Diffusion and adaptation of the regulatory agency model in Brazil

Abstract This article examines the diffusion of independent regulatory agencies (IRAs) in Brazil, demonstrating how important domestic factors interplayed with traditional mechanisms of diffusion lead to an expressive process of agencification. Although top-down, bottom-up and horizontal mechanisms played an important role in the Brazilian agencification process, they fail to explain the creation of IRAs in unexpected sectors or the frequent modifications that occurred at subnational levels in a short period of time. To understand how local political actors adapted the regulatory agency model to the Brazilian institutional legacies, field research was conducted, based on bibliographical, documental, and interviews with key political actors. The specificities of Brazilian federalism and the strategic role of the film industry, bureaucrats and politicians in (re)interpreting the agency model helped to boost the diffusion of IRAs in Brazil.


INTRODUCTION
The diffusion of independent regulatory agencies (IRAs), or the agencification process, observed in both developing and developed countries in recent decades was followed by an extensive literature that identifies the potential mechanisms at play in this process. Despite the different motivations involved in the establishment of IRAs worldwide, these entities represent, at least discursively (Pollitt, Bathgate, Caulfield, Smullen, & Talbot, 2001), an attempt to separate politics and administration. Agencies characterized by being independent of political influence represent a profound change in the structure of the state and in the division of authority between elected and unelected officials (Levi-Faur & Jordana, 2005).
While most of the literature is concentrated in the European or North-American contexts, new contributions seek to advance research in developing nations, particularly those located in the southern Hemisphere (Dubash & Morgan, 2012). In fact, the diffusion of IRAs followed a central to peripheral axis, from developed to developing nations, and found propitious conditions in several Latin American countries that have been involved in market-oriented processes since the late 1980s, pressured by international organizations such as the World Bank and the OECD (Dubash & Morgan, 2012;Levi-Faur & Jordana, 2005).
This article focuses on the diffusion of IRAs in the Brazilian context based on the premise that important domestic factors have influenced the "centre to periphery" diffusion process. The Latin American version of the Regulatory state (Levi-Faur, 2003;Majone, 1997) influences the manner in which IRAs are adopted, interpreted and implemented by domestic political actors.
In this sense, Brazil is an excellent example of the diffusion of IRAs. Brazil's agencification process resulted in the establishment of more than sixty IRAs since 1996, and this topic has been thoroughly discussed by local authors. Yet, those studies are concentrated in the common aspects of this process, focusing mainly on bottom-up arguments, such as credible commitment, political uncertainty and blame-shifting (Melo, 2002;Melo & Pereira, 2013;Mueller & Pereira, 2002). Far less understood are the specificities of the Brazilian agencification process, usually considered an "unduly extension" of the IRA model (Salgado, 2003).
Accordingly, the present study is motivated by the following question: what forces explain the diffusion of IRAs in Brazil? In order to focus on both commonalities and variations of IRAs diffusion process, this paper focuses on two "unexpected" results of Brazilian agencification: the creation of Ancine (Brazilian Film Agency) and the extinction of Asep (Regulatory Agency of the Rio de Janeiro State) and its division into two new sectoral regulatory agencies within a short period of existence. When analyzing the process of agencification in Brazil, two aspects attract the attention of the observer. The first is the so-called unduly extension of the agency model to economic sectors without natural monopolies to regulate or concession contracts to monitorate (Salgado, 2003). The second is related to the process itself. At the state level particularly, the process of agency creation was marked by delays and reformulations (Pó & Abrucio, 2007). The exceptionality of Ancine and Asep influenced a bottom-up methodological research strategy based on bibliographic and documental data, as well as in-depth interviews with key political actors involved in the creation of these IRAs and their subsequent modifications. This bottom-up approach supported the integration of generic, nation-specific and sectorial-specific factors, and revealed that -whilst traditional mechanisms of diffusion can explain the adoption of IRAs in different regions -they only tell part of the story. The Brazilian process of agencification can only be understood through the analysis of both common patterns and specificities of local context.
In practice, understanding key factors influencing the diffusion of IRAs in Brazil is highly relevant to subsidizing public policies aimed at refining institutional design. Furthermore, studying the way in which the "regulatory agency concept" has been adapted and reinterpreted in the three governmental spheres of Brazil can aid in avoiding "cookie-cutter" regulatory/institutional improvement policies. Lastly, this paper contributes to more recent research that examines how the State is expanding -and not retreating -via regulation and regulatory agencies (see Haber, 2011;Levi-Faur, 2014).

REGULATING THE EXPECTED: COMMON RATIONALES FOR ADOPTING AN IRA
The literature that approaches IRAs' adoption as a diffusion process (Gilardi, 2005;Levi-Faur, 2005) suggests three main classes of explanations for this process: bottom-up, top-down and horizontal. These classes of explanations, called here "mechanisms of diffusion", group distinct rationales that can explain the decision to create an IRA, integrating the contributions developed by the rational choice theory (in the case of bottom-up mechanisms) and the sociological institutionalisms (in the case of top-down and vertical mechanisms).
Bottom-up mechanisms explore why, and under which conditions the Legislative Branch decides to delegate power to autonomous bureaucracies and may be summarized by the need to (i) reduce the cost of decision-making, and (ii) ensure the credibility of commitments (Majone, 2001). The creation of an IRA could be justified when the policy in question demands high levels of technical information, but the potential political benefits are low (Epstein & O'Halloran, 1999) or when it is interesting to avoid responsibility for negative consequences (Fiorina, 1982). The political insulation of bureaucracies allows greater flexibility of action, blame-shifting and the acquisition of technical knowledge, representing a highly effective way of reducing decision-making costs. The importance of credible commitments is underlined by Kydland and Prescott (1977) and is characterized by temporal inconsistency. In other words, the expectation of rational agents -such as the fear of future expropriations -affects current investment decisions, so that it is preferable to reduce policymaker discretion. Spiller (1994, 1996) brought the credibility issue to the forefront of the debate involving the capacity to promote private investment in regulated markets. The authors redirected the attention of research efforts from regulatory incentives to the importance of an adequate regulatory governance structure. In this way, the fit between institutional endowment, regulatory governance and incentive designs became the determinant in explaining the performance of the regulated sector. This explanatory mechanism has predominated in the Brazilian literature (Melo, 2002;Mueller & Pereira, 2002;Salgado, 2003).
The second class of explanations, or top-down mechanisms of diffusion, interprets the diffusion of IRAs as a response by policymakers to the external pressures of international organizations, national policy communities and support agencies (Gilardi, 2005). Dubash and Morgan (2012) call this process "institutional transplant" and recognize the relevance of these mechanisms in the context of developing countries, as a consequence of the pressures brought to bear by international organizations such as the World Bank, particularly in the case of foreign debt refinancing. Top-down mechanisms are equivalent to the idea of coercive isomorphism (DiMaggio & Powell, 1983), where organizations become similar due to formal or informal pressures of other organizations which they depend on or the local society.
Lastly, horizontal mechanisms translate the essence of diffusion. Diffusion is about uncoordinated and interdependent processes where the policies' choice of one unit is influenced by the choice made in a different unit (Shipan & Volden, 2012;Simmons & Elkins, 2005). In practice, this means that the creation of an IRA increases the probability of another IRA being created. This type of explanation is similar to the concept of mimetic isomorphism developed by DiMaggio and Powell (1983), in which organizations model themselves in accordance to peers that are regarded as successful and legitimate. In this way, similarities in institutional design and functioning in regulatory agencies occur not because a certain model is more effective or adequate, but because it is seen as legitimate and successful, setting off a process of diffusion by mimicry. Box 1 summarizes the main contributions of each mechanism. The need to establish credible commitments to attract private investment. "Tying the hands" of political opponents, in order to maintain the preferences of the elite in power when the chances of re-election are low. Flexibility of action and, with it, greater responsiveness. Political benefits stemming from the possibility of blame shifting. High technical information requirements. Majone (2001); Moe (1990); Levy and Spiller (1996); Epstein and O'Halloran (1999); Fiorina (1982); Pereira and Mueller (2002); Gilardi (2005a;2005b); Melo, Pereira, and Werneck (2010)  Considering the astounding number of IRAs created in Brazil since 1996, horizontal mechanisms become an important class of explanation to be investigated. While previous empirical research has demonstrated its importance (Gilardi, 2005;Yesilkagit & Christensen, 2009), these mechanisms alone fail to explain what aspects of IRAs were considered "successful" or "legitimate" to local political actors, and why they were interpreted this way.

BOX 1 TRADITIONAL MECHANISMS OF DIFFUSION
In practice, there were a significant number of local solutions, such as the adoption of three different IRAs for the transportation sector at the federal level (land, waterway and civil aviation) and the different dynamics observed at the subnational level. It is precisely the local adaptation of the IRA model that this paper seeks to elucidate.

REGULATING THE UNEXPECTED: A "CONTEXT MATTERS" APPROACH
Agencification processes are commonly portrait as an international fad in the field of public administration (Pollitt et al., 2001). While it can be considered a convergent phenomenon, given the popularity of the 'agency' form and its global dissemination, what each locale understands as an 'agency' , and the observed differences in their degree of autonomy indicate that the outputs of the process are quite divergent -since agencies go through unique trajectories given their countries and sectors (Levi-Faur, 2006;Nakano, 2004;Pollitt et al., 2001;Verhoest & Laegreid, 2010;Verschuere & Barbieri, 2009;Yesilkagit & Christensen, 2009). As a result, agencies not only differ in the shapes they assume and the trajectories they follow, but also in the rhetoric that upholds them (Smullen, 2010) in a process of "divergent convergence" (Tenbucken & Schneider, 2004).
Once one recognizes that "context matters", the first question that comes to mind is how it matters. Levi-Faur (2003) discusses the relevance of an actor-centred historical institutionalism in demonstrating how actors behave in different institutional environments while accommodating the process of learning to their advantage, and how different historical legacies constrain their actions. Moynihan (2006), defends that context matters to interpreting lessons, considering that policy lessons can be understood in different ways. In this case, context affects policy selection indirectly, by influencing how policy makers interpret the policy lessons. This approach, therefore, emphasizes the role of interpreters and their discretion in matching policy doctrines and context. This perspective reveals the dynamics of decision-making as a garbage can process. Local political actors interpret policy lessons in their own way because they depart from different problem-definitions which are, in turn, constrained by particular institutional trajectories. At the same time, they try to adjust "international" solutions to their peculiar needs. As March and Olsen have noticed thirty years ago, administrative reorganization processes are likely to become a highly contextual combination of people, choice opportunities, problems and solutions" (1983, p. 286) whose main motivation does not lie in the search for greater performance. Nakano (2004), for example, demonstrated how Japanese policymakers reinterpreted and redefined the British concept of agencification. The rationale behind Japan's agencification process was not related to the possibility of improving the efficiency of public service but aimed to reduce the role and size of central departments. The difference between the British and the Japanese processes of agencification is not due to a misunderstanding; but it represents the way that Japanese policymakers found to resolve local issues: by adapting subjective political concepts, such as agencification.
This work adopts just such a view. In other words, the Brazilian agencification process can be explained by the different diffusion mechanisms interplaying with the specificities of local context. It holds that agencification is ambiguous enough to be understood and adapted by local needs (Moynihan, 2006) and historical legacies.

EMPIRICAL STRATEGY
The present study entailed two phases: a) a cross-cutting analysis of the IRAs created, based on documental and bibliographic research; b) the analysis of two cases, chosen based on their exceptionality: Ancine, at the federal level, and ASEP, at the state level. From all the federal agencies created since 1996, Ancine is by far the most unusual. Alvarenga (2010) mentions three aspects that make Ancine a peculiar IRA: (i) it resulted from demands of the industry, and not the State; (ii) the lack of definition about which Ministry the agency should be associated with and (iii) its status as an independent regulatory agency. Its status as an IRA differs from other organizations that support the film industry (Fornazari, 2006), as well as agencies that regulate the audiovisual sector, but do not provide financial or regulatory support to the film industry, like the US Federal Communications Commission.
The case selection at the state level followed a different rationale. One of the most striking characteristics the agencification process in the states are the recurrent delays in the implementation and reformulations of these IRAs (Pó & Abrucio, 2007). The case of Rio is emblematic not only because illustrates this dynamic but because -different from other states -the reformulation process involved the transition from a multissetorial model -the most popular at the state level -to the unisectoral model -the rarest. As Box 2 indicates, about 27% of all the IRAs created at the state level were extinguished or reformulated. ASES, in the state of Sergipe, was replaced by AGRESE, in 2009, but some of its fundamental procedures -as tariff revision and readjustment, public consultation and the establishment of an ombudsman -were included only last year, by the state Law 8.442/2018. The dynamics at the Southeast -and by far the richest region in Brazil, in terms of GDP -are quite interesting. This is the only region with sectoral agencies at the state level, and the interchange between the two models is a notable feature of this region. Minas Gerais went from a multisectoral non-existent agency, ARSEMG, to a sectoral one ARSAE, created 11-years after the first attempt to establish an IRA in that state. Espírito Santo followed a sinuous path. It went from a multisectoral model, with AGESP in 1998, to a sectoral one, creating ASPE and ARSI, in 2004 and 2008, respectively, and back to a multisectoral model, with the merging of these two agencies in the current ARSP multisectoral one. São Paulo, on the other hand, followed a reasonable trajectory in terms of IRA creation: its first two IRAs were unisectoral and the energy regulatory agency was transformed after 10-years to include the sanitation sector. The case of Rio, which will be analyzed in this paper, resembles the case of Espírito Santo, and its first multisectoral agency was dismembered into two sectoral agencies, AGETRANSP and AGENERSA. Yet, differently from that state, the agencies were created at the same time, and, despite an attempt to merge them -by the PL 2.750/2009 -both agencies are still operating. The data was collected from three different sources: (i) official documents (the Official Gazette, minutes from congressional assemblies, open-access documents of public bodies, legislation pertaining to regulatory agencies); (ii) bibliographic review (previous research on the topic) and (iii) field research (semi-structured, in-depth interviews). This research was complemented by observation, and seven Directors of both federal and state IRAs were interviewed during the VII and VIII Brazilian Regulatory Congress, hosted by the Brazilian Regulatory Agencies Association, (ABAR), in 2011 and 2013. The documental research encompassed a total of forty-six IRAs, ten at the federal level and thirty-six at the state level. Subsequently, a bibliographic and documental research was conducted, regarding: i) privatization at the Union and state-level, ii) the renegotiation of states' debt, and iii) Union-level norms that could potentially result in the creation of a regulatory agency. The goal was to understand the top-down mechanisms at play in the agencification process.

BOX 2 IRAS ANALYZED
To explore the process of creation of Ancine, Asep, Agetransp and Agenersa, a fieldwork was conducted, including nine in-depth semi-open interviews with durations varying from one to three hours. Five of these interviews contributed deeply to the comprehension of the regulatory environment of Rio de Janeiro and the remaining four were pivotal in understanding the process that led to the creation of Ancine. The interviewees are to remain anonymous, and a summary of their professional experience can be found in Box 3. Additionally, two previous studies (Alvarenga, 2010;Fornazari, 2006), and many media newspapers remounting the time of the agencies' creation were extremely helpful in guiding the fieldwork.

BOX 3 INTERVIEWEES
Identification Considering the lack of data, the selection of respondents was crucial to understand these two cases. It must be emphasized that the majority of interview subjects had extensive participation in several phases of the Brazilian agencification, playing diverse political roles (often simultaneously) during this process. Hence, given the extensive and relevant trajectories of the interviewees, they sometimes contributed in more than one aspect to the research, allowing for a better understanding of the peculiarities of the Brazilian context and the diffusion mechanisms at hand, notably top-down and horizontal ones. Moreover, the interview protocol was based on the concepts identified in the literature review, and open questions were introduced to capture contextual factors. The interviews -except one -were recorded and transcribed. As the author itself conducted the transcriptions, all documents were first read entirely and in depth and then, in second and third readings, the document was coded according to the concepts identified in the literature, following a deductive approach to content analysis. Box 4 presents an example of how the data was coded.

Categories Interview extracts, emphasis added
Horizontal Mechanisms (at Ancine) "But, at that time, it was fashionable to adopt regulatory agencies. [...] Creating a new public body at that time, when the government's attitude was toward reducing the size of the state, was not a good idea. As independent regulatory agencies (IRAs) were being created, then creating one more was easier. I believe the IRA model made it possible to put into practice the filmmakers' idea of having one public body exclusively to the movie industry" -I3 Bottom-up Mechanisms (at ASEP) "For a simple reason. Both governments were from PSDT, both governments made public concessions, the first [to do that]. And public concessions assume the existence of an independent regulatory agency to regulate. This is in theory, that's how it is supposed to be, but this is not the reality. At that time, I had the illusion that it would be the reality. You need an IRA because a private firm will provide a public service, and it needs to be regulated by the State, which is the granting power." -I8

THE BRAZILIAN AGENCIFICATION PROCESS: TRADITIONAL MECHANISMS OF DIFFUSION
In Brazil, explanations focusing on bottom-up, top-down and horizontal mechanisms have prevailed in literature, which seems natural since the period in which IRAs were created coincided with a national process of desestatization that marked the opening of the infrastructure sector to private capital. In that sense, important research on the subject have attributed the establishment and the level of independence of Brazilian IRAs to the need of credible commitments (Mueller & Pereira, 2002), political uncertainty (Melo et al., 2010) and blame-shifting (Melo, 2002). Top-down mechanisms related to international pressures have also played an important role in the Brazilian agencification process, particularly in the case of infrastructure IRAs (Melo, 2002). This research also revealed the importance of internal or national top-down mechanisms. Specifically, two federal laws were of great importance in the diffusion of the agency model at the state and municipal levels. The first is federal law number 9.496/1997, which deals with goals and commitments as to "privatization, permission or concession of public services, administrative and patrimonial reform" as a counterpoint to debt renegotiation. Every Brazilian state, except Tocantins and Amapá, adhered to the fiscal adjustment program of 1997/98 and, with it, to the goals and commitments established in the aforementioned law, at least to some extent. The role of the Union was emphasized during interviews on several occasions. One of the interviewees, I9, also revealed that more than simple incentives, the Union exerted severe pressure over state governors, who were left with no alternative but to accept the package offered by the federal government.
The second law is the new regulatory framework for sanitation, established by federal law number 11.445/2007. Article 11 of this law attaches the validity of contracts to the existence of a regulatory entity. Since its promulgation, at least twenty-six sanitation IRAs have been established, and they continue to grow at fast pace. Such "new" surge of IRAs diffusion leveraged by the sanitation law has already revealed important institutional innovations, such as the creation of IRAs organized as consortiums of different municipalities (Souza, 2013).
Brazilian federalism may also be seen as a top-down mechanism. This can be illustrated by the distinct role that the federal regulatory agency of energy, Aneel exercised alongside the state agencies throughout the decentralization of activities and financial resources. The law that creates the agency is responsible for establishing the conditions under which decentralization is possible. In practice, decentralization requires the existence of a regulatory agency, in the terms established by Aneel. Therefore, if any Brazilian state wishes to have some power over the electricity distributors, it must have an IRA with an organizational structure and function similar to that of Aneel.
The decentralization of activities, besides bringing state government and local society closerand potentially benefiting consumers/voters -restores the power of local governors. Although the privatization of electricity distribution was the responsibility of Federal government, the approval depended, to a large extent, on the consent of governors, who were losing their influence on the state-owned enterprises (Melo, 2002;Olivieri, 2006). With decentralization, state governors were able to maintain relative power over the concessionaires without burdening the state treasury -since Aneel transfers the financial resources required to implement the activities. In exchange, they had to establish state regulatory agencies with acceptable levels of regulatory governance. Currently, ten state agencies are affiliated with Aneel, three hold an agreement of interest and fourteen are not affiliated (see Box 5). Some of the 14 non-affiliated states are relevant in economic terms, such as Rio de Janeiro, Minas Gerais and Bahia. One possible reason for the absence of decentralization in these states, beyond the official explanation (i.e., that state regulatory agencies do not have acceptable levels of governance) is the lack of interest of some state governors, as mentioned by interviewees I6 and I9. For the state agencies that do not need the financial resources of Aneel, the blame-shifting strategy is quite attractive, particularly in sectors with high political costs. The case of manhole explosions in Rio de Janeiro illustrates the success of this strategy, as highlighted by interviewee I9. The fact that Agenersa, the agency that could have power over electricity concessionaires, doesn't need Aneel's financial 'help' endorsed the governor to blame Aneel for the explosions. The role of Aneel in furthering IRAs at the subnational level is a typical powerful contextual factor.

AGENCIFICATION IN CONTEXT: THE IRA MODEL REINTERPRETED BY LOCAL ACTORS
The unexpected outputs of the Brazilian agencification process -i.e., extension of the agency model to economic sectors without natural monopolies to regulate or concession contracts to monitorate and the delays and reformulations in the IRAs adopted at the state level -can be interpreted as a result of the distinct and creative ways in which local actors -bureaucrats, film industry, and politicians -(re) interpreted the concept of agencification. This (re)interpretation -or the adaption of the agency model to the local needs -will be discussed next.

Bureaucrats' Interpretation: Administrative Flexibility vis a vis Regulation
The horizontal mechanisms of diffusion can offer a richer understanding of Brazilian agencification when related to the interpretation of the IRA model by the existent bureaucracy. Such interpretation is associated with a likely "confusion" between the bureaucracy's need for greater administrative flexibility and the acceptance of IRAs as a more autonomous model. It has been argued that some local actors (bureaucrats, in this case) interpreted IRAs as the only available opportunity to enhance management flexibility in a context where public organizations were historically limited by extreme rigidity and controls (Peci, 2007). The importance of a greater administrative flexibility will be resumed in the analysis of the creation of Ancine, in the next subsection.
Brazilian IRAs, based on a more flexible organizational model due to certain financial and administrative autonomies, became very attractive for other branches of existing bureaucracy. This apparent confusion is due, in part, to the strength of the ideas of the The Guiding Plan to Reform the State Apparatus (PDRAE). The main feature of the PDRAE was the need to reduce (or skip) the rigidity imposed by the Constitution of 1988, with greater emphasis on administrative flexibility (Bresser-Pereira, 2010). Although the plan delineates three types of organizations -the policy formulators, the executive agencies and the regulatory agencies -each having a different level and type of autonomy (Pacheco, 2006), in practice, executive agencies were scarcely implemented, while regulatory agencies proliferated (Costa, 2002), based on a sector-channeled regulatory reform (Levi-Faur & Jordana, 2005).
Utilities agencies initially fueled by market-oriented strategies came to proliferate in other social or non-utilities sectors mainly due to their attractive organizational model, which provides greater financial and managerial autonomy. Agencies have independent sources of revenue; they are not hierarchically subordinate to the ministries and have flexibility to bid and contract differently from the rest of the administration (Farias & Ribeiro, 2002). As a consequence of such interpretation by local bureaucracy, IRAs extended beyond the infrastructure sectors.

Industrial Actors' Interpretation: the "movie people" and the creation of Ancine
A second (re) interpretation of the IRA model was made by industrial actors, and, together with horizontal mechanisms, resulted in the creation of Ancine. One of the key interviewees made sure to state the following in the first minute of the interview: [...] first of all, let me tell you something: every public policy of the film industry, throughout the past several years, was conducted by the industry. It was not something from the Government to the Industry, but the industry has always proposed (policies) to the Governments. (I4) 1 The initial articulations by industrial actors, or the "movie people", occurred in the late 1990s and were spearheaded by Gustavo Dahl, possibly the most important individual actor in the process of establishing Ancine. Gustavo Dahl represented the industry interests during his professional trajectory and was the director of Embrafilme, a state-owned enterprise dedicated to the production and distribution of Brazilian movies. Since the late 1990s, industry demands contemplated the necessity of "a governmental supra-ministerial entity", such as "National Council of Audiovisual Policy" or a "Social Organization" (CBC, 2011, p. 170), since their diagnosis indicated the absence of a "systematic and ongoing policy for the development of film activity " (CBC, 2011, p. 169). The Brasilia Letter of 1998 also mentioned the need for the State to assume "its role as a regulator of economic activity" and that to perform this function, "an integrated performance of various ministerial levels, as well as its coordination" was required (CBC, 2011, p. 177). This integrated action, as highlighted by an interviewee, would be addressed with the creation of a Council, the "Superior Council for the Film Industry": At first, we didn't think about creating any public body. The idea was to establish a Council, which is the Superior Council for the Film Industry, and then identify the best way to deal with the public policy 2 (I4).
As supported by all interviewees and the industry documents above mentioned -especially the book of the Congresso Brasileiro de Cinema (CBC), Ancine was born of pressures by industrial actors in order to enhance the role of the State in supporting and promoting the national movie sector. Coincidentally, the moment when industrial actors finally rearticulated and took their demands to the Executive is precisely the "time of agencies" (I3), the moment when "government was making regulatory agencies" (I3). The agency model was seen as a legitimate form of state intervention at that time. In fact, the agencies were a fad -in September 2001 about 26 regulatory agencies had been established in Brazil, between the federal, state and municipal levels.
Besides the presence of horizontal mechanisms and pressures from industrial actors, the desire for greater management flexibility may also have influenced the creation of ANCINE. This becomes clear with the testimony of Carlos Diegues at the Senate Subcommittee on Film (Brasil, 2000): The Ministry's own structure impedes agility, something the cinema industry would require of it. In the mid-term, in a relative short horizon, we would need an agency or a secretariat dedicated to the audiovisual and connected to the Presidency. I'm not talking about the old type of vertical administration [...]. I'm talking about horizontal administration -like that of the National Energy Agency and of the National Oil Agency -capable of articulate the needs of cinema in several areas of the Executive (Emphasis added).
The creation of an IRA for cinema was intimately connected to a moment when the Brazilian State was rethinking its role and size. Attacks on the bloated state apparatus and the success of IRAs, which were "the face" of this new State, were immediately associated. One interview subject stated that "creating a new public entity at that stage, when the government stance was reducing the size of government, was a taboo" (I3), corroborating the notion that an agency was the most legitimate way to manage the public thing.
Therefore, the creation of an IRA for cinema was the end result of a mixture of (i) a mimetic process that granted legitimacy to this "new" organizational form and (ii) contextual variables stemming from the interpretations of the dominant industrial actors of the time. Industrial actors wanted an encompassing and coordinated support from the State, and found the agency model the most likely way to put through their demands.

Politicians' Interpretation: The Case of Rio de Janeiro
The change in the regulatory environment of Rio de Janeiro illustrates the importance of another political actor, politicians themselves, in the creation of Brazilian IRAs. Rio de Janeiro's first IRA, Asep, was established in 1997, in a moment of privatization of public services and renegotiation of the state debt. The State Desestatization Program (PED), established by the state law number 2.470/1995, had as its main goal to reduce the public debt and improve finances. Thus, the creation of ASEP and its institutional design seemed a natural consequence of the state privatization process, which in turn, reflected what was happening at the federal level and at most of the other states of the federation (I5, I6, I7, I8 and I9).
However, Asep did not last long. In 2005, the agency was extinguished, and two others were created: Agetransp and Agenersa. Apparently, these two agencies did not result from the same diffusion mechanism responsible for the creation of Asep. This reformulation occurred in 2005, during the administration of Governor Rosinha Garotinho, in a time when ASEP was finally operating again after almost being "closed down" during the Anthony Garotinho administration, as stressed by one of the interviewees (I7): He (Garotinho) won in 1998, took office in 1999 and shut down the agency. Shut down, exonerated everyone who was there [...]Suffices to say that the concession contracts disappeared. I was in this agency until the beginning of 1999. After the agency was shut down, the counselors were still appointed by the governor and confirmed by the state's legislative assembly. They had mandates, they continued doing nothing, because no one could get them out of there. 3 (I7) While this represented a significant change in Rio's regulatory environment, the extinction of ASEP and the subsequent creation of AGETRANSP and AGENERSA remain scarcely explored or understood in the literature. An initial intuition concerning this reformulation is the passage of the multi-sector to the single-sector model. It is worth reminding that multissectoral agencies predominate at the state level -80% of all IRAs created at the state level are multissectoral, as Box 2 indicates. Possible reasons for this predominance are (i) the need for economies of scale, (ii) the scarcity of human resources and (iii) cost reduction requirements (Melo, 2002). In spite of this fact, the rationale is that states that have to deal with more concessionaires require more specialized organizations. São Paulo -Brazilian richest state in terms of GDP -for example, opted for the sectoral model, and after some years extended the scope of one of its sectoral agencies, having now one sectoral and one multisectoral agency. The belief that the single-sector model, which is more specialized, would be more effective in dealing with the concessionaires is a potential explanation for the change in 2005. Interviews indeed indicated that this is the "official explanation" (I5 and I6).
Yet, the apparent concern with the levels of regulatory governance in the state was not translate to practice. The interview subjects were asked about possible improvements to the levels of regulatory governance and they all share the opinion that it did not happen. Curiously, AGETRANSP was highlighted by two of the interview subjects (without ever being asked about it directly) to have even lower levels of regulatory governance than AGENERSA. Not surprisingly, while AGENERSA's structure of political appointment positions is available on its website, AGETRANSP refused to provide this information even after a formal request. Moreover, AGETRANSP has been the target of much criticism within civil society and state legislature 4 . Another indicator of a governance improvement would be the strengthening in the quality of the technical body, by competitive tendering, which did not occur. After more than twenty years since the establishment of the first regulatory agency in the state, only one tender was held, and it happened before the extinction of ASEP and creation of AGETRANSP and AGENERSA.
While two interview subjects emphasized the "official explanation" (that the observed change was motivated by a concern with the regulatory quality), all interview subjects, save one 5 , mentioned the creation of political appointment positions as the de facto explanation for the dismemberment of ASEP.
Summing up, the regulatory re-organization in the state of Rio de Janeiro is justified by the [apparent] concern with the levels of regulatory governance of the state and by the "discovery" by governors that the director-level positions could be used as an effective instrument of bargaining with the state legislature.
The lack of improvement in the levels of regulatory governance with the dismemberment of ASEP reinforces the alternative explanation mentioned by the interview subjects. That is, the reformulation of Rio de Janeiro regulatory environment is related to political motivations, mainly the importance of director-level positions. Considering that the privatization process and the subsequent creation of regulatory agencies were driven by Union incentives, it is natural that, in the long run, the governors, who are historically the most important actors at the state level (Pó & Abrucio, 2007), would alter the initial model according to their political interests.
The example of Rio de Janeiro is extremely curious, since it illustrates the presence of bottom-up and top-down mechanisms of diffusion, as well as and contextual variables. The agency was created as a natural consequence of the privatization process in Rio, inspired on the international experience, notably the utilities commissions, and with the encouragement and sympathy of the Union and the World Bank, as highlighted by interviews (I5, I6, I7, I8, I9) and verified in the available documents.
Conversely, the change observed in 2005, i.e. the creation of two agencies and the extinction of ASEP, seems uniquely influenced by contextual factors related to the need to assure governability in multiparty presidential systems. The use of trade-worthy coalition goods between the Executive and Legislative -such as political appointment positions in the public service -has been highlighted as one of the key factors responsible for the 'surprising success' of Brazil's multiparty presidentialism (Pereira & Melo, 2012). This paper shows that such coalition goods also played an important role in the Brazilian agencification process, explaining the regulatory reorganization operated in Rio de Janeiro.

CONCLUDING REMARKS
This paper investigated the diffusion of Brazilian regulatory agencies at different levels of government, showing that, although Brazil has experienced a "boom"of IRAs' creation, institutional and cultural values of the local political-administrative apparatus shaped their adoption. In practice, the Brazilian agencification is best understood as a result of the different mechanisms at work, i.e. different rationales for the adoption of regulatory agencies, and associated contextual variables. Therefore, anomalies in the diffusion of IRAs can be attributed to both the different theories that justify the adoption of an agency, as to how the local political actors interpreted and appropriated the agency model.
The three diffusion mechanisms appointed in the literature, namely bottom-up, top-down and horizontal are presented in the diffusion process of IRAs in Brazil. In general, the first two are highly correlated, because the fiscal adjustment program, the privatization processes and the need for credibility went hand-in-hand. States that had their debts renegotiated also started the process of privatization and, almost as a natural consequence of it, established their first regulatory agencies. In practice, Brazilian IRAs were not a legal innovation within the existing institutional landscape. IRAs were (re)invented, by adapting an existing legal model to the new rhetoric of regulatory reforms.
Reinforcing Moynihan's (2006) argument, this paper identified that a key contextual factor that boosted the Brazilian agencification process was the (re) interpretation of the agency model by local actors, such as bureaucrats, politicians and industrial actors. The possibility to overcome administrative rigidity via regulation, the presence of strongly articulated industrial actors, such as the "movie people" and the importance of trade-worthy coalition goods in Brazilian multiparty presidentialism are the main drivers for these interpretations.
Ancine's case represents a good example of the interpretation by bureaucrats and industrial actors. While the creation of Ancine may be explained by horizontal mechanisms of diffusion and a strategic 'confusion' between administrative flexibility and regulation, the agency would not have been created were it not for the pressures of the "movie people". These actors, and not the government, have been conducting the public policies of the sector for a long time, as highlighted in the interviews. This may explain the recent proposals made by industrial actors for a separation between the functions of regulation and support. Federal law n. 12.485/2011 strengthened the sectoral fund that now 'has an amount never before available to the industry' (I1). So, with the objective of eliminating the bureaucratic bottlenecks of government agencies, industrial actors have already mobilized towards removing the support function from Ancine. The agency would act alongside Anatel, being moved to the Ministry of Communication, and the sectoral fund would be transformed into an autonomous support body 6 . This endorses the statement made by one of the interviewees, that "one of the sources of the permanent crisis in the film organizations in history" is that "the filmmakers have always wanted state agencies dedicated to cinema, but only if they could direct them" (I3).
The interpretations by politicians have not yet been associated to the creation of IRAs in Brazil. As discussed here, one of the prominent features of the Brazilian agencification process is the dynamic at the state level, either by their extinction, non-implementation, extension of their responsibilities or ultimate dismemberment. Each reformulation may have a different motivation and can highlight a different contextual variable. This research chose to analyze the regulatory environment of Rio de Janeiro, more specifically the decision to extinguish Asep and create Agetransp and Agenersa. Field work indicated that this decision was motivated by the possibility of creating more political appointment positions and, consequently, to increase the Executive's bargaining power with the Legislative assembly. An alternative hypothesis was explored, which is a possible attention to technical aspects, since a multi-sector agency would have greater difficulty in overseeing the distinct concessionaires. However, no improvement was observed in the levels of governance of the state agencies, especially Agetransp, which strengthens the belief that the change in Rio's regulatory environment was motivated by aspects of political retail, as emphasized in the interviews.
This work analyzed the Brazilian agencification process from a temporal perspective and from two distinct theoretical standpoints. Future research can explore how the variety of forces behind the Brazilian agencification process affects the outputs and outcomes of the regulatory activity. This is important because part of the literature that deals with diffusion argues that such processes can lead to suboptimal outcomes, since (i) the policy in question was designed for the needs of others and/or (ii) the policy selected is superior to the local capabilities (Elkins & Simmons, 2005). This is especially relevant for developing countries, which seems to be always trying to catch up with the developed world, mixing outside ideas with its local endowment. In this sense, the Brazilian agencification process can be seen as an effective process that balanced the need for credible commitments, the pressure of the Union and the World Bank for privatization and regulation by independent bodies, and contextual features, increasing the legitimacy of State intervention while embracing local needs. On the other hand, and in a more pessimistic interpretation, one can consider that the diffusion and adaptation of IRAs to local needs inhibited the emergence of innovative practices more adequate to the national context, creating distortions and instability and, therefore, contributing to the persistence of the status quo of a developing country.