“This book could not have been written in Europe.” That was Roger Bastide, writing in Cahiers Internationaux de Sociologie in 1962, on the doctoral thesis of a former student of his at the Universidade de São Paulo. The thesis was Capitalismo e escravidão no Brasil meridional. The student was Fernando Henrique Cardoso, then 31. Bastide’s point (that nowhere in Europe would a sociologist mix Weber, Marx, and Sartre in the same paragraph) was meant as praise, with the hint of condescension that always lurked in his praise.

The description of the eclectic register would become, over the next twenty years, a method. He would later reveal, in a 2022 documentary, that the reception of the book in Brazil wasn’t as positive as Bastide might have expected. In a presentation of the book at the Palácio do Itamaraty, the headquarters of the Brazilian diplomacy body, some diplomats complained to him that promoting research on slavery and contemporary racism would paint Brazil in a bad light to the international community.

Two years after the review, on May 1, 1964, Cardoso boarded a flight from São Paulo to Buenos Aires with a small suitcase. He had refused to depose at one of the new military regime’s police inquiries, hidden for several weeks in friends’ apartments, and now he was leaving. He left his family behind. He has called the moment of departure, in interview after interview, the worst of his life.

He was 32, and his second book (on the political culture of São Paulo industrialists, in many ways an oblique prediction of the coup that had just happened) had appeared two months earlier. He would spend the next three years in Santiago, in what the Argentine sociologist Fernanda Beigel has called the world capital of the Left, where the Spanish exile José Medina Echavarría (one of Weber’s first translators into Spanish) gave him a job at the CEPAL’s planning subsidiary, and where he met the Chilean sociologist Enzo Faletto in the ILPES corridors. The two of them wrote, in Spanish, at the kitchen tables of various Santiago apartments, the book that would become the most internationally influential of his career.

Five years after the flight, in April 1969, he was back in São Paulo. He had returned the previous year, against the advice of his Chilean and French friends, to compete for the chair of Political Science at the USP after Lourival Gomes Machado’s sudden death. He had won the chair in October 1968 by competitive examination and had taught for six months of one course on political theory.

But, in the 13th of december, 1968, AI-5 (Institutional Act No. 5, 1968), the most repressive decree of Brazil's military dictatorship, which granted extraordinary powers to the executive branch, suspended habeas corpus for political crimes, and marked the beginning of the regime's harshest period, known as the "Years of Lead”, was passed. And so, his name was scribed off the university payroll by administrative decree.

He went to collect his time-of-service compensation the next week. The clerk at the social-security counter looked at the paperwork and said, “este aqui está morto.” He pointed out that he was sitting in front of her, alive. She corrected the file and handed him the money. “Coitado, tão jovem e já aposentado”. The sociologist who would, in the next decade, build the most influential account in Portuguese of how state bureaucracies capture power from the governments that nominally run them had just been processed by exactly such a bureaucracy as a corpse.

He’s 95 today. For about twenty years now, commemorative essays about him have been about the eight years in Brasília. The Real Plan, the privatizations, the long quarrel with Lula. The biographies all stop at 2002 and read backwards through the politics. The Fernando Henrique Cardoso who existed before he was FHC has been almost entirely covered over by the brand, even in the universities where his books once organized the curriculum. I’d like to dig him back out.

The story the orthodox Brazilian left tells about that work, and has been telling for fifty years, runs as follows. A young Marxist sociologist, trained at the USP under Florestan Fernandes, drifts slowly rightward across the 1970s. He abandons Capital for a watered-down Weberianism, calls his old comrades stagnationists, ends up in the early 1990s presiding over privatizations and constitutional amendments, and spends his retirement writing op-eds against the workers’ party that inherited his old constituency.

What actually happened, in the record between 1962 and 1984, is more peculiar. The eclectic register Bastide noticed in 1962 turned, across two decades of exile, censorship, and forced retirement, into something more rigorous. A young dialectical Marxist slowly drew Weber deeper into his theoretical apparatus, producing by the late 1970s a sociology of political domination in dependent capitalism that anticipates most of what would later become respectable in Anglophone historical institutionalism.

Burguesia de Estado, he called it. Anéis burocráticos. Regime burocrático-autoritário. We say state capture now, and regulatory rents, and bureaucratic insulation, and path dependence. When we write it in English in our papers, with footnotes and citations to Skocpol and Evans and to North, Acemoglu, and Robinson. The terminology may be younger and flashier, but the analytical object is the same. The Anglophone political science of the 1980s and 1990s rediscovered most of Cardoso’s objects without rediscovering Cardoso. By then, he had become the President of Brazil, and the work was easier to read through the politics than to read on its own terms.

I want to try to read it on its own terms.

São Paulo, between the lab coats and the seminar

In one of the long oral-history interviews Cardoso gave to the FGV team in 2011, he tries to explain what it was like to enter the Faculdade de Filosofia, Ciências e Letras of the USP in 1949, wanting to “do socialism” and finding Parmenides instead, the pre-Socratics, and Kant in German. The professor of Kant was a Frenchman from the Collège de France who assigned the bibliography in German, and Cardoso’s first paper on the subject was on Parmenides. He’d wanted to change the world, but he got five out of ten on the assignment.

The Faculdade de Filosofia in 1949 was, intellectually speaking, a French province. Roger Bastide. Claude Lévi-Strauss, who had passed through and was still being read. The philosopher Paulo Arantes would call the college at the time an “overseas French department”. The Cardoso who emerged from it in 1953 with his degree in Social Sciences, and who took up his first assistantship under Florestan Fernandes, was a creature of this late-1940s French sociology. His doctoral thesis, Capitalismo e escravidão no Brasil meridional, defended in 1961 and published by Difel in 1962, would carry the marks: long references to Lukács and Sartre, and an awkward, polemical attempt to graft a Marxian dialectic onto a functionalist substrate that Florestan was still teaching out of Parsons and Robert Merton.

The famous division was Florestan’s: each saint on its own altar. Marx for slow structural transformations across centuries, Weber for ideal-typical comparisons of forms of authority, and Durkheim for supra-historical regularities…. Parsons was, in Cardoso’s later phrase, “Florestan’s bible.” It was scientific sociology with a capital S. The men literally wore white lab coats to lecture; Cardoso confirms this in the FGV interview, with what reads on the page as embarrassment four decades later. The lab coats were a way of marking distance from the older Brazilian tradition of the ensaio de interpretação nacional, the classic essays of the Gilberto Freyres, the Sérgio Buarques, the Caio Prados, and so on… whom the USP sociologists regarded as too literary, too ideological, insufficiently empirical.

For the paulistas, the proper sociologist did fieldwork and large ethnographies. Florestan and Bastide ran a long research project on the integration of Black populations into the class structure of São Paulo, and Cardoso did the fieldwork. He went into the cortiços downtown (there was one called the Buraco Quente, a few blocks from where the FGV interview was being conducted sixty years later, as the interviewer noted), and he ran surveys, and he got himself trained, badly, in early American social statistics.

The atmosphere was militantly anti-ISEB. The Instituto Superior de Estudos Brasileiros in Rio was, in those years, the institutional embodiment of national-developmentalist ideology: knowledge as a tool for state intervention and sweeping interpretations of Brazilian history were required for producing a coherent national project. The USP sociologists thought this was ideology with a thin academic veneer. They wanted science, and the young Cardoso did too.

And then, in 1957 or 1958, José Arthur Giannotti came back from Paris, where he had been studying philosophy with Lucien Goldmann, and proposed something different: A small reading group on Das Kapital every fortnight. The original participants included Cardoso, his wife Ruth Cardoso, Octavio Ianni, Paul Singer, Fernando Novais, and a slightly younger Roberto Schwarz. Over the years, others joined, such as Bento Prado Jr., Francisco Weffort, Leôncio Martins Rodrigues, and Michael Löwy. Florestan, however, was not invited. He found out anyway, ran into Cardoso on the street one day carrying a book, asked who that old man was on the cover (he meant Lukács), and registered his displeasure. The seminar was, from his point of view, a regression. A retreat from scientific empirical sociology back into the kind of German philosophical ethereal and abstract mumbling he had spent twenty years trying to shake off.

Florestan was not entirely wrong, however, as the Marx Seminar was a peculiar reading of Marx. It was not Stalinist. It was not Trotskyist. It was not, despite Giannotti’s eventual move toward an Althusserian-structuralist reading, ideologically aligned with any of the Parisian Marxisms of the time. Roberto Schwarz, in his famous 1995 essay “Um seminário de Marx,” would describe it as “a Marxism without slogans, equal to the level of contemporary university research, open to reality, with no skeletons in the closet and no authoritarianism to conceal.” It was a Marxism read from Capital outwards: philological, rigorous, refusing the easier categories of either national-popular front politics or armed-struggle voluntarism. They read in five languages and argued over the translations. The participants included card-carrying communists and Trotskyist sympathizers, and one or two socialists with no party, but the political differences did not affect the textual work. They sat down with the book.

This Marxism, it turns out, did not make it into the canon of Brazilian Marxism. The six-volume História do marxismo no Brasil, published by Editora da Unicamp between 1991 and 2007, the most ambitious attempt to historicize Marxist thought in Brazil, has no chapter on Cardoso. Guido Mantega, the future Minister of Finance of many of the Workers Party governments, gives the Marx Seminar a page and a half of polite recognition in the 1995 edition of A Economia Política Brasileira, his tedious and half-baked attempt to summarize a history of Brazilian economic thought, and moves on. Many left-leaning humanities scholars would classify it as “Armchair Marxism (Marxismo de Cátedra)”, a derogatory term on the Brazilian left meaning roughly “Marxism that was too academic and too liberal-democratic to count as real.”

Pedro Luiz Lima, in a long, careful essay published in Novos Estudos CEBRAP in 2022, calls Cardoso “the accursed share” of Brazilian Marxism. Lima claims that the exclusion has very little to do with Cardoso’s later political career; it would have happened anyway. It has to do with four things internal to the early Marxism itself: a refusal of discipleships (no Lukácsian school, no Althusserian, no Gramscian, just Marx); a skepticism about the imperative of revolution; an explicit critique of nationalism as a category of political analysis; and what Lima calls, with some precision, “a Marxism of reconciliation” — a Hegelian instinct to balance the contradictions of the real rather than to resolve them by negation. None of that fit the prosodic standard of mid-century Brazilian Marxism, which was nationalist, with revolutionary pretentions, and increasingly Cubanist from 1959 onwards. Fernando Henrique Cardoso would be left of the Brazilian Marxist canon precisely because he didn’t follow his contemporaries’ fashions and fetishes.

But the Marx reading was real and deeply influenced his intellectual production. Capitalismo e escravidão no Brasil meridional is the work of a young Marxist who has decided to argue, against the orthodox Brazilian Communist Party (PCB) line of the late 1950s, that there was never any meaningful feudalism in Brazilian history, and that the PCB’s anti-feudal-alliance strategy is therefore politically meaningless.

By the late 1950s, the leaders of the PCB decided that the agrarian economy of Brazil at the time was to be characterized as a feudal economy with servitude and so on; therefore, Brazil required a bourgeois capitalist revolution that would create the proletariat, and then, as Marxists usually believe, the capitalist contradictions of this new economy would trigger the workers’ revolution. As such, they decided that the official party line at the time would be to support an interventionist and nationalist agenda that would industrialize the country, creating a new bourgeoisie that would trample the agrarian basis of power and form an urban worker class.

Cardoso, however, would argue that what you would find in nineteenth-century Rio Grande do Sul state was a specific variant of the capitalist mode of production, organized around slave labor, articulated to a particular pattern of regional landed-military domination. Therefore, the main consequence of the book’s thesis is that the form of labor (slavery) is internal to the capitalist totality, not a vestigial pre-capitalist remainder. If the thesis was right, the PCB’s strategic argument would fall apart, and so would the easier developmentalist argument that Brazil simply needs to “catch up” to a supposedly universal European pattern of bourgeois development.

The thesis is, in important respects, a polemic. It is also the first published book in which Cardoso’s distinctive method appears: the dialectical movement between general categories drawn from Capital and the historical specifics of a peripheral, slavery-based capitalist formation. The eclectic register, on Bastide’s reading, was the peripheral sociologist’s productive license, the freedom of someone working at a distance from the metropolitan academic policing that would have made the mixture unthinkable in Paris or Frankfurt. He may have been condescending; that was Bastide’s house style, but he was also right: This method would carry Cardoso through every major book he wrote for the next twenty years.

Cardoso’s original preface attacked functionalism so openly that Florestan, who was the supervising professor and would chair the defense committee, refused to accept it. He thought the criticism was personal. Cardoso, lying at home with measles, agreed to soften the language but keep the substantive argument. In the Tempo Social interview that Marcelo Ridenti conducted with him in 2013 (only published in 2025), Cardoso remembers the scene with real affection for the old man Florestan, who had been more than a mentor; he had been his neighbor on Nebraska Street in the Brooklin neighborhood of São Paulo, his early professional patron, and his friend. The fight over the preface was tough, but the friendship survived it… the functionalism influences on FHC’s work, however, did not. From 1962 onwards, Cardoso’s published work would be more openly Marxian than anything coming out of Sociology 101 at USP.

Attacking the consensus on the Brazilian industrialists’ political behavior

The famous-among-specialists, almost-unknown-otherwise book is Empresário Industrial e Desenvolvimento Econômico no Brasil, defended as a livre-docência thesis in 1963 and published two months before the April 1964 coup by Difusão Européia do Livro. I want to spend some time on it, as it is my favorite book by FHC, but it is also the book in which the political analysis Cardoso would carry through the rest of his career first comes into focus.

The empirical material came from the fieldwork sponsored by the Cesit, the Centro de Sociologia Industrial e do Trabalho (Center of Social and Labor Sociology), which Cardoso and Florestan had organized in 1961 with funding cobbled together from the recently created Fapesp (the public São Paulo Research Foundation) and the Confederação Nacional das Indústrias (National Confederation of Industry, CNI). The CNI funding came through a personal connection: Fernando Gasparian, an old friend and the husband of one of Ruth Cardoso’s college classmates, was running the CNI as an interventor for the João Goulart government, and arranged the grant.

The reader will notice this was an unusual arrangement even by Brazilian standards: a Marxist sociologist taking money from the national industrialists’ federation to write a sociological study of the political mentality of industrial entrepreneurs. Cardoso has not denied any of this in his interviews; his methodological argument was that you cannot understand the political behavior of a class without sitting in front of its members and asking them questions, and to do that, you need to be let in the door. Gasparian opened the door.

The fieldwork involved long interviews with São Paulo industrialists about their views on the labor question, the role of the state, import-substitution industrialization, the agrarian sector, and (the politically hot question in 1962 and 1963) the relationship between Brazilian and foreign capital. The book’s published form is heavily theoretical; Cardoso was, after all, presenting a thesis for the livre-docência, the qualification immediately below the full cátedra, and the conventions required a substantial theoretical apparatus. The interview material is woven into a long argument about the structural conditions of late industrialization in dependent capitalism. But the empirical anchor is real, and the book’s strongest claims come from it.

The central claim cuts simultaneously against the PCB, against ISEB, and against the developmentalist economic-policy establishment around the Goulart government. There were no “national bourgeoisie” in Brazil, in the strong sense in which the PCB and ISEB used the term, but rather a heterogeneous, recently-formed industrial fraction of Brazilian capital, operating in conditions of imperfect competition, with prices indexed to imported substitutes, without serious organized labor pressure, with low rates of technological reinvestment and weak innovation incentives.

Its interests are structurally contradictory. To pursue a political project of national-popular development, the industrial bourgeoisie would have to ally with urban popular sectors against the agrarian elite and against foreign capital. But that alliance would threaten the very rents that the existing market structure generates. So the industrialists renounce the political project of leadership to protect the conditions of accumulation. Daniela Costanzo and Rafael Marino, in a sharp 2022 essay for the Boletim Lua Nova, put it well: the analytical move is essentially Marx’s analysis of the French bourgeoisie in The Eighteenth Brumaire. A class that accepts political nullification to preserve its private power.

This was a grenade thrown at the prevailing political strategies of the Brazilian left. If there was no national bourgeoisie capable of leading a developmentalist alliance, then the PCB’s national-democratic strategy was dead on the beach. The ISEB’s vision of an autonomous, state-led national development project was dead on arrival. The Goulart government’s reform program, which depended on cobbling together exactly such a coalition, was structurally vulnerable.

The book was published in February 1964, but, even before it could be trashed around by the communist hardliners, the military removed Goulart in the April coup. Cardoso has always insisted, plausibly, that the timing was a coincidence; the book had been written between 1961 and 1963, before the political crisis came to a head. But the analytical fit is unmistakable. The industrialists, faced with a popular-democratic mobilization that threatened the prevailing accumulation pattern, did what the book had said they would do: they accepted military intervention as the price of preserving the conditions of their profit.

André Singer’s recent work (the long 2015 essay Cutucando onças com varas curtas and the 2018 book O lulismo em crise) returns to Empresário Industrial to argue that the same structural pattern explains the impeachment of Dilma Rousseff in 2016. The 1964 industrialist and the 2016 industrialist, on Singer’s reading, behave the same way for the same structural reason. They are not strong enough to lead; rather, they are too weak to govern themselves. Singer’s argument has the additional merit of taking the early Cardoso seriously as a theorist whose categories survive the political career of his author. Whether Cardoso himself would have endorsed the application in 2016 is a separate question; I do believe, however, that the structural diagnosis stands.

Two things about Empresário Industrial matter for the later trajectory, and they have to do with the parts of the book that are not about the empirical industrialists at all.

The first is that the book is, conceptually, already a Marx-Weber hybrid. The categories of class, accumulation, and dialectical mediation are visibly Marxian, drawn from Capital and from Cardoso’s careful reading of The Eighteenth Brumaire. But the long theoretical chapters on the modern bureaucratized firm, on the distinction between traditionalism and the “spirit of capitalism,” on the typology of leadership orientations in late capitalist economies, are visibly Weberian; in fact, he ostensibly quotes Weber and Manheim. This book must be read in conversation with the German and Anglo-American sociologies of industrial enterprise (Berle, Sweezy, Wright Mills, Dahrendorf) of the 1950s and must be read, in retrospect, as the first published work in which Cardoso’s later theoretical fusion takes shape. He is using Marx to explain the historical specificity of peripheral industrial capitalism, and Weber to explain the bureaucratic form that economic enterprise takes inside it.

The second point is that the book introduces, almost in passing, the conceptual vocabulary that would become central to the theory of authoritarianism a decade later. The idea that in the bureaucratized modern firm “capital, not the capitalist, holds the secret that explains the system.” Therefore, it is the recognition that economic decisions in late capitalism are necessarily political decisions: it was an early version of the diagnosis that there was now a new managerial-bureaucratic elite that fuses with the state bureaucracy to produce a new pattern of domination, something that would be deeply expanded later by Guillermo O’Donnell and David Collier.

Santiago, or the accident of the CEPAL

The coup happened on April 1, 1964. The flight from São Paulo to Buenos Aires that I opened this essay with took off a month later, on May 1. The four weeks in between are a small but useful biographical episode that the published Cardoso scholarship has paid less attention to than it deserves. He spent the first weeks of the dictatorship moving among friends’ apartments in São Paulo, mostly that of Fernando Gasparian and his wife, with intermittent visits to Rio.

The Military Police Inquiry (the legal instrument used by the military to investigate and arrest “subversives”) summons had been delivered by hand to him right after the coup. He had refused it on the advice of a lawyer who told him that complying would mean an immediate detention with no clear legal endpoint.

The decision to leave the country was made, in the end, when the security police arrived at the Cardoso apartment looking for him while he was hiding two kilometers away… he has talked about the moment of departure as the worst of his life. The thirty-year political career that began with his Senate run in 1978 makes more sense if you assume that the four years he spent in Chile and France were, for him, a period of forced separation from the country he most wanted to fix.

After a quick stop at Buenos Aires, where he stayed a few days with José Nun, he moved to Santiago, where Nuno Fidelino de Figueiredo, an older friend who was already working at the CEPAL, met him. Figueiredo told him that José Medina Echavarría, the Spanish exile who ran the social-research division of the ILPES (the Latin American Institute for Economic and Social Planning, a CEPAL subsidiary), was looking for sociologists. Cardoso went to see him. Medina, who, as one of Weber’s first Spanish translators, had a reading of the German sociology that no one else in Santiago at the time had, hired him on the spot. Within a year, Cardoso was the de facto deputy director of the division.

It is hard to overstate how much the Santiago of 1964 to 1967 mattered to Latin American social science. The Argentine sociologist Fernanda Beigel has called the period “the world capital of the Left.” Almost everyone passed through… Celso Furtado was there, Aníbal Pinto, Osvaldo Sunkel, Theotônio dos Santos, Vânia Bambirra, André Gunder Frank, Ruy Mauro Marini, all of them were exiled there, and Francisco Weffort had been living in Santiago since 1962. Adolfo Gurrieri, Edelberto Torres Riva, and the young Vilmar Faria… They were all in the same Santiago apartments, the same CEPAL cafeteria, the same ILPES seminar rooms. They argued about industrialization, the failures of import-substitution, the role of the state, the future of the populist developmentalist alliances that had just collapsed in Brazil and Argentina. They argued, increasingly, about whether the CEPAL’s old prescriptions were still valid.

Cardoso, who had arrived from São Paulo with a Marx-Weber hybrid analysis of dependent industrial bourgeoisies, walked straight into this argument. By 1965, he was teaching at FLACSO and at the Faculdade de Economia da Universidade do Chile (where Vânia Bambirra was his teaching assistant). By 1966, he and the Chilean sociologist Enzo Faletto, whom he had met in the ILPES corridors, were writing what would eventually be the most internationally influential book of his career. They wrote it together, in Spanish, at the kitchen tables of various Santiago apartments. Medina edited the manuscript and, as Cardoso put it in his 2001 interview with Bastos, Abrucio, Loureiro, and Rego, “corrected the interpretations” page by page. The text was finished by late 1967, but CEPAL refused to publish it.

The reasons were mostly petty administrative nonsense. The CEPAL would not publish the manuscript because it mentioned countries and individuals by name, which CEPAL documents as UN documents are not supposed to do, so they decided to publish it independently. The book came out through Siglo XXI in Mexico City in 1969, and the first Portuguese edition appeared in 1970 through Zahar in Rio. The English translation, by Marjory Mattingly Urquidi, was not published by the University of California Press until 1979. By the time the English edition appeared, the entire field of Anglophone “dependency theory” had organized itself around competing books, principally Frank’s Capitalism and Underdevelopment in Latin America (1967). Cardoso and Faletto’s book entered the American debate as a delayed correction to a literature that had already taken shape in its absence. We will come back to this.

The book nobody read

What does the book actually say? It is a useful question, because Dependência e desenvolvimento na América Latina is one of the most cited and least read texts in Latin American social thought.

The first thing the book says is that peripheral capitalism is capable of producing development, if you understand development as real economic growth, industrialization, urbanization, and expansion of an internal market. Cardoso and Faletto were writing against a specific theoretical position, sometimes called “stagnationism,” that was widely held on the Brazilian left at the time, especially around some neo-Marxists and parts of the CEPAL itself.

The stagnationist view was that dependent capitalism was permanently incapable of producing growth in the periphery, because the imperialist core would always extract surplus and prevent autonomous industrialization. Therefore, the only route to development was the socialist revolution. Cardoso and Faletto thought this was wrong as an economic theory and disastrous as a political strategy. Read Empresário Industrial, they said in effect: there was industrial bourgeois accumulation happening in São Paulo at the moment. The structure of dependent capitalism was producing real economic transformation, but the transformation, however, was not the same as the transformation of nineteenth-century European cores. It was uneven, conditioned on foreign capital and foreign technology, and integrated into a world capitalist system rather than autonomous from it. Calling it stagnation is empirically wrong.

The second thing the book says is that there is no single path of dependent capitalist development. There are at least three structurally distinct patterns: The first is the enclave economy (mining, plantation agriculture), in which a small foreign-controlled extractive sector exists alongside a largely subsistence domestic economy. The second is the economy of national control, in which a domestic bourgeoisie controls the leading sectors and accumulates within national borders before slowly integrating into the international system. The third, and this is the conceptual innovation of the book, is what they called the “internationalization of the internal market.”

In this third pattern, foreign firms establish productive operations inside the dependent economy, using the internal market as their target. There would be manufacturing for the local middle class, producing in São Paulo for sale in São Paulo. The pattern is qualitatively different from the older imperialist pattern of raw-material extraction and finished-goods export. And it has specific political consequences: a new alliance between foreign capital, a local bourgeoisie that supplies and partners with foreign firms, and a state bureaucracy that regulates the whole arrangement. Cardoso’s later term for this pattern was “associated-dependent development.” The word “multinational” had not yet been coined. The phenomenon was something he and Faletto were watching come into being in real time.

The third thing the book says (and this is the methodological heart of it) is that the dynamics of any particular peripheral capitalism cannot be explained from external relations alone. The internal class structure, the political coalitions that organize accumulation, the political form that the state takes, the conflicts and alliances among different fractions of the dominant class: none of this is derivable from the country’s position in the world system. They are the result of historical struggles inside the country. Cardoso and Faletto called this the dialectic of “internal” and “external” determinations, and they insisted that the internal moment is decisive. The position in the world system sets the parameters; the politics inside the country determines what actually happens. To analyze a peripheral country, you have to do historical sociology. You cannot deduce its political form from its position in the global division of labor.

This methodological move had a clear polemical target: André Gunder Frank’s then-famous thesis of “the development of underdevelopment,” which held that the periphery is necessarily and permanently underdeveloped because the core extracts surplus from it. Frank developed this thesis in two books published in 1966 and 1967. By the early 1970s, in the United States and Western Europe, it had become the canonical statement of “dependency theory.” Cardoso and Faletto’s book was, among other things, a sustained argument against Frank.

Their account was not stagnationist, nor was it deterministic, neither posited the necessity for a revolution. It was historical, structural, and open about its own political indeterminacy. The countries of dependent Latin America, in their reading, had alternatives… some of those alternatives could be democratic, but none of them could be revolutionary in the Cuban sense, because the Cuban path was conditioned on a specific historical configuration that was not generalizable. The book was, in this respect, an early and rigorous demolition of focoism. Régis Debray’s Révolution dans la révolution came out in the same period, and Cardoso, reading it, was furious. They were operating from exactly the deductivist methodology that he and Faletto had spent two years arguing against.

The misreading is the famous part. When the English translation finally appeared in 1979, the book entered an Anglophone debate that had been organized for a decade around Frank’s much shorter, much more accessible, and ideologically more attractive theses. The American “dependency school” was already constituted, with its own internal factions (West Coast neo-Marxists, East Coast modernization theorists), and its definitional terms had been set largely by Frank. Cardoso, who by 1979 sat on the boards of half a dozen American academic institutions and had spent five years giving polemical lectures at LASA conferences explaining what he and Faletto had not been saying, watched the book get absorbed into a “dependency theory” he had spent a decade attacking. He has been publishing footnotes and afterwords to correct this misreading ever since.

The polemics from the 1970s are some of the best things he ever wrote. “Teoria da dependência ou análises concretas de situações de dependência?” (1971, published in six different versions between 1970 and 1974) is a brisk refusal of the theoretical-system pretensions of the dependency literature. What he and Faletto had produced, Cardoso insists, was not a theory of dependency but a method for concrete analyses of dependency situations. “As contradições do desenvolvimento associado” (1973) is the working out of what associated-dependent development actually looks like in practice.

“O consumo da teoria da dependência nos Estados Unidos” (1976), originally a LASA conference paper in Atlanta and then published in the Latin American Research Review the next year, is a sustained satirical attack on the way the American academy had received and dehistoricized the dependency literature. And “As desventuras da dialética da dependência,” the 1978 essay Cardoso co-authored with José Serra in response to Ruy Mauro Marini’s theoretical work, is a methodological demolition of the orthodox Marxist-dependentist tradition. These essays, taken together, constitute Cardoso’s most sustained intellectual self-defense; on the other hand, they are also the texts that the orthodox left has held against him ever since.

Aposentado compulsoriamente

Cardoso returned to Brazil in 1968. The choice surprised his Chilean and French colleagues. Alain Touraine, who had brought him to teach at the Université Paris-X Nanterre in 1967, thought he was crazy to go back, as Paul Ricoeur, the rector of Nanterre, had offered him a permanent position. The American sociologist Richard Morse had offered him a job at Yale. He turned them all down. The reason was a vacancy at the USP, as Lourival Gomes Machado, the professor of Political Science, had died. A cátedra was being opened by competitive examination. Cardoso wanted to come home, and the cátedra was the way back in. He won it in October 1968.

Six months later, in April 1969, he was forcibly retired. The AI-5, the Institutional Act of December 1968 that closed Congress, imposed direct censorship, and gave the military regime the power to remove civil servants by administrative decree, was the legal instrument. Florestan Fernandes was removed in the same purge, along with most of the senior figures of the FFLCH-USP. Cardoso had taught for six months a single course on political theory before his name appeared on the list.

The scene at the social-security counter that I opened this essay with came a few days after. It is the only public moment in which Cardoso has made the AI-5 sound funny. The serious account, in the same FGV interviews and in the 2021 memoirs, is darker. He had spent eight years building back the academic career the 1964 coup had taken from him, won the chair in October 1968 by competitive examination against considerable resistance, and within six months had been processed out of the university by a decree he never saw and that named no specific charge. The forced retirement was not a punishment in any institutional sense. It was an administrative annulment.

The forced retirement was the founding moment of CEBRAP. With seed money from the Ford Foundation, arranged through Kalman Silvert and Peter Bell (the regional representatives in Rio), after political vouching from Paulo Egídio Martins, the latter governor of São Paulo, and Severo Gomes, a longtime friend of the Cardoso family, Cardoso, Paul Singer, Octavio Ianni, José Arthur Giannotti, Cândido Procópio Ferreira de Camargo, Francisco de Oliveira, Elza Berquó, and Juarez Brandão Lopes founded an independent research center in São Paulo. The model was new for Brazil. The institute would be funded by international philanthropy (later joined by the Swedish, Canadian, Dutch, and German foreign-aid agencies), would offer salaries and research space to exiled intellectuals, and would operate outside the university system. There was no precedent in the country.

There was also no precedent for the institution that CEBRAP rapidly became. Bernardo Sorj, in his 2001 book A construção intelectual do Brasil contemporâneo, makes the case that CEBRAP managed three things at once that none of the other Latin American institutions of the period quite achieved. It became a refuge for the intellectuals who were politically persecuted by the dictatorship. I

t became a bridge between the pre-1964 generation of social scientists and the new generation being trained in graduate programs that were opening up at IUPERJ in Rio, in Brasília, and at the new programs at UFMG. And it became a node in the international circulation of social-scientific ideas, with continuous visiting fellowships, collaborative research projects, and translation programs. The Ford Foundation, which had been accused in the late 1960s of being a vehicle of American Cold War cultural influence in the periphery, was, in this period, the principal external funder of an institute whose research output was openly critical of the Brazilian military regime, of the American hegemonic system, and of the comparative social science (American modernization theory) that the Foundation itself had been thought to promote. Whatever you think about the geopolitics of this arrangement, the concrete outcome was that a group of deposed Brazilian Marxists got the salaries and the research infrastructure to write a decade and a half of consequential social science.

Censorship in São Paulo through the early 1970s was severe, however. CEBRAP was raided by the political police at least once. Cardoso was taken to the Oban, the São Paulo military counter-insurgency unit famous for the torture it conducted, and interrogated with a hood over his head. He has talked about this in interviews with some understatement; he was not himself tortured, but he was made to listen to the screams of someone in the next room. The Oban officers thought CEBRAP was a cover for an armed-struggle group. It wasn’t. Cardoso came home that night. Other people did not.

What CEBRAP produced in this period was the body of work I want to discuss next. The 1970s Cardoso is not a dependency theorist. He is the theorist of the bureaucratic-authoritarian state.

The burguesia de Estado and the anéis

The central piece of writing I want to talk about is “Da caracterização dos regimes autoritários na América Latina,” which exists in two main versions: the long English version published in David Collier’s edited book The New Authoritarianism in Latin America (Princeton, 1979) under the title “On the Characterization of Authoritarian Regimes in Latin America,” and the Portuguese version included in the 1982 Paz e Terra edition of the same volume. The argument had been worked out across about six years of essays for Estudos CEBRAP and Cadernos CEBRAP: “Estado e sociedade na América Latina” (1971), “Notas sobre Estado e dependência” (1973), the long “Estado capitalista e marxismo” (1977), and the 1977 “Estado y proceso político en América Latina.” The thinking is dispersed across these pieces. The Collier essay is the synthesis.

Three theses do most of the work. I’ll take them in order.

The first thesis is the distinction between the state and the political regime. Cardoso uses these as analytically separate categories. The state, in its usage, is the “pact of domination”, the underlying coalition of class fractions, bureaucratic interests, and foreign economic interests that defines the basic pattern of who has power over whom and over what. The regime is the institutional form through which that pact organizes itself: the formal rules linking executive, legislative, and judiciary; the party system, if there is one; the rules of representation and competition.

Two countries can have the same type of State (the same underlying coalition of capitalist-bureaucratic interests organized around associated-dependent development) and have very different regimes. Brazil and Mexico in the 1970s are the standard comparison. Both are associated-dependent capitalist economies, urban-industrial, internationalized in their leading sectors. The Brazilian state was administered by a bureaucratic-authoritarian regime, with a military presidency, restricted electoral competition, and a tightly controlled party system. The Mexican state was administered by a single-party-corporatist regime, with elections that ran on schedule but did not change the outcome.

This sounds like a small terminological move, but it isn’t. What it actually does is refuse what Cardoso called “economicism”, the assumption that a given pattern of capitalist accumulation produces, by some functional necessity, a single political form. The Brazilian, Argentine, Chilean, and Uruguayan military regimes of the late 1960s and 1970s all served associated-dependent capitalism. They served it very differently. The Chilean military regime decimated the industrial economy in the name of monetarist orthodoxy, while the Brazilian military regime expanded the state-owned enterprise sector and accelerated import-substitution industrialization in the name of national developmentalism. The economic pattern was similar, but the political form was wildly different. While Economicism could not explain this, the Cardoso framework could.

The second thesis was the one that gave the framework its political bite, and that has survived best. The bureaucratic-authoritarian regime is not a simple instrument of the dominant economic class. Cardoso did not believe, even in the early 1970s when the orthodox Marxist analysis insisted on this, that the Brazilian military government was the “armed branch of the oligarchy” or the “executive committee of the multinational bourgeoisie.” His reading was different.

The regime had been put in place in 1964 by a coalition that included the industrialists who feared popular mobilization. Once in place, the regime developed a logic and a set of interests of its own. The civilian technocracy that ran the economic ministries (the team around Delfim Netto, Roberto Campos, Mário Henrique Simonsen) had emerged as an autonomous power center inside the state, with interests of its own in the expansion and consolidation of state power. The military hierarchy had emerged as another autonomous power center. By the mid-1970s, the state-owned enterprise sector had the largest single block of the Brazilian economy, accounting for a third of the whole GDP. Together, these formed what Cardoso called the burguesia de Estado, the state bourgeoisie. It was a new social formation, generated by the expansion of the state itself, whose interests were not reducible to the interests of any of the older private-capitalist class fractions.

If you read that sentence in 2026 and think “state capture” or “regulatory capture” or even just “bureaucratic insulation,” you are in the same conceptual neighborhood. Cardoso is making exactly the move that 1980s American historical institutionalism would later make: he is treating state bureaucracies as autonomous actors with interests of their own, not as transmission belts for the interests of underlying social classes. Theda Skocpol’s Bringing the State Back In (1985) is the canonical American reference, but Da caracterização preceded it by six years. Whether this counts as “early historical institutionalism” or as “late Marxism corrected by Weber” is a question of nomenclature. The analytical object is the same, while I prefer the first interpretation.

The third thesis is the conceptual move that has had the largest influence in Brazilian political science. Cardoso called it the theory of anéis burocráticos — bureaucratic rings. The argument is this: The old populist republic of 1945 to 1964 had organized class representation through formal institutions, with corporatist labor unions, populist parties, and the Vargas-era apparatus of regulated negotiations between capital and labor.

The bureaucratic-authoritarian regime abolished or hollowed out most of these formal institutions, but it did not abolish the underlying need for the dominant economic interests to organize themselves inside the state. What replaced the old corporatist structures was a system of informal, ad hoc, mobile circles of interest that cut across the public-private boundary, linking sections of the state apparatus (a particular ministry, a particular regulatory agency, the leadership of a particular state-owned enterprise) with sections of capital (a particular industrial sector, a particular banking conglomerate, a particular construction firm) around specific policy issues. These were the anéis burocráticos.

Different rings operated simultaneously… Sometimes they cooperated, and sometimes they fought each other. They were not formal institutions. They were not visible in the official organization charts. But they were the actual mechanism by which the political economy of the Brazilian military regime was managed.

Anyone who has paid attention to Brazilian political journalism over the last twenty years has seen this analysis in action. The Lava Jato investigations of 2014 to 2018, the construction-firm corruption scandals of the 1990s and 2000s, the petroleum sector under Petrobras and the state-bank sector around BNDES, the long history of regulatory rents in agriculture and telecommunications: every one of these is a description of an anel burocrático in operation. Cardoso did not predict any of these specific scandals. He could not have. What he did was build the analytical framework that makes them legible, in 1975, before any of them had happened. The vocabulary he used is older than the vocabulary the American political scientists would later use. The analytical object is the same.

One more thing about Da caracterização deserves to be said. The essay was published in English in 1979, when Cardoso was still under a formal political ban in Brazil, in a Princeton University Press volume edited by David Collier, alongside contributions by Guillermo O’Donnell, Albert Hirschman, Robert Kaufman, and José Serra. It was, at the moment of its publication, the canonical statement of the new comparative political sociology of authoritarianism. And then it was almost completely forgotten in the Anglophone literature. O’Donnell’s category of “bureaucratic-authoritarianism” continued to be cited; Cardoso’s contribution did not.

By the late 1990s, when historical institutionalism had become the dominant framework of American political science, Cardoso’s anticipatory framework was almost never cited. The vocabulary had moved on. The analytical objects were rediscovered without their original genealogy. And by then, the original Portuguese-language sociologist who had built the framework was the president of Brazil, and the work that mattered was read backwards through the politics. The intellectual history got lost.

How the Weberian crept in

I keep saying “Weberian” and I should explain what I mean by it.

The transition from a primarily Marxian to a primarily Weberian framework happens, in Cardoso’s published work, between roughly 1967 and 1977, but it wasn’t a clean break. It was gradual, sometimes invisible to Cardoso himself, and registered most clearly in the conceptual apparatus that the late-1970s essays ended up using. Three things shifted.

The first is the influence of José Medina Echavarría at the ILPES. Medina was a major Spanish-language Weber scholar; he had translated Economy and Society and the Soziologie der Herrschaft into Spanish, and his German formation gave Cardoso access to a reading of Weber that was unavailable in the Brazilian university system of the 1950s, where the standard introduction to Weber was through Raymond Aron and the older French sociology. In Santiago, Cardoso was reading Weber daily, in the editions Medina was preparing, and discussing him at the kitchen table with people (Faletto, Sunkel, the older Argentine economists who had read Weber in German in the 1940s) who took the German political sociology seriously. The Weber that mattered to Cardoso from this period on was the political Weber: the Weber of the typology of regimes, the Weber of bureaucratic domination, the Weber of “Politik als Beruf.”

The second is the cátedra of Political Science at the USP. When Cardoso won the cátedra in 1968, he had to read systematically through a political-science literature he had largely ignored as a sociologist. This meant Juan Linz’s then-very-recent work on authoritarian regimes (the 1964 paper on Franco’s Spain that had introduced the category to the international literature), Reinhard Bendix’s comparative-historical studies of state-building, Barrington Moore Jr.’s Social Origins of Dictatorship and Democracy (1966), and the early comparative-politics literature on political development.

Most of this literature was written in a broadly Weberian register; even when it was not explicitly Weberian, it took the institutional form of the political regime as analytically separate from the underlying economic structure. Cardoso absorbed this literature in the few years just before the AI-5 cassated him, and the absorption is visible in the Política e desenvolvimento em sociedades dependentes thesis (defended in 1968, published in 1971), which is the first of his major books to organize the analysis around the political form as such. The book is also the first in which Weber is cited not as one author among others but as the methodological anchor of the political analysis.

The third was the simple analytical necessity of explaining the form of the Brazilian state in the early 1970s. The orthodox Marxian framework, which derived the form of the state from the structure of accumulation, could not explain why Brazil was bureaucratic-authoritarian while Venezuela was a restricted democracy, or why Mexico had a single-party-corporatist regime while Chile (until 1973) had a multiparty system that included a Marxist president.

The economic structures were too similar; the political forms were too different. To explain the differences, you needed a theory of the relative autonomy of the political. Cardoso, like Poulantzas in France a few years earlier, reached for the concept. Unlike Poulantzas, however, he did not want a structuralist version of it; he wanted a historical-empirical version. The Weberian distinction between regime and state — the same distinction I discussed in the previous section — is exactly that.

The key point about this Weberianization is what it does not do: It does not abandon Marx. Cardoso, in the 1977 essay “Estado capitalista e marxismo,” is at pains to make clear that the framework he is using is still Marxian in its account of capitalism as a global system, in its account of the accumulation imperative, and in its account of class formation. What he wants Weber for is the analysis of political domination as such, the analysis of the regime, the bureaucracy, the elites, the institutional form of the state. The two registers run in parallel, with Marxism doing the structural-economic theoretical work and Weber doing the political-institutional theoretical work. As Cardoso put it to Marcelo Ridenti in a 2013 interview, only published in Tempo Social in 2025: “the historical-structural method, I don’t know how to do another one. When I analyze globalization, I analyze it that way, not in any other.” But the political analysis, by the late 1970s, is openly Weberian.

The orthodox Marxists noticed, and, boy, they didn’t like it. Maria Goreti Juvencio Sobrinho Frizzarini, working in the radical Marxist tradition of José Chasin, has written what is probably the most analytically rigorous critique from the orthodox side. The 2008 essay “O pensamento político de Fernando Henrique Cardoso,” published in the journal Ponto-e-Vírgula, names the move precisely. Cardoso, on Frizzarini’s reading, has detached the political from its material base.

The “relative autonomy of the political” is, on her account, a sleight of hand that turns Weberian institutional differences into the basis of an entire theoretical framework, with the consequence that the regime form (democratic, authoritarian, semi-authoritarian) comes to look like a matter of political will rather than a determinate effect of class structure. The horizon of this framework, Frizzarini argues, can only be a liberal-democratic one: a politics of regime-design and institutional engineering inside the existing pattern of capitalist accumulation, with no theoretical room for the question of whether capitalism itself can be transcended. It is a serious critique. It is also more or less right about what Cardoso was doing.

He was building a theoretical framework in which the political form is genuinely independent of the economic substrate, and in which the political horizon is the design of institutions inside dependent capitalism, rather than the supersession of dependent capitalism. The disagreement here is not about what Cardoso was doing. It is about whether what he was doing was a good idea.

For what it’s worth, my own view is that the Weberianization made the analysis more empirically useful at the cost of making it less politically radical. The framework that comes out of “Da caracterização” and the surrounding essays explains Brazilian political reality in the late 1970s, in the 2010s, and in the 2020s, better than any of the orthodox Marxist analyses that compete with it.

It does so because it takes the autonomy of institutional logics seriously, the role of bureaucratic interests, the way private economic interests embed themselves in regulatory apparatuses, and the way regime forms reproduce themselves across changes of government. The cost is that the framework has no theoretical room for the transcendence of capitalism. It is, on its own terms, a sociology for the management of dependent capitalism democratically. Whether that is a virtue or a vice depends on whether you think the transcendence of dependent capitalism is, in the foreseeable historical time, a real political horizon. The orthodox Marxists thought yes, and FHC disagreed. He was, in the event, the one whose framework predicted the next forty years more accurately.

The democratic horizon

The other side of the late-1970s Cardoso is the practical political work, and this is, again, intellectual work rather than the work of the presidency that came later. The shift in his published output between 1974 and 1980 is striking. The political theory of authoritarianism gives way, by stages, to a political theory of democratization. The titles are explicit: Autoritarismo e democratização (Paz e Terra, 1975); Os partidos e as eleições no Brasil, co-edited with Bolívar Lamounier (Paz e Terra and CEBRAP, 1975); “A democracia na América Latina” (Novos Estudos CEBRAP, 1984); and the long essay “A questão dos partidos” (Contexto, 1978).

The shift in the work tracks a shift in the political situation. The November 1974 MDB victory in the parliamentary elections opened a possibility that none of the analyses produced in the previous decade had imagined. The bureaucratic-authoritarian regime could be defeated electorally. It had built itself an electoral facade as a way of managing internal divisions in the ruling coalition; the facade now turned out to be exploitable from the opposition side.

Ulysses Guimarães walked into the CEBRAP offices in late 1974 to ask the cassated sociologists to help write the MDB’s programmatic statement. Cardoso, Francisco de Oliveira, Weffort, Singer, and Lamounier wrote what came to be called “the little red book,” which became the template for MDB’s 1974 campaign and, arguably, the founding programmatic document of redemocratization-era center-left politics in Brazil. By 1978, Cardoso was running for the Senate himself. He lost the seat in the direct count but came in as substitute and took office in 1983 when Franco Montoro became governor of São Paulo.

The intellectual interest of this phase, leaving aside Cardoso’s eventual decision to enter electoral politics, lies in his sustained attempt to build a political theory of democratization compatible with his earlier analysis of authoritarianism. The architecture of the bureaucratic-authoritarian state, on his reading, was held together by the anéis burocráticos, the burguesia de Estado, and the relative autonomy of military and technocratic interests inside the regime.

Democratization, on this analysis, could not be a simple matter of removing the military and holding elections. It had to be a slow institutional process of building new channels of representation between civil society and the state, of multiplying the political spaces available to social movements, parties, unions, and the alternative press. Autoritarismo e democratização is, in this respect, a programmatic book disguised as an analytical one. It is the working text of an intellectual who has decided that the route out of the dictatorship runs through the institutional, electoral, and party-political channels rather than through armed struggle.

The distance this represented from the dominant Brazilian left position of the early 1970s is hard to overstate. The armed struggle had been crushed by 1973-74, but its remaining sympathizers, much of the exiled intelligentsia, and the still-clandestine PCB continued to imagine the end of the dictatorship as some kind of revolutionary rupture. Cardoso did not. In the 1978 essay “A questão dos partidos,” he sketched what he called a partido dos assalariados, a broad confederation of organized labor, salaried professionals, and civil-society movements, capable of contesting elections within the existing institutional framework and slowly pushing it toward democratic openings.

This was not the program that the PT would adopt under Lula’s leadership four years later, and Cardoso did not join the PT. The two political projects diverged from the first moment of their construction, and the divergence is exactly the divergence between an electoral-democratic, civil-society-based, institutionally gradualist strategy and a syndicalist, mass-mobilizational, popular-democratic strategy. The two strategies would compete for the next forty years.

A point that the sympathetic critics (Limongi, Belinelli, Helayel) and the more hostile ones (Frizzarini, the TMD authors) actually share is that the conceptual move that allowed Cardoso to imagine a democratic horizon within dependent capitalism produced, over time, a kind of analytical concession to the inevitability of dependent-associated development. If you have argued against the stagnationist left that capitalist development is possible in the periphery without revolution, and if you have argued against the economicist Marxists that the form of the political regime has relative autonomy from the form of accumulation, then it follows that the political horizon shrinks to one of managing dependent-associated capitalism democratically, rather than transcending it.

The 1984 essay “A democracia na América Latina” makes this explicit. And it is, in retrospect, the template for the political project that Cardoso carried, two presidencies later, into the 1990s. The relationship between the late-1970s sociology and the 1990s politics is, therefore, neither the relationship of betrayal that the esqueçam o que escrevi reading would have it (the apocryphal line is something Cardoso has spent thirty years denying ever saying), nor the relationship of straight-through application that the more hagiographic reading would have it. It is something more complicated. The sociology made certain political conclusions easier to draw, and Cardoso eventually drew them.

The Americans, finally

The international career deserves its own section, partly because Cardoso has been candid about its protective function during the worst years of the dictatorship, and partly because it is one of the rare cases of a Latin American social scientist being received in the United States and Europe as a theoretical authority rather than as a source of empirical material.

The basic chronology, drawn from Lidiane Rodrigues’s careful 2022 essay in the Novos Estudos dossier is at follows: 1962, first trip abroad to Paris and Cambridge after the doctoral defense; 1964 to 1967, he was stationed at ILPES in Santiago; 1967 and 1968, Université Paris-X Nanterre at Touraine’s invitation, in the run-up to Mai 68; 1968, return to the USP, cátedra, AI-5, forced retirement; from 1969 onwards, CEBRAP, with regular visiting positions at Stanford, the University of California Berkeley, Cambridge, the École des Hautes Études en Sciences Sociales, the Institute for Advanced Study at Princeton; vice-presidency of the International Sociological Association from 1978 to 1982 and presidency from 1982 to 1986; doctor honoris causa at Rutgers in 1978; Wilson Center fellowship; Helen Kellogg Institute committee; LASA lectures; eventually the John W. Kluge Prize at the Library of Congress in 2012.

What Rodrigues argues, against more triumphalist readings of this trajectory, is that the success was made possible by a set of conditions not reducible to the intrinsic quality of the work. The Cold War philanthropic networks (the Ford Foundation, the Rockefeller Foundation, the Inter-American Foundation, the Social Science Research Council) were, in the 1960s and 1970s, actively constituting Latin American social science as a transnational field, and Cardoso, with his fluent French, his comfortable relations with elites, and his family connections, was an ideal interlocutor.

The very controversy over “dependency theory” in the United States gave him a polemical platform from which to distinguish himself. His strategy, in Rodrigues’s reading, was to “belong to dependentism to distinguish himself within it” — to claim founder status while attacking the canonical readings. The translation lag of Dependência e desenvolvimento into English (twelve years, again) had the unintended effect of producing, by 1979, an active North American demand for the canonical statement, which Cardoso could then proceed to clarify, complicate, and partly disown.

Rodrigues judges that the success was real, but it was also produced by careful management of correspondence with Kalman Silvert and Peter Bell, by strategic publication of the same essays under different titles in different languages (the piece that became “O consumo da teoria da dependência nos Estados Unidos” was first published in French in 1976, then in English at the 1976 LASA conference, then in Portuguese in 1977, with subtly different framings for each audience). By skilled use of the institutional resources that the Ford Foundation made available to CEBRAP. None of this is scandalous, but rather it is what an internationally successful peripheral intellectual is required to do to be heard at all in a field whose structural asymmetries point in exactly the opposite direction. In this, FHC was the prototype for the successful Latin American intellectual in the following decades.

But it does mean that the Cardoso who became, in the late 1970s, one of the most cited Latin American social scientists in the world is not simply the author of the books. He is also the author of a sophisticated reception strategy, executed in three languages over a decade, with the institutional resources of CEBRAP and the personal networks that Florestan Fernandes had taught him to build in his twenties. The work was excellent, but the work was also made canonical, against the structural inertia of an asymmetric global field, by a great deal of skilled labor that the work itself does not register.

At ninety-five

So what do you take away from this body of work, in 2026, in a country that has just spent four years under another quasi-authoritarian government, in a region where the relationship between dependent capitalism and democratic stability remains, to put it gently, unsettled?

The first thing is the most obvious. The vocabulary of “state capture” that has dominated discussions of democratic decline in Brazil since 2016 is the contemporary form of Cardoso’s burguesia de Estado. The Lava Jato investigations, the perennial corruption in the Bolsonaro government, and the recent Banco Master scandal described, in journalistic terms, an array of anéis burocráticos that Cardoso had described, in conceptual terms, forty years earlier. The current arguments about the operational autonomy of the federal police, the prosecutorial corps, and the supreme court inside an elected government are, structurally, the same arguments Cardoso was making in the 1970s about the autonomy of the technocratic and military bureaucracies inside the military government. Different actors, naturally, but similar analytical objects.

I’m trying to state here that the majority of the FHC’s intellectual framework still works. It does not predict specific outcomes… nothing predicts specific outcomes… but it organizes the empirical data in a way that makes the political situation legible.

The second thing is the methodological move that made the framework possible: the disentangling of state from regime, of accumulation pattern from political form. This is the move that allowed Cardoso to argue against the orthodox Marxists that democracy was possible within dependent capitalism. It is the move that, in the contemporary political science of historical institutionalism, is the conceptual basis for almost every interesting comparative argument about why similar economic structures produce different political outcomes.

Whatever you think about the political conclusions Cardoso drew from it, the analytical move is one of the more consequential things a Brazilian social scientist has done in the last seventy years. It has happened that the move has been rediscovered, in English, by people who do not know that they are rediscovering it. The story of how the original gets forgotten is the story of how peripheral intellectual production gets metabolized into the metropolitan canon. Cardoso is, in this respect, both an example and a counter-example. He got into the metropolitan canon. The categories he used to get there got translated out of his vocabulary on the way in.

The third thing — and I am being more openly normative here than I was through most of this — is the example of what it looks like to do intellectual work of lasting consequence from a peripheral institutional position. Cardoso did not retreat into specialist scholarship. He produced his work in continuous argument: with the Brazilian Communist Party, with the ISEB, with the CEPAL, with the Cuban revolutionaries, with the North American development theorists, with the French structural Marxists, with the orthodox dependentists, and eventually with the social-democratic parties of southern Europe.

He produced it in continual translation across at least three intellectual languages. He produced it under censorship, under forced retirement, under hooded interrogation. And he produced it, throughout, by talking. With the carcereiro, as his father had taught him, with the dependentists he disagreed with, with the industrialists whose ideology he was anatomizing, with the generals who had persecuted him.

The intellectual lesson, if there is one, is that work of lasting consequence is rarely the product of a single national tradition, a single ideological camp, or a single methodological orthodoxy. It is the product of long, patient, polemical conversation across boundaries that the participants are usually trying to keep closed.

Cardoso, at his most productive, between 1962 and 1979, was that kind of intellectual. The presidency that followed is a different story. The sociologist who existed before the president is one of the most important thinkers of the second half of the twentieth century in the Americas, and one of the very few peripheral intellectuals who managed, against the structural odds of an asymmetric global field, to make the periphery audible on terms that the periphery itself set.

References

BELINELLI, Leonardo. Marxismo como crítica da ideologia: um estudo sobre os pensamentos de Fernando Henrique Cardoso e Roberto Schwarz. 2019. Tese (Doutorado em Ciência Política) – Faculdade de Filosofia, Letras e Ciências Humanas, Universidade de São Paulo, São Paulo, 2019.

BELINELLI, Leonardo; HELAYEL, Karim. Teoria, história e política em Fernando Henrique Cardoso (1969-78). Novos Estudos CEBRAP, São Paulo, v. 41, n. 2, p. 253-271, maio-ago. 2022.

BELINELLI, Leonardo; LIMA, Pedro Luiz; HELAYEL, Karim (org.). Fernando Henrique Cardoso: modos de ler. São Paulo: Hucitec, 2025. (Coleção Pensamento Político-Social).

CARDOSO, Fernando Henrique. Capitalismo e escravidão no Brasil meridional: o negro na sociedade escravocrata do Rio Grande do Sul. São Paulo: Difel, 1962.

CARDOSO, Fernando Henrique. Empresário industrial e desenvolvimento econômico no Brasil. São Paulo: Difusão Européia do Livro, 1964.

CARDOSO, Fernando Henrique. Política e desenvolvimento em sociedades dependentes: ideologias do empresariado industrial argentino e brasileiro. Rio de Janeiro: Zahar, 1971.

CARDOSO, Fernando Henrique. O modelo político brasileiro e outros ensaios. São Paulo: Difel, 1972.

CARDOSO, Fernando Henrique. Notas sobre Estado e dependência. Cadernos CEBRAP, São Paulo, n. 11, 1973.

CARDOSO, Fernando Henrique. As contradições do desenvolvimento associado. Estudos CEBRAP, São Paulo, n. 8, p. 41-75, 1974.

CARDOSO, Fernando Henrique. Autoritarismo e democratização. Rio de Janeiro: Paz e Terra, 1975.

CARDOSO, Fernando Henrique. The consumption of dependency theory in the United States. Latin American Research Review, Pittsburgh, v. 12, n. 3, p. 7-24, 1977a.

CARDOSO, Fernando Henrique. Estado capitalista e marxismo. Estudos CEBRAP, São Paulo, n. 21, p. 5-31, 1977b.

CARDOSO, Fernando Henrique. A questão dos partidos. Contexto, São Paulo, n. 5, p. 1-20, 1978.

CARDOSO, Fernando Henrique. On the characterization of authoritarian regimes in Latin America. In: COLLIER, David (ed.). The new authoritarianism in Latin America. Princeton: Princeton University Press, 1979. p. 33-57.

CARDOSO, Fernando Henrique. As idéias e seu lugar: ensaios sobre as teorias do desenvolvimento. Petrópolis: Vozes, 1993.

CARDOSO, Fernando Henrique. Fernando Henrique Cardoso [Entrevista concedida a Elide Rugai Bastos, Fernando Abrucio, Maria Rita Loureiro e José Márcio Rego]. In: BASTOS, Elide Rugai et al. (org.). Conversas com sociólogos brasileiros. São Paulo: Editora 34, 2006. p. 66-105.

CARDOSO, Fernando Henrique. Fernando Henrique Cardoso IV (depoimento, 2011). Rio de Janeiro: CPDOC/FGV, 2019. Entrevista concedida a Celso Castro e Helena Bomeny, 19 out. 2011.

CARDOSO, Fernando Henrique. Um intelectual na política: memórias. São Paulo: Companhia das Letras, 2021.

CARDOSO, Fernando Henrique. Os anos vermelhos de um sociólogo: entrevista com Fernando Henrique Cardoso [Entrevista concedida a Marcelo Ridenti]. Tempo Social, São Paulo, v. 37, n. 2, e2025236636, 2025.

CARDOSO, Fernando Henrique; FALETTO, Enzo. Dependencia y desarrollo en América Latina: ensayo de interpretación sociológica. México: Siglo XXI, 1969.

CARDOSO, Fernando Henrique; FALETTO, Enzo. Dependência e desenvolvimento na América Latina: ensaio de interpretação sociológica. Rio de Janeiro: Zahar, 1970.

CARDOSO, Fernando Henrique; FALETTO, Enzo. Dependency and development in Latin America. Berkeley: University of California Press, 1979.

CARDOSO, Fernando Henrique; LAMOUNIER, Bolívar (org.). Os partidos e as eleições no Brasil. Rio de Janeiro: Paz e Terra; São Paulo: CEBRAP, 1975.

CARDOSO, Fernando Henrique; SERRA, José. Las desventuras de la dialéctica de la dependencia. Revista Mexicana de Sociología, México, v. 40, p. 9-55, 1978.

COLLIER, David (ed.). The new authoritarianism in Latin America. Princeton: Princeton University Press, 1979.

COSTANZO, Daniela; MARINO, Rafael. As leituras do empresário industrial de Fernando Henrique Cardoso. Boletim Lua Nova, São Paulo, 12 set. 2022. Disponível em: https://boletimluanova.org/as-leituras-do-empresario-industrial-de-fernando-henrique-cardoso/. Acesso em: 18 jun. 2026.

FRIZZARINI, Maria Goreti Juvencio Sobrinho. O pensamento político de Fernando Henrique Cardoso. Ponto-e-Vírgula, São Paulo, n. 4, p. 53-70, 2008.

GARCIA Jr., Afrânio. A dependência da política: Fernando Henrique Cardoso e a sociologia no Brasil. Tempo Social, São Paulo, v. 16, n. 1, p. 285-300, 2004.

GONÇALVES, Rodrigo Santaella. Teoria e prática em Fernando Henrique Cardoso: da nacionalização do marxismo ao pragmatismo político (1958-1994). 2018. Tese (Doutorado em Ciência Política) – Faculdade de Filosofia, Letras e Ciências Humanas, Universidade de São Paulo, São Paulo, 2018.

HELAYEL, Karim. Um sociólogo na periferia do capitalismo: a sociologia histórico-comparada de Fernando Henrique Cardoso. 2019. Tese (Doutorado em Sociologia) – Universidade Federal do Rio de Janeiro, Rio de Janeiro, 2019.

LAHUERTA, Milton. Intelectuais e transição: entre a política e a profissão. 1999. Tese (Doutorado em Ciência Política) – Faculdade de Filosofia, Letras e Ciências Humanas, Universidade de São Paulo, São Paulo, 1999.

LEME, Alessandro. La centralidad de la política para pensar lo económico en Fernando Henrique Cardoso. Revista Mexicana de Sociología, México, v. 77, n. 3, p. 357-384, 2015.

LIMA, Pedro Luiz. As desventuras do marxismo: Fernando Henrique Cardoso, antagonismo e reconciliação (1955-1968). 2015. Tese (Doutorado em Ciência Política) – Instituto de Estudos Sociais e Políticos, Universidade do Estado do Rio de Janeiro, Rio de Janeiro, 2015.

LIMA, Pedro Luiz. A parte maldita: Fernando Henrique Cardoso e as histórias do marxismo no Brasil. Novos Estudos CEBRAP, São Paulo, v. 41, n. 2, p. 231-250, maio-ago. 2022.

LIMA, Pedro Luiz; BELINELLI, Leonardo; HELAYEL, Karim. Apresentação: novas abordagens da obra de Fernando Henrique Cardoso. Novos Estudos CEBRAP, São Paulo, v. 41, n. 2, p. 223-228, maio-ago. 2022.

LIMONGI, Fernando. Fernando Henrique Cardoso: teoria da dependência e transição democrática. Novos Estudos CEBRAP, São Paulo, n. 94, p. 187-197, nov. 2012.

NATALINO, Enrique Carlos. A construção do pensamento cardosiano. blogNEC – Novos Estudos CEBRAP, São Paulo, 21 jul. 2023. Disponível em: https://novosestudos.com.br/a-construcao-do-pensamento-cardosiano/. Acesso em: 18 jun. 2026.

O’DONNELL, Guillermo. Modernization and bureaucratic-authoritarianism: studies in South American politics. Berkeley: Institute of International Studies, University of California, 1973.

PACKENHAM, Robert A. The dependency movement: scholarship and politics in development studies. Cambridge: Harvard University Press, 1992.

PÉCAUT, Daniel. Os intelectuais e a política no Brasil: entre o povo e a nação. São Paulo: Ática, 1990.

RICUPERO, Bernardo. Caio Prado Jr. e a nacionalização do marxismo no Brasil. São Paulo: Editora 34, 2000.

RODRIGUES, Lidiane Soares. A produção social do marxismo universitário em São Paulo: mestres, discípulos e “um seminário” (1958-1978). 2012. Tese (Doutorado em História Social) – Faculdade de Filosofia, Letras e Ciências Humanas, Universidade de São Paulo, São Paulo, 2012.

RODRIGUES, Lidiane Soares. Fernando Henrique Cardoso nos Estados Unidos da América: a obra de um scholar, um scholar como obra. Novos Estudos CEBRAP, São Paulo, v. 41, n. 2, p. 273-293, maio-ago. 2022.

SCHWARZ, Roberto. Um seminário de Marx. In: SCHWARZ, Roberto. Sequências brasileiras: ensaios. São Paulo: Companhia das Letras, 1999. p. 86-105.

SCHWARZ, Roberto. Seja como for: entrevistas, retratos e documentos. São Paulo: Editora 34, 2019.

SINGER, André. Cutucando onças com varas curtas: o ensaio desenvolvimentista no primeiro mandato de Dilma Rousseff (2011-2014). Novos Estudos CEBRAP, São Paulo, n. 102, p. 39-67, 2015.

SINGER, André. O lulismo em crise: um quebra-cabeça do período Dilma (2011-2016). São Paulo: Companhia das Letras, 2018.

SORJ, Bernardo. A construção intelectual do Brasil contemporâneo: da resistência à ditadura ao governo FHC. Rio de Janeiro: Centro Edelstein de Pesquisas Sociais, 2008.

Disclaimer: Claude was used for the illustration on the splits, as my design skills are subpar

A new working paper by Dahis and Szerman (2025) documents what happened when Brazil added 1,383 new municipalities in eight years. The results push back against the standard fiscal critique of decentralization, without quite settling whether the country should have done more of it.

The textbook debate about decentralization assumes the political units already exist, and the only question is how much power to push down to them. That is the world of Wallace Oates writing about American federalism (Oates, 1972), of Charles Tiebout sketching his suburban voters (Tiebout, 1956). Step into most of the developing world and the prior question becomes inescapable: how many local governments should there be?

Brazil ran an unusually clean version of that experiment between 1988 and 1996. Coming out of military rule, the new constitution made municipalities into real governments. They collect taxes, run schools, deliver services, and elect their own mayors. It also handed states the authority to write their own rules for creating new municipalities, and most states wrote permissive ones. Small towns and rural districts started petitioning to leave their parent municipalities, and the state legislatures said yes. As per the new constitution, federal transfers kicked in for the new units, which incentivised new polities to emerge. Over the next eight years, Brazil added 1,383 new municipalities, a 34% expansion of the country’s political map.

To anyone trained in the canon of fiscal federalism, the obvious worry was overreach. More units mean more bureaucracies, more demand on shared transfers, and more openings for narrow interests to capture small governments that lack the scale to deliver. Alesina and Spolaore (1997) and Boffa, Piolatto, and Ponzetto (2016) laid out the case carefully. Smaller jurisdictions forgo economies of scale and survive only because the rest of the country subsidizes them. By 1996, the Brazilian Congress had heard enough and approved a constitutional amendment to shut the process down.

That window is what Dahis and Szerman (2025) work with. Armed with decades of follow-up data on municipalities, such as fine-grained fiscal records, matched employer-employee files, satellite night lights… they studied Brazil during the boom as one of the better-instrumented natural experiments in the decentralization literature… Well, guess what? Turns out the warnings may have been misguided.

The political economy behind the boom of new municipalities

The mechanism that turned permissive rules into a thousand-plus new municipalities was the Fundo de Participação dos Municípios. The FPM is the federal transfer that supplies between thirty and 60% of revenue for the typical Brazilian municipality, and the way it is allocated did most of the heavy lifting.

Each year, 22.5% of federal revenues from income tax and the industrial product tax flow into the FPM pool. The pool is divided by state into block grants, and within each state, municipalities receive shares according to a convex step-wise formula on population. The formula has a floor that disproportionately benefits municipalities below roughly 10,188 inhabitants (Tomio, 2002). A small new municipality of, say, 8,000 people gets a per-capita transfer well above what an established municipality of 50,000 receives. The fund is zero-sum within each state. When a new municipality is created, the state block grant gets resliced, and the existing municipalities each lose a sliver of their previous share.

This generates an obvious arbitrage. If a peripheral district with 8,000 residents splits off from a parent municipality of 100,000, the new municipality moves from a per-capita FPM share calculated on the parent’s bracket to its own much higher per-capita share calculated on its own bracket. The total state-level transfer pool does not change. The redistribution from older to newer municipalities is built into the formula.

Layered on top of this, 15% of FPM transfers are earmarked for education and 15% for health (Brollo, Nannicini, Perotti, and Tabellini, 2013). The rest are unmarked. Local taxation and fees average around five percent of total municipal revenues. The decentralized Brazilian municipality is, on the revenue side, mostly a vehicle for spending federal transfers.

The decision to split also responded to non-fiscal pressures. A 1993 survey of Brazilian mayors found that the two most common reasons given for seeking emancipation were neglect by the parent local government (63%) and the territorial size of the original municipality (24%) (Bremaeker, 1993). The districts that wanted out were peripheral, poorer than their parent municipalities, more remote from the old town hall, and politically marginalized by the elites in the denser headquarters of the municipalities.

Before going to the empirical work it is worth taking a closer look on the conceptual model the authors sketch in their appendix, because it disciplines the empirical predictions and clarifies what each result is and is not testing.

The setup is a minimal Tiebout-style framework in the tradition of Bolton and Roland (1997) and Dur and Staal (2008). Imagine a municipality made of two districts, A and B, where A holds the headquarters and decision-making authority. The headquarters chooses public-goods provision across both districts by maximizing a Pareto-weighted sum of utilities, with weight λ on district B and (1 − λ) on district A. Lower λ means more capture and neglect of the periphery by the headquarters. Revenues come from a uniform income tax τ and from federal transfers T(·), where T is weakly increasing and concave in population. Concavity is the FPM floor logic in shorthand.

In the integrated regime, the headquarters chooses public goods in A and B and a tax rate. In the split regime, district B chooses its own provision and tax rate, subject to a smaller transfer T(αB). Comparing solutions yields two predictions worth singling out. First, the benefits of splitting for the seceding district are larger when it was more captured and neglected by the headquarters (low λ), and when it has a high comparative gain in transfers if split. Second, conditional on the rest of the country being much larger than the splitting district and on the splitter being small and neglected, the welfare change for the parent district is small, the welfare gain for the seceding district is large, and the welfare loss for the rest of the country is small.

The model gives the empirical work three targets. Splitting should benefit the applicants most when they look neglected and peripheral at baseline, the parent headquarters should not see much, and the rest of the country should not see much either. The empirical results confirm all three hypotheses.

Solving the Selection Problem

The hardest thing about studying municipal splits is that… well, the units choose to split. Whatever drives a district to push for emancipation, poverty, distance, political marginalization, and neglect also shape its economic trajectory. Comparing splits to non-splits just gives you back the underlying selection.

Dahis and Szerman (2025) get around this by going to the state archives. Between 1988 and 1996, every district that wanted to split had to file a formal petition, win a local referendum, and get past the state legislature. Plenty of petitions failed: vetoed by committees, blocked by governors, defeated in referenda, left pending when the 1996 amendment hit. The authors hand-collected eleven states’ worth of these records and built a control group of “almost-split” districts. Places that wanted out, started the process, and got stopped for reasons unrelated to their economic prospects.

That is the identification strategy. It works because the failed petitions look a lot like the successful ones statistically (Table 1). Both come from poorer-than-average municipalities, both are larger and more rural, and both feel the same federal transfer formula. The only thing that separates them is essentially procedural luck.

The empirical model is a difference-in-differences specification estimated on two waves of splits, in 1993 and 1997. Both waves followed Brazilian municipal election cycles, which is the moment at which newly elected officials in the recently created municipalities took office. The baseline regression in Dahis and Szerman (2025) takes the form

where y_mst denotes the outcome of interest for municipality m in state s in year t, α_m absorbs time-invariant municipality characteristics, α_st captures state-specific shocks in each year, Split_m is an indicator equal to one for municipalities that ultimately split, W_m is the wave-year in which municipality m split (1993 or 1997), and the indicators 1[t − W_m = τ] index event-time relative to the split. The coefficient β_{−1} is normalised to zero, so that the post-event coefficients β_τ trace out the dynamic effect of splitting relative to the year immediately before the wave. Standard errors are two-way clustered at the state and split-wave levels1.

The control group is composed of almost-split municipalities, which contain at least one applicant district whose petition ultimately failed for reasons unrelated to local economic potential. Importantly, these control municipalities are never treated within the sample window. The design, therefore, relies on a clean, never-treated comparison group rather than on a contrast between earlier-treated and later-treated units.

The canonical concerns about two-way fixed effects estimators with staggered treatment, raised by Goodman-Bacon (2021), Callaway and Sant’Anna (2021), Borusyak, Jaravel, and Spiess (2024), de Chaisemartin and D’Haultfœuille (2020), and Sun and Abraham (2021), are mitigated by construction in this setting. Those concerns arise when staggered treatment timing leads earlier-treated units to be used as effective controls for later-treated units, contaminating the estimated average treatment effect under heterogeneous individual effects.

With only two waves of treatment and a stable never-treated control group, the variance decomposition of the difference-in-differences estimator places negligible weight on forbidden comparisons. Robustness exercises reported in the paper re-estimate the principal results separately for the 1993 and 1997 waves and recover quantitatively similar magnitudes, a pattern that would not be expected if the staggered-treatment problem were materially distorting the estimates.

The identifying assumption is that, in the absence of splitting, treated and almost-treated municipalities would have evolved along parallel trajectories. The event-study coefficients β_τ for τ < −1 provide a direct visual test of this assumption. Across the principal outcomes, those pre-event coefficients sit statistically and substantively close to zero, which is the empirical analogue of parallel pre-trends.

Dahis and Szerman (2025) complement the difference-in-differences design with a regression-discontinuity strategy that exploits a distinct source of identifying variation. Before the 1996 constitutional amendment, every district seeking to split was required to obtain at least a simple majority of valid votes in a local referendum. Minas Gerais is the only Brazilian state for which referendum vote shares are publicly reported, but this isn’t a problem as the state is broadly comparable to the rest of the country on dimensions that matter for external validity. Minas Gerais is the second most populous and third wealthiest state in Brazil, with an area larger than that of metropolitan France, and its ethnic composition and geographical features are close to the national averages.

The authors implement a difference-in-discontinuities specification in two stages. The first stage models the probability of splitting as a function of crossing the 50% vote threshold,

where RV_d is the referendum vote share in favour of splitting in district d and g(RV_d) is a linear distance function from the cutoff. The second stage estimates the effect of splitting on outcomes,

with district and year fixed effects, an interaction of the running variable with the post-wave indicator, and a vector X_dt of pre-determined controls. The preferred specification uses a 15% bandwidth around the 50% cutoff in order to balance bias against precision.

The first stage delivers a sharp jump at the threshold: crossing the simple-majority requirement raises the probability of splitting by 0.96, which corresponds to the expected mechanical effect of a binding electoral rule. The reduced-form effect of crossing the threshold on log luminosity for applicant districts is 0.16 and statistically significant at the one-percent level. This implies a Wald estimate of approximately 0.27 (≈ 0.16 / 0.96), or a twenty-eight percent gain in luminosity attributable to splitting, almost identical to the difference-in-differences estimate of 26% obtained when the sample is restricted to Minas Gerais.

The continuity test of pre-referendum characteristics around the threshold is clean for log area, log baseline luminosity, and log distance to the parent town hall. The single exception is log population, which exhibits a modest discontinuity at the cutoff. The authors address this by including interactions of baseline population with year fixed effects, which allow for differential trends by initial population size and absorb the only visible threat to the local exogeneity assumption.

The two designs rest on distinct identifying assumptions. The difference-in-differences strategy relies on parallel counterfactual trends between treated and almost-treated municipalities. The regression-discontinuity strategy relies on continuity of potential outcomes at the 50% vote threshold. No obvious source of confounding would generate parallel violations of both assumptions with comparable magnitudes, which is why the convergence of the two estimators around a twenty-six to 28% luminosity gain is more reassuring than either point estimate considered in isolation.

What Happens When a District Becomes a Municipality?

The first thing it does is build a bureaucracy. Capital expenditures jump by about 40% in the year of the split and settle at 27% above the counterfactual over the next 15 years. Current spending, the payroll and day-to-day operations bucket, rises by 17%. Municipal headcount grows by 16%. Average wages do not budge. New municipalities expand by hiring, not by paying existing workers more.

The pattern of capital expenditures rising faster than current expenditures is consistent with the institutional reading. Capital spending captures one-time outlays on machinery, vehicles, and buildings, the basic setup costs of a new local government. Current spending captures the ongoing payroll and operations. The initial jump in capital is a setup burst, and the steady plateau in current spending is the new equilibrium of running a municipality. Lima and Silveira Neto (2018) noted the same broad pattern in earlier work on Brazilian secessions, but without the credible counterfactual that Dahis and Szerman (2025) construct, the estimates were harder to interpret.

However, the expansion does show up in services. Trash collection rises 4.4%, sewage 1%. Piped water and electricity do not show any significant change. This should not be interpreted as a minor result or an econometric kink, because trash collection and primary education are exclusively municipal mandates, while water, sanitation, and electricity are shared with state and federal governments. When accountability is shared, no level of government takes ownership, and investment falls between the cracks (Kresch, 2020).

For education, they use a different identification strategy. The authors borrow the cohort exposure logic from Duflo (2001) on Indonesian school construction. If splitting builds out schooling infrastructure, the cohorts young enough to use the new schools should show the biggest gains. They formalize this as:

where i indexes individuals, k indexes age at the census year, m and s index municipality and state, and t indexes census year. The fixed effects α_km and α_kt absorb baseline differences across age-municipality and age-time cells. The coefficient β_τ on the interaction of split status and age dummy identifies the differential effect of splitting on cohort τ relative to the omitted age category, comparing split and almost-split municipalities before and after.

The results trace out an exposure gradient. Kids under fifteen in the post-split census show literacy gains up to four percentage points and school attendance gains up to five. Older cohorts, who had already cycled through, see much smaller effects. The crowding-out tells a coherent story. Nonprofit jobs in education shrink while public-sector jobs in education expand. Splitting reorganizes the supply of schooling from third-sector provision toward direct municipal provision.

When we go beyond the public sector, however, the story gets murkier. Private-sector jobs and establishments tick up, but the confidence intervals are wide. Retail and services drive most of the new establishments, but agriculture, manufacturing, and construction barely move. Nighttime luminosity, validated as a proxy for local GDP by Henderson, Storeygard, and Weil (2012) and Henderson, Squires, Storeygard, and Weil (2018), rises sharply for five years and stabilizes at 8% above the counterfactual. Because household access to electricity does not change, the luminosity increase is not just something like streetlights or bonfires.

Cui bono?

Here is the result that should change how people read the rest of the paper. Within each new municipality, there are typically three kinds of districts:

i) the applicant district that initiated the split,

ii) sometimes other peripheral districts that came along (the "remaining" districts),

iii) and the headquarters district of the original parent municipality.

The luminosity gains are not spread evenly across them. They are concentrated almost entirely in the applicants.

The applicants’ luminosity rises about 34 log points (40%) over 15 years. The remaining districts show no detectable change. The old headquarters districts pick up a small gain of around six percent. The extensive-margin measure (the share of pixels lit) tracks the intensive measure. It rises about four log points in applicants and barely moves elsewhere. Measuring luminosity outside a 5km radius of the new town hall, to test whether the gains are concentrated around the seat of government as in Bluhm, Lessmann, and Schaudt (2023), gives essentially the same estimate. The growth is spread across the applicant district, not concentrated in the new political center.

Table 3 unpacks the heterogeneity. The benchmark coefficient on log luminosity for applicant districts is 0.35. Interacting that coefficient with baseline characteristics shows where it comes from. The gain is larger in districts with lower baseline urbanization (the interaction coefficient is −0.01 per percentage point of baseline urbanization, significant at one percent), larger in districts further from the old town hall (the interaction with log distance is positive at 0.24), and somewhat larger in districts further from the state capital. The interaction with the baseline population is statistically insignificant. The interaction with district area is negative and significant at −0.13, meaning the per-district gain is larger when the new municipality is geographically smaller.

This heterogeneity is one of the key results of this paper. A pure income shock would raise every component of the municipality's spending in proportion to the size of the shock. What we see instead are tracks distance from the old town hall and low urbanization, which are precisely the proxies for capture and neglect by the parent government identified by Mansuri and Rao (2012) in their survey of decentralization in the developing world. The gains are largest in the districts that were structurally worst served before the split. In other words, we see that the improvement in participation and constituents’ voices truly matters for local development.

Money Versus Power

The temptation, once you see the revenue numbers, is to attribute everything to federal transfers. Total municipal revenues jump 15% after a split, almost entirely from federal transfers, which is, naturally, a real shock, and it explains part of the result. But only a part.

Dahis and Szerman (2025) test the proposition directly with a horse race. The benchmark specification on outcome y is re-estimated controlling for log municipal revenues:

If the gains were just an income effect, β would shrink toward zero once revenues are absorbed. It shrinks, but it lingers as a substantively large and significant effect across outcomes. For trash collection, the coefficient drops from 4.45 to 3.38 percentage points. For sewage, from 1.12 to 0.76 percentage points. For luminosity (in a parallel district-level horse race using imputed district revenues), from 0.35 to 0.32 log points. Income explains a slice of the result. It does not explain most of it.

The asymmetric service pattern is the more persuasive piece of evidence. Gains show up where new municipalities have unilateral authority and vanish where they share authority with higher levels of government. A pure cash injection would not behave that way, and the geographic heterogeneity points in the same direction. And after splitting, applicant and headquarters districts elect mayors from different parties about 75% of the time, rising to 85% two decades later. The applicants were not politically aligned with their old headquarters. They had distinct preferences and, once they could express those preferences in their own elections, they did. This is the premise of Oates (1972) tested literally, in line with the formal representation logic of Persson and Tabellini (2002) and the accountability mechanism of Seabright (1996).

The political-economy alternatives mostly do not fit. Hassan (2016) and Gottlieb, Grossman, Larreguy, and Marx (2019) model splitting as an endogenous distributive choice by incumbent politicians who benefit from carving away opposition voters. The Brazilian data show no such pattern. Political alignment between local mayors and state governors does not predict either the filing of split requests or their success (Appendix Table D.9 of Dahis and Szerman 2025). Left-wing mayors are somewhat more likely to file split requests but no more likely to succeed. The boom does not look like a state-level partisan project.

Tiebout-style migration also fails to show up. A cross-sectional regression using the 2000 census, which is the first to record municipality-of-residence five years earlier, finds no effect of splitting on inward migration (Appendix Table D.10 of the same paper). People did not vote with their feet for the new municipalities. Whatever was driving the gains, it was not residents reshuffling across jurisdictions in search of better services. The Tiebout (1956) mechanism is absent from this case, while the Oates (1972) mechanism of better-aligned local policy is doing real work.

Does the Rest of the Country Pay

The FPM is zero-sum within each state. When new municipalities draw their share, the municipalities that did not split lose a bit of theirs. Dahis and Szerman (2025) take this seriously and run a state-level analysis. States with more splits lose more transfers, so if the policy creates losers, we should see them concentrated in high-fragmentation states. The data is clean enough to compute the share losses precisely within their sample period. Municipalities that split increased their share of federal transfers by about 20.3% on average. Municipalities that did not split saw their share fall by about 13.7% on average. The mechanical redistribution is large.

What happens to the non-splitters? The slopes from a state-level binned scatter, weighted by population and residualized for region dummies, are essentially flat. The coefficient on public jobs is 1.06, with a standard error of 7.91. On private jobs, 1.56 (s.e. 11.61). On establishments, −4.25 (s.e. 2.95). On luminosity, 2.43 (s.e. 2.44). None of these moves systematically with the size of the transfer loss. The confidence intervals are wide enough to rule out very large negative effects but not modest ones. With 25 effective states, the inference is genuinely limited, but the population coverage is essentially complete, which weakens the standard amostral critique.

The most plausible reading borrows from Liebman and Mahoney’s (2017) work on year-end federal procurement and the marginal value of public spending. Liebman and Mahoney document that federal agencies systematically spend down their budgets in the last weeks of the fiscal year on lower-value projects to avoid future budget cuts. The implication is that the marginal value of public spending can sit well below the social cost of funds in many institutional contexts. If pre-split parent municipalities looked like Liebman-Mahoney federal agencies, then reallocating FPM toward smaller new municipalities (where the marginal dollar bought more) raised aggregate welfare even at constant transfers. This is not directly tested. It is consistent with what we see.

The authors are duly cautious. The reader who wants to insist the boom was a net loss can find room to do so in the wide confidence bands, but they will have to find that room somewhere other than the data on visible economic damage to non-splits, because the data on visible economic damage is clean and shows none.

How Much of a Deal Is This

Two summary statistics tell you whether the magnitudes are large. The first is the cost per public job. The aggregate effect of splitting on municipal employment is roughly 15.8% against a baseline of 604 jobs in the year before splitting, so 95.65 additional jobs per municipality. The aggregate effect on FPM transfers is 36.75%. Translating to constant 1998 reais at 2016 dollar equivalents (the convention used by Corbi, Papaioannou, and Surico (2019) to make their numbers comparable), the additional FPM is about US$ 347,738 per municipality per year. That gives a cost per public job of around US$ 3,635 per year.

Corbi, Papaioannou, and Surico (2019) estimated US$ 8,000 per job for the same outcome from FPM transfers without an accompanying split. Splitting produces public-sector employment at roughly half the cost per dollar of transfer. For reference, Gerard, Naritomi, and Silva (2024) report a cost per job of US$ 9,799 for the Bolsa Família cash transfer program in Brazil, so splitting also dominates cash transfers on this margin.

The second statistic is the implied output multiplier. With no municipal GDP data before 2002, the authors estimate the GDP gain indirectly through the elasticity of nighttime luminosity to GDP in their sample. The luminosity gains imply about US$ 717,164 in additional GDP per municipality per year. Dividing by the US$ 347,738 in additional FPM gives an output multiplier of 2.06.

An alternative methodology from Chodorow-Reich (2019) ties the output multiplier to the employment multiplier through the production function:

where (1 − ξ) is the labor share in production, (1 + χ) is the elasticity of hours per worker to total employment, Y/E is output per worker, and μ_E is the employment multiplier (the inverse of cost per job). Using the same parameter calibration as Corbi, Papaioannou, and Surico (2019) of (1 − ξ) = 0.666, χ = 0.12, Y/E = 21,152, the splitting employment multiplier translates into an output multiplier of 4.34.

The question is that Chodorow-Reich (2019) surveys fiscal spending multipliers across mostly rich-country studies and reports a median of 1.9, but Brazilian splitting comes out at the upper end of that distribution. The interpretation is that splitting combines a transfer (which has its own multiplier) with an institutional change (which raises the productivity of the transfer). The fact that the splitting multiplier exceeds the pure transfer multiplier from Corbi, Papaioannou, and Surico (2019), which ranged from 1.1 to 2.6, is the headline summary measure of how much extra economic activity the institutional change is producing. The institutional component cannot be a null contribution if the combined multiplier is larger than the pure-transfer multiplier.

Now, none of this tells you the optimal number of Brazilian municipalities. A multiplier above the literature median says the marginal real was put to productive use during this particular boom, but it does not mean at all that further fragmentation would have the same results. The benefits Dahis and Szerman (2025) document apply to a specific configuration of unusually large pre-existing municipalities, demonstrably underserved peripheral populations, lenient state-level rules, generous federal transfers, and voluntary applications from below. Strip those conditions out and, well, everything changes.

The limitations of the paper

The textbook fiscal critique of fragmentation, that small new units cannot finance themselves, that they survive only on transfers, that the transfers impoverish the rest of the country, that the new units do not deliver services worth the money, does not survive the Brazilian data. New municipalities funded themselves through federal transfers, built bureaucratic capacity quickly, expanded the services they were uniquely accountable for, attracted private retail activity, and grew their local economies.

The municipalities that did not split do not show visible damage. The structural gradient of the gains, concentrated in peripheral low-urbanization districts far from the old town hall, fits the decentralization theory of Oates (1972) and Bardhan (2002) better than it fits any pure income-effect alternative.

The harder questions are still open. Revenue windfalls in Brazilian municipalities have been linked to higher corruption and worse politician quality (Brollo, Nannicini, Perotti, and Tabellini 2013, Boffa, Piolatto, and Ponzetto 2016). Small jurisdictions can fail to coordinate on pollution, infrastructure, and regional planning (Lipscomb and Mobarak 2017).

Nothing in the Dahis and Szerman (2025) data speaks to these costs directly, however. The aggregate question of whether Brazil should have fragmented more or consolidated requires evidence that goes beyond this paper. Recent work by Narasimhan and Weaver (2024) on splits in Uttar Pradesh and by Cohen (2022) on Uganda gives slightly different readings in those settings, which suggests that the conditions that made Brazilian splitting productive are not universal.

In a country with very large municipalities and visible peripheral neglect, subsidized voluntary splitting generated durable gains for the seceding peripheries, did not visibly drain the rest of the country, and produced new public employment at a cost per dollar that beats most fiscal interventions in the literature. The Brazilian boom is one careful data point on a long-running argument. It is worth taking seriously.