious because they appear to challenge the autonomy of sub-units already operating policies in this sphere, or because subnational policies circumscribe options open to the central government (Pierson, 1995, p. 456). On the flip-side, subnational units have often been sites of policy innovation. In both Brazil and India, the major new social assistance programmes of the 2000s -Bolsa Família, and the MGNREGAbegan their life within individual municipalities or states within the federations.
While we are primarily concerned with the political effects of federalism on social policy in this article, the fiscal determinants are of course important too. As by now classic texts in the field of fiscal federalism suggest, fiscal decentralisation can increase horizontal competition between sub-units, generating pressures to 'race to the bottom' by cutting taxes and thereby social expenditure in order to attract mobile capital and avoid becoming 'welfare magnets' (Tiebout, 1956). However this dynamic is mediated by the design of fiscal federalism, and the division of taxation and expenditure responsibilities between the central government and states. Where control over expenditure but not taxation is devolved to the subnational level, horizontal tax competition is likely to be reduced and subnational governments may increase expenditure by over-spending common resources (Rodden, 2003).
Federalism may also affect the ability of central states to raise revenues to finance social expenditure. As Lieberman (2003) showed of Brazil, federalism impeded the ability of the poor majority to make national, class-based demands for progressive taxation because elites in wealthier regions perceived demands for greater direct taxation to imply redistribution to poorer regions. But degrees of centralisation within federal systems change across time. Brazil underwent a period of recentralisation in inter-governmental relations under President Cardoso, which has been sustained under his successors. The tax capacity of the central Brazilian state, its ability to impose subnational fiscal discipline, as well as to administer new forms of social policy were all enhanced in this period. The central government also used earmarked taxes called 'contributions', not shared with state and local governments, to raise revenues for social sector expenditure buoyed by the cross-class social coalition of working class, social movement and middle class actors that supported the PT (Partido dos Trabalhadores or Workers' Party) government (Schneider, 2015). Revenues from such taxes increased from 9% of GDP in 1994 to 13% in 2010 (Schneider, 2015, p. 6). Thus it is important to remember that degrees of centralisation and coordination within federal systems change over time, with implications for taxation and expenditure.
One feature that has an important influence on the degree of centralisation, beyond the constitutional inter-governmental allocation of powers and responsibilities, is the nature of the party system. Literature on elections in multi-level settings suggests that there are substantial differences in the degree of political coordination in federal settings which are caused by different structures of political competition. Of particular relevance to this discussion is the extent to which one tier of government is pre-eminent over the other in the electoral arena. The extent to which either the national or subnational tier of government predominates in the minds of voters varies substantially across federal systems. In some federal systems, subnational elections are clearly subordinate (such as Germany and Brazil). In such systems, voters use information about national policies, parties or presidential candidates as a heuristic for making decisions about how to vote in subnational elections (Borges & Lloyd, 2016;Rodden & Wibbels, 2011). In other multilevel electoral systems (such as Canada or India), subnational elections have greater autonomy from the national layer. Factors that have been demonstrated to increase the extent to which subnational elections are dominated by national politics, and the extent to which parties coordinate with each other at the national level, include fiscal and political centralisation (Brancati, 2008;Chhibber & Kollman, 2004;Diwakar, 2010), holding concurrent elections to both tiers of government (Borges & Lloyd, 2016;Brancati, 2008), the absence of territorial cleavages (Schakel, 2013) and presidentialism (Hicken & Stoll, 2011). In more decentralised parliamentary federal systems such as India and Canada with strong territorialised regional and ethnic identities, voters are more likely to attribute policy responsibility to their regional or subnational government. They can also be more likely to vote in subnational elections based on the performance of the subnational tier, rather than the national level (on India see Yadav & Palshikar, 2009).
In the following section we look at how federal and party systems intersect in Brazil, where presidential coattails are important for subnational elections and in India, where subnational elections have a first order character in electoral competition. 6 We then move on to analyse the implications of these contrasting dynamics for social policy.

Comparing federalism and party systems in Brazil and India
Brazil's electoral system is unusual and has led to an extreme form of multipartyism, with 32 political parties represented in Congress and one of the highest number of effective parties in the world. In 2010 the effective number of parties in the lower house of Congress was over 11 by votes, and over 10 by seats (Epstein, 2012, p. 47). There is no threshold for parties to obtain seats in Congress, and politicians find it easy to create new parties. Most parties are amorphous in ideological and programmatic terms and many politicians switch parties several times during the course of their careers. In the lower house, candidates win election on the basis of open list proportional representation (a first past the post system is used for the Senate races.) This diminishes the incentives for party discipline, as candidates in the same party essentially compete against each other. The parties that compete for power, and the coalitions that form at the state level, are different across states. In this sense Brazil has a high level of political fragmentation, and a lower degree of party system nationalisation than other Latin American countries.
However certain features of the Brazilian system strengthen the 'first order' character of presidential elections, making Brazil's party system far more nationally integrated than India's. Since 1994 presidential, gubernatorial, and legislative elections at the state and federal level have concurred in Brazil, strengthening the alignment between national and subnational election outcomes (Borges & Lloyd, 2016, p. 107). With the exception of small groups of indigenous people, Brazil does not have the same territorialised ethnic identities that help to increase the primacy of subnational election contests in multi-ethnic federations such as India (Borges & Lloyd, 2016;Schakel, 2013). Furthermore, the Workers' Party (PT), in power at the federal level from 2003 to 2016, is a truly national party with representation in almost all of Brazil's municípios (counties). Presidential politics shape state politics because, as Borges and Lloyd (2016, p. 104) argue: 'No candidate for subnational executive office receives as much media attention or as many campaign contributions as the top presidential contenders do.' At the national level, a system of 'coalitional presidentialism' helps to achieve coordination within Congress despite the high number of parties (Melo & Pereira, 2013). This has elements of a parliamentary system, in that presidents are elected from parties that never control more than 20 per cent of the seats in either house, and therefore have to stitch together complicated multiparty coalitions in order to govern. As the party system has fragmented in recent years, those coalitions have grown in size. In the government of Dilma Rousseff (2011Rousseff ( -2016, the president impeached in 2016, the coalition was made up of 10 different parties. Presidents placate these parties and keep them in the governing coalition by awarding them ministerial posts (sometimes with the right to appoint large numbers of people at lower levels of the ministry) and federal resources for their home states.
In recent years the coordination of parties at the national level has been supported by a trend of fiscal recentralisation. Under the new constitution that was promulgated at the end of the military dictatorship in 1988, Brazil entered an era of decentralisation. Because the dictatorship had centralised power in many areas, there was popular pressure and regional political interest in decentralising. Decentralisation became associated with democratisation. However, this process began to be partially reversed in the 1990s, in part because state spending got out of control, jeopardising the fiscal equilibrium of the federal government. Under President Fernando Henrique Cardoso (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002), a fiscal responsibility law reined in state spending. In the same period, Cardoso's government decentralised expenditure responsibilities to the municipal level in order to make it more difficult for governors to divert revenues towards patronage (Dickovick, 2006). New centre-local linkages helped the federal government restrain governors who had been empowered in the process of democratisation. Significantly for the story in this paper, social policy was the most substantial area of expenditure responsibility to be transferred to the municipalities (Dickovick, 2006;Fenwick, 2016). The federal government thus reduced the spending capacity of the estados while also enhancing fiscal space at the federal level by expanding taxation. 7 New revenues collected by the federal government were increasingly directed towards the social sector from Cardoso onwards (Melo, 2008, p. 176;Schneider, 2015), and were directed via municipalities rather than the estados (Fenwick, 2016). India also has a high level of political fragmentation though not as extreme as Brazil's. A total of 35 political parties found representation in the Lok Sabha (lower house of parliament) in the 2014 elections. Each state today effectively has its own party system in which different parties and political leaders are significant. Most regional parties have a strong base in only one state. Regional parties are dominant in many states and in some have completely sidelined national parties. From 1984 onwards, this increasingly complex regional scenario was reflected in a marked increase in the effective number of parties by seats won from about two in the period 1952-1984, to above six by 2004, although the number has been steadily decreasing since 2004 to just over three today (Tillin, 2015, p. 187). And despite the high overall number of parties in the Lok Sabha, at the state level most districts tend towards two party competition (Chhibber & Murali, 2006). The present-day scenario of regionalisation reversed the trend of the first four decades after Independence in which India's party system was relatively nationalised with the Congress Party predominant, and represented across the breadth of the country. From 1989 onwards, India entered a 'post-Congress' polity in which the political landscape was markedly decentralised and electoral competition was no longer organised around the Congress Party (Palshikar, 2013;Sridharan, 2002;Tillin, 2015).
Until the early 1970s, national and state elections were held concurrently but have since been held according to separate five-year timetables in each state. 8 Holding state elections according to separate timetables increases the primacy of state-level political leaders, parties and issues in election debates. State-level issues loom large even during Lok Sabha elections. With the partial exception of the more presidential style of the 2014 elections in which the Narendra Modi-led BJP came to power, Lok Sabha elections since 1989 have been interpreted as the aggregation of multiple state-level contests with state-level issues and actors filtering the choices that voters make in national elections. Yadav and Palshikar (2009) argued that there are many ways in which the state level should in fact be seen as the principal electoral layer. The menu of political choices in Lok Sabha elections, the relevant social cleavages and cultural-ideological context is determined by the statelevel party system. Furthermore, voters are often focused on the performance of their state government even when voting in national elections (Yadav & Palshikar, 2009, p. 57). This assumption is borne out by survey data collected in the National Election Studies conducted by CSDS (Centre for the Study of Developing Societies)-Lokniti, as shown in Table 1. Taken together over half of all voters in the 2009 and 2014 general elections said that they gave more or equal importance to the state government, as compared to the central government, when deciding how to vote, although the number identifying the state government alone fell in 2014. This does not mean that the performance of the central government is completely absent from the calculations of Indian voters in general elections, nor that there are no patterns in the social support base of national alliances that transcend state boundaries (Chhibber, 2009). Rather the key point is that, especially compared to Brazil, voters have been more likely to read national elections through the filter of a state-level lens.
In India, by contrast to Brazil, the prevailing pattern of the period since the early 1990s has been one of political and fiscal decentralisation but from a starting point of a more centralised system of fiscal federalism. With the onset of economic liberalisation in 1991, the central government's earlier overriding role in directing regional flows of public and private investment was dismantled and states have become the primary gatekeepers for business investment. In the fiscal realm, as in Brazil, India's states also faced a debt crisis in the late 1990s but implementation of fiscal responsibility and budget management legislation helped stabilise state finances. The fiscal position of the states also improved as a result of new powers to raise VAT that were introduced in 2005, higher devolution of central taxation as a result of improved revenue collection, overall high economic growth and low interest rates (Government of India, 2015). Overall, the proportion of taxes collected by central and state governments (before centre-state transfers take place) looks similar to Brazil with 62 per cent of tax revenues collected by the central government, and 38 per cent by the states. 9 These parallel processes of regionalisation in India have enhanced the first order character of subnational elections and the decentralised nature of credit claiming for policy initiatives by state governments, compared to Brazil. In Brazil, as we delineate in the next section, national policy initiatives in the field of social policy have supported gradual processes of party system nationalisation, albeit in a context of continued political fragmentation.
It may be noted that these contrasting patterns might not be predicted by looking at the formal institutional structures of federalism alone. Stepan (1999) described Brazil and India as sitting on opposite ends of what he called a demos constraining to demos enabling continuum. By this he meant that institutional features of Brazilian federalism such as the overrepresentation of small states in the Senate, the strong powers of the Senate vis-à-vis the lower house, and the embedding of the allocation of inter-governmental powers in the constitution undermined the ability of the central government to pass reforms on the basis of popular majorities and increased the powers of minority interests. By contrast, India's federal institutions appeared demos enabling because states are represented in the upper house on the basis of their population, the territorial chamber has inferior powers to the directly elected Lok Sabha and important changes can be made on the basis of simple parliamentary majorities. In important ways, India's federal system is relatively centralised in constitutional terms. For instance, the central government has special powers to redraw state boundaries on the basis of a simple parliamentary majority or to suspend elected state governments by imposing 'President's Rule'. Residual power resides with the central government rather than the states. In practice, however, as we have started to detail in this section and will explore further in the following section, what has transpired in Brazil and India is rather different than might be expected based on a reading of institutional structures alone. In Brazil, successive presidents have succeeded in managing presidential coalitions that allow them to pass social policies responding to the interests of popular majorities and that have had broadly redistributive effects. In India, on the other hand, despite the apparently demos enabling characteristics of the Indian constitution, the central government has not been able to override the state governments in the field of social policy. We cannot predict the impact of federalism on policy processes in either country without examining the interaction between formal institutions and multi-level elections.

Centralisation of credit claiming in Brazil
Brazil, historically, was a good example of a federal system in which empowered subnational tiers of government were able to stymie the achievement of national policy goals including welfare provision (Fenwick, 2016, p. 8). In Brazil's 1988 constitution, the Union, states, Federal District of Brasília and the municipalities are jointly allocated the power to 'fight the causes of poverty and the causes of marginalisation, promoting the social integration of the unfavoured sectors' (Article 23X), and education, health and social security are areas where there is concurrent legislative jurisdiction between the federal and state governments. In practice, as a result of the decentralisation that occurred alongside democratisation, state governors were able to capture flows of patronage and exercise much discretion in social sector expenditure. The municipalisation reforms undertaken by the Cardoso administration in 1994 routed more social expenditure, especially in the health and education fields, directly to the municipalities, sidestepping the estados and reducing their fiscal autonomy (Dickovick, 2006). This process supported the federal government in exerting more effective control over social policy, and in claiming political credit for such policies. During Cardoso's administration (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002) and that of his successor Luiz Inácio 'Lula' da Silva (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010), spending on new social policies was scaled up and centralised. This harked back to an earlier period under President Getúlio Vargas (1930-1945; 1950-1954) who was known as the 'father of the poor' because under his rule the first national labour and social benefits were established (Levine, 1998). The flagship social programme Bolsa Família (Family Allowance) is clearly identified as a federal policy and national level politicians have been able to claim credit for it. In the section that follows, we first review recent initiatives by the central government in the social arena and then examine the issue of credit claiming for these policies by national and subnational political actors. Melo (2008, pp. 163-173) describes the extraordinary upsurge in spending on social programmes in Brazil during the presidencies of Cardoso and Lula. Pre-existing cash transfer schemes for non-contributory rural medical, dental and pension services and for the elderly, poor and disabled were supplemented by innovative preventative health care schemes, new funding for basic education, and Bolsa Escola (School Allowance). Bolsa Família, which became the linchpin of the social policy regime under Lula, is a CCT programme begun at the city level (Campinas and the Federal District of Brasília) in the 1990s. It supplemented the incomes of poor families on the condition that they kept their children in school and allowed them to have the mandatory innoculations required by the national health service. During the presidency of Cardoso, Bolsa Escola had been launched as a national programme, and most subnational CCTs were gradually phased out for lack of independent fiscal resources (Fenwick, 2016, p. 63). The Lula administration expanded Bolsa Escola and merged it with three other CCT programmes to create the flagship Bolsa Família (BF). By 2011-2012 Brazil was spending 23.5 per cent of its GDP on social programmes, a relatively high percentage for a middle-income country. Bolsa Família accounted for 0.5 per cent of GDP, unemployment insurance and salary supplements another 0.9 per cent of GDP, public education 5.5 per cent, health 4.5 per cent, and pensions 12 per cent (Almeida, 2013).
Politicians at the federal level, most notably Lula, have been able to claim credit for Bolsa Família and use it to augment their electoral appeal (Niedzwiecki, 2016). The 2010 Brazilian Electoral Panel Survey found that over 80 per cent of people attributed Bolsa Família to the president, which the survey architects found 'even more impressive when one considers that beneficiaries enrol at their municipal government offices' (Ames et al., 2013, p. 36). Lula changed the name of the CCT from Bolsa Escola to Bolsa Família in order to help him claim credit for this policy, even though it built on earlier initiatives.
The fact that the cash transfer is made directly into beneficiaries' bank accounts by the federal government has reduced the involvement of states and municipalities to relatively simple administrative functions, particularly the registration of beneficiaries. Families are eligible for the BF if their income falls below a certain threshold (R$150 per capita per month or about £39 at the rate of exchange of 3.86 Brazilian reais per pound). A representative of the family must present documents to local authorities in order to qualify for the programme. Once the family passes the means test, cash is transferred electronically onto ATM cards issued by the Federal savings bank, the Caixa Econômica Federal. As Fenwick notes, municipalities do see benefits from administering CCTs such as new revenues to cover administrative costs, increased social sector employment in areas that were linked to fulfilling recipients' conditions as well as other economic benefits for the local economy. Such features have enhanced the cooperation of mayors with the federal government (Fenwick, 2016, p. 73).
Because the programme is clearly identified as a presidential initiative and was implemented via the municipalities, it has been argued that it may provide incentives to governors to compete with or otherwise subvert the programme at the subnational level. Niedzwiecki (2016) shows that enrolment in Bolsa Família is lower in states run by governors opposed to the president, because subnational governments can make it harder for the federal government to access databases of potential beneficiaries and/or run their own competing CCTs. However, their ability to undermine the programme may diminish once bureaucratic routinisation sets in, and the systems necessary for cooperation between state and federal authorities for the implementation of Bolsa Família have been put into place. In the state of Goiás which Niedzwiecki highlights, which had its own CCT (Renda Cidadã, or Citizen Income) prior to the creation of Bolsa Família, there is no longer a major competition between the state's programme and the Federal one as was the case in the early 2000s. The two programmes now share informationand the Goiás state programme only covers 40,000 families (Straiotto, 2016), compared to 320,749 families receiving Bolsa Família in the state (Freitas, 2016). Even in its early years, the territorial dispersal of Bolsa Família across major regions of Brazil mapped fairly closely onto levels of poverty suggesting that political factors were not driving its distribution and that the policy was being fairly implemented across the country (Fenwick, 2016, p. 77).

Decentralisation of credit claiming for social policy in India
In India, the constitutional allocation of responsibility for social policy has historically given states a strong role, especially as the primary layer for implementation. Like Brazil, most areas of social policy are concurrent subjects in the Indian constitution, giving the central government and state governments shared jurisdiction. The main exception is public health, which is exclusively a state subject (see Deshpande, Kailash, & Tillin, 2017). State governments, and in the colonial period provincial administrations, have long designed local schemes to respond to local needs or more often crises such as famines or worker unrest. However, as political centralisation has ebbed and flowed, so has the locus of debates about social policy. The first prime minister to seek personal credit for a poverty reduction agenda at the all-India level was Indira Gandhi in the early 1970s, who countered the opposition attacks on her with a promise to 'eradicate poverty' (garibi hatao). Her substantive achievements were quite limited however and welfare initiatives were often at the heart of platforms adopted by regional parties challenging Congress pre-eminence. Examples include mid-day school meal programmes in the southern states of Tamil Nadu and Andhra Pradesh introduced by chief ministers from regional parties. Furthermore, in the context of economic liberalisation, federal decisions to implement 'below poverty line' targeting in the Public Distribution System (through which subsidised foodgrains are distributed) effectively pushed the responsibility for maintaining major subsidy programmes to the states. Since states feared a backlash from voters for retrenchment in the food programme, many found their own resources to maintain or expand existing levels of provision. As Jenkins (1999, p. 182) argued, one reason why the central government was able to cut back food subsidies in the 1990s was because it was 'state governments, the first line of public defence, who must face the irate public. Electorates vent their frustration at the most accessible level of government, not necessarily the one most responsible for their problems'.
It was in the 2000s, as economic growth soared to over 8 per cent per annum and the Congress-led United Progressive Alliance (UPA) came to power on a social democratic platform that India witnessed renewed central government activity in the field of social welfare, with a major expansion of both social expenditure and the scale of government interventions. The thrust of policy goals also changed with an increasing focus on embedding social rights through legislation that underpinned the entitlements of citizens and their right to demand social provision (for an overview see Harriss, 2013;Ruparelia, 2013).
The biggest initiative of the first term of the UPA government, during which it was supported in parliament by left parties, was the passage of the National Rural Employment Guarantee Act (NREGA) in 2005 which in theory guarantees up to 100 days of unskilled employment paid at the minimum wage to any rural household that demands it. The work is conducted on public works schemes designed by panchayats (village councils). The NREGA built on an older Employment Guarantee Scheme in the state of Maharashtra dating back to the early 1970s, but was rolled out on a national scale and became the flagship social policy of the UPA's first term. Initially it covered the 200 most 'backward' districts, but was extended to cover all rural households in 2008. It was primarily federally financed with the central government providing 100 per cent of the funding for wages under the act which are paid directly into workers' bank accounts, and 75 per cent of material costs of the public works schemes carried out under the act. The Congress Party's manifesto for the 2009 Lok Sabha elections included a promise to legislate on a Right to Food. Although the Congress-led UPA won a second term in 2009, progress on social policy making slowed down in the tailwind of the global financial crisis and with growing concern in parts of government about the fiscal implications of further expanding subsidy schemes. However a NFSA, which again expanded the Public Distribution System, was eventually passed by executive order in late 2013, on the insistence of Congress Party President Sonia Gandhi, and retrospectively approved by parliament. 10 The expenditure on social policy by the central government saw a five-fold increase between 1991-1992 and 2012-2013 (Kapur & Nangia, 2015, p. 79) but overall social spending is lower than in Brazil. At its peak in 2009-2010, spending on MGNREGA accounted for approximately 0.6 per cent of GDP (Zepeda, McDonald, Panda, & Kumar, 2013, p. 14), while public health expenditure accounted for 1.4 per cent and education for just 3.7 per cent of GDP in 2014. 11 When designing these new social policies, the Congress Party at the Centre had a clear eye on the potential political dividends they would yield. As a leading architect of the Congress manifestos in 2004 and 2009 admitted, the Congress Party clearly saw the rights-based policies of the first UPA term in political terms, whereas their allies on the leftwhose support for the policies in parliament was importantsaw them more programmatically. In the second term of the UPA, the respondent noted that the NFSA was passed so late that it made it more difficult for the Congress Party to get political credit for it. 12 The party was also aware that credit for social programmes could be claimed by state governments led by opposition parties, especially because state-level governments are the primary layer for the implementation of most social policy.
A concern that opposition-ruled state governments might seek to appropriate credit for NREGA led to its renaming after Mahatma Gandhi in October 2009, a few months after the general election that returned the Congress to power. Indian newspapers widely reported that this initiative was taken to make it more difficult for opposition parties at the state level to steal the credit for the programme. As the Times of India explained: While Congress had political hopes from it, BJP [Bharatiya Janata Party] has been quick to brand NREGA as its own. With state governments being its implementing agencies, [a] few ruled by BJP prefixed their names to it while some CMs [chief ministers] even put their pictures on the job cards. The aggressive BJP bid to appropriate the job scheme sent a wave of concern in [sic] Congress camp. 13 Other schemes of the UPA government, such as the Jawaharlal Nehru National Urban Renewal Mission, Indira Gandhi National Widow Pension Scheme and Indira Gandhi National Disability Pension Scheme were named at their conception after figures from the Gandhi-Nehru Congress dynasty. The National Old Age Pension scheme that had been introduced by the earlier BJP-led National Democratic Alliance central government in 1995 was relaunched under the new name Indira Gandhi National Old Age Pension scheme in November 2007, with a lowered age of 60 rather than 65 for eligibility.
While there is some evidence that, like the Bolsa Família in Brazil, MGNREGA also contributed to the Congress' re-election in the 2009 general elections, the effects were likely on the margins (Manor, 2012). It was state governments that appeared to garner much of the electoral credit. Some state governments worked creatively to rebrand central schemes as state government programmes, or dovetailed national policies with state-level initiatives which complemented or extended the scope of central schemes. In addition, many popular chief ministers introduced their own schemes branded by use of the chief minister's name (such as 'Amma' canteens in Tamil Nadu) or by using the word 'Chief Minister' in the title (such as the Mukhyamantri Khadyann Sahayata Yojna, the Chief Minister's Food Security Campaign in Chhattisgarh which is effectively an extended and over-hauled version of the central government's Public Distribution System).
Election survey data confirms that state governments have been reasonably successful in claiming credit in the eyes of the electorate for social policies that had been initiated by the central government, in stark contrast to Brazil. In the run-up to the 2014 elections, when asked which level of government was responsible for the MGNREGA, 42 per cent of those who had benefited from the programme identified the state government as responsible compared to 27 per cent who identified the central government. The attribution of responsibility to the state governments was even higher in the fields of health and pensions (see Table 2). 14 Thus in India, attribution of policy responsibility for the major social protection programmes is much less clear to voters than for a programme such as Bolsa Família in Brazil. As state governments are the primary layer of implementation for most programmes initiated by the central government and in some instances add to or extend central programmes with initiatives of their own, it is difficult for voters to accurately parse where responsibility for different initiatives lies. The chains of electoral accountability in India's federal system are quite opaque. Furthermore, the importance of state governments in implementing central policies has ensured that there is substantial territorial unevenness in their implementation. Poorer states have long spent less than they are entitled to under federal schemes as a result of lower state capacity (Mathew & Moore, 2011), meaning that central programmes are not well targeted to the poorest citizens. As a wide literature attests there is also substantial variation among poorand richstates in terms of how well centrally sponsored social programmes, such as MGNREGA, are implemented (see, for instance, Chopra, 2015;Deshpande et al., 2017;Drèze & Khera, 2014;Drèze & Khera, 2015;Maiorano, 2014;Roy, 2015;Tillin, Deshpande, & Kailash, 2015).

Epilogue: federal dynamics and social policy in an era of retrenchment
The bulk of this article has considered the dynamics of social policy during a period in which provision was expanded in both countries in the course of the 2000s. But in both Brazil and India, the political and fiscal climate has altered in the 2010s. In this final section we reflect on signs that changes in national regimes in both Brazil and India, and increasing fiscal pressures (especially in Brazil) might alter the federal dimensions of social policy.
The election of a BJP-led government with Narendra Modi at the helm in India in 2014 saw the emergence of a more polarised debate about social policy at the federal level, with some supporters of Modi openly critical of the emergence of what they described as an 'entitlements' culture under the UPA regime. After making disparaging noises about MGNREGA in the early months of his premiership, 15 in the 2016 budget in the face of drought conditions in many rural areas, Modi moved to reassert his administration's commitment to the programme. But there have been signs of a change in the dynamics of claims making and centre-state relations in the field of social policy that could move in two alternative directions. On the one hand, the implementation of the Fourteenth Finance Commission recommendations in 2015 devolved substantial fiscal responsibility for social policies to the states. This gave rise to fears that the central government would retreat from the social sector leading to increased regional inequalities in provision and a 'race to the bottom' as social expenditure is squeezed out by other fiscal pressures within the states. At a minimum, the increase in centre-state fiscal transfers, and reduction in the number of federally financed 'centrally sponsored schemes', will cement the centrality of the states for social policy. On the other hand, the Modi government is making a big push towards its own form of cash transfer (building on initial exercises under the previous government), known in India as 'direct benefit transfers' (DBTs). The government is seeking to unify existing DBTs under one platform, and many would like to see such cash payments replace the in-kind subsidy regime (for a recent overview of these arguments see Joshi, 2016). Modi, an arch centraliser whose image is already omnipresent across government publicity materials, will be well aware of the opportunities to claim credit personally for a unified DBT programme that he could use to establish an unmediated connection to voters across India in ways not dissimilar to Lula in Brazil.
In Brazil, Bolsa Família had enhanced the Workers' Party's image as a progressive pro-poor party, at least until the economic recession that began in 2015 and corruption allegations that targeted senior PT leaders. Anti-corruption investigations and an anti-PT movement against Lula's successor President Dilma Rousseff led to the latter's impeachment in 2016. Under Rousseff's successor, President Michel Temer (2016-present), there has been increasing criticism of the scope of social policy amidst a climate of fiscal austerity and economic liberalisation. The Temer administration has hinted that Bolsa Família includes too many people and should be temporary, even though in its early days it promised to preserve the programme, for fear of the electoral consequences of not doing so. The Temer government also passed a controversial constitutional amendment that froze government spending in real terms for 20 years, and this has already begun to squeeze spending on social programmes (Sátyro 2017). While consensus in Congress in favour of social inclusion is still robust, as detailed by Pereira and Bertholini in this issue, that belief is combined with an opposition to tax increases. Therefore in periods of recession the scale of the commitment to social policy will be reduced. This has happened under Temer, and the trend is likely to continue beyond the 2018 presidential election, given the size of Brazil's federal budget deficit and the conservative shift in Brazilian politics. Whether this opens up new space for subnational activism in the sphere of social policy remains to be seen.

Conclusion
In this article we have demonstrated that the effects of federalism for social policy cannot be read off institutional structures alone. It is important to pay attention to degrees of party system nationalisation and fiscal centralisation in order to understand how federal constitutional settings influence the shape of social policy design and implementation in periods of social policy expansion. We have shown that in Brazil, greater federal initiative in social policy from President Cardoso onwards was made possible by expanded revenues at the federal level, and used particularly by President Lula to secure the prestige of his Workers Party. The vast majority of Brazilian voters see Bolsa Família as a presidential policy, and also typically vote in elections based on their assessment of national candidates and policies. By contrast, in India we demonstrated that much of the credit for the expansion of social provision by the central government was claimed by state governments in the eyes of voters. This took place in a regionalised party system in which states remain the primary level of identification for most voters.
Brazil and India are contrasting cases that illustrate the importance of two neglected features of federalism in shaping social policy: inter-governmental competition and the degree of nationalisation of party systems. In Brazil, where parties are more national than they are in India and politicians at the federal level capture voter attention to a far greater extent than their state counterparts, the expansion of CCTs has largely benefited national leaders, and has reinforced party system nationalisation. In India, by contrast, where state elections have more autonomy from the federal level and parties are more regionalised than in Brazil, social welfare has become an increasingly prominent part of state-level electoral contests. The effectiveness of social policy implementation across states is highly variable and depends upon the initiatives of chief ministers. While regional political autonomy has been a bulwark of federal stability in India helping to prevent internal conflict, it helps to perpetuate territorial unevenness in the functioning of social policy. Notes 1. The view that multicultural policies, which include but are not limited to federal arrangements, restrict the development of welfare states has been challenged (see Banting & Kymlicka, 2006). 2. The attribution of responsibility across areas of social policy is likely to differ.
Whereas survey data clearly shows the large majority of voters identify Bolsa Família as a presidential policy, they attribute responsibility for health care and schools to local governments (Niedzwiecki, 2016). 3. The BJP-led Modi government since 2014 has sought to drive reforms from the top down using the biometric technology platform Aadhar, which signals an aim to move in a more nationalised direction, however the major implementation challenges remain at the state level.
4. In making this observation, we do not necessarily imply that social policies determine election outcomes. Rather we are interested in the process by which voters attribute policy responsibility to one level or the other, and ultimately what the implications of this are for the development of welfare provisioning. 5. It should be noted that the overall tax take remains much lower in India than in Brazil, which has reached near OECD levels of almost 37 per cent tax/GDP ratio. In India, general government revenues increased from 17.73 per cent to 19.72 per cent between 2002 and 2012 (see Maiorano and Manor, this special issue), nevertheless even this was a period of substantial revenue growth in relative terms for India. 6. The concept of 'first order' elections comes from the literature on European politics, in which national elections are seen as having precedence over 'second order' elections to the European parliament (Reif & Schmitt, 1980). 7. In 2016, the federal level of government collected 60 per cent of all tax revenue, with the states getting 23 per cent, and the municipal governments only 17 per cent (Fucs, 2016). 8. The Modi government has proposed moving back to a system of concurrent elections, although the absence of fixed term parliaments in India complicates such a proposal. http://timesofindia.indiatimes.com/india/NREGA-gets-named-after-Mahatma-Gandhi/articleshow/5081715.cms. Accessed August 19th 2016. This renaming departs from an earlier decision not to name the scheme after a political leader. The architects of the original NREGA wanted other parties to embrace the scheme should the UPA lose power (Manor, 2012). 14. A separate survey conducted by CSDS in 2001-2002 asked a similar question about perceptions of public goods provision. It found even higher attribution of responsibility to state governments: when asked which level of government was responsible for different areas of public goods provision, 72 per cent identified the state government as responsible for education facilities, 70 per cent identified it as responsible for medical facilities, and 67 per cent credited it with responsibility for ration supplies (subsidised wheat and rice under the Public Distribution System) (Chhibber, Shastri, & Sisson, 2004, p. 350). 15. In February 2015, Modi described MGNREGA as a 'living monument of [Congress] failures'. He said, My political acumen tells me that MNREGA should not be shut down. I will not make such a mistake because MNREGA is a living monument of your failures. 60 years after independence you still have to send people to dig holes in the ground. So I will celebrate this with pomp and splendour. I will tell the world that these holes you are digging are for your own sins. (Joshua & Sriram, 2015) Disclosure statement No potential conflict of interest was reported by the authors.

Funding
This work was supported by the Economic and Social Research Council (ESRC) under Grant ES/J012629/1