WORKING PAPER. Gustavo Andrey de Almeida Lopes Fernandes 2 Fundação Getúlio Vargas EAESP Departament of Public Management

Transcription

1 The limits of horizontal accountability in Brazil: transparency of oversight institutions of subnational governments in the largest country of the south hemisphere 1 WORKING PAPER Gustavo Andrey de Almeida Lopes Fernandes 2 Fundação Getúlio Vargas EAESP Departament of Public Management King s College London Visiting Researcher King s Brazil Institute ABSTRACT Participation and social control of the State by civil society are seen as fundamental to the improvement of democracy. In newly established democracies, horizontal accountability tends to be more fragile than vertical accountability. In Brazil, Federal and state legislation maintained almost the same institutional design of oversight institutions from dictatorial periods, despite the huge progress towards democratization. In this paper, we investigate how the head of the state executive is held accountable, based on the functioning of those institutions. Our findings show significant evidence of a limited horizontal accountability. Oversight institutions are not a reliable supplier of information about the government for the Brazilian Society, lack of transparency is disseminated along these institutions. All in all, income inequality and technical expertise seem to play the main role to explain lack of transparency. 1 This working paper is part of a large research Agenda jointly done with Gabriela de Brelaz, Marco Teixeira, Ivan Filipe and Luciana Elmais. Part of research was done during my stay at King s College London as a visijng researcher in the Brazil InsJtute. I am very thankful for all the assistance. 2 For a more recent version, please me at gustavo.fernandes@fgv.br

2 1. Introduction Latin American countries seem to be embroiled in corruption and inefficiency. In Brazil, scandals of corruption and poor provision of public services have been found in every level of government. According to Transparency International, Brazil ranks the 64 th position in the Corruption Perception Index. Lately, recurrent cases of overpricing in public procurement as well as political involvement in illegal activities led to a feeling of distrust in the political system (Hagopian, 2016). Large demonstrations since 2013 and in the following years underlined the decrease of social confidence in the public sector, including main figures in Executive, Congress and political parties (Mello, 2016). Clearly, despite political and economic stability after democratization in 1988 oversight mechanisms undeniably need to hold the public sector more accountable. Hence, the objective of this paper is to analyze how Tribunais de Contas in Brazil are transparent, since they are the key institutions to hold subnational politicians accountable. To understand how they work as well as how they deal with the disclosure of information is an essential investigation to grasp why inefficiency and corruption are still widely seen in the country. We pay special attention to how the chief of the subnational Executive is held accountable. Governors have been called in the Brazilian literature (Abrucio, 1994), the Barons of the Federation, and have huge political power inside de state. In São Paulo, for instance, the same party (Partido da Social Democracia Brasileira PSDB) has dominated the political

3 life of the state for more than 20 years. Mayors are also audited but governor are the key players when the members of Tribunais are chosen. In short, what is behind this analysis is to understand the aspects of oversight institutions as mechanisms of horizontal accountability in the subnational level, a prerequisite for democratic institutionalization. On a broad perspective, our aim is to shed some light on the quality of democracy in the 27 states of the country. As it will be shown in the next sections, according to Brazilian experience, vertical and horizontal accountability can be highly uncorrelated from each other if we consider national and subnational levels of democratization. The importance of this research object is related with the priority given by Brazilian society to the problems concerning the lack of accountability in the behavior of political authorities. Winters and Weitz-Shapiro (2013) shows that vast majority of voter in Brazil have a willingness to punish corrupt politicians if there is enough information available. Pereira and Mello (2015) found that access to information provided by radio and possession of television in households is also positively correlated with punishment of corruption in the local level. However, at a political scenario with lack of accountability, information about the government can be stonewalled. Hence, at an environment of poor transparency, public resources cannot be efficiently allocated accordingly to public demand. Insufficient accountability seems to be related to an inadequate institutional design; an issue

4 addressed by a large literature such as North (1990), Acemoglu et al (2001) among others. In a representative democracy, information about how the government deals with daily problems and also how outlays are used are equally important as free and competitive elections. Nevertheless, almost three decades after the new Brazilian constitution, there is still a tension between strong vertical accountability, the presence of participatory budgeting and high levels of electoral turnout, and a very weak horizontal accountability network in the country. Successful accountability institutions have a crucial role for better social results with the use of public spending, regardless of its nature. When grassroots democracy is not observed, especially with respect to associative government, control agencies become even more important, since they held accountable key players in the political system in between elections. On average, how these institutions can improve government outcomes is very heterogeneous in Latin America (Santiso, 2009). In Mexico, for instance, all members of the federation are overseen by the Auditoría Superior de la Federación. In Brazil, however, states and municipal governments, which are constitutionally recognized as subnational autonomous unities and are in charge of basic health and education, are hold accountable by a special group of state level institutions, called Tribunais de Contas. The Brazilian framework follows the court model, which is most common in Roman law countries such as France s Cour des Comptes, Italy s Courte dei Conti, Spain s Tribunal de Cuentas and the European Union

5 Court of Auditors. Each one of the 27 Brazilian states has one Tribunal de Contas which is responsible for holding accountable subnational governments, which include more than municipalities, besides, of course, the state government itself. In Brazil, municipalities are full member of the federation. In a retrospective view, horizontal accountability has improved mostly as a result of oriented participatory policies, driven mainly by the Executive after democratization (Vera and Lavalle, 2012). At the federal level, some important advances were achieved byoversight institutions, especially in the judicial system (Hagopian, 2016). Nevertheless, in spite of advances on participation, in the Brazilian subnational level, the provision of horizontal accountability through institutions that can question and punish decisionmaking actors is more fragile. As Diamond (2002) points out, the design of those institutions affects their action and, in the Brazilian case, there were no significant changes since the military period. Some advances have been made at the federal level (Santiso, 2009, Hagopian (2016), however, in the state and municipal level, evidence is not so positive. The bureaucratic structure of (no)oversight control, erected during dictatorial times, remains very stable. This evidence indicates that the problem of accountability is not related to the lack of strong institutions, but on how they are operated inside the political system. Furthermore, in spite of some minor modifications in the way members were chosen, Hidalgo, Canello e Lima-de-Oliveira (2015) and Weitz-Shapiro, Hinthorn, and Moraes

6 (2015) also have shown substantial evidence that subnational oversight institutions are deeply political oriented; a signal of weakened accountability. Claudio and Finan (2008, 2011) focusing on the role of Controladoria Geral da União, an agency of the federal government which audit non mandatory grants to subnational governments, finds that corruption affects negatively student s outcomes in public schools. However, the mandatory grants, which includes categorical and block grants, are audited only by Tribunais de Contas, a fact that may have underestimated corruption and the delivery of poor public goods. For instance, in Sao Paulo, the largest city of the country, which is governed by an ally of the federal government, only 5.24% of the whole budget in 2016 is audited by the federal oversight institution. The vast bulk of outlays are audited by the local Tribunal de Contas. 2. Accountability in Brazil: back and forward historical process. Contrasting to the African experience which was decolonized after the second world war, Latin American countries have older state structure since they became independent in the early XIX century. In the Brazilian case, state building is directly related to the Portuguese kingdom whose capital from 1808 through 1820 was Rio de Janeiro. Actually, between 1808 and 1822, Brazil was already independent as it was the seat of the United Kingdom of Brazil, Portugal and Algarve. Not surprisingly, Emperor Pedro I, the first Brazilian monarch was son of the Portuguese king. As a result, most basic

7 political institutions of the country are rooted in Europe without the great social reforms that took place in the XX century in the North Atlantic countries (Skidmore, 1999). The Constitution of 1824, the first in the country newly independent, was imposed by the Brazilian Emperor, D. Pedro I following the tradition of the Ancien Régime. As a result, it had little to say about accountability since only the emperor had the right to be informed about the use of treasure money. The second Brazilian constitution, promulgated in 1891, two years after the Proclamation of the Republic in 1899, had important institutional advances, as the new regime introduced representative mechanisms, although the results of these arrangements were not so effective, given the numerous episodes of electoral fraud (Fausto, 2014). As part of the attempt to modernize the national structures, the republican constitution established in 1891 the Tribunal de Contas da União, the first Brazilian oversight institution. Members of this institution were appointed by the President with Senate approval and could lose their seats because of a judicial sentence. During Republica Velha ( ) no important advances were made. The Constitution of 1934 was the one that showed major changes concerning the accountability process. The main reason for that is probably a context of deep political struggle: São Paulo upraised in 1932 for a constitutional government after Getúlio Vargas coup d état in Reforms set up a system of checks and balances where subnational governments must have an audit institution designed as technical agency; that is to say, a loose form of control body beyond the federal level. The oversight institution

8 would prepare an assessment of the accounts of the Chief of the Executive while the local Legislative would judge him or her based on this report. Remarkably, this is the model still in use in Brazil, so the state representatives are responsible for holding the Executive branch accountable. The constitution also defined that members of oversight institutions were be chosen by the president of the republic with the approval of the Senate. However, the Constitution of 1934 had a very short life, as in 1937, a new coup d état inaugurated the Estado Novo regime, with centralism and weakened accountability institutions. In 1946, under the political opening after the end of Vargas dictatorship ( ), a new Constitution revived the institutions created in It also reinforced the power of the Legislative to examine the accounts of the President, with the assistance of the Tribunal de Contas da União. Besides that, it established that subnational oversight institutions would be designed by the state Constitutions mirroring the federal one. After that, the Constitution of 1967, promulgated in the midst of another authoritarian age, repeated many points of the previous constitutions, although the political context had changed radically. A series of changes in the legislation in subsequent years, without any popular consultation, the so called Institutional Acts, deepened the authoritarian character of the Brazilian regime. Nevertheless, the bureaucratic reforms in this period also increased the concern about legal compliance of the use of public resources. Finally, in 1988, with the end of military rule, a new constitution was enacted, with very ambitious targets regarding a social welfare state as well as the institutionalization of

9 participation mechanisms. With respect to accountability, the Constitution prohibited the creation of new oversight institutions, besides those already established. Only internal bodies were allowed, such as CGU in the federal government, which was created only in Nevertheless, little was added to the constitutional text of the military regime, besides the emphasis on supervision of the budget and operation of government agencies. For example, oversight institutions must review the legality of acts of admission of civil servants, but they are not held accountable on their productivity. Actually, efficiency became part of the Brazilian legal framework only in 1998, in the FHC government. 3. Theory and Hypotheses 3.1 The Brazilian notion of accountability: some remarks. According to Dye and Stapenhurst (1998), Santiso (2009), among others, Brazil follows the court model which is a collegiate court with quasi-judicial power in administrative matters. The quasi-judicial tradition usually implies a closer link with the judicial branch, although oversight institutions are independent from it. As they must send their reports to the Legislative branch, there is a high level of uncertainty about who oversees Tribunais de Contas in the country. Furthermore, because of the judiciary influence, the Brazilian court model emphasizes legal and financial compliance over performance auditing. Blume and Voigt (2011) point out recurring problems encountered as obstacles to the advancement of the effectiveness of audit agencies. (The incentive system used in these

10 institutions, for instance, leads to a bureaucracy that is not rewarded when assessing if public policy results meets social demands. In some states, for example, civil servants are reward according to a simple time passing rule: after every 5 years wages increases in 5%, after every 20 they are increased in 1/6. This simple rule has no link with enhancing the quality of the public sector and it is still widely used in important states such as Sao Paulo, the richest in the nation 3. Most importantly, when the control systems are analyzed at the subnational level, although the federal State and Constitutions provided some basic determinations for the Brazilian audit institutions, there is no clear regulation. This leads to a lack of uniformity that allows members of the control bodies to define internally the rules used to the judgment of public accounts as well as how information is disclosed. According to Rocha (2013), the absence of regulation on what should be a national standard decreases the quality of the opinions issued by the courts and weakens the effective control. His findings also highlight excessive formalism as well as a lack of audit performance, limiting therefore the information received by citizens. Regarding how oversight institutions members are appointed, Blume and Voigt (2011) and Melo Pereira and Figueiredo (2009) converges on the idea that in spite of being an autonomous body, the external control is heavily influenced by political and economic actor and interests groups... 3 In São Paulo, this rule about remuneration of the civil servant service is established in the state constitution.

11 Abrucio (1994) in a seminal paper shows that Tribunais de Contas are key institutions to enforce political control by governors over subnational elites. One hypothesis is that members of the institution would act in favor of political interests of those who appointed them. Similar findings are also shown by Hidalgo, Canello e Lima-de-Oliveira (2015) and Weitz-Shapiro, Hinthorn, and Moraes (2015), indicating that political ties still play a fundamental role in Brazilians TCs. In fact, although it is forbidden to oversight institution members engage in political activities, it is very usual the presence of former politicians as members of these institutions. A comprehensive survey about members of Tribunais de Contas carried by Transparencia Brasil, a local NGO, found that most counselors have been part of the political game or related to politicians (Paiva and Sakai, 2014). On average, out of ten directors, six were former politicians. The need for political background to occupy such a position exists. As a result, 17% of the counselors are relative of a local politician; half of them are children or parents of politicians. Former governors and deputies are the positions with greater kinship, precisely those who are supposed to appoint them. Loureiro, Teixeira and Morais (2009) showed that institutional development of the Brazilian Audit Courts after 1988 was partially blocked by counselors. This can be an important factor reducing the control capacity of oversight institutions, as counselors are key players in how accountability is done in the country. Besides the political connection, 47 out of the 238 advisers have some judicial problem, with fourteen being sued by improper conduct with public funds. In eleven Brazilian

12 oversight institutions counselors are being sued by various irregularities, e.g. embezzlement, undue accumulation of jobs or wages, nepotism, bribery, electoral crimes or against the financial system, tax evasion and, finally, a counselor is being sued due to money laundering according to the survey carried by Transparencia Brazil. 3.2 Hypotheses Oversight institutions influence directly the information available to legislative branch and, on a broader perspective, to the citizenry. They can be important tools for transparency or, conversely, power mechanisms of political control. However, the historical process of formation and consolidation of Brazilian oversight institutions had given a mixed nature to the Tribunais de Contas. As a result, the main question that we want to answer in this paper is what drives higher or lower level of transparency of the oversight system. Brazil is a country of high level of income inequality, despite important improvements along the last years. When there is a large economic divide, the wealthy minority has greater motivation and more opportunity to control information about the state. On the other hand, the vast majority of the population is less able to push for better monitoring of the state. Jong-Sung and Sanjeev (2005) shows evidence in a comparative analysis of 129 countries that the more difference is found about the top and lower brackets of income, the more corrupt is the country. In the Brazilian subnational level Pereira and Mello (2015) found

13 that the presence of media increases the likelihood of punishment. In the case, more radios and having television are the key factors. Both are related to income and political competitiveness. The latter increases the chance that radios and TV channels broadcast critical views. The former is directly related to having the tools to receive and use the information. There is no reason to think that both does not affect how Tribunais de Contas disclosure information about politicians and the civil service. More transparency of the oversight institution influences how wrongdoing is detected since it improves information from outsiders of the public sector. Besides that, counselors are often chosen among important politicians, e.g. a former member of the state Legislative. As ex-professional politicians, they still have political links and interests, so they usually keep on playing a relevant role with political elites. Therefore, if the level of political competition in the subnational level is low, political elites can keep their strong ties with the group of counselor, reducing horizontal accountability. We propose two factors that are possible drivers of higher levels of transparency: political and economic competition within society. Hence: H1 More competitive political system in the subnational level is positive correlated with transparency and more access to information. H2 More competitive economic relations (less economic inequality) is positive correlated with transparency and more access to information.

14 Finally, the participation of civil service cannot be disregarded. Conselor have the final word, however, civil servants produces the reports which bases the final opinion. With respect to them, the International Development Bank launched in 2006 the External Control Modernization Program States Promoex a US$ 64.4 million program whose overall objective was rebuilding the Brazilian Tribunais de Contas. One of its specific targets was to introduce performance audit techniques to civil servants of all oversight institutions of the country. Courses were offered and also many discussion forums were made among audit institutions (IDB, 2016). Thus, this set of policies developed by IDB may have affected Tribunais de Contas civil service leading to our sixth hypothesis. H3: The spillover of autonomous technical expertise increases transparency, i.e. the existence of participatory or results evaluation increases transparency. To check specifically the spillover of autonomous technical expertise, we investigate the nature of the audit procedure. According to the World Bank (2001) audit models can be summarized into three main types: financial audit, which monitors the balance of funds invested, compliance audit, which verifies compliance with the rules on the application of these resources, and performance audit, which government's operational efficiency and effectiveness of the programs implemented. While the first two in principle can also be seen in non-democratic countries, we focused on identifying performance analysis, since it is directly connected to the delivery of public goods and so to democracy (Lonsdale, Wilkins, Ling, 2011).

15 One final question, we should be concerned with respect to the outcome of the accountability process. In fact, the disclosure of information of the process is important but, it is also desirable that the final opinion is also known. Most important than being able to find information about the whole process is to have information concerning the final decision of the TC. To do that, a decision must be taken regardless if positive or negative towards the governor. Given the political background of counselors, our last hypothesis it that the same forces that influence the divergence of transparency will also impact the decision, in the same direction. Son on, as last hypothesis we propose: H4: The spillover of autonomous technical expertise, the competitiveness of the political system and the competitiveness of economic relations are positive correlated with the taken of decision.

16 4. Research Design and Methods We analyze the opinion issued by Tribunais de Contas for the period from 2003 through 2012, in all Brazilian states. As it is stated by the Brazilian constitution, every governor must have their accounts reviewed by Tribunal de Contas. The procedures are usually broadly defined by state legislation and detailed in the Internal Rules of Tribunal. At section 5, we present the main details of the process by which each governor has its accounts controlled by the oversight system. At the third step of the research, our unity of analysis became the accounts of each Tribunal at each year. A survey was conducted on all accounts of the chief of the executive from 2003 through 2012, identifying the procedures adopted in the accountability process. Variables After presenting the overall procedures, we identify some important variables related to the nature of the oversight control process for each account/year. As main characterizations, we investigate the easiness of the disclosure of information as a measure of Transparency, if any decisions were taken Decision Taken, and if there was information about performance techniques in the account report Report wih Perfomance Analysis. Transparency and Decision Taken are our main quantitative dependent variables. Regarding transparency, if there were no immediate data availability on the institution

17 website, several attempts were made to get the information. First, we sent an requesting information, if no answer was giving to the , we made a phone call to ask for the information. Using this research strategy e are able to verify if there was information on the website, if there was delivered of information with just an and how many days it took to be released, if information was delivered after a phone call, and, finally, if no information was delivered whatsoever. Decision Taken is just a dummy variable indicating if any decision was taken, no matter if positive or negative towards the governor. An important limitation of our research approach is that we were able to measure Decision Taken just for the cases where information was available. Report wih Perfomance Analysis is a third dummy variable that indicates if there was any performance audit in the review of governor s accounts. After collecting data about the accounts reports, we collected data about social-economic conditions and also about the structure of each Tribunal de Conta. We use data of Transparencia Brasil about characteristics of counselors, for instance, political ties or crime conviction, as well as control variables related to idiosyncratic features of Brazilian states. As we do not intend to discuss the level of political competition in the subnational level, we use a minimal definition: the share of votes of the elected governor in the first round of elections. Although, this variable is far from being perfect, it is undeniable correlated with political competition. Besides, we use the Gini Index in 2000 to control for income inequality as proxy of economic competitiveness.

18 As Brazilian society has a high degree of variation regarding socioeconomic development among its states we use a set of control variables: population, state gross product, state area in square kilometers, infant mortality, the number of mobile phones per 100 inhabitants in each state and also the share of population with access to high speed internet. The description of the quantitative variables as well controls is further depicted in Section 6. Estimation Techniques To estimated our hypotheses, we we use OLS and also techniques for dummy depend variables, following Wooldridge (2011) for both Transparency and Decision Taken. For robustness check, we estimated the selection of Decision Taken following a selection procedure as proposed by Heckman (1979). The rightness of this model is related with the fact the information regarding whether the Tribunal has already taken a decision depends on getting the information before. We estimated both regressions together, following a selection procedure (Heckman, 1979). As identifying variables for the selection model, we use the number of mobile phones per 100 inhabitants in each state as well as the access of high speed internet. We assume that both variables affect the degree of transparency but at the same time are not correlated with the decision to judge the account. It is not a strict assumption since the disclosure of information can be made more easily when modern methods of communication are available. On the other hand, there is no reason to consider that both identifying variables will influence when an account will be judged.

19 Nevertheless, despite the fact that qualitative results are very similar in all models, the independence of the two equations in model 6 is not rejected so, they can be estimated independently as we cannot reject the null hypothesis of independence. Consequently, the decision about transparency and about when to judge seems to be not related. 5. How governors are held accountable: procedures for accountability In this section, we analyze the opinion issued by Tribunais de Contas for the period from 2003 through 2012, in all Brazilian states. As it is stated by the Brazilian constitution, every governor must have their accounts reviewed by Tribunal de Contas. The procedures are usually broadly defined by state legislation and detailed in the Internal Rules of Tribunal. In most states, the Governor s Cabinet must send a report, sixty days after the opening of the Legislative termwhich is usually in early February. It consists of a general assessment, comparable to the State of Nation in the US tradition, though no speech is delivered by the Chief of the Executive. Every year, one counselor is chosen by rotation, which is the most common or, in some states, by random assignment to be the relator that will conduct the evaluation of the accounts. The whole process must be finished usually after sixty days the report arrives at the Tribunal. In between this period, a technical committee composed by members of the Secretary of External Control or equivalent body is formed to help the relator formulate another technical report. They usually analyze all the documentation presented by the Governor,

20 as well as data and information on the achievement of goals set out in law. If there is any evidence of irregularity, the Governor may present a defense in thirty days. With the technical report ready, the relator sends all documents to the Public Attorney who must issue an opinion: favorable or unfavorable. In some states, the External Control Secretariat also issues an opinion after the Public Prosecutor. Finally, a session in the plenary is called so as to the other counselors may issue their opinions based on the report presented by the relator. The final report and the opinion, i.e. favorable or unfavorable, are sent to the Legislative which can be change it only if there is qualified 2/3 majority of the state representatives. A negative decision suspends political rights and can lead to an impeachment. Diagram 2 illustrates how the process occurs.

21 Diagram 1 Contas dos Governadores Governor (i) Audit Institution (ii) Counselor is chosen randomly or by rotation (iii) Legislative Branch (viii) Full Chamber (vii) President Auditors Councelors Public Prosecutors (vi) Tecnical Report (

22

23 6. Results: reasons for divergence. The first variable measure in our paper is Transparency as a measure of the easiness to the disclosure of information. Despite the fact that legal procedures are very clear, there is no complete access of information about the final decision sent to the Legislative as it should be in a high accountable democratic environment. The information was available for the public only in 79.30% of cases, i.e. 214 out of 270 observations. This result is further investigated in Table 1. The Transparency variable was grouped into three groups. The most accessible are those whose final opinion was available on the website. When the document was no available online, an inquiry was made by and, if there was no success, a telephone consultation with the ombudsman. In the last two columns, is exposed, respectively, whether there was no response or, if so, the amount of working days long to answer the inquiry. It is noteworthy that in some states that electronically provide any kind of accountability information, it is sometimes difficult to get it in the website, despite search tools, because a specific random assigned number identifying the process is required. For the regression models, we consider transparency as a dummy variable whether or not we had found information about the account report on each year for each state. Table 01 Transparency: Access to technical reports and prior opinions Type of Response Number of Reports Number of States Relative Frequency Available on the website ,7% Available after request ,0% Available after phone request ,7% No return ,5%

24 Only in 11 of the 27 audit the opinion about the governor s account is available on the website. In other cases, it was necessary to request it by , with an average of 7.2 days for the response and a maximum delay of 29 days, when there was response. In some states, there was no return, despite several attempts. Acre and Amapá, two poor states in the Amazon region underline the lack of horizontal accountability. In both cases there were not scanned documents, and everything must be consulted personally, in the institution's headquarters; something impractical in a country with continental size. The Tribunal de Contas of Espírito Santo in the Southeast region of Brazil also does not provide their reports or opinions online, requiring a submission of a formal request by . After 29 days of the request, we had an answer and, two months later, a CD-ROOM was mailed with the Technical Reports. Moreover, in Court of the State of Alagoas that is located in the Northeast region, it was informed that the issue was confidential and the disclosure of information was allowed only for those living inside the state. Finally, 5 audit institutions which were located in the North and Northeast did not respond 45 days after the request. The second variable, Report with Performance Analysis, is related with the technique of accountability. We analyzed each account review to identify if there was any performance technique. Performance techniques were found in 62.90% of reports analyses, usually as an overall chapter of assessment. This outcome mirrors the efforts of the Brazilian government and the IDB with PROMOEX. Participation as a measure of estimating demand of public services if the government holds public meetings, for example was seen in only 33.34% of cases. In a second step, we looked closely the information that was available about 214 cases. Within this group, in 87.90% cases an opinion as issued. Therefore, despite the maximum legal time for

25 the analysis of the accounts, in 5 Tribunais de Contas for 50 reports, deadlines were totally disregarded, especially in the northern states of the country. Although formally all have no later than 60 days for the completion of the whole process, either enforced by the respective basic law or due to internal regulation, in many cases, the oversight institutions do not respect the time limit. In Roraima, Amapa, in the North region, for example, all accounts prior to 2007 have not yet been analyzed. Furthermore, an important finding is the lack of negative opinions, which suggests a trend towards automatically issuing a favorable opinion. In fact, out of 27 states, there were negative opinions only in Goiás, Tocantins and the Federal District. The description of all quantitative variables is depicted on Table 2.

26 Table 02 Statistical Description Variable Obs. Mean Std. Error Min Max Information Available Decision Taken Gini Counselor with Criminal Conviction Report with Performance Analysis Political Competition Counselor with Political Involvement Population 270 7,044, ,316, , ,300, State Gross Product* , , , ,408, State Area Km , , , ,559, Infant Mortality Rate Density of mobile phones High Speed Internet Use - % of population * In millions of reais. 7. Weakened horizontal accountability Tables 03 and 04 show our main quantitative empirical results. Regarding Transparency, outcomes are in Table 03 and regarding Decision Taken, outcomes are in Table 04. At both table, Models 01 and 03 are estimated using OLS, while at models 02 and 04 we deploy probit model. Finally, as the information regarding if the Tribunal has already taken a decision depends on getting the information before, we estimated in the model 05 both regression of each table together, following a selection procedure (Heckman, 1979). At Table 05, income inequality seems to play a key role for explaining the disclosure of information. The more unequal is the state, the less transparency is found. It is important to consider that inequality remains high in Brazil ranging from in the Federal District to

27 0.448 in Santa Catarina, so it is truly a salient channel. This result does not come along with respect to political competition. There is no effect at all. As a result, it seems that income inequality reduces the amount of information to the general public. On the other hand, the game played by the political elite does not have any relation at all. This result is not surprisingly as politicians make a specific, powerful and very well informed group. Moreover, the use of performance audit is statistically significant in all models. In Brazil, compliance audit is done in every Tribunal de Contas since they were set up with this purpose. So, as we claim in hypothesis H6, the use of new methods may be an evidence of modernization. This result again shows that the efforts of IDB, with the Promoex initiative may have enhanced transparency on oversight institutions in the country. When we add controls to our baseline model, our findings remain the same. More population decreases transparency levels, as it seems to be more difficult to communicate given that we also control for internet and mobile phone use. The opposite impact is seen with respect how wealthy is the state measured by gross product. As expected, the richer the state by income per capita levels, the more transparent it is. Consequently, growth along with more equality is a salient factor to improve the disclosure of information. The spread of new technologies is also important. Interestingly, characteristics of counselors seem to be statistically irrelevant to transparency. As data from Transparencia Brasil shows 52.30% of counselors had in the past formal political involvement, however, this number may be underestimated since the remaining counselors may have informal ties too. In fact, given the clear political criteria in the choice of counselors, as

28 uncovered in the historical analysis, this outcome suggests that the oversight system as a whole works politically and not because of the will of an individual member. When we move to Table 04, however, about whether a decision was taken, regardless if it was favorable or not, our results shows that political competition and more expertise in performance audit are the most important factors. In all models, the share of votes of the winner seems to be positive correlated. So, the less competitive is the state, the more likely it will be a decision. It may seem a puzzle, but it is important to remember that a negative decision will make the present governor not eligible to run for reelection. Thus, instead of issuing a negative decision or a favorable but technically fragile decision, one may think that the audit institution could decide to postpone the final decision. So, when there is more political dispute, there seems to be a pattern in which the TCE s decision is delayed and so there is no political loss for the chief of the state executive. On the other side, again the use of performance tools seems to be relevant. It is interesting to observe that this variable was built assessing the audit reports so it mirrors the technical capacity of each Tribunal civil servants. So, at the same time, Promoex enhanced internal quality increasing the quality and response of overseeing institutions. However, they are still headed mostly by former politicians. All things considered, there is strong evidence that Tribunais de Contas are politically controlled by state elites and they may be an actual important agent in the subnational politics. Besides that, our findings show the lack of an integrated control system in Brazil, as well as the absence of a superior control body such as the National Council of Justice, or a national standard, did not prevent a high degree of isomorphism among the oversight institutions. Tribunais de Contas are

29 undeniable quite similar all around the country in spite of Brazilian continental size, following close the quasi-judicial court system. Nevertheless, there is also significant evidence of a limited horizontal accountability, especially because oversight institutions are not a trustable supplier of information about the government for the Brazilian society. Lack of expertise, low income and high inequality seems to be the most relevant factor driving to explain lack of transparency in the Brazilian subnational level.

30 Gini [1.850] [11.061] [2.438] [12.996] [1.003] Counselor w/ Criminal Conviction [0.156] [0.802] [0.141] [0.985] [0.073]** Report w/ Performance Analysis [0.111]*** [0.471]*** [0.086]** [0.678]*** [0.060]*** Political Competition [0.377]** [1.746]*** [0.324]** [1.887]*** [0.217]*** Counselor w/ Political Involvement [0.189] [0.807]** [0.170] [1.512] [0.109] Log Pop Table 03 Transparency Results [0.083] [0.525]** [0.055]** Log Gross Product [0.075] [0.563] [0.057] Log State Area Km [0.026] [0.114] [0.017] Infant Mortality Rate [0.013] [0.068] [0.007] Constant [1.107] [6.206] [1.365] [7.196] [0.600] Statistic F(5,23)= W(5)= F(9,23)= W(9)= W(9)= Prob>F or Prob>X Number of Observations

31 Table 04 Decision Results Gini [1.725]*** [10.972]** * [2.024]* [13.632]* * [9.035]** Counselor w/ Criminal Conviction [0.228] [0.977] [0.201] [0.959] [0.595] Report w/ Performance Analysis [0.120]*** [0.497]*** [0.162]** [0.727]** [0.491]** * Political Competition [0.452] [2.529] [0.405] [2.343] [2.440] Counselor w/ Political Involvement [0.252] [1.093] [0.289] [1.294] [0.613] Log Pop [0.082]** [0.717] [0.622]** Log Gross Product [0.097]** [0.711]** [0.576]* Log State Area Km [0.034] [0.204] [0.185]

32 Infant Mortality Rate [0.011] [0.060] [0.050] Internet [0.048]** Mobile Phone [0.038] Constant [0.965]** [5.821]** [1.384]* [6.708]** [5.639]** * Number of Observations F(5, 26)= W(5)= F(9,26)= W(9)= W(1)= Prob>F or Prob>X

33 8. Conclusion Accountability is a necessary condition to democracy. In newly established democracies, horizontal accountability tends to be more fragile than vertical accountability, specially regarding the subnational level. In Brazil, federal and state legislation maintained almost the same institutional design of oversight institutions from dictatorial periods, despite the huge progress in the last thirty years towards free and competitive elections. The aim of our paper is to analyze oversight institutions and their role the role in holding the state government accountable during the period between elections. Our findings show significant evidence of a limited horizontal accountability, especially because oversight institutions are not a reliable supplier of information about the government for the Brazilian society. Consequently, we identified a democratic deficit in the Brazilian subnational level, associated with a low level of transparency. Factors as income inequality and political competition in the subnational level play a major role in defining transparency levels. As intuitively expected the more equal and the more competitive are subnational elections, oversight institutions seem to be more transparent. Furthermore, a more equipped civil service also seems to affect positively the disclosure of information regarding the chief of the Executive. All in all, inputs collected over the course of this research strongly support the lack of accountability in the subnational states, especially concerning transparency. In fact, what is observed is a fully institutionalized system performing compliance audit with low levels of transparency and results. The improvement of horizontal accountability is a challenge to make

34 democracy work in Brazil, especially at the levels of government closer to the daily life of the citizenry. Corrupt incumbents may be elected simply because enough information is not available. And inefficiency may be widespread because efficient is not a big concern for the Brazilian subnational accountability system. 9. References Abrucio, F. L. and Loureiro, M. R. (2005), Finanças públicas, democracia e accountability in: Biderman, C. and Arvate, P. (Eds). Economia do Setor Público no Brasil, Elsevier, p Bac, Mehmet. "Corruption, connections and transparency: Does a better screen imply a better scene?." Public Choice (2001): Barreto, I., and C. Baden-Fuller To Conform or To Perform? Mimetic Behavior, LegitimacyBased Groups and Performance Consequences. Journal of Management Studies 43 (7): doi: /j x Barzelay, M. (1997), Central audit institutions and performance auditing: A comparative analysis of organizational strategies in the OECD. Governance, v. 10, n. 3, p Beckert, J Institutional Isomorphism Revisited: Convergence and Divergence in Institutional Change. Sociological Theory 28 (2): Blume, L. and Voigt, S. (2011), Does organizational design of supreme audit institutions matter? A cross-country assessment. European Journal of Political Economy, v. 27, n. 2, p Blume, L. and Voigt, S. (2007), Supreme audit institutions: supremely superfluous? A cross

35 country assessment. ICER Working Paper. Brasil. Constituição (1988). Constituição da República Federativa do Brasil. Brasília: Senado, Brasil. Lei complementar no 709, de 14 de janeiro de Lei Orgânica do Tribunal de Contas do Estado de São Paulo. Tribunal de Contas do Estado de São Paulo, São Paulo, SP, Brasil. Lei no 8.443, de 16 de julho de Lei Orgânica do Tribunal de Contas da União. BTCU no 08, Edição Especial, Brasília, DF, 4 fev Brasil. Lei no , de 28 de maio de Dispõe sobre a organização da Presidência da República e dos Ministérios, e de outras providências. Diário Oficial [da] República Federativa do Brasil, Brasília, DF, 28 mai Brasil. Medida provisória no , de 02 de abril de Altera dispositivos da Lei no 9.649, de 27 de maio de 1998, que dispõe sobre a organização da Presidência da República e dos Ministérios, e dá outras providências. Diário Oficial [da] República Federativa do Brasil, Poder Executivo, Brasília, DF, 2 abr Carvalho, M. C. and Takahashi F. (2014), Tribunal rejeita contas da USP por reitor ganhar mais que governador, available at: tribunal-rejeita-contas-da-usp-por- reitor-ganhar-mais-que-governador.shtml > (accessed 26 March 2014). Clark, C, De Martinis, M. and Krambia-Kapardis, M. (2007) Audit quality attributes of European Union supreme audit institutions. European Business Review, v. 19, n. 1, p Coutinho, F. (2014). Estádio de Brasília foi superfaturado, aponta Tribunal de Contas, available at: estadio-de-

36 brasilia-foi-superfaturado-aponta-tribunal-de-contas.shtml (accessed 16 March 2014). Dagnino, E.; Tatagiba, L. (2007). Democracia, sociedade civil e participação. Chapecó: Argos. Dagnino, E. (2002), Sociedade civil e espaços públicos no Brasil. São Paulo: Paz e Terra. Diamond, Jack The Role of Internal Audit in Government Financial Management: An International Perspective. IMF Working Paper. DiMaggio, Paul, and Walter Powell The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in Organizational Fields. American Sociological Review 48: Dye, Kenneth, and Rick Stapenhurst Pillars of Integrity: The Importance of Supreme Audit Institutions in Curbing Corruption. World Bank. Elster, J. (1998), Deliberative democracy. Cambridge: Cambridge University Press.. Eisenhardt, K. M. (1989), Agency Theory: An Assessment and Review. Academy of Management Review, 14(1): Fausto, Boris. A Concise History of Brasil Cambridge University Press. Fernandes, G., Brelaz, G. and Elmais, L. P (2014). O papel limitado dos órgãos de controle no aperfeiçoamento da accountability horizontal: o caso dos governadores paper presented at Encontro Nacional de Administração Publica e Governança, November, Belo Horizonte, available at: Ferraz, Claudio, and Frederico Finan Exposing Corrupt Politicians: The Effects of Brazil s Publicly Released Audits on Electoral Outcomes. The Quarterly Journal of Economics 123 (2):