Electoral Politics in Brazil with Evidence from the State of Bahia: State-Led Social Funds versus Federal-Led Social Policies

Transcription

1 Electoral Politics in Brazil with Evidence from the State of Bahia: State-Led Social Funds versus Federal-Led Social Policies Preliminary Version Celina Souza Research Fellow at the Center for Human Resources (CRH), Federal University of Bahia, Brazil Prepared for delivery at the 2009 Congress of the Latin American Studies Association Rio de Janeiro, Brazil June 11-14, 2009

2 Electoral Politics in Brazil with Evidence from the State of Bahia: State-Led Social Funds versus Federal-Led Social Policies * Celina Souza Research Fellow at the Center for Human Resources (CRH), Federal University of Bahia, Brazil Abstract The paper explores the dynamics of continuity and change in subnational electoral politics. It analyzes the role of a distributive policy (Social Funds) implemented by the state government of Bahia in poor municipalities in the survival of a conservative and authoritarian political group in the state of Bahia. The paper tests the hypothesis that one of the electoral strategies for the continuing success of this political group was the concentration of resources from multilateral organizations in small, less urbanized and poorer territories, in which it has always achieved a high percentage of votes in successive elections since the re-introduction of the popular vote in However, in the 2006 state election the group was defeated. The paper asks whether this distributive policy played a role in the group s successive electoral victories and why the group was defeated in The argument put forward is that the defeat of the authoritarian and conservative group in Bahia can be partially explained as a result of the continuous federalization of social policies in Brazil, thus creating incentives for the poorest to switch their vote preferences from competitive and uncertain bids for community projects (Social Funds) to federal-led social policies. Introduction Two decades since the return to democracy, several competitive and popular elections for the executive and improvements in the country s social indicators, in particular in basic education and health care, in some Brazilian states party politics have been less competitive and improvements in social indicators have been slower than the national average. Furthermore, in some states political elites established or consolidated during the military regime continued to gain successive electoral victories both for the state executive and legislature. This has been the tale of the state of Bahia in Brazil s Northeast until the 2006 state election. Although the state of Bahia is used as a narrative in which the social and political promises of redemocratization are far from being reached, it is not a peculiar case. Not only have other Brazilian states experienced low competitive elections but they have also kept political elites consolidated during the military regime in power. 1 In other states the electorate has also kept * Support for this research is provided by the CNPq. I am thankful to Priscila Caires and Sara Fadigas for data collection and Luciano Damasceno, Flávio Fontanelli, and André Greve Pereira for statistical support. 1 Borges (2008) created an index of electoral dominance to measure electoral competition in Brazilian states. According to this index, in the states of Amazonas, Bahia, Ceará, Goiás, Paraíba, Maranhão, and Tocantins only one party achieved electoral success in a period equal to or over three elections. 2

3 politically active politicians with close ties to the authoritarian regime and there are welldocumented studies showing similarities in countries such as Mexico and Argentina. In Brazil, however, strikingly but not surprisingly, most of these political elites lost ground in electoral politics in the 2000s. The paper analyzes the role played by different types of social policies in the electoral results for governor in Bahia over the course of four state elections. The main policy focused on in this paper is a distributive, therefore discretionary policy implemented by the PFL (now DEM), a party to which Bahia s authoritarian and conservative group was affiliated. This policy, known in Bahia as Produzir, is financed by a loan from the International Bank for Reconstruction and Development, also known as the World Bank (WB). With different labels and formats, Produzir exists since the military regime and survives until today. In the early 1990s the WB decided to change the management of Produzir into a Social Fund (SF). SF is a broad label used by multilateral organizations to finance certain projects. Although they vary across countries and receive different labels in each country and in each Brazilian state, they share the following characteristics: (a) funds are made available to communities who choose among a list of projects; (b) project design and construction are decentralized, involving local actors private firms, NGOs, or local governments; (c) community groups are responsible for making contracts, monitoring project execution and after conclusion their operation and maintenance; (d) local contribution is often required, roughly 10 to 15% of project costs (Tendler, 2000: 89). The WB and the Inter-American Development Bank have together spent a total of more than $2.6 billion on SFs in Latin America and Africa since the late 1980s, and European donors roughly half that amount (Tendler, 2000). There is widespread enthusiasm in multilateral organizations for SFs and this enthusiasm is shared by the current government of Bahia. 2 While the main policy focused in on the paper is Produzir, federalized social policies, in particular Brazil s conditional cash transfer (CCT) program known as Bolsa Família (BF), are also discussed. The paper poses two questions. Firstly, whether there is any relationship between a distributive policy since 1993 named Produzir - implemented by the state government in poor rural communities with the financial backing of the WB, and the survival of a conservative and authoritarian political group in the executive and in the legislature in the state of Bahia. This group, affiliated to the PFL, ruled the state during the authoritarian regime and for over three decades after redemocratization until it was defeated in Secondly, why the PFL was overthrown from the state government in 2006 and whether the party s defeat was influenced by the consolidation of federalized social policies. The paper claims that one of the electoral strategies for the continuing success of this political group was the concentration of resources from multilateral organizations in small, less urbanized and poorer territories, in which it has always achieved high percentage of votes in successive elections ( ), with only one intermission in The analysis 2 A press release by the government of Bahia on November 11, 2007 praises Produzir with enthusiastic words on the part of both WB and state officials. The former stated: other countries wish to copy the success of Bahia s program, in particular its decentralized and community participatory features ( The release treats Bahia s SF as if these two features are peculiar to Bahia and not part of every SF. 3

4 of the effects of Produzir on the voting behavior of the poor attempts to explore this claim further, although it will not answer the research questions definitively. As data below suggest, the electorate living in small and less urbanized municipalities and with lower literacy rates has traditionally supported Bahia s conservative elite however it also shows that they have been slowly changing their electoral preferences. This change might point to the importance of policies for electoral results, as suggested by the literature. But why did this happen? The argument put forward is that the defeat of this political group in 2006 for the state executive can be partially explained as a result of the consolidation of federalized social policies. This created incentives for the poorest to decrease their support for the party associated with the policy of Social Fund, Produzir, and to switch their vote preference to PFL s major contender in the state the PT a party seen as committed to income transfers to the poor and to the improvement of their social conditions through redistributive policies. The reason for this, I argue, is that Produzir is an unstable, uncertain policy and too dependent on the interests of local bosses whereas federalized social policies are relatively free from local political interests either because of their universal coverage or clear targeting. Furthermore, Produzir, like all SFs, implies competitive bids among municipalities and among a wide array of community associations whereas federalized social policies are either universal (e.g. health care) or have clear targets (BF and retirement entitlements, for example). But how can we assert that distributive or redistributive policies influence the poorest vote preferences? At this point it is important to clarify this paper s modest ambition. It does not make straightforward connections between politicians policy choices and voter preferences nor does it follow the avenues opened by Mayhew s (1974) landmark work on electoral connection or the contributions of the literature on how partisan differences lead to different distributive paths in order to benefit particular constituencies. 3 It also does not discuss the important contributions of the literature on electoral targeting or, more specifically, the debate between the core voter model versus the swing voter model (Dixit and Londregan, 1996, 1998; Cox and McCubbins, 1986; and Lindbeck and Weibull, 1987). Although, as stated by many, connections between voters, elected officials, and policies matter considerably to the study of voting behavior, there are several constraints mainly data availability and methodological obstacles that stand in the way of developing appropriate empirical analyses. 4 Despite the existence of several studies showing the linkages between policy preferences and actual behavior of voters in elections and statistical results suggesting that policy preferences shape the behavior of voters, it has also been acknowledged that (1) a wide variety of preferences operate simultaneously and (2) several other political and socioeconomic factors also influence voters decisions. In the light of this, this paper does not search for a connection between distributive and redistributive policies Produzir vs. federalized social policies - and electoral results by applying econometric techniques alone, rather it provides a more detailed and qualitative view 3 For a review of this literature and its empirical application to policies in Argentina, see Calvo and Murillo (2004). 4 Brooks (2006) discusses several new research programs that have advanced scholarly understanding of these linkages providing insights into the understanding of the micro-foundations of voting behavior, coupled with new analytical models of aggregate preferences. 4

5 to this connection. Qualitative contributions of this kind have been made by scholars analyzing the dynamics of subnational politics and policies in Latin America. Tendler (1999), in her paper on SFs in Brazil s Northeast, acknowledges that the pattern of association between vote-seeking and SF expenditure is richly complex. The intensity of the political courtship through SFs varies from one period to the other, depending not just on the electoral cycle, but on other factors like the strength of each opposition party at a particular moment and how much of a challenge it represents, on whether the elections are midterm or not, and on the balance struck between rewarding those who vote or voted in favor, withholding from those who vote against, and courting the fence-sitters. In a similar vein, if one associates the design and the management of Produzir and of federalized social programs to what could be labeled as traditional and modern policies, one might be able to investigate territorial-level voting by aggregating the electorate according to the socio-demographic characteristics of their municipalities in association with policy preferences. Research using this association, both quantitatively and qualitatively, has shown the existence of a complementarity between modern and traditional politics and policies. Gibson (1997), for instance, has convincingly shown how populist parties in Mexico and Argentina have been able to build a national-level strategy based on this complementarity, which he labels metropolitan and peripheral sub-coalitions, to overcome constraints on the adoption of policy reforms. Following Gibson s typology, Souza (2007) has discussed the policy of hard budget constraints in three Brazilian states and shown that this policy was made possible because in some states their political coalitions were able to keep their territorial bases of electoral support both in the metropolitan and in the peripheral sub-coalitions. Souza (1997) also analyzed the strategy of Bahia s conservative elite for consolidating and furthering the policy of industrialization and economic modernization in Bahia by isolating the bureaucracies operating planning and finance portfolios from those in charge of social policies. However, and as Tendler (1999) reminded us, these findings are neither novel nor unique to the so-called developing world. Although there is strong evidence showing this complementarity in certain issues, this paper analyzes a case of competition between political parties - the PFL and the PT - in a given territory the state of Bahia - where they have been sponsoring and implementing different formats of social policies: on one hand, Produzir, a distributive and discretionary program, and, on the other, federalized and redistributive social programs designed as universal or clearly targeted. Despite the importance of subnational politics and policies in a federal and redemocratized country, scholarly work on state politics, institutions and policies after redemocratization are in their infancy in Brazil in comparison to national and local politics and policies. Only very recently scholars have started to look at the states in such a way that can contribute to a cumulative research agenda. 5 Filling this research gap is important, in particular for testing institutional hypotheses given that the states in Brazil are characterized by a high level of institutional and policy uniformity but also by different outcomes as regards electoral, 5 Diniz s (1982) pioneering work on the political machine of a state governor of Rio de Janeiro in the 1970s did not open the way for further inquires until the late 1990s. 5

6 political and policy results. 6 This paper also goes beyond previous studies by discussing the possible reasons for the change in the pattern of voter preferences among the poorest in an electoral district of large magnitude the state of Bahia by confronting distinctive features of social policies. It also makes an attempt to link the role played by subnational elites and federal institutions and policies. 7 Furthermore, most studies linking voter preferences and policies focus on national elections, whereas this paper investigates a subnational setting. To support the claims above the paper makes use of the following tools. Firstly, it maps the existing literature on SFs, in particular those focused on the experiences of the Northeastern states of Brazil. Secondly, interviews were carried out with government officials and statelevel politicians in 2006, soon after the PFL s defeat. Thirdly, electoral results are correlated with socioeconomic and demographic variables. This latter tool allows for the analysis of the relationship between electoral results, i.e. changes in the electoral bases in each of the four state elections for governor 1994, 1998, 2002 and and the selected socioeconomic and demographic variables. Fourthly, econometric models are applied to test the effects of certain variables, mainly Produzir funds on candidates running for the state executive in the 2002 and 2006 elections. The paper is organized as follows. Section 1 analyzes the socioeconomic and political contexts of the state of Bahia in recent decades. Section 2 introduces the discussion about the evolving trajectory of distributive and redistributive policies and, more specifically, the electoral influence of state-led versus federal-led social policies. The main focus of this section is Produzir. Section 3 presents data and statistical analyzes on the impact of Produzir and BF on the percentage of votes received by candidates for governor in 2002 and The final part concludes. The state of Bahia: Socioeconomic and political contexts Socioeconomic background As it also happens in several developing countries, the trajectory of Bahia s socioeconomic performance presents a paradox: since the 1970s Bahia has been recovering from a long period of economic stagnation which started at the end of the 19 th century, to reach the 6 th position in the states ranking of GDP, a position maintained ever since, whereas social indicators have lagged behind: in 2000, Bahia ranked 22 nd on its Human Development Index (HDI) and in 2003 it ranked 17 th on GDP per capita. Table 1 displays how Bahia is below the national average in every indicator. 6 If there are differences in political processes and policy results, a study by Sátyro (2006) using different statistical techniques showed that as regards social spending there are no differences among the states, meaning, that variables such as ideology, party affiliation, electoral competition, coalition governments, previous policies and social, demographic and economic conditions do not make any significant difference in the patterns of social spending in Brazilian states. This is because, I argue, of the high degree of federalization of social policies in Brazil. 7 Mccoy and Stroupe (2001) suggest that there is an endogenous and ambivalent relationship between political elites and federal institutions generally neglected in studies of federalism. 6

7 Table 1 Bahia and Brazil: Socioeconomic indicators Indicator Bahia Brazil Poverty rate (2000) Income inequality index (income of the richest 10% / income of the poorest 40%) (2000) Illiteracy rate (population > 15 years old) (2007) Rate of urban households with sanitation (2007) GDP per capita (in R$ of 2006) 4,111 7,530 Sources: IBGE ( Ipeadata ( If one takes the human development index by municipality (HDI-M - Human Development Index-Municipality), although Bahia has advanced two positions in the last decade, its municipalities still register one of the lowest positions in the ranking, 22 nd, meaning that municipalities of 21 states are better positioned and only five are worse than Bahia. This applies to all HDI-M sub-indicators: income per capita, literacy, and longevity. In order to advance the knowledge on Bahia politics, it is also important to show its demographic pattern. This is because Bahia still has a high rural population as compared to other states (Table 2). Furthermore, Bahia has only one city, except Salvador, with more than 500,000 inhabitants, Feira de Santana. Indeed, 309 of Bahia s 417 municipalities have between and inhabitants and only 11 between and inhabitants ( Table 2 - Bahia: Urban and rural populations, Total population Urban Rural Urbanization rate - Bahia (%) Urbanization rate Brazil (%) ,070,250 8,772,348 4,297, ,867,991 7,016,770 4,851, Sources: Atlas do Desenvolvimento Humano ( and IBGE ( On the economic realm, however, Bahia is pointed out as an economic success. Bahia s economy was, in 1975, strongly concentrated on agricultural activities (30.7%) and the manufacture of intermediate industrial goods accounted only for 14.7%. In 2005, however, the former represented 10.3% of the GDP and the latter 36.1%, being the state s main economic activity. Financial activities, strong in the 19 th century, represented close to 15% in the 1990s and have been declining steadily, reaching 3.3% in The state is also loosing position in services, from 8.3% in 1975 to 6.2% in These figures show the success of a policy of leveraging the state s economic position through industrialization. Freeing Bahia from economic stagnation is a rich and interesting tale but beyond the objectives of this paper. The success of the strategy to manufacture capital-intensive intermediate goods and lately a car manufacture can be seen in the state s GDP as 7

8 compared to other states in the Northeast (Table 3). This strategy is responsible for GDP s positive performance since the late 1970s, most of the times above the national average. 8 Table 3 Northeast: GDP participation, 2001 Bahia Sergipe Alagoas Pernambuco Paraíba Rio Grande do Norte Ceará Piauí Maranhão , Bahia is also appointed as one of the most successful cases of fiscal adjustment. It is important to note that despite implementing a severe fiscal adjustment after 1997, which involved several privatizations and a reduction in payroll expenditure, the PFL won its least contended victory, in 1998, as discussed elsewhere (Souza, 2007). Summing up, Bahia is a state in which a less uneven distribution of public goods, both territorially and socially, has not yet come close to the expectations brought about by redemocratization, but it is a case of economic success as compared to other Brazilian states. Economic achievements are the result of a well-planned strategy pursued by its political elites since the mid-1950s. Bahia s political dynamics Bahia s recent political history is intertwined with that of Antônio Carlos Magalhães (ACM) who governed the state or supported its governors for the last 25 years, with few intermissions. After the political opening, ACM joined the PFL, which later had its name changed to DEM. He governed the state twice during the authoritarian regime and in 1990 he was elected by popular vote. If he was not governing the state, he occupied appointed federal offices both during the military and in the democratic regimes. In the post-authoritarian period and prior to 2006, the PFL suffered only one electoral defeat, in ACM died in 2007 while he was in the Senate. Table 4 lists the names and the parties of state governors during the military and the democratic regimes. Table 4 - Bahia Governors by term and party, Governor Term Party Luis Viana Filho ARENA Antônio Carlos Magalhães ARENA Roberto Santos ARENA Antônio Carlos Magalhães ARENA/PDS João Durval Carneiro PDS Waldir Pires /Nilo Coelho PMDB Antônio Carlos Magalhães PFL Paulo Ganen Souto PFL Cesar Borges PFL Paulo Ganen Souto PFL Jacques Wagner PT Notes: (1) Election by popular vote was reintroduced in the states in (2) Two rounds of voting when no candidate achieved a majority was introduced in Data on economic and social indicators of Bahia are available at 8

9 As the table shows, in only two elections 1986 and 2006 was the PFL overthrown from the state government. The reasons for its defeat in 1986 have been analyzed elsewhere (Souza, 1997) but two stand out. Firstly, an external-national factor: the election was influenced by the popularity and the overall, albeit temporary, success of a stabilization plan named the Cruzado, which was credited to the PMDB, a party to which the elected governor, Waldir Pires, was then affiliated. 9 Secondly, a domestic factor: opposition parties were able to put together a wide coalition and this coalition was able to name a credible candidate. As shown below, both reasons prevailed in the 2006 election, although the external factor was no longer a stabilization policy, rather, as I argue, the consolidation of federalized social policies. The following table shows the electoral results for the governorship in the 1994, 1998, 2000, and 2006 elections. 10 Bahia is home to Brazil s 4 th largest electorate, with 9,109,353 registered voters in Table 5 - Bahia: Results of the 1994, 1998, 2000 and 2006 election for state governor Year Candidates and Null/Void/Abstentions Party Number of Votes % 1994(1) Paulo Souto João Durval Void/Null Abstentions 1998 Cesar Borges José Ribeiro Others Void/Null Abstentions 2002 Paulo G. Souto Jacques Wagner Others Void/Null Abstentions 2006 Jacques Wagner Paulo G. Souto Others Void Null PFL PMN PFL PT PFL PT PT PFL 2,235,659 1,577,043 2,418, , ,297 2,871,025 2,057, ,895 1,052,887 2,168,773 3,242,336 2,638, , , ,875 Abstention 1,884,249 Notes: (1) There were two rounds of voting because no candidate achieved a majority in the first round. Source: Tribunal Superior Eleitoral ( Table 5 shows that the increase in the number of votes of what became PFL s main opposition party in Bahia, the PT, started in More importantly, the figures above show that PFL s candidates for the state executive have traditionally maintained a relatively 9 The main evidence of this is that the PMDB elected 22 governors of the then existent 24 states. 10 Data on elections previous to 1994 are not available on line. In the 1990 election, when ACM was elected for the first time by popular vote, he won with 30% of the votes and there was no second round because the rule was only adopted in the following election. 9

10 constant number of valid votes, around 2.5 million. The stability in the number of votes indicates that the PFL has managed to sustain a captive and faithful electorate. Unlike other similar cases of one party control over politics in democratic regimes, figures above demonstrate that the PFL never achieved easy or unchallenged victories, with the exception of the landslide election of The fact that PFL s candidates had always captured around 2.5 million valid votes raises two questions: (1) what sectors of the electorate had been voting for the PFL and (2) are they territorially concentrated? Table 6 Votes for candidates and parties by municipal size (% of valid votes)) Municipality Size (inhabitants) Souto (PFL) Durval (PMN) Borges (PFL) Ribeiro (PT) Souto (PFL) Wagner (PT) Souto (PFL) Wagner (PT) Up to 10, ,001 to 20, ,001 to 50, ,001 to 200, Above 200, Sources: Tribunal Superior Eleitoral - TRE; IBGE 1991 and 2000 Censuses and Population Calculations for 2002 and 2006 Table 6 shows that in 1994 there was no relationship between the percentage of votes for the candidates and the size of the municipality. In 1998 there is a slight relationship, with PFL s candidate performing better in middle- and small-size municipalities, and votes for the opposition candidate more concentrated in more urbanized municipalities. In 2002 Souto s votes declined as the size of the municipality increased. In 2006 Wagner increased his percentage of votes in municipalities of all sizes when compared to 2002 but the most dramatic increase in the percentage of votes given to the PT in 2006 was in smaller municipalities, i.e. those with up to 50,000 inhabitants. As the table shows in this sub-group the PT s candidate increased his percentage of votes by almost 80%, with the highest increase in the sub-group made up of municipalities with up to 20,000 inhabitants. Statistical analyses also confirm the importance of small municipalities to the PFL s victories and the relative fall in PFL s support from voters from small municipalities (up to 50,000 inhabitants) between 1998 and It also shows that the party s control of votes in small municipalities was achieved in the 1998 election given that the percentage of votes to the PFL in 1994 was much smaller than in 1998 and The table also reveals the decrease in PFL votes in small municipalities after 1998 (Table 7). 10

11 Table 7 Percentage of votes for PFL s candidates and size of municipality Inhabitants Number of municipalities pop2000 total % Pop Tot % of votes to PFL sout06 sout02 borg98 sout94 Up to 10, ,0% ,001 to 20, ,3% ,001 to 50, ,3% ,001 to 200, ,2% Above 200, ,1% Total 415* 13,039, % *Two municipalities created in the early 2000s were excluded Table 8 shows the variation in the percentage of votes won by the PFL between elections. It also displays how the party started to lose support in 2002 in middle-size and in large cities but the greatest variation found in 2006 was among municipalities with fewer than 50,000 inhabitants. Table 8 Variation in the percentage of votes for PFL candidates by group of municipalities Group Var94-98 Var98-02 Var02-06 Up to 10,000 40,7% -13,6% -23,3% 10,001 to 20,000 44,5% -12,6% -23,1% 20,001 to 50,000 49,3% -17,2% -23,9% 50,001 to 200,000 52,9% -31,1% -22,8% Above 200,000 29,1% -35,4% -6,8% Graph 1 introduces other demographic data the percentage of urban population over the total population in relation to the percentage of votes for PFL candidates. As in Table 6, no relation was found in 1994 but in the following elections it is in municipalities with less than 50% of urban population that the PFL reached absolute majorities, 1998 being a landslide victory when 76.5% of the electorate of municipalities with an urbanization rate of less than 25% voted for Borges. 11

12 Graph 1 Percentages of votes for PFL s candidates for governors and urbanization rates 1994, 1998, 2002, % of votes < 25 = 25 - < 50 = 50 - < 75 =75 - < 95 >= 95 Urbanization Rate A common assumption among political analysts is that voters living in municipalities with lower HDI would support PFL candidates. To test the assumption, municipalities from Bahia were divided into five groups: Group 1 encompasses 79 municipalities with HDIs lower than 0.50; in Group 2 there are 84 municipalities with HDIs between 0.59 and 0.61; in Group 3, 73 municipalities with HDIs between 0.61 and 0.63; in Group 4, 89 municipalities with HDIs between 0.63 and 0.655: and in Group 5, 90 municipalities with HDIs above Graph 2 displays the average number of votes for PFL candidates in each of these groups. As expected, PFL candidates were electorally stronger in municipalities with lower HDIs, particularly in 2002 and Nevertheless, and as shown below, the overall electoral results are far from being explained by HDIs. Graph 2 - Percentage of votes for PFL governors and HDIs % of votes HDI Groups Despite the relative decline in PFL electoral support after 1988, the party remained in office until Some reasons to explain the electoral success and the longevity of the PFL in 12

13 Bahia have become the focuses of political analysts. Most concur that its success lies in internal characteristics of the party in Bahia. Many saw this political group as an oligarchy (new and old) and/or as playing the political game only upon patron-client relations, local bossism, pork-barrel, etc. These analyzes, however, are acknowledged by some (Dantas Neto, 2006 and Souza, 1997) as an oversimplification of a much more complex political phenomenon. Furthermore, many analysts tend to credit the PFL s successive victories to the personage of ACM alone leaving aside what others see as a style of doing politics and of governing that no longer requires the existence of the personage (Dantas Neto, 2006; Souza, 1997). This is why it is important to investigate whether policy strategies used by ACM and his group (the carlistas) first and the PFL later have played a role in the group`s electoral victories and in keeping a captive and steady electorate. In the light of the descriptions and analyses above I now return to the research question above: what was the PFL s strategy for keeping a steady number of absolute votes? Or better, was there a policy that helped to keep the loyalty of its electorate? And more, why did this electorate shift its vote to the PFL s main contender, the PT, in 2006? These are the foci of the following sections. Distributive and redistributive policies for Bahia s poorest The analysis of social policies in connection with electoral results to a certain extent follows Pierson s (2004: 31) claim that a fundamental feature of politics is its focus on the provision of public goods. Although none of the policies analyzed here can be labeled as pure public goods, neither are they private goods. As we know, between the two extremes are many shades of grey. Both Produzir and the federalized social programs discussed here can be classified as club goods, which are enjoyed by members of a certain subset of the population, such as poor communities receiving Produzir funds and those entitled to be the beneficiaries of universal or targeted federal programs. 11 Even if these policies can be said to fall into this category of goods, their design is quite different. This difference allows politicians room for maneuvering the selection of beneficiaries of Produzir and to use the program for their electoral survival, as discussed below. The use of distributive policies for electoral survival is by no means odd because, as we know, politicians play strategically and are goal-seekers. However, one of the discussions here is that the PFL did not foresee the increasing importance of federalized social policies and kept its electoral strategy to attract the votes of the poor by pursuing its traditional distributive policy. Because of the party s lasting control of the votes of the poorest, they relied on a distributive mechanism alone to fulfill their expectations and preferences, ignoring the leverage of federalized social policies. Furthermore, voters did not need to fear an end to Produzir resources because they are attached to a WB loan and therefore not threatened by changes in the governing party. In that sense, Produzir is not a typical discretionary policy as those analyzed by the literature on the political influences of discretionary funds given its forced continuity. However, communities receiving more resources than others could lose but this might be compensated by the tendency of mayors of small municipalities to switch parties to get closer to the winning party coalition at state level, as discussed below. 11 Some projects in Produzir can be classified as local public goods because they can be used by the entire population of the municipality. However, the list of projects financed by Produzir shows that most are restricted to the receiving community. 13

14 Produzir In Bahia there are currently 16 programs financed by multilateral organizations destined to rural areas and small municipalities. Of these, the most important in the number of municipalities reached and in the amount invested is Produzir, which, since 1993 has invested resources in rural areas and in urban centers with under 7,500 inhabitants. Today it reaches 407 municipalities, 450 thousand families and US$ million have been invested through a WB loan. Produzir has been the object of several loan agreements and it is now in its third phase, which expires in The first phase covered 190 municipalities but in the following phases and starting in 2002, the number of municipalities increased dramatically. It might not be a coincidence that the increase in the number of municipalities occurred when state elections became more competitive. Produzir is the continuation of several other loans from the WB destined to rural development programs in Brazil s Northeast. A 1995 report by WB staff summarizes the program s trajectory as follows. Since the early 1970s federal and state governments and the WB have been involved in financing schemes to alleviate rural poverty in the Northeast. They started in the 1970s with Polonordeste projects in each of the nine Northeast states and northern Minas Gerais. Nevertheless, the evaluation of Polonordeste by the WB in the early 1980s showed that, in the aggregate, they had had a limited impact on poverty. The microregions supported by the WB projects were relatively small areas in the context of the whole Northeast; poverty targeting mechanisms were weak; and delivery mechanisms relied exclusively on federal and state bureaucracies. These considerations, according to the report, prompted the WB and the federal government to change the program in the mid-1980s. As a result in 1985 the Northeast Rural Development Program (NRDP) replaced Polonordeste. However, with the exception of one component supporting small community projects (known as the APCR), implementation of the NRDP lagged behind schedule and performance in meeting basic objectives was weak. Some of the main causes were strikingly similar to the causes of the poor performance shown under the earlier Polonordeste projects. One of the problems identified in the report is that decision-making, administration and financing arrangements were generally non-participatory, and the sense of project "ownership" at both state and local levels was limited. On the other hand, the innovative APCR communitybased components, which relied on local planning and implementation, stressed community organization and contracted out needed technical expertise, were noticeably more successful. In 1993 on the WB's initiative the NRDP was reformulated and transformed into a community-based development program drawing both on the experience of the APCR small community projects and on lessons learned with similar schemes elsewhere in Latin America, particularly the Mexican Solidaridad program. Characteristic of WB all reports, this one also praises the demand-driven and participatory features of the program, ignoring both qualitative and quantitative evidence revealing its several problems. As one of my interviewees put it, WB officials carry on with their typical behavior: we like to be deceived (this is a rough translation of a saying in Portuguese: me engana que eu gosto). Another major change is that the state government has become fully responsible for counterpart funding. Produzir s stated objective according to the WB is to increase social and economic opportunities for the rural poor by improving their access to basic socioeconomic infrastructure. The project supports community matching grants for investment in about 14

15 2,500 small scale socioeconomic infrastructures, education, health, cultural, environmental and other productive projects aimed at raising the HDI. Loan agreements also assign resources for technical assistance and training to support the mobilization and strengthening of community associations and capacity-building for municipal councils. The most common investments funded by Produzir in Bahia are for water provision (wells), energy, bridges, tractors, sewerage - mainly toilets -, and grain mills. Non-agricultural small businesses such as bakeries, handcraft, and laundries have also been funded (Leite et al., 2007). According to the design of SFs, the decision making about which projects are to be funded is granted to members of poor communities gathered in a community council known in Bahia as FUMAC. FUMAC is made up of community representatives, members of civil society and the municipality. A survey by Leite et al. (2007) shows that on average each council is made up of 20 community organizations but few have more than 40. As will be discussed later, this empowerment of the community as decision makers has proven to be difficult to put into practice. SFs in general and Produzir in particular have been subjected to several evaluations by multilateral organizations. 12 An audit of Produzir was carried out in 2005 by the State Accounts Tribunal based on several sources including fieldwork in 62 rural communities of 13 municipalities. 13 As far as I know, three studies by academics have focused on SF experiences in the Northeast: one by Judith Tendler (1999; 2000), covering four states - Ceará, Maranhão, Bahia, and Pernambuco -, one by Arretche et al. (2006), discussing the functioning of community councils in Bahia, Ceará, and Pernambuco, and another by Leite et al. (2007), focusing exclusively on Bahia. 14 Another two studies by Schady (1998) and Graham and Kane (1998) focused on the Peruvian experience of SFs. A brief summary and discussion of the main findings of these studies are presented below. Tendler (2000; 1999) convincingly argues that evaluations carried out or funded by the multilateral organizations themselves provide more grounds for skepticism than for enthusiasm. The focus of Tendler s comprehensive analysis of SFs in general and of their performance in the Northeast is the decision making role of community councils coupled with the demand-driven feature of SFs. In her own words, what was this community that, in the SF descriptions and donor literature generally, seemed to be speaking with a single voice about the kind of project that would serve its interest best? Wouldn't there normally be divergence of opinion? How was that settled? The evaluation literature on SFs and demanddriven approaches does not deny such divergence, but is almost silent on how it is resolved. Nor does it explore whether the ultimate choice and the process by which different views were transformed into that choice, enhances or sets back the public good, at least as embodied in the stated goals of such programs. This is partly because the approach assumes, as so much of this literature states, that a choice made by "the community" is better than one made by "the government." 12 Tendler (2000) provides a throughout list of these evaluations. On the specific evaluation of the program in the Northeast, see Pozzoni (2006). 13 The audit report is available at 14 There was also an evaluation carried out by a team from the University of Campinas. However, I was unable to obtain a copy of the report, despite several attempts. 15

16 Tendler (1999) found out in her fieldwork that many association officials and other community members frequently reported how others had made the choice for them or had forced a particular choice. These other actors fell into three categories - private firms (design firms, building contractors, and equipment suppliers); politicians (mayors, state legislators, and governors, as well as community leaders); and state government staff. Tendler discusses the roles played by each actor in detail. As regards the role of politicians, she shows that politicians also shaped and limited community demand - mayors, city-council members, and ward heelers, as well as state and national legislators. Community leaders, or small powerful factions within communities, also made or induced choices without consulting the community. Sometimes the mayor (or legislator) even imposed his choice over the opposition of a community association, thereby marginalizing it, or brought over the association president to his side, or simply circumvented any associations altogether. State legislators interviewed by Tendler and by myself at different moments candidly acknowledged that SF projects that they "arranged" for their constituents were crucial to get them elected. Tendler concludes that an induced or imposed decision does not necessarily mean that they were not the best in terms of some vision of the public good. Rather, the decisions simply had more in common with supply-driven than demand-driven approaches. Whether this produced better decisions and outcomes on average is not clear - and probably depended on factors not fitting into the dichotomy of demand- vs, supply-driven. Tendler s general conclusion is that there is substantial evidence that the SFs are richly embedded in politics, from the local to the national level, and often an old-fashioned clientelistic politics at that. A survey by Arretche et al. (2006) supports Tendler s (1999) main findings. Comparing the functioning of SFs community councils in Bahia, Ceará and Pernambuco, the survey shows that a great number of mayors also acted as heads of councils but in Bahia the number was much higher. The study puts forward two arguments. First, the presence of mayors is a rational choice by councilors given their material and political resources to pressure state legislators voted for in their municipalities who then also put pressure on the governor to release funds. The argument is that this strategy is a result of the formal and informal rules guiding distributive programs designed as participatory. In other words, SFs by their very nature are likely to create incentives for the leadership of actors rather than the communities. However, one aspect deserves further consideration. Why this is stronger in Bahia? My view is that the answer lies in the lasting control of one political group and its dependence on voters direct or indirectly receiving funds now, or expecting to receive funds in the future, hence the need for stronger control of councils as compared to the other two states. The second argument put forward by Arretche et al. (2006) is that councils are not insulated from party-politics given politicians role in selecting municipalities and communities and in releasing funds. This claim requires a more nuanced interpretation. I argue that the dependence of councils on party-politics is rather a matter of local political dynamics and not of party-politics per se. I base this claim on evidence of party switching, quite a common feature in Brazilian small and poor municipalities. In the case of Bahia, in the 1996 municipal elections the PFL and small parties of its coalition won 75% of municipal executives, in %, in %. However, data collected by Borges (2008) shows that in mayors of 278 affiliated to the PFL or to its coalition changed parties, most of them to the PMDB, a party of the coalition of the winning candidate for the state executive. As I have argued elsewhere (Souza, 1997) and a decade ago, mayors of small and 16

17 middle-size cities are marked by pragmatism and although they might be elected or affiliated to the PFL or to its satellite parties they can betray their parties if they foresee the victory of a candidate from the opposition. Analyzing Brazilian elections between 1945 and 1964, Lavareda (1991) has also shown that voters in large cities exhibited fairly stable partisan identification but in the interior parties formed opportunistic alliances in local elections with ideologically noncontiguous parties. Despite anecdotal and empirical evidence showing a trend of change at the national level, politicians of small and poor municipalities are still playing the old game. Hence, the claim that community councils are not insulated from partypolitics should be reformulated: they are not insulated from parties (any party) governing the state. If there is some anecdotal and empirical evidence that political parties are becoming more associated with the policies they implement and that the electorate, poor or otherwise, are tending to take decisions more based on party policies at the national and the state levels, the same cannot be said when it comes to local political leaders governing small municipalities. 15 It is important to note that neither in the survey by Arretche et al. (2006) and by Leite et al. (2007) nor in my own interviews were federal legislators said to be related to the functioning of Produzir. This stands in contrast to Tendler s (1999) findings in which federal legislators were still said to influence decisions. This signals important changes in less than a decade and confirms one of the conclusions in the survey by Hagopian et al. (2009): the emergence of national legislators more oriented by their party policies and less by distributive benefits to their constituencies. The study by Leite et al. (2007) was commissioned by the government of Bahia to map out the performance of all rural development programs focusing on social inclusion. Their main findings are the following. First, investment decisions about projects, communities and municipalities are subjected to strong politico-electoral influences. This, they argue, has other consequences besides the maintenance of clientelist relations. It allows for badly conceived projects and for wrongdoings in resource allocation because of lack of transparency and monitoring. There were also cases in which the state agency decided on 15 A survey with federal deputies conducted in 1999 and 2000 by Hagopian et al. (2009) showed that the value of patronage [sic] as an electoral strategy declined during the 1990s and that among the factors politicians believed most contributed to their party s electoral success at state level were the party s image whereas one of the least important factors was the resources the party delivered to the state in the preceding electoral cycle and those anticipated by voters in the future. Using a different research method, Desposato and Scheiner s (2007) study on party affiliation and party switching in Brazil between 1991 and 1998 focusing on national and state legislators, the latter from the states of Bahia and São Paulo, argue in a different direction: that pork [sic] remains as a substitute for policy proposals. Although discussion about party switching is not central to this paper, two issues are worth mentioning. First, the controversial conclusion of these two studies probably as a consequence of the use of different research methods. Second, although I concur with the claim by Desposato and Scheiner (2007) that for state legislators, the state governor is the dominant player, I doubt whether party switching in the case of Bahia in that period was driven by pork [sic] alone. This is because exactly in that period an important key for understanding party switching among state lawmakers was the strategy used by PFL s leader, ACM, to accommodate disputes among local politicians. The strategy was to make the party s most loyal state politicians leave the PFL and to affiliate to what I have called PFL s satellite parties (PL, PTB, PP etc) in order to accommodate fierce disputes among local politicians. Therefore, in the case of Bahian legislators, pork [sic] was not the incentive for party switching, at least among party members from the right of the political spectrum, which does not mean that they are not interested in distributive resources accruing from the state government. This highlights the importance of combining different methods of political analysis and of putting the results of quantitative analysis into context. 17