I. INTRODUCTION

It is difficult to argue with success. Brazil's public steel firms - once the crown jewels of a state-led model of development⁽¹⁾ - were, by the late 1980's, heavily indebted, unproductive and dependent upon the financial support of a state in fiscal crisis. As if by a wave of the proverbial magic wand, their privatization in the 1990-94 period made them exponentially more productive, financially sound, and competitive. Of the eight major Brazilian steel enterprises, seven were returning profits by the end of 1994.⁽²⁾ One of the seven profitable steel firms, Usiminas, became the most profitable private firm in Brazil in 1993, two years after it was privatized.

Popular economic periodicals such as U.S./Latin Trade (now Latin Finance) and The Economist touted such transformations as "a lesson in free enterprise" and evidence that free markets "win" when "nanny state" stays out (U.S./Latin Trade, 1994; The Economist, 1995). According to these views, steel privatization in Brazil was the result of a new commitment to neoliberal principles among economic policymakers and the retreat of a bankrupt developmentalist state with few options.

Although such arguments are intuitively appealing to the advocates of liberal economic reform, they are ultimately facile and uninformed. This essay will argue against such interpretations on two levels. First, I will argue that the privatization of steel in Brazil was a profoundly political process that was not primarily led by either fiscal constraints or "neoliberal" policy ideas. Instead, Brazil's privatization of steel during the 1990-94 period represented a deliberate strategy by a neodevelopmentalist state to implement an industrial policy. Second, I will argue that in order to make the new industrial policy politically sustainable, state managers used various policy mechanisms during the privatization process to forge alliances with segments of business and labor. These strategies satisfied state interests by supplying private resources to serve public purposes in restructuring the public steel firms.

The first argument is based on the study of the Fernando Collor administration (1990-92), which initiated the Programa Nacional de Desestatização (National Program of Destatization - PND). The Collor government saw privatization as part of a larger program to reorient Brazil's industrial policy along more efficient and sustainable lines. The result would be a stronger state rid of patrimonial interests and endowed with a strategic sense for retooling the economy. Such ideas were neither "neoliberal" in content nor designed by neoliberal "technopols." The technocrats who conceived privatization policy came from the ranks of the National Economic and Social Development Bank (Banco Nacional de Desenvolvimento Econômico e Social - BNDES), the state agency most responsible for Brazil's developmentalist policies from the 1950's through the 1980's.⁽³⁾ These policy leaders advocated strategic forms of public sector intervention that would enhance rather than replace market forces.

Collor's support for the PND as part of his campaign to restructure the Brazilian state provided the program's managers with extraordinary protection from opposition groups led by nationalists and leftists who charged that the "patrimony of the Brazilian people" was being sold to the highest bidders. In order to make privatization policymaking autonomous from these voices, Collor created a Directorate of Privatization to manage the PND and he assigned the president of the BNDES as its head. Along with the support of the Ministries of Economy and Planning, the Directorate, led by Eduardo Modiano, planned the first wave of steel privatizations with blistering speed and through confidential negotiations with business and the BNDES on sale prices.

The second argument is based on the fact that dependence upon Collor's support was insufficient for guaranteeing the success or survival of the PND. In order to find enough capital and political support for privatization, policymakers had to create durable alliances with business and labor. By restructuring the debt of the steel mills, subsidizing potential private buyers, and distributing ownership of the firms to workers, the managers of the PND were able to forge alliances with key economic actors by reducing their risk of investment and feeding their expectations of increasing gains from future privatizations. These alliances produced a growing constituency for privatization that outlived Collor's presidency.

Despite the key role of business and labor, this argument does not lose sight of the fact that these ties were driven by state interests. The PND's various mechanisms were designed to create a sufficiently financed private market for the steel mills - one capable of restructuring the firms according to industrial policy goals established by the BNDES. Since tapping into private resources was essential to implement privatization as industrial policy, the PND could not be conceived in isolation by an autonomous coterie of BNDES elites. The program had to create a web of associations with business and labor.

Peter Evans' (1995) has argued that structural reforms of industry are best handled when the state is sufficiently autonomous to be independent of societal demands, but well connected to economic actors to maintain access to information and other material resources crucial to policymaking. Following Evans' notion of "embedded autonomy," this study argues that the PND and its supporting policy constituency enjoyed both the internal coherence of BNDES technocrats and key government supporters (autonomy) and the external connectedness of ties to private sector interests (embeddedness).⁽⁴⁾ The combination of Weberian bureaucratic cohesiveness and associations between policy elites and economic actors signaled business and labor of the government's commitment to industrial reform while it provided the material means (finance and political support) to consolidate structural adjustment.⁽⁵⁾

The crucial test of the durability of the PND's embedded autonomy came after Collor was forced from office in 1992 under mounting evidence of corrupt political practices. He was replaced with his vice-president, Itamar Franco, a politician with strong ties to the opponents of privatization. Despite such challenges to the PND, its administrators proved well-connected and flexible enough to shift authority to other ministries and rely on supporters in other state agencies. The PND also survived due to a formidable constituency of supporters created during the Collor privatizations: large conglomerates with sunk costs in privatized companies, workers with money at stake in future sales, and an increasing pool of interested foreign investors. Without the flexibility embeddedness afforded the PND's architects, bureaucratic autonomy would not have been enough to guarantee the program's survival.

The case of Brazilian steel privatization as discussed above challenges several established views in the political economy of structural reform literature. First, by emphasizing the political interests of the Brazilian state, the approach followed here highlights the weaknesses of arguments which see the pace and content of reform as the result of economic crises. In these formulations Latin American states are seen as being pushed and pulled by external and fiscal necessities. Usually market-oriented reforms are seen as the inevitable responses to eroding import-substitution industrialization models (Balassa et al., 1986). Such reforms are also treated as the products of political pressures on debtor states by international financial institutions (Stallings, 1992).

Although economic crises often catalyze policy change (Gourevitch, 1986; Collier & Collier, 1991), even the best comparative studies are not always clear on how particular crises determine the specific content of policy (Corrales, 1997-98: 618-620). Much recent evidence from Latin America shows that the major external shocks caused by the debt crisis led not to the reign of a neoliberal policy "consensus" but to diverse roads through economic adjustment. Comparative surveys of economic policy have revealed that both the most neoliberal reformers in Latin America (Mexico and Chile) and the most hesitant and gradualist cases (Brazil), have formulated and sought to implement a diverse array of interventionist development policies during the 1990's (Peres, 1997a & 1997b; Pérez-Alemán, 1997; Montero, 1997). Such variance among so many countries facing severe economic constraints suggests that domestic political interests shape reactions to external pressures (Kahler, 1992).

The possibility for diverse responses to economic crises is evident in the case of privatization. Privatization is often simplified as "any withdrawal of government from economic activity," but state interests always intervene to configure the reform process into something more than a mere retreat of the state from the market. The development goals that privatization is meant to serve are potentially legion, ranging from efforts to resolve macroeconomic problems to concerted programs meant to promote microeconomic restructuring of industry (Bartell, 1994: 72-74). These goals are not determined exclusively by economic exigencies. They are shaped by politicians, policymakers, and societal agents.

The current study argues that the interests of state and societal actors in Brazil mediated the economic pressures on privatization policy. But it also attempts to address a political economy literature that is not always clear on how to characterize the way that political interests affect policymaking. On the one hand, many political economists argue that the main influences on the content of structural adjustment emanate from autonomous coteries of technocratic elites capable of implementing policies despite determined opposition (Waterbury, 1992). These "strong states" are well known in the study of East Asian development (Johnson, 1987) and efforts to apply these "lessons" to Latin America are now commonplace (e.g., Gereffi & Wyman, 1990; Wade, 1990: 345-381). On the other hand, in recent years, numerous observers of structural adjustment in developing countries have argued that such policies have not been primarily shaped by autonomous state technocrats but by the interests of business (Frieden, 1991; Silva, 1993 & 1996) or, more broadly, by elites in response to distributive conflicts that threaten the power base of reformers (Olson, 1982; Manzetti, 1993a; Hellman, 1998). Therefore, analysis of economic reform has tended to stick in between these two irreconcilable extremes of seeing policy as the product of autonomous state elites or the result of determined societal interests.

Lost in the equation are the interests of politicians and bureaucrats and their relationship to economic actors. In a broad review of the literature, Barbara Geddes (1995) has recently noted that there is little on-the-ground evidence that either completely autonomous state managers or self-interested societal actors have proven capable of determining the content of structural adjustment. Geddes finds a weak relationship between societal interests and government policies.⁽⁶⁾ Instead, the interests of politicians and bureaucrats, in concert with the economic strategies of segments of business and labor, have guided reform.

Even when political outsiders have assumed power and produced reform behind an institutionally autonomous policymaking bureaucracy, the survival of these reforms has required the forging of strategic alliances with societal actors. These ties have emerged not only to satisfy the electoral interests of reforming politicians, but also to assure technocrats and signal economic actors that the reform process will persist after the politicians that inaugurated them vanish from the scene. Without an enduring constituency for reform (privatization in this case), policymaking becomes politically vulnerable to opposition actors who seek to slow down or stop the process (Manzetti, 1993b: 435). Without substantial commitments by the state to the reform process, economic actors who are required to bear the costs and risk investment in privatized companies will not see official adjustment efforts as credible.⁽⁷⁾ Therefore, by focusing on the organization of the state and its relations with societal actors, embedded autonomy traverses the gap between neopluralist and statist arguments, providing a more useful analysis of how structural reforms are conceived, implemented, and survive.

The dual purpose of this essay is to correct economically determinist arguments about the adjustment process in Brazil and to provide further evidence for the utility of embedded autonomy through a case study of steel privatization. In the following section, I will discuss how the Collor administration packaged privatization as part of a larger program to restructure the role of the Brazilian state in the economy. I will demonstrate that Collor's role proved crucial to moving new industrial policy ideas mostly composed within the BNDES to the foreground of economic policymaking. In the subsequent section, I will show how these ideas were put into practice before, during, and after the auctioning of the steel mills. The evidence will prove that the state's fiscal crisis was only a secondary policy consideration. Privatization was implemented in ways that made the competitive restructuring of the steel mills policy priorities. The final two sections will flesh out how privatization mechanisms were designed to create sustainable political alliances among state agencies, businesses, and workers. The endurance of these ties imbued the PND with enough political flexibility to survive popular opposition to privatization and even the concerted efforts of some segments of the Itamar Franco administration to slowdown or scuttle the process. The evidence will show that Collor's initial support was crucial for getting the PND moving, but it was not sufficient for explaining its survival through repeated political attacks after Collor abandoned the presidency. The embedded autonomy of steel privatization policy explains these unexpected outcomes.

II. COLLOR AND THE FORMULATION OF THE NEW INDUSTRIAL POLICY

The unlikely rise of Fernando Collor to the Brazilian presidency in 1989 was the most important political catalyst for advancing the country's structural reform agenda. The PND would not have been possible without Collor and it was clear that the BNDES' ideas about using privatization as a means of implementing a new industrial policy satisfied the president's plans for restructuring the Brazilian state. Collor's embrace of the PND gave BNDES managers the necessary protection they needed from nationalist opponents to privatization. At the same time, the new industrial policy promised to respond to the demands of public steel firms eager for a new infusion of capital required of productivity-enhancing restructuring. With its focus on steel, the new industrial policy promised to hand Collor his first major successful structural adjustment, one that would send strong signals to the multilateral agencies and the business world that Brazil was serious about economic reform.

Collor's sudden rise to the presidency in the late 1989 presidential elections was a consequence of a popular backlash against the failed policies of his predecessor and fears of what his most powerful opponents on the left would do if they were elected. José Sarney and his "New Republic" (1985-90) proved incapable of reversing Brazil's steep economic decline. Megainflation and slow growth doomed Sarney's numerous anti-inflation plans and all efforts to keep Brazil from slipping into a fiscal crisis that would seriously erode the capacity of the central state (Bresser Pereira, 1993). The explosion of clientelist politics under the "New Republic" further fragmented Brazil's political society and paralyzed all attempts to reform the state (Weyland, 1996; Schneider, 1987-88). These failures also strengthened the organized left, led most prominently by Luis Inácio "Lula" da Silva, the fiery head of the Workers' Party (Partido dos Trabalhadores - PT). By the late 1989 presidential elections, Brazilians were demanding a political outsider who would revamp the country's politics. In response, Fernando Collor promised middle class and upper class voters a seachange in Brazilian politics without the wrenching sociopolitical upheavals many feared Lula represented.

Central to Collor's electoral campaign was the idea of reforming the state in order to strengthen it (Weyland, forthcoming). Collor vowed to jettison the indolent Marajás ("Maharajas") of the state who reaped untold rewards from the corruption and patronage of public office. Intent on excluding the parochial interests that penetrated the public firms, Collor justified privatization as a tool for sweeping away patrimonialism from the economic bureaucracy and enhancing the efficiency of public policy (Collor, 1989). Along with liberalization, deregulation, and aggressive plans to "kill inflation with a single shot," Collor's campaign made privatization of public firms a priority of his government (Schneider, 1992: 231).

At no point, however, was Collor's reform package a carbon copy of the neoliberal structural adjustments championed by other Latin American leaders. Collor was not simply a Brazilian version of Carlos Menem (Argentina), Alberto Fujimori (Peru) and Carlos Salinas de Gortari (Mexico). Collor and his reform-minded ministers did not see the state as the problem and the market as the solution. Assuming otherwise glosses over the very active role these policymakers envisioned for the Brazilian state in restructuring the public steel firms. It also obscures the priority for state managers and Collor himself of using structural adjustment to strengthen the state's role in the economy.

First, Collor was no textbook neoliberal and neither were the political technocrats in his government. The new president advocated neoliberal reform with the avowed intention of empowering the state, not forcing the state into retreat in favor of market forces (Collor, 1990a; Collor, 1990b).⁽⁸⁾ Unlike Mexico, the Brazilian state was not captured by a neoliberal technocracy.⁽⁹⁾ Collor's economic team did not generally share a common educational background as was true of Chile's "Chicago Boys" during the 1970's and 1980's. Indeed, few held graduate degrees from foreign universities (Schneider, 1992: 228-229).

Unlike Menem and Salinas de Gortari, Collor was not compelled to adopt neoliberal policies by right-wing parties or business associations. In fact, Brazilian businesses offered almost no clear signals of any support for privatization (Bartell, 1994: 88-89). Brazil's major business associations, especially the Federation of São Paulo Industries (FIESP), remained skeptical of the benefits of privatization. Smaller business groups gave hesitant support for structural reforms. Yet no widespread right-wing or business sentiment called for privatization before or during the Collor administration (Kingstone, 1994; Schneider, 1992: 226). Also, unlike Margaret Thatcher in the United Kingdom or General Augusto Pinochet in Chile, Collor did not implement privatization policy as a response to the actions of a strong political left.

Rather than an embrace of neoliberal ideas, key policy statements during the Collor administration evinced considerably more support among state managers for new forms of 'sustainable' industrial policies. In early 1991, the Collor government launched the Programa de Competitividade Industrial (Industrial Competitiveness Program - PCI) as a mix of federal tax incentives, subsidized credit, and tariff adjustments that would reduce the price of investment in a select group of agroindustrial, manufacturing, and high-technology sectors. Observers remarked that the PCI had pretensions of returning Brazil to its developmentalist industrial policy of the 1950's when Juscelino Kubitschek organized "sectoral groups" of business leaders that, along with state technocrats, designed and implemented investment plans (Erber, 1992).⁽¹⁰⁾

The PCI, however, was not purely developmentalist. It focused on promoting industrial sectors with comparative advantages, industries viewed as "generators and diffusers of innovation and technological progress." As such, the plan envisioned an "active industrial policy" designed to "catalyze" private production around performance targets established by the market (Ministério da Economia, Fazenda e Planejamento, 1991).

The neodevelopmentalist ideas behind the PCI were general and, in the end, were not implemented to the ambitious extent first envisioned in the plan. Nevertheless, the importance of these ideas was not lost on technocrats operating within Brazil's national development bank. With the decline of state-led developmentalism during the 1980's, BNDES leaders had become more focused during the "lost decade" on alternative mechanisms of industrial policy that would, as the PCI argued, "catalyze" market forces. Gone was the emphasis on public production. Instead, by 1989, the bank had privatized nearly all the public firms it had in its holding company, BNDES-Par.

The shift in the BNDES's bureaucratic mission was the result of several factors. First, the development bank became more market-oriented as it fell under the influence of bank presidents with a private sector background. Marcio Fortes, the BNDES' president under the Sarney administration, favored policies designed to enhance the competitiveness of private firms in foreign and domestic markets (Schneider, 1987-88: 99). During this time, public funds for industrial automation became a prominent issue for BNDES directors (e.g., BNDES, 1988).

Second, continued circulation of bank elites between the BNDES and the private sector during the 1990's reinforced market-oriented approaches (Lima, 1996; Castanheira, 1995). Several PND heads during the Collor and post-Collor privatizations came from and later returned to the private sector. This was true of Eduardo Modiano and Elena Landau, the head of the PND during the first year of President Fernando Henrique Cardoso's term (1995-?).

Finally, the bank's own homespun experience with privatization during the 1980's convinced many of its executives that the same could be done with the rest of the Brazilian state, but in ways that would also reorganize the BNDES' (and more broadly, the Brazilian state's) capacity to intervene in private markets. Bank technocrats embraced new directives circulated during the latter half of the decade that focused on creating a role for the bank in facilitating long-term finance for large firms eager to restructure their production and orient themselves toward the external market (Amarante de Andrade et al., 1994). Increasingly, support for infrastructure projects and specialized finance programs replaced wholesale support for public firms (Meilman, 1996; Carneiro, 1996; e.g., BNDES, 1996). These represented qualitative shifts from developmentalism to a mixed approach at the BNDES.

Privatization became a centerpiece of the new thinking at BNDES but in ways that were quite opposed to neoliberal ideas that sought the minimization of state intervention in the economy. BNDES managers saw privatization as a mechanism for making state intervention in the economy more efficient. Eduardo Modiano, the first head of the PND under the Collor government, articulated the general intention more directly when he noted shortly after announcing the first plans for privatization in early 1991:

The reform of the state [through privatization] is not intended to debilitate the state. On the contrary, the objective of privatization is to strengthen the state and make it more capable of performing those functions that are a part of its undelegatable attributes; functions that have been neglected in recent years (Jornal do Brasil, 1991b).

Despite the change of government and Modiano's departure in 1992, BNDES officials in charge of the privatization process a year later articulated the same goals. André Franco Montoro Filho, director of the commission for the PND under the Itamar Franco administration (1992-1994) would argue:

I think that we have demonstrated that this program is very important for the reformulation of the Brazilian state; for the capacity of the state to recuperate its power to coordinate, articulate and implement policy (Estado de São Paulo, 1993).

In the case of the public steel firms, BNDES leaders saw privatization as essential for making the mills competitive. Although Brazilian steel had achieved some export growth during the late 1980's, productivity was low and the sector was too specialized in low-end steel. Unlike the world's leading steel producers, most steel manufacturing in Brazil used antiquated methods with low levels of productivity. Only about half of Brazilian steel production employed continued stamping and hot laminating processes whereas more than 80 percent of the European and 94 percent of the Japanese and South Korean producers employed this more productive method. Proposed investments for restructuring the firms with new production techniques required an average of $3 billion per firm, an amount well beyond the financial capacity of either the state or the firms themselves. Additionally, most Brazilian steel producers focused on the export of low-end steel, not the specialized steel most in demand in the advanced markets of Europe, Asia, and North America. As the supplier of only 3 percent of the world's steel, Brazilian steel would have to advance several costly restructuring programs at its major enterprises to really compete in the global steel market (Amarante de Andrade et al., 1994).

The primary impediments to systemic restructuring at the steel mills were financial. Excessive protection and low export activity were not the culprits of Brazilian steel's problems. Quite on the contrary, by the early 1990's, due to the country's recession and the decline in the internal market for steel, more than 50 percent of steel production in Brazil was being exported (Coutinho and Ferraz, 1994: 265). Brazilian economists consistently argued that steel was one of the few industrial sectors in Brazil that showed good chances of becoming more competitive in world markets, if it could be restructured (e.g., Coutinho and Ferraz, 1994: 261-281).

Decades of developmentalist policy, however, left the public steel mills heavily indebted.⁽¹¹⁾ As a group, the steel firms ran heavy losses due to their financial commitments (Figure 1). Siderbrás, the state holding company for the steel firms, was hopelessly mired in more than $7 billion of debt - a condition that led to prolonged freezes in its plans to expand public investments.

BNDES leaders agreed with steel firm managers that public steel was too "over-bureaucratized" and "under-financed" to improve their rates of productivity. The status of these firms confirmed views widely held by BNDES leaders and steel firm executives that state ownership was insufficient to make the steel firms internationally competitive.⁽¹²⁾ For BNDES technocrats, the fiscal crisis of the Brazilian state severely limited the extent to which the administration could spend on restructuring the steel firms. By mid-1990, most new investments by the public steel firms were halted for lack of government funding. As a result, the Second National Steel Plan (Pedesid), an elaborate roadmap of directives that foresaw investments of more than $10 billion by the public steel firms, was relegated to the attic of planning by Ministry of Infrastructure officials.

More strategic forms of state intervention, however, along with an infusion of private capital, could provide the necessary resources Brazilian steel needed to boost productivity and restructure production. The Institute of Brazilian Steel (Instituto Brasileiro de Siderurgia - IBS), the major business association of the steel sector, embraced such a mixed approach.⁽¹³⁾ Arguments for privatization among IBS members included references to the need for public finance. Many steel executives noted that, as private firms, the steel mills would be eligible for an array of official credits from the national development bank, still the chief financier of productive restructuring investments in Brazil. In this way, BNDES officials and steel firm managers viewed privatization as part of a larger state-initiated process of productive restructuring of the steel sector that would persist with an important state component even after the sell-offs (e.g., Landau, 1995).

Support for privatization had been growing for some time among the ranks of management in the public steel firms. Saddled with debt, each of Brazil's major public steel enterprises was running annual losses close to $2 billion. In early 1990, several managers of the public steel firms announced their support of a plan first proposed by the outgoing president of Siderbrás, Moacélio Mendes, to convert the holding company into a "development agency."⁽¹⁴⁾ The essence of the proposal was a call to reduce public liabilities in the steel sector while at the same time provide enough capital to allow the steel firms to restructure and become more competitive on international markets.⁽¹⁵⁾ During the policy debates that followed the Siderbrás proposal, BNDES leaders, Collor administration ministers, and the heads of the state steel companies agreed that privatization would be the preferred course of action.

Actually transforming these ideas into policy required, first and foremost, a political catalyst supplied by the Collor administration. The architects of privatization policy were concerned with losing control over the process to legislators and other political interests. For this reason, the Directorate of Privatization of the PND, led by Eduardo Modiano, defended its independence from the congress that initially sanctioned its creation and approved its members (Modiano, 1991). Modiano and the Directorate also defended the presidency's continued control over the "minimum price" of the firms to go on sale. This would prove a crucial authority since the minimum price would become the greatest target of anti-privatization forces and, at the same time, the most important lever for subsidizing domestic buyers.

Support by Collor and his closest economic advisors proved essential for protecting the autonomy of the PND Directorate from opposition voices. Zélia Cardoso de Mello, Collor's first Minister of Economy, insisted that all government ministries support the "destatization" program. At times, her determination to keep other members of the government in line led to public squabbles with other ministers and public firm executives.⁽¹⁶⁾ More than a year later, after "Zélia" left the Collor administration, her replacement proved to be a more ardent defender of the PND. Marcílio Marques Moreira, the new Minister of Economy, reassured Modiano and BNDES officials that the new economic team would stick to the established schedule of privatization auctions. Moreira also defended the PND Directorate's authority over the selection of purchasing modalities and other mechanisms crucial to the sales.

Although the PND's managers received the benefit of a high level of political autonomy, they were also aware of the dangers of becoming too insulated from economic actors. Despite ministerial shake-ups in 1991, Eduardo Modiano remained the only high-ranking economic official of the Collor administration who survived and persisted in his job through much of 1992. During May of 1992, however, Modiano's position as the head of both the PND and the BNDES was threatened by a proposed plan circulated among government leaders that would replace privatization auctions with a system of direct negotiation between a new Secretariat of Privatization and buyers. Several government observers admitted that the proposed reform was in fact an attempt to use the BNDES in the routine exchange of political appointments and patronage for political support common in Brazilian politics (Estado de São Paulo, 1992a).⁽¹⁷⁾ Daniel Dantas, president of the Icatu bank and a personal friend of the very symbol of patronage in Brazilian politics, Bahia governor Antônio Carlos Magalhães, was said to be behind the move. A few days later, however, Moreira's support of Modiano made it clear that the BNDES president would not be replaced and no substantial adjustment would be made to the PND. Accusations that Dantas' proposed reform would eliminate "transparency" from the privatization process forced the bank president to cancel his attempt to upstage Modiano. Such events suggested that the PND required the political support of key ministers such as Moreira, but insulation of policymaking would invite arbitrary and parochial interests.

Based on the points made above, the PND was made possible by two complementary processes. First, the existence and increasing sophistication of sustainable industrial policies in the BNDES made industrial reform the centerpiece of the PND. This satisfied the interests of both state technocrats and public steel firm elites who believed that only a mixture of public and private financial support could provide the resources needed for productive restructuring of the mills. These were technical grounds to support the PND as industrial policy. However, Collor supported the PND as part of a larger political campaign to strengthen the state. BNDES' industrial policy ideas fit neatly into the Collor government's plans for restructuring the state as they provided the means for reorganizing state intervention in ways that would enhance official goals in industrial reform. In turn, the Collor administration provided the key political catalyst that got the PND moving as part of a larger package of economic reforms.

In part, the original goals of the PND were preserved due to the coherence of its management under Modiano and the autonomy the Directorate enjoyed. In no small way that autonomy was protected by the political support of consecutive Ministers of Economy during the Collor administration. However, it was also clear that the PND relied on numerous contacts in Brazil's private financial and industrial markets. By the end of the first wave of privatizations, support for the PND could be found among influential industrial, financial and even labor groups. This support was engineered by the very mechanisms of privatization policy in the steel sector. Financial restructuring of the steel mills before their sale, subsidies to private buyers involved in the official auctions, and the distribution of parcels of the privatized companies to workers reduced the risk of investment and enhanced the perceived gains from future privatizations. Over time interested business and labor groups created a constituency for reform that proved crucial in moving the PND forward. Mere reference to economic and fiscal pressures surrounding the Brazilian state do not explain either the content or the political organization of these policies.

III. USING PRIVATE RESOURCES FOR PUBLIC INTERESTS IN STEEL PRIVATIZATION

Public firm managers, BNDES technocrats linked to the steel sector, and private financial and industrial elites provided an emerging constituency for using privatization policy as a mechanism of industrial policy. Policymakers used a diverse array of fiscal, financial, and informational resources to restructure the state steel firms, facilitate their auctioning, and preserve the evolving dominance of a small group of industrial-financial conglomerates in the sector. These mechanisms satisfied both the interests of state technocrats in the BNDES who were eager to show that privatization could create a base for a model of more efficient state intervention in the economy and steel industry elites who saw the continuation of strategic public support as crucial for improving the competitiveness of Brazilian steel. Rather than representing a rolling back of state intervention in Brazil's economy, the new industrial policy sought to strengthen the Brazilian state's ability to contour industrialization in a way that mixed market incentives with a new public role.

The popular view of privatization in Brazil is that fiscal crises forced state managers to pursue privatization. This perspective is not completely misleading. After all, the original executive decree that initiated the PND stipulated that proceeds from the sales of public firms would be applied to the federal debt. The PND's architects continued publicly to articulate this principle as a key goal and Collor embraced it as central to privatization.⁽¹⁸⁾ A cursory look at the aggregate performance of privatization policy as a whole would even suggest that the Collor government succeeded to a certain degree. It sold 17 firms between October 1991 and October 1992, including some of the large steel mills, generating $3.9 billion to the federal government's coffers (BNDES, 1992).

A closer look at the privatization process, however, reveals that fiscal arguments and fiscal mechanisms were manipulated for political purpose. On the surface, the Brazilian privatization process reduced much medium-term and some long-term debt. These gains, however, were partly offset by dozens of public financial mechanisms that were designed to make candidate firms "privatizable." Adding to the $11.3 billion spent on the Brazilian steel mills between 1986 and 1992, the state spent more than $757 million upgrading and restructuring production and reducing debt at its steel enterprises between 1990 and 1993. During privatization, state managers employed an array of other financial mechanisms to make the steel firms more attractive and to subsidize their buyers. The design and use of these financial levers were driven by a political raison d'être rather than a fiscal one.

In order to make candidate firms "privatizable," Brazilian state managers spent millions to reform their debt and productive structures. During the Sarney administration, more than $7.6 billion of debt was transferred from the steel firms themselves to Siderbrás. As a result, firms such as CSN and Cosipa that faced perennial losses due to heavy indebtedness, had their finances stabilized months before their privatization. The restructuring of debt cleared the way for public investment in the restructuring of production (Amarante de Andrade et al., 1994: 87-88). In justifying such public investments in the restructuring of these firms, Eduardo Modiano echoed the arguments of his predecessors at BNDES when he argued, "There is not a firm that must be privatized that must not first undergo a process of restructuring" (Jornal do Brasil, 1992).

Among the steel firms to go to sale, Usiminas, Cosipa, Açominas, CSN, Acesita, and Piratini all underwent significant restructuring before privatization; all paid for by public sector expenditures. Concerns for completing restructuring programs determined which steel mills would be first to stand on the auction block. The first privatization under the Collor administration, Usiminas, focused on the steel firm with the most advanced program of productive and financial restructuring in the sector. Usiminas had never posted the large losses of Cosipa and CSN, but it did produce only minimal profits and modest losses in the years leading up to its privatization. The steel mill's inadequate levels of productivity were not ignored by firm managers who began a restructuring program in 1986 that foresaw investments of up to $745 million. These new investments significantly improved automation and quality control, boosting productive capacity and efficiency (Correio Braziliense, 1991a; Jornal de Brasília, 1992). By the time it was sold in October 1991, Usiminas had dramatically improved its competitiveness in both the domestic and the international market, making it a far more valuable firm at the time of sale than it was at the beginning of its restructuring program in 1986 (Corrêa, 1991).

Another example of restructuring is the case of Companhia Siderúrgica Nacional (CSN), the venerable National Steel Company that initiated the Brazilian state's involvement in the sector in 1941. By the time BNDES officials placed CSN on the auction block in 1992, it had undergone extensive debt and productive restructuring. CSN's total debt was reduced by 73 percent ($1.9 of $2.6 billion). The firm's tax debt with the state of Rio de Janeiro was cut from $220 million to $160 million through a combination of write-offs and shifting of liabilities from the state government to Brasília. CSN's labor force was downsized from 24,000 workers to just less than 17,000 and its production system was significantly retooled to reduce costs and increase output (Netto, 1993).

Another noteworthy example is that of Acesita (Aços Especiais Itabira). Acesita started as a private non-flat steel plate firm in 1944. In 1950, the Banco do Brasil, a national public bank, bought up the steel mill. In 1974, the military government led by General Ernesto Geisel imported Japanese technology and employed Banco do Brasil resources to transform Acesita into a specialized producer of rust-proof steel plate, making it the only Latin American producer of this product. However, as a result of its transformation, the firm assumed heavy debts with the Banco do Brasil, totaling $1.1 billion by 1982.⁽¹⁹⁾ Like its sister public firms on the eve of privatization, Acesita engaged in an extensive program of restructuring its production, inventory systems, payroll and external debt in anticipation of being sold off. In July 1991, the company initiated a Plan of Demobilization that sold more than 880,000 square meters of the company's land to the municipality of Timóteo (Minas Gerais) as payment for local taxes (IPTU) and the use of other municipal services and properties. More than 10% of the workforce was laid-off. These changes generated a total $30 million in cost savings (Gazeta Mercantil, 1991c).

On the financial side, Acesita restructured its debt with the Banco do Brasil, which held 97% of the firm's stock. The Bank employed a complex international transaction to swap Acesita's foreign debt of $220 million for Brazilian debt at a discounted price, an operation that effectively reduced the firm's total debt with the Banco do Brasil to $90 million. As a result of the firm's restructuring and simply its placement on the privatization block, its stock values rose markedly.⁽²⁰⁾

The BNDES' view of privatization as industrial policy justified such examples of official facilitation of productive restructuring and revaluation (Tourinho & Vianna, 1993: 68-69).⁽²¹⁾ In justifying these measures, fiscal concerns were made secondary to the interests of making the firms more productive. Luis Chrysostomo, one of the founders of the PND, minimized the fiscal effect of privatization, claiming that the sell offs would affect less than 20 percent of Brazil's external debt, while the benefits of privatization for the firms on the block would more fundamentally "reorient the state's role in Brazilian development" (Gazeta Mercantil, 1991b). The fact that the PND started with Usiminas, arguably the steel firm in the best financial position of the major public producers, suggests that consolidating improvements in production, and not minimizing the state's fiscal costs, were foremost amongst the interests of BNDES managers.

The political and financial resources BNDES poured into the steel firms emphasized to the development bank's managers the importance of making certain that these firms would receive the best possible opportunity in the market. After all, there were many reasons to believe that the process could go terribly wrong for the steel firms. Until this time, Brazil had only a very modest experience with privatization that was limited to the sell-off of firms with serious problems. These privatizations were not designed to make firms more efficient, they were meant to simply unload fiscal weights on the state. That experience was not an appropriate benchmark for restructuring steel firms. Another hazard was the poor financial state of the private market. With few buyers with sufficient capital and in the midst of a recession, BNDES officials feared that the steel firms would be sold to groups that would be unable to meet "minimum price" (the official 'asking price') requirements. Worse still, even if such requirements were met, buyers would not have the capital to make the necessary investments that would complete the productive restructuring of the steel firms.

Perceived limits to the size and financial ability of the private market for public firms encouraged state managers to use fiscal mechanisms to provide incentives to buyers. The early months of planning for the implementation of the PND were dominated by concerns that the program had to use financial mechanisms engineered to "accommodate the characteristics of demand."⁽²²⁾

An array of fiscal mechanisms was used during the implementation of privatization policy. In January 1991, the government announced that certificates of public external debt (Deposit Facility Agreements - DFA's) would be accepted as "currency" during the sale of the first wave of firms that included the public steel giants, Usiminas and Companhia Siderúrgica Tubarão (CST). Prospective buyers would have to purchase these certificates and then use them as "privatization currency" in public auctions. Soon other forms of public debt paper, including certificates of agricultural debt, were included in the expanding pool of "privatization monies." The DFA's became a staple currency in future privatizations and were preferred to the use of "live" money (financial liquidity). During the first full wave of privatizations under the Collor government, between 1991 and mid-1992, the use of DFA's minimized the use of "live" money to only $10-15 million.⁽²³⁾

The employment of DFA's and other assorted "privatization currencies" were important fiscal mechanisms designed to guarantee a sufficiently financed pool of potential buyers for Brazil's public enterprises. Early in the process, Collor administration officials, and particularly those in the BNDES, feared that privatization would be stalled indefinitely as it was so often during the 1980's. Ostensibly, therefore, the government approved the use of public external debt as currency in the sales. Without prior approval by the legislature, the Collor administration's unilateral decision to accept DFA's instantly freed up more than $50 billion in Central Bank accounts.

Other financial and fiscal maneuvering secured a potentially larger buyer's market. The government approved the creation of new funds in the currency of the government's macroeconomic stabilization package, the Collor Plan's cruzados novos. Small investors gradually purchased these accounts and jumped into the first wave of privatizations. This provision freed up more than $36 billion in cruzados novos that had been frozen by the Collor Plan in Central Bank accounts and made them an additional source of capital in the privatization auctions. Like the DFA's, the debt of the public firms up for sale, including recent Siderbrás debentures acquired under restructuring plans, could be purchased for "Certificates of Privatization" (CP's) by prospective buyers. Banks were forced to purchase CP's and hold them for client-buyers. The conversion of debt paper (both DFA's and CP's) and the use of cruzado accounts rapidly infused $109 billion into the market of potential buyers. Brazilian stock market authorities then created funds of DFA's and CP's whose value would fluctuate daily.⁽²⁴⁾

The complex financial transactions that underpinned Brazilian privatization represented more than simply a direct swap of equal value between public debt and equity. The DFA's and other assorted Certificates of Privatization depreciated as they were held in the months prior to purchase. International financial markets and creditors concerned with the overall pace of debt negotiation and structural reform in Brazil exerted the most pressure on the value of Brazilian debt paper. The real value of the notes dropped as international investor confidence in the reform process flagged. Government leaders had anticipated these problems by guaranteeing domestic buyers of privatization certificates the difference (the so-called deságio) between the face value and the real value. As a result, buyers of DFA's and CP's would pay the depreciated market value for this debt paper but it would be the face value of DFA's and CP's that would be counted during the actual privatizations against the minium price set by the state.⁽²⁵⁾ Since the face value of the DFA's and CP's no longer reflected their value on international and domestic financial markets, the money actually paid by buyers was considerably less than the face value of the DFA's and CP's they held. Nevertheless, the final price fetched by privatized firms was often above the minimum price (denoted by the face value of DFA's and CP's) since bidding within the small pool of CP-holding buyers tended to increase the final bid (again, denoted by the face value of debt notes). This practice effectively subsidized the purchase of the public firms while preserving the minimum price requirements. Rules which allowed buyers to purchase DFA's and CP's with other debt paper compounded the positive financial effect for holders of privatization monies. In this way, Usiminas sold for $1.1 billion but the buyers paid only $800 million for the firm.

The conversion of debentures held by Siderbrás into CP's provided added incentives to buyers who knew they were purchasing firms with a much-improved debt structure. In this way, the state relieved some of the privatized firms' debt burden during the sell-offs and it reduced or eliminated its own future financial obligations to the steel firms. In certain cases, such as the sale of Cosipa, the BNDES added financing. Bozano Simonsen, an industrial-financial group, received $278 million in BNDES loans to finance the purchase of CP's in the sale of Cosipa.

Both the pre-privatization process and the actual sales involved an important, articulating role for the state and, increasingly, the other members of the policy constituency. Yet even after the sell-offs, BNDES officials continued to make the privatized steel firms privileged recipients of the development bank's array of subsidized finance. For example, BNDES continued to play a role in restructuring Usiminas after its privatization. In 1992, the steel firm received $75 million in low-interest development loans to continue its modernization. The development bank added $200 million the next year for the same purpose. In October 1993, the BNDES announced that it would invest more than $2.5 billion in the now private and soon-to-be-private steel companies. The lion's share of this money would go into automation and other productivity-enhancing reforms. Similar commitments were made by national fiscal authorities who guaranteed the steel firms the continuation of public credits based on exemptions from the national tax on industrialized products. Such credits amounted to an estimated $484 million by 1996. Although no longer strapped with going concerns in the sector, the Brazilian state nonetheless remained a key player in shaping the investment strategies of the steel firms.

Privatization did reduce short-term debt, but when the cost of restructuring the firms and employing subsidized purchasing modalities is figured in and placed in the context of an expanding internal and external debt during the 1990's, the fiscal gain from privatization was marginal (Mello, 1994; Pinheiro & Giambiagi, 1993). The fiscal impact was too modest and came too late to improve public accounts (Pinheiro & Schneider, 1995; Werneck, 1991).⁽²⁶⁾

Privatization of steel in Brazil was driven by much more than fiscal pressures. It was actively guided by a state interested in restructuring the steel mills. The prior financial restructuring of debt and production at the steel firms, the subsidization of purchases, and the use of public finance after the sales sought to diminish the perceptions of risk in investment among prospective buyers and maximize the competitive potential of the public enterprises (Tourinho & Vianna, 1993: 75).

Like other industrial policies in developing countries (e.g., Evans, 1995), the PND relied on more than the autonomy of state technocrats and reformist politicians. The mechanisms of privatization policy followed a political logic of constructing a constituency for the sell-offs that would guarantee the survival of the PND.

IV. ALLIANCE-BUILDING THROUGH PRIVATIZATION

Purchasing instruments such as DFA's, CP's, Siderbrás debentures and direct BNDES loans made possible the sale of the steel firms by subsidizing an underfinanced private market. Such fiscal and financial sleight-of-hand also helped to silence nationalist voices that predicted the "sell-off of the public patrimony at bargain prices" by artificially overclaiming the final selling price. The official rationale for this array of "privatization monies" (popularly known as "rotten money" - moedas podres) underscored the necessity of starting the privatization process with state-directed mechanisms. BNDES leaders believed that the use of a diverse set of fiscal mechanisms would create the necessary pool of buyers to make the first wave of privatizations possible. Modiano and other PND officials made this point consistently.⁽²⁷⁾ The moedas podres, then, were a necessity born of market constraints.⁽²⁸⁾ In the initial sale of Usiminas, for example, economists note that the auction of the steel mill would not have been possible without the use of moedas podres (Prado, 1993: 109).

Nevertheless, the purchasing modalities used in the privatization of the steel mills satisfied more than the economic requirements for the sales. These mechanisms enhanced the credibility of the government's commitment to structural reform and created a political base of support for the auctions that became broader and increasingly more important to the PND's survival than the protection afforded by Collor and his economic ministers.

The members of the constituency of business and financial supporters for the PND was carefully selected by the program's planners. That was evident with the concentration of purchases in the hands of large conglomerates. The Bozano Simonsen group, a conglomerate with extensive holdings in finance and extractive industries, became a major buyer during the steel privatizations, taking over the presidency of Usiminas and holding significant influence in Cosipa (12.4 percent of shares) and controlling shares in Acesita, Brazil's only major producer of stainless steel. Bozano Simonsen exerted additional influence on Cosipa through Usiminas, which became the principal investor in the privatized steel firm. Such influence was not isolated to private sector investors as Companhia Vale do Rio Doce (CVRD), the gigantic public mining group, bought up outstanding shares of Usiminas, CSN, Companhia Siderúrgica de Tubarão (CST), and Açominas. Both Bozano Simonsen and CVRD teamed up along with a second private financial group, Unibanco, in the purchase of CST, Brazil's chief producer of semi-completed metal plates for export. Grupo Gerdau, the largest private steel producer on the eve of privatization, continued as a major manufacturer of metal plate with its majority purchase of Cosinor and Aços Finos Piratini and its ownership of Usiba, the largest steel producer in Brazil's poor Northeast. Mendes Jr., one of Brazil's largest groups with financial, industrial and public contracting interests, purchased Açominas. The PND's selective mechanisms might have avoided the concentration of ownership as depicted in Table 1.1. However, that interest was not more important for BNDES officials than restructuring the firms and creating a powerful business constituency for continued privatization.

The economic potential of this concentration of ownership was tremendous. By September 1994, six industrial and financial groups dominated the Brazilian steel sector: Usiminas/Cosipa, CSN, CST, Gerdau (Piratini), Mendes Jr./Açominas and Acesita. Usiminas and Cosipa alone controlled 61 percent of the market for flat, unalloyed steel plate used by the major producers of Brazil's extensive automobile manufacturing complex. CSN commanded the other 39 percent. Considering interconnections among the six groups through common ownership of shares, the industrial-financial conglomerates behind these groups exerted a formidable influence on the working of the steel sector. Bozano Simonsen's steel empire controlled 90 percent of semi-completed metal plates and 100 percent of large plates. The Gerdau group came to control 41 percent of the market for laminated metal plate and 14 percent of the market for metal beams and spikes.

By defining the industrial groups that would control these firms through the prior sale of moedas podres, BNDES managers were able to create additional incentives for the participation of foreign and domestic firms linked to these private groups during the second wave of stock market sales of shares. As several of Brazil's large industrial and financial groups became involved in the purchase of moedas podres, affiliated and competing industrial and financial interests were drawn into the Brazilian market. This tactic constituted an additional guarantee on the appreciation of the value of outstanding minority shares of privatizing firms while it provided added assurance that the newly private firms would be adequately financed.

The use of moedas podres also attracted the support of Brazil's largest banks for the continuation of the PND. After being forced to purchase certificates of privatization in 1990 and eager to unload them and the government debt they represented, Brazil's largest banks embraced privatization as a means for expanding their own interests in Brazilian industry and, at the same time, reducing their outstanding debt, particularly the $3 billion in Siderbrás debentures they held. As a result, Brazil's largest private banks became key owners of newly privatized firms. Bamerindus, Itaú, and Unibanco added their capital to emerging industrial-financial groups led by Bozano Simonsen and CVRD.

Why didn't such apparent collusion among the Brazilian political and economic elite inspire the opposition to privatization policy? One reason was that the true nature of the purchasing modalities was hidden through secret negotiations and the red herring of having the auctions occur "in public." The use of the stock market to organize sales helped rebut claims by the opposition to privatization, particularly the Workers' Party, that the process lacked "transparency." The PND's supporters, although ignoring the selective ways that privatization initially empowered a particular group of buyers, employed references to the stock market sales as "evidence" that all was being done above the table (Gazeta Mercantil, 1991d).

Despite these spins on the PND, the opposition to privatization was not entirely fooled. The auctions of state firms continued to be scenes for often violent demonstrations including the rare bombing by leftist unions and political parties. The PT and the populist governor of Rio de Janeiro, Leonel Brizola, never stopped their attacks on privatization as an illicit and corrupt attempt to create private monopolies.⁽²⁹⁾ Brazilian unions, particularly those led by the Central Única dos Trabalhadores (Workers' Singular Peak Association - CUT) and the Central Geral dos Trabalhadores (General Workers' Central - CGT) blasted privatization as the "giving away of the public patrimony" (Neto, 1993).

In response, BNDES leaders attempted to undercut the opposition by adding workers as allies to the privatization policy coalition. In early 1992, Eduardo Modiano proposed the use of $1.5 billion owed by the federal treasury to a forced savings scheme for severance pay called the Guarantee Fund for Time Worked (Fundo de Garantia por Tempo de Serviço - FGTS) in the first wave privatizations.⁽³⁰⁾ Modiano's proposal called for the government's debt to the FGTS to be converted, at the discretion of workers, to certificates of privatization. The federal government agreed not to tax this money for two years and the BNDES arranged subsidized finance for additional purchases of stock by workers using FGTS monies. The proposal sought to inject more than $20 billion of buying power into the market for public firms.

The unions, which controlled the council that oversaw the FGTS system, strongly opposed the bank's initial proposal for absolute freedom of investment. The BNDES' second proposal to allow workers discretion with 30 percent of their FGTS monies was also struck down. The unions argued that any use of FGTS funds should give workers a voice in firm decision-making. BNDES officials vehemently rejected the union's approach on the grounds that it would politicize the financing of newly privatized firms. Faced with another stalemate, the bank's officials went on the offensive. They marketed earlier proposals for the use of FGTS monies by employing the idea of "popular capitalism," an idea first coined by Margaret Thatcher during the first wave of privatizations in Great Britain. As in the United Kingdom, the intent was to fragment union opposition to privatization. BNDES officials proposed to have FGTS monies converted into moedas sociais ("social money"). They argued that such an infusion of "social money" would "democratize" the privatization process. Workers would be able to convert their FGTS accounts held at banks to CP's at a 70 percent discount. As a further incentive, BNDES would provide 10-year loans with subsidized interest on additional purchases of CP's.

The moedas sociais idea presented BNDES leaders with a powerful political weapon against the union and political opponents to privatization. Development bank leaders argued that the investment by workers in newly privatized firm stock would perform well above the 3 percent annual output of FGTS accounts. This was enough to convince many workers to part from their union's resistance to the use of FGTS funds. In the steel sector, it created a larger wedge between the rank-and-file of the metallurgical unions and their confederations, the CUT and Força Sindical, a rival confederation to the CUT. Local union leaders, along with their members, systematically defected from their confederations' opposition to privatization. Even before FGTS funds were infused into the process, workers at Usiminas had split from their confederation's opposition to the sale and formed a club of investors which eventually gained 10 percent of the shares of the steel mill.⁽³¹⁾ By June 1995, more than 105,000 workers in the public sector had purchased shares with moedas sociais averaging 10 percent of total shares for each public firm set for auction. The financial gains for workers who participated were remarkable. Labor purchased $290 million in shares that, by 1995, were worth more than $1 billion (Veja, 1995: 60)!

At the same time, the infusion of FGTS money into future privatizations would expand the range of "privatizable" firms. BNDES officials correctly hypothesized that the most likely buyers for public firms in the worst financial shape would be their own workers.⁽³²⁾ This was clearly the case with Cobra, a troubled public computer firm, and Embraer, once the chief public aeronautics firm. Labor (cum capital) from previous privatizations added to this growing constituency for sustaining the policy of sell-offs. Indeed, auctions in the steel sector continued to confirm the importance of the buying power of labor. The Club of Usiminas Investors, a group composed of the 17,000 Usiminas workers who owned ordinary and preferential shares in the steel giant, became one of the key players in the privatization of Acesita.

Again, the use of the FGTS and the expected gains for workers served to divide union opposition. During the privatization of Embraer, for example, the union's rank-and-file threatened to abandon the union when their leaders opposed the sale of the aeronautical firm in 1994. During the Acesita privatization, local leaders of the metallurgical workers union championed the cause of the Club of Usiminas Investors to the chagrin of the national union opposition to privatization of the steel companies.

The piecing together of the "new triple alliance" for privatization among the state, private industrial-financial groups and the privileged "labor aristocracy," lacked a member of the "old triple alliance": foreign capital. To avoid kindling the opposition to privatization, state managers who designed the rules governing the use of DFA's, CP's, FGTS funds and other "privatization monies," placed restrictions on when investors could sell their shares. Foreign investors were dissuaded from participating in the Usiminas sale and other auctions in 1991 due to the rule that investments would have to remain in Brazil for 12 years. Domestic investors could not sell their shares on the Brazilian stock exchanges for two years. By early 1992 Central Bank and BNDES officials recognized that such rules, although important for refuting opposition claims that the public patrimony would be sold and moved abroad, also reduced the interest of the private market for ownership of public firms.⁽³³⁾ Ironically, such limitations would only be reversed under the administration of Itamar Franco, who at least in principle, opposed the acceleration of privatization. Meanwhile, the negative effects from the absence of a noteworthy participation by foreign capital in the first wave of sales were attenuated by policy mechanisms such as the moedas podres that financed the domestic pool of buyers.

The policy constituency of business, finance, and labor created by the various instruments of the PND sought to achieve nothing less than the implementation of a new industrial policy. After Collor's ignominious exit from the presidency and the emergence of his vice-president, Itamar Franco, the survival of the PND would become even more dependent upon the support of these allies and the capacity of program administrators to shift their authority to like-minded ministries and politicians.

V. TESTING THE PRIVATIZATION COALITION

As a result of a corruption scandal and his imminent impeachment, Fernando Collor was forced from the presidency in late 1992. Collor's dramatic exit from power created problems for the PND. Itamar Franco, Collor's successor and vice-president, had none of Collor's enthusiasm for the program. As a long-time defender of public steel and mining companies in his native state of Minas Gerais, Itamar Franco was suspicious of privatization. For example, he had opposed the privatization of Usiminas as "an act that would weaken the state." Franco's unlikely rise to the presidency of Brazil thus empowered privatization opponents and other like-minded forces both within and outside the Franco cabinet. Chief among the insiders were Paulino Cícero, Franco's new Minister of Mines and Energy and a former president of Usiminas; Maurício Correa, Minister of Justice and a fervent opponent of the PND in congress; Junia Marise, senator from Minas Gerais and author of an executive decree designed to stop the Usiminas sale; and Pedro Simon, senator from the state of Rio Grande do Sul and one of the notable leaders of congressional opposition to the steel mill sell-offs.

Collor's exit and Itamar Franco's rise also marked important changes at the highest levels of the PND. Modiano left his post as head of the PND Directorate. Franco installed Antônio Barros de Castro, a gifted economist who was sympathetic to Brazil's history of state-led growth, as president of the BNDES. Castro's vice-president, Marcos Vianna, took over the PND. Vianna, like Castro, was suspicious of privatization and for good reason. Vianna was a past president of the BNDES and a key architect of the second National Development Plan (ironically, an acronym that, in Portuguese, is "PND") during the military governments of the 1970's.Both Castro and Vianna talked vaguely about a "new industrial policy" for Brazil, but their initial concerns focused on "rethinking" privatization policy (Castro, 1995).

The new policy discourse focused on the moedas podres. True to his developmentalist background, Itamar Franco and Castro questioned the exchange of equity for debt that left nothing for "public investments" and social policies. Moreover, Cícero and other privatization opponents argued that the PND's lack of transparency disguised the wholesale subsidization of buyers that many in the policy opposition suspected was built into the moedas podres purchasing modality (O Globo, 1992). Critics argued that privileged industrial and financial groups were conniving to create a de facto oligopoly in the steel sector. Not surprisingly, the Bozano Simonsen group was singled out for castigation.⁽³⁴⁾

The Itamar Franco administration made it more difficult for the BNDES to sell-off the public firms according to the schedule first established by Collor. The new president's strategy was a dual one: he would decrease the attractiveness of the public firms by making them more costly to purchase and he would increase the political influence of privatization opponents. On December 23, 1992, Itamar Franco signed a decree reforming the PND in ways that satisfied these goals. First, the reform required that a large portion of any bid on a public firm be composed of non-debt paper, "live" money. Moedas podres would no longer be accepted as the chief financial mechanism in selling public firms. The new policy required that an amount of "live" money equal to the value of firm assets and profits be introduced by any prospective buyer. Without access to moedas podres, potential buyers were left without the subsidized mechanisms that had once made privatization financially possible and very attractive for domestic industrial-financial groups. The December reform also forced firms placed on the auction block to expend millions more on restructuring to meet more stringent environmental standards. This provision placed additional financial burdens on the public firms, thus reducing their attractiveness to buyers.

The December reform empowered societal opponents of privatization with more information. The Franco government made the technical and financial data of the firm viability reports available to the public. Previously, these were produced by private consulting firms before public auctions and were for the eyes of a select group of financial, industrial, and BNDES elites only. In this way, the phasing out of the moedas podres provision was made a greater deterrent to potential buyers as financial and technical figures from the economic viability studies became more widely available. Prospective buyers who normally prized the secretiveness surrounding their everyday financial deals were faced with the prospect of having their actions subject to greater public scrutiny. Privatization opponents who were previously without access to such confidential yet politically explosive information about the ingredients of auction pricing, were now armed with details to feed their claims that the public patrimony was being sold on the cheap.

The political gain for the administration was clear. The reform played to the president's pseudo-populist message (perhaps a page taken from Collor's own rhetoric about the "Maharajahs"). By injecting transparency into privatization policymaking, Franco was seen as pulling the curtain on the wizards of state and financial interests that were selling the patrimony of the nation undemocratically, without the say of the people. As José de Castro Ferreira, Franco's Counsel General, made clear in defense of the December reform:

[Information about these public firms] should not be protected as if they were state secrets as they have until now. In reality, a privatization auction is a commercial act to which not just some supposed buyers are privy. The people must also oversee the process (Gazeta Mercantil, 1992b).

Although the ambivalence of the Franco government toward privatization succeeded in slowing the process, the new president did not attempt to stop privatization altogether. In part, this was due to the continued activities of proponents of privatization in the BNDES, the state firms, private investors, and key members of Itamar Franco's cabinet.

Paulo Haddad, the new Minister of Planning, was, like Vianna, a member of the developmentalist old guard. Given Brazil's intense fiscal crisis, however, Haddad doubted whether the country could return to its developmentalist past. Instead, the planning minister embraced privatization as a viable mechanism of industrial policy and he viewed the management of privatization policy as being in line with the BNDES' classic role as a "development bank" (Gazeta Mercantil, 1992a). Similarly, the other major economic ministry during the Itamar Franco administration came to be commanded by individuals who did not embrace neoliberal ideas but agreed with BNDES officials who saw privatization as a means of improving the financial stability of both the state and privatized firms. Eliseu Resende, Minister of Finance during the early part of 1993, supported the unlimited inclusion of foreign capital in the privatization process. His successor in May 1993, Fernando Henrique Cardoso, a sociologist and senator from São Paulo, announced his intention to accelerate the privatization of the steel and energy sector during mid 1993.

Like Haddad and Resende, Cardoso was no neoliberal. His arguments reflected those of BNDES managers. For example, Cardoso embraced the use of FGTS monies in the privatization auctions. By September 1993, the Ministry of Economy, the head of the PND, André Franco Montoro Filho, and the new BNDES president, Pérsio Arida,⁽³⁵⁾ agreed on a plan that would inject $40 billion from FGTS and workers' pension funds into the privatization auctions and expand the scope of the PND to strategic, and more politically sensitive, sectors such as energy, infrastructure, and mining. In October 1993, Cardoso spearheaded the preparation of a presidential decree. Decree 362 would require that all "live" money generated by future privatizations be used to reduce public debt. The decree also allowed foreigners to buy up to 100 percent of shares of public companies up for sale.

BNDES' leadership continued to have much control over the forward thrust of privatization. Despite political pressures from above and below to alter the schedule for privatization first established by the Collor government, BNDES officials were successful in keeping the schedule of firms roughly on track. The process was slowed, but not significantly altered or derailed. As part of the October 1993 reform, the Ministry of Economy took over the administration of privatization but preserved the PND and the BNDES as the implementors of policy. The move was a calculated one. Arida and other BNDES leaders sought to inject additional capital through swaps of National Treasury debt into the privatization process. Since the development bank lacked authority over the Treasury, Arida advocated the shifting of control over the policymaking functions of the PND to the Ministry of Economy.⁽³⁶⁾

The BNDES' strategic support of the Ministry of Economy's new command over privatization only made sense in the context of newfound political support for new moedas podres. Despite the December 1992 reform's intentions to reduce the use of moedas podres, several additional provisions in 1993 only expanded their availability. In June 1993, Itamar Franco, following prompting from BNDES and Ministry of Economy leaders, sanctioned the use of National Treasury Notes (Notas do Tesouro Nacional - NTN), public debt held by private contractors, as "privatization money." The purpose of the NTNs was to finance the purchases of state firms by private contractors, many of whom had already jumped into the first wave of sales.

With the creation of NTNs, major groups such as Odebrecht, Bozano Simonsen, Andrade Gutierrez, PIC, PPE and Safra expanded their interests in the privatization process. Labor groups who already had a stake in privatized firms, campaigned for future privatizations which guaranteed substantial earnings for pension funds and FGTS accounts. By the end of 1993, the reforms initiated by the Ministry of Economy and implemented by the BNDES and the PND Directorate, including the use of NTNs, FGTS and other monies, and the inclusion of larger amounts of foreign capital in privatization, substantially expanded the policy constituency for public auctions of state firms.

Even reforms that intended to reduce the use of moedas podres had salutary effects on the overall course of privatization. The emphasis placed on "live" money did not halt the PND. On the contrary, new privatizations during the Franco government generated $2 billion, 72 percent of which was now in liquid currency. The use of live money efficiently funded government without the deságio that subsidized firm buyers and penalized public accounts. Additionally, the removal of many restrictions on the factors determining firm minimum pricing allowed BNDES managers to employ "book building," the practice of floating the price when demand was high to take advantage of marginal increases in value on the eve of auctions. For Cosipa, a public steel company based in São Paulo, book building in the internal market generated $200 million in live money during the firm's purchase, an amount more than 10 percent above the original value first predicted by BNDES officials.

Of course, book building was a more viable strategy now that greater international confidence had been established in the value of privatized steel firms. The good performance of Usiminas after privatization appreciated the value of the minority shares the state continued to hold in the enterprise. The sale in foreign stock markets of 16 percent of the state's minority shares in Usiminas generated $480 million in 1994. Usiminas' good performance also appreciated the value of other steel firms on the auction block.

Thus by the beginning of 1994, the privatization program in the steel sector had been firmly established with a series of ongoing policy mechanisms that linked the interests of state managers in the Ministry of Economy, the BNDES, and the PND with an array of powerful financial and industrial groups and labor interests. Despite significant political challenges to this policy constituency, the embedded autonomy of Brazilian privatization guaranteed the continuation of the process.

VII. CONCLUSIONS

Privatization policy in Brazil was not governed by a weak state forced to sell-off decaying public industries due to fiscal and political pressures. Rather, the process was part of a larger, very active restructuring of the public steel firms before, during, and after their sale. Far from ceding to private interests, public policymakers actively molded the financial capacities of potential buyers and facilitated purchases and further productive restructuring.

The modalities of Brazilian privatization forged a supporting constituency for reform within the state apparatus and between state agencies and societal actors. The PND's managers established ties with like-minded ministries and politicians that preserved the autonomy of privatization policy but also made the pool of political support broader and more flexible to changes in government. The mechanisms of the PND extended the resources of the state among networks of financial, industrial, and labor elites who provided crucial material resources and political support for the program. This embedded autonomous policy structure proved sufficiently resistant to pressures from privatization opponents - some from below (the national labor confederations and opposition leaders such as Brizola and Lula) and some with a significant potential for derailing the process from above (the presidency of Franco and key politicians in opposition to the public auctions). The result was a structural reform process that proved credible for the business and labor members of the privatization constituency.

Brazil's experience with steel privatization demonstrates that embedded autonomy is an effective means for solving credibility gap problems in structural reforms. Through financial incentives, firms and workers received positive signals as well as capacitating supports that convinced them that participation in privatization would bring future rewards. By investing heavily in the restructuring of debt and production at the steel mills, the PND's managers also developed sunk costs, commitments indicating that the government would follow through on efforts to improve the performance of the steel firms even after their sale. The PND followed a consistent strategy of industrial reform both ex-ante and ex-post and, as a result, it was met with an expanding pool of buyers as the program matured.

By employing the embedded autonomy concept, this study reinforces Evans' (1992) notion that the study of economic policy must focus on the state apparatus and its relations with society - ideas central to the "third wave of thinking about the state." The practice of embedded autonomy in the case of the privatization of Brazilian steel demonstrates that structural reform requires an association between state and economic interests rather than insulated ("strong") states. At the same time, however, the case shows that the state is not an empty vessel that is filled with the interests of privileged societal actors. Rather, state interests direct the reform process in ways that employ private sector resources and political support. Embedded autonomy in this case also illustrates the simplistic fallacies of popular perceptions of structural reform as the result of economic necessities and fiscal exigencies. Politics shapes the content and credibility of reforms, but understanding what kind of politics matter is a task for those who are willing to rethink the state and its relations with society.

1. The role of public steel in Brazil's development is discussed in a number of important studies including Evans (1979), Baer (1965) and (1969), Schneider (1991), and Shapiro (1994).

2. The seven profitable steel firms in 1994 were Companhia Siderurgica Nacional (CSN), Usiminas, Companhia Siderurgica de Tubarão (CST), Acesita, Açominas and Piratini. Cosipa, the only steel firm showing a loss at the end of 1994 was also the last to be privatized (Exame, 1994).

3. Among the most important studies of the BNDES which analyze the interests, structure, and efficiency of the bank during its developmentalist period are Willis (1986) and (1995) and Martins (1985).

4. My study differs from Evans' work in two fundamental ways. First, I focus on structural adjustment, not the developmentalist period of industrial transformation that Evans analyzes. Doing so, however, verifies Evans' (1992: 142) argument that "an understanding of industrial transformation may not be irrelevant to the eventual construction of an analysis of the state's role in successful adjustment." Second, the study applies embedded autonomy as a proximate cause for economic policy outcomes, not an intervening or dependent variable as it sometimes is in Evans' work.

5. Schneider (1997: 200-201) makes a similar argument regarding the credibility of reform, except his focus is on "concertation" - the routinized negotiation between state managers and business of the details of policy implementation. I prefer the term "embedded autonomy" because of the centrality it affords state interests. In the case of Brazilian steel privatization, state interests were dominant. The goals and purchasing modalities of privatization policy were not negotiated with business, but private sector resources and political support proved crucial to maintaining the PND. These were employed according to a strategy devised by state managers. The centrality of state interests is apparent in Schneider's own Mexican and Brazilian cases, as he points out, "concertation collapsed quickly...especially when the state withdrew from it" (p. 210).

6. Geddes (1995, 204-206) finds that even when reforms impose heavy costs on society, incumbents tend to stay in office.

7. For a theoretical treatment of "signaling" and the "credibility gap" in economic reform processes, see Rodrik (1989).

8. Unlike Menem and Fujimori, but like Salinas de Gortari, Collor did not hide his advocacy of neoliberal reform before being elected president. Therefore, Collor was not a "stealth" neoliberal candidate. I thank an anonymous reviewer for this point.

9. The role of a neoliberal technocracy in Mexico explains, at least in part, why that country shifted to market-oriented structural reform policies several years before Brazil. For more on the Mexican technocracy and neoliberal reform, see Centeno (1994).

10. The ideas undergirding the formation of the Grupos Executivos da Política Setorial (Executive Groups of Sectoral Policy - GEPS) as part of the PCI envisioned a return to the executive councils employed by Kubitschek's Plano de Metas for the auto industry and heavy manufacturing during the 1950's (Oliveira, 1994). For the official view of the PCI, see the column by João da Silva Maia in Visão (1990). These ideas were also applied to new industrial policies designed to promote science and technology. See Comissão Especial (1990).

11. The indebtedness of the public steel firms was caused by a combination of using state enterprises in anti-inflation efforts and as mechanisms for bringing in foreign exchange. The latter came in the form of debt which financed chronic balance of payments deficits. See Werneck (1991) and Baer and Villela (1994).

12. Luiz Chrysostomo de Oliveira Filho, one of the original architects of the PND, argued that Brazil's public firms suffered from "over-bureaucratization" and poor productivity. He compared the performance of private and public firms to justify his position that privatization would raise the performance of the public firms to private sector standards. Elena Landau argued that the state could not make the financial commitment it would take to consolidate the restructuring of steel (O Globo, 1994). These views were shared by the management of the steel firms. See the testimony of the president of Usiminas, Rinaldo Campos Soares, in Gazeta Mercantil (1991e), the view of the former Secretary of Planning of Companhia Siderúrgica Nacional (National Steel Company - CSN), Fernando da Silveira Cotrim, in Gazeta Mercantil (1990b), and the testimony of former head of CSN, Roberto Procópio de Lima Netto (1993: 187-188).

13. See the comments of IBS leaders in Gazeta Mercantil (1993).

14. The steel firms represented in the Siderbrás plan in 1990 were Usiminas, Açominas, Companhia Siderúrgica de Tubarão (CST), Companhia Siderúrgica Nacional (CSN) and Cosipa. Moacélio Mendes was the last Siderbrás president of the Sarney administration.

15. See comments by Moacélio Mendes in Gazeta Mercantil (1990a).

16. One notable example of Zélia's policing role within the government was her semi-public feud early on in the privatization process with Evilásio Soriano, the last public executive of Petroquisa, a large state petrochemical interest. Soriano spoke for several government leaders who clung tenaciously to the idea that Brazil's industrial development still required public firms. For more on these events, see Estado de São Paulo (1991).

17. For a discussion of patrimonialism and Brazilian politics, see Hagopian (1996).

18. See comments by Marcílio Marques Moreira, Collor's second Minister of Economy, and Eduardo Modiano in Gazeta Mercantil (1991a).

19. For an historical analysis of Acesita, see Veja (1989).

20. Between July 1991 and January 1992, Acesita's stock price had risen 572% nominally, 272% over the average São Paulo stock market rate for the same period (Gazeta Mercantil, 1992c).

21. Eduardo Modiano, for example, argued before the sale of Usiminas, that privatization was only the most politically sensitive portion of a firm's several years of productive restructuring. See the interview with Modiano in Correio Braziliense (1991b).

22. This position was consistently articulated in public by Modiano himself. As an example, see the PND chief's comments in Gazeta Mercantil (1991d).

23. The figure used here is an approximation. The exact amount of "live money" during the Collor privatization was much disputed subsequently among ex-Collor and Itamar Franco administration technocrats. Nevertheless, the consensual view is that the total amount of "live" money used during the first stage of the Collor administration's PND was extremely limited. For a perspective from post-Collor managers of the PND, see the interview with Elena Landau in Folha de São Paulo (1994) and Estado de São Paulo (1994). For an assessment of the use of debt paper and liquidity during the Collor privatizations, see Baer & Villela (1994).

24. In the case of DFAs, the Central Bank had control over these funds while the CPs were administered by commercial and investment banks and fund management companies.

25. For a complete explanation on how state managers set minimum price and then calculated the deságio, see Prado (1993: 106-107, fn 18).

26. Among the Latin American cases, only in Argentina did privatization have an appreciable effect on public accounts. See Pinheiro & Schneider (1995) and Manzetti (1993b).

27. In February 1991, Eduardo Modiano gave a talk before the U.S. Chamber of Commerce in Brazil and was greeted by the publication of a poll which showed that the majority of the organization's members (54%), believed that the privatization program would stall due to bureaucratic intransigence. Modiano's primary response was to emphasize that the PND's rich array of financial mechanisms would stimulate substantial business demand for the state firms (Gazeta Mercantil, 1991d).

28. Modiano admitted as much in hindsight: "We never accepted these notes as part of some dogma. It was always a practical question created by the circumstances and by the market that we had to face" (Modiano, 1992).

29. Due to ongoing and deeply held mutual suspicions and political rivalries, Brizola and Lula failed to "team up" publicly on the anti-privatization campaign. Each assumed the stewardship of their own particular movement against the PND. Later, Brizola gave his qualified support for the privatization of CSN, a decision that raised the ire of the PT (Netto, 1993). For the PT's view of privatization, see the interview with Lula in Jornal do Brasil (1991a).

30. The federal treasury owed this amount to the Fundo de Compensação de Variações Salariais (Compensation Fund for Salary Variations - FCVS), a fund linked to FGTS.

31. Prior to the actual sale of Usiminas, polls showed that 45 percent of the steel mill's workers supported privatization and 35 percent opposed the sale (Folha de São Paulo, 1991).

32. See the interview with Elena Landau in Folha de São Paulo (1994).

33. See comments by Armínio Fraga, Director of International Affairs of the Central Bank, in Estado de São Paulo (1992b).

34. For an exposition of the accusations lodged against the Bozano Simonsen group during this time, see Cícero's comments in Jornal da Tarde (1992).

35. Adding to the emergence of pro-privatization forces within the cabinet, several members of the anti-privatization camp did not get much of a chance to oppose privatization. Castro was president of the BNDES for only a couple of months when he was replaced in January 1993 due to a disagreement with Counsel General Ferreira over minimum pricing of a fertilizer firm.

36. See the rationale as depicted by Arida himself in O Globo (1993).