Decentralization as a Process

Decentralization is a process of state reform composed by a set of public policies that transfer responsibilities, resources, or authority from higher to lower levels of government in the context of a specific type of state. Compared to previous definitions, this one poses four important restrictions. First, decentralization is conceived as a process of public policy reforms and not as a description of the state of being of the political or fiscal systems at a point in time. Second, lower levels of government are the recipients of the transferred responsibilities, resources, or authority. Reforms such as privatization or deregulation, which target nonstate actors, are not included in this definition (cf. Cheema and Rondinelli 1983, 24–5). Third, because decentralization is a process of state reform, a transition to a different type of state necessarily implies the commencement of a new decentralization sequence. The contents of decentralization policies and their interaction with the broader political and economic systems are highly determined by the type of state they seek to reform. Hence, in order to compare decentralization policies across countries as part of analytically equivalent processes, we must compare policies taking place within the same type of state. Finally, in studying the downward reallocation of authority, much is gained from a clear taxonomy of decentralization based on the type of authority devolved, such that three types of decentralization can be distinguished: ⁴

• Administrative decentralization comprises the set of policies that transfer the administration and delivery of social services such as education, health, social welfare, or housing to subnational governments. Administrative decentralization may entail the devolution of decision-making authority over these policies, but this is not a necessary condition. If revenues are transferred from the center to meet the costs of the administration and delivery of social services, administrative decentralization is funded and coincides with a fiscal decentralization measure. If subnational governments bear these costs with their own preexisting revenues, administrative decentralization is not funded.
• Fiscal decentralization refers to the set of policies designed to increase the revenues or fiscal autonomy of subnational governments. Fiscal decentralization policies can assume different institutional forms such as an increase of transfers from the central government, the creation of new subnational taxes, or the delegation of tax authority that was previously national.⁵

  Unlike other definitions of fiscal decentralization that collapse decentralization of revenues and expenditures, in this definition fiscal decentralization refers to revenues, whereas expenditures are part of administrative decentralization. This analytic separation makes it easier to evaluate the consequences of decentralization processes where the transfer of revenues and expenditures do not go hand in hand, allowing the disentanglement of seemingly contradictory outcomes such as “centralization via decentralization” (see Wibbels 2004, 220–21)

  .
• Political decentralization is the set of constitutional amendments and electoral reforms designed to open new—or activate existing but dormant or ineffective—spaces for the representation of subnational polities. Political decentralization policies are designed to devolve political authority or electoral capacities to subnational actors. Examples of this type of reforms are the popular election of mayors and governors who in previous constitutional periods were appointed, the creation of subnational legislative assemblies, or constitutional reforms that strengthen the political autonomy of subnational governments.⁶

  Although political decentralization and democratization can be mutually reinforcing, the two processes need to be distinguished analytically. For example, the return to free and fair elections at all levels of government after an authoritarian regime does not necessarily constitute a political decentralization policy. The transition to democracy may simply be reinstating the electoral norms and rules of the pre-authoritarian period, with no negotiation of a policy reform that specifically targets the subnational level. Similarly, if an electoral reform that is designed to augment political competition in the political system as a whole were to have the unintended consequence of increasing the power of subnational political actors, it cannot be considered a political decentralization measure because it was not planned, designed, or negotiated with the explicit goal of empowering subnational polities. To qualify as political decentralization, the reform in question must explicitly address the devolution of political authority or capacities to subnational polities

  .

Regarding the consequences of each type of decentralization, I expect administrative decentralization to have either a positive or a negative impact on the autonomy of subnational executives. If administrative decentralization improves local and state bureaucracies, fosters training of local officials, or facilitates learning through the practice of delivering new responsibilities, it will increase the organizational capacities of subnational governments. Nevertheless, if administrative decentralization takes place without the transfer of funds, this reform may decrease the autonomy of subnational officials, who will be more dependent on subsequent national fiscal transfers or subnational debt for the delivery of public social services. Similarly, fiscal decentralization can have either a positive or a negative impact on the degree of autonomy of the subnational level. The result will depend largely on the design of the fiscal decentralization policy implemented. Higher levels of automatic transfers increase the autonomy of subnational officials because they benefit from higher levels of resources without being responsible for the costs (political and bureaucratic) of collecting those revenues. On the contrary, the delegation of taxing authority to subnational units that lack the administrative capacity to collect new taxes can set serious constraints on the local budgets and increase the dependence of the local officials on the transfers from the center. Prosperous subnational units prefer to collect their own taxes, but poor states or municipalities are negatively affected every time the collection of taxes is decentralized and, as a consequence, the horizontal redistribution of transfers from rich to poor subnational units is affected. Finally, political decentralization, by the definition provided previously, should almost invariably increase the degree of autonomy of subnational officials from the center. The only case when political decentralization could have a negative effect on the power of governors and mayors vis-à-vis higher level authorities is when, by augmenting the separation of powers at the subnational level (such as through the creation of subnational legislatures or municipal councils), it leads to divided subnational governments. In such instances, the subnational political opposition could undermine the authority of governors and mayors vis-à-vis the national executive.

By unpacking decentralization in this way, we see that, depending on their institutional design, decentralization policies can actually decrease the power of subnational officials with regard to the national executive. As shown in the following, the institutional design of decentralization policies is highly dependent on when those policies take place within the sequence of reforms. Political and fiscal decentralization policies that take place early in the sequence tend to increase the power of governors and mayors, whereas early administrative decentralization reforms tend to negatively affect their power.

Territorial Interests of Bargaining Actors

The territorial interests of presidents, governors, and mayors are defined by the level of government (national, state, or municipal) and the characteristics of the territorial unit (e.g., rich or poor province, big city, or small town) they represent. Drawing from the literature on decentralization and in-depth interviews with national and subnational politicians and public officials,⁷

I describe the set of preferences of national and subnational actors with regard to decentralization types.⁸

The national executive prefers administrative decentralization (A) to fiscal decentralization (F), which in turn is preferred to political decentralization (P), or A>F>P. The rationale of this ordering is that the national government seeks to divest itself of expenditure responsibilities first and foremost. Administrative decentralization is greatly preferred over the other two types of decentralization. As Garman, Haggard, and Willis (2001) say: “[W]e would expect the president to be more inclined to transfer responsibilities than the resources to meet them” (209). If the center is forced to choose between surrendering fiscal and political authority, it will choose to give away fiscal authority and to retain political control, which may serve to influence the expenditure decisions made by subnational officials.

The same reasoning applies to explain the reverse order of preferences of the subnational governments: P>F>A. Their preference, first and foremost, is political decentralization. If the president does not control the appointment and removal of governors and mayors, they can push forward the issues and concerns of their territorial units without fear of retaliation from above. If governors and mayors have to choose between fiscal and administrative decentralization, they will choose the transfer of revenues over responsibilities, particularly if the unions representing the public sectors to be decentralized are large and strong. That is, subnational executives prefer political autonomy, money, and responsibilities, in that order.

Sequences of Decentralization: Origins, Timing, and Mechanisms

The origins of the process of decentralization are important both theoretically and methodologically. On the one hand, the main argument of this article incorporates elements of path dependence, for which the issue of origins is crucial. On the other hand, the method of process tracing requires specifying when the process starts. Scholars have adopted different approaches to answer the question of when a path-dependent process starts, such as critical junctures (Collier and Collier 1991) or contingent events (Mahoney 2000). I define the origin of the decentralization process by the state context in which it takes place. As stated earlier, the contents of decentralization policies and their interaction with the broader political and economic systems are largely determined by the type of state they seek to reform. In Latin America, for example, in the context of the oligarchic states, decentralization reforms sought to consolidate or balance power among regional elites (Ansaldi 1992, 17). In the context of the developmental states, meanwhile, decentralization policies sought to strengthen certain regions to make them more adequate for private investment (González 1990, 75); whereas in the context of market-oriented states, decentralization policies largely sought to reduce the size of central governments. Of course, these were not the exclusive goals of decentralization reforms in each of these periods. Nonetheless, it is evident that in different historical periods the policies that transferred responsibilities, resources, or authority to subnational governments were part of state reform projects that had largely different overarching political and economic objectives. For this reason, when comparing across countries, the researcher should qualify the processes or sequences of decentralization by the type of state in which they take place, in order to assure the analytic equivalence of the compared policies.

For the purposes of this article, I focus on the process of decentralization that began with the transition from a “developmental” to a “public-goods” type of state (Block 1994). In Latin America, this was the transition from a desarrollista to a market-oriented type of state. During this period, decentralization policies were part of what became known as “second-generation” reforms (Camdessus 1999). Prior examples of centralization and decentralization existed in the region (see Eaton 2001, 2004; Montero and Samuels 2004, 14) and constitute the background against which the policies analyzed in this article took place. However, because they occurred in different types of state contexts, they form part of prior sequences of intergovernmental reforms.

Regarding the origin of the sequence analyzed here, although it is difficult to pin down the exact date when the desarrollista state ended in each case (Schneider 1999, 293), it is nonetheless possible to identify the first administration in each country that applied orthodox measures of economic adjustment and moved the state away from intervention in the economy. The market-oriented sequence of decentralization reforms starts with the first decentralization policy successfully implemented by the first administration that made the transition from a developmental economy and state toward a market-oriented economy and state.⁹

Failed decentralization attempts do not constitute original moments because they do not have an impact on intergovernmental relations. They are analyzed as part of the process, as they may reflect on the distribution of power among bargaining actors, but do not constitute key transformative or original moments.

Using Skowronek's (1993, 9) terminology, we may conceive of intergovernmental relations as a layered structure of institutional action. Decentralization policies affect the fiscal, administrative, and political layers of intergovernmental relations. Rarely does a decentralization policy simultaneously affect all three intergovernmental layers (although it is possible). More often, different types of decentralization (as well as different policies within each type of decentralization) are negotiated and enacted at different points in time. Hence, the timing of each reform determines the particular sequence of decentralization that a given country undergoes. If the three types of decentralization defined previously take place (which is not theoretically necessary, but is a common occurrence), we can identify six sequences of decentralization according to the timing of the first decentralization policy within each intergovernmental layer. This does not mean that posterior decentralization policies in each layer do not happen or should be overlooked (see the analysis of empirical cases in the following section). However, the sequencing of the first decentralization policy in each layer is particularly important because it sets constraints on what is feasible in the remainder of the sequence and allows us to establish a basic model of the impact of different sequences of decentralization reforms on the intergovernmental balance of power.

The level of government whose territorial interests prevail at the origin of the decentralization process is likely to dictate the first type of decentralization. The first round of decentralization, in turn, produces policy feedback effects that account for the order and characteristics of the reforms that follow. If subnational interests prevail in the first round of negotiations, political decentralization is likely to happen first. Political decentralization is likely to produce a policy ratchet effect (Huber and Stephens 2001, 10): a group of supporters who will continue to push in the direction of further decentralization. The formation of associations of governors, mayors, or similar instances of coordination of subnational politicians is an example of such policy ratchet effect. Lobbying through these associations, governors and mayors will enhance their power and capacities for the next rounds of decentralization. Even if this coordination mechanism is not in place, governors and mayors will find themselves in a better position to advance their preferences in the second round of reforms because they will enjoy greater political autonomy from the national executive. The president, moreover, may become more dependent on elected governors and mayors for the mobilization of votes in national elections. Thus, in the second round of decentralization, governors and mayors will most likely demand fiscal decentralization and influence its terms. Administrative decentralization, which after fiscal decentralization is likely to follow to compensate for the previous decentralization of resources (Haggard 1998, 217), will be the last type of reform. Administrative decentralization will therefore be funded and will not have a negative impact on the power of governors and mayors. The final outcome of this trajectory of decentralization (P→F→A) that conforms to the preferences of the subnational officials is likely to be a high degree of autonomy for governors and mayors with respect to the president (see Table 1). I show below that Colombia followed this path from 1986 to 1994.

If, instead, national interests prevail at the beginning of the process, administrative decentralization is likely to occur first. If fiscal resources do not accompany the transfer of responsibilities, the national executive will strengthen its power vis-à-vis subnational officials, who will become more dependent on transfers from the center. If the process of decentralization continues, the president will choose fiscal over political decentralization. But due to a power reproduction mechanism (Stinchcombe 1968, 117–18), the national executive will control the timing, pace, and contents of the reform. Governors and mayors, under the fiscal strain of the first round of unfunded administrative decentralization, will be in no position to reject those terms set by the center—unless exogenous circumstances were to change their relative power vis-à-vis the president. Following this trajectory, political decentralization, if it happens, will be the third type of reform. The outcome of this trajectory of reforms (A→F→P) that conforms to the preferences of the national executive is likely to be little or no change in the redistribution of power to the subnational authorities. I show in the following that Argentina followed this path of reforms from 1978 to 1994.

It is also possible that exogenous changes (such as midterm elections, a context of fiscal expansion, fiscal crisis, or a process of democratization) could produce reversals on the distribution of power between national and subnational executives once the process of decentralization has started. This would lead to the alternative sequences P→A→F and A→P→F. In the first scenario, subnational interests prevail at the beginning of the sequence, triggering political decentralization. However, reactive mechanisms (such as a fiscal crisis that undermines subnational demands for fiscal decentralization) lead to a prevalence of national interests in the second round and, thus, to administrative decentralization. The last stage (should it happen) is fiscal decentralization. This trajectory may be disastrous for subnational officials if administrative decentralization is unfunded. If they are granted political autonomy, and soon after that they receive unfunded responsibilities, their subnational constituencies will blame them for poor performance. Most likely, this trajectory will lead to a low change in balance of power. If administrative decentralization is instead funded, this trajectory may lead to a medium degree of change in intergovernmental balance of power. In the second scenario, national interests prevail at the beginning of the sequence, but reactive mechanisms (such as a process of democratization that undermines centralized power) afford subnational executives the possibility of pushing political decentralization forward in the second round. In this situation, subnational actors (due to the political power they now have) are in a better position to set the terms of fiscal decentralization. The overall outcome of this trajectory would be a shift in the balance of power in favor of subnational authorities, but not as significant as in the first aforementioned trajectory (P→F→A).

Finally, we could also conceive of a tie between national and subnational interests at the outset of the reform process, such that no side is capable of achieving its most preferred outcome. In this situation, either the status quo will prevail or bargaining actors will compromise in their second most preferred outcome: fiscal decentralization. If this happens, the way in which the sequence continues will depend on the effects of this reform on the relative power of national and subnational executives. If the national executive prevails, administrative decentralization should follow, with political decentralization happening last (F→A→P). This trajectory should lead to a medium to low change in balance of power. The crucial issue here is the time lag between the first and second rounds of reforms. If subnational officials receive money without strings, and they can apply it to strengthen their support base and popularity for a considerable amount of time before they receive new responsibilities, this trajectory may lead to a medium increase in balance of power, even if political decentralization only takes place at the end of the trajectory. In contrast, if money and responsibilities are decentralized practically at the same time, this means that subnational officials are receiving new responsibilities without political autonomy. The impact of decentralization of funded responsibilities on balance of power will then be highly dependent on how successful the subnational governments are in efficiently delivering the newly transferred services. Considering, however, that the subnational officials are probably more accountable to the national executive than to their local constituencies (recall that political decentralization does not take place until the end of the trajectory), administrative performance will likely be poor, and the result will be small change in intergovernmental balance of power. Alternatively, if after a tie of territorial interests and a first round of fiscal decentralization subnational executives prevail, political decentralization should be next, with administrative decentralization taking place last (F→P→A). This sequence should lead to a high change in balance of power in favor of the subnational governments. In this type of trajectory, subnational governments gain fiscal capacities, then political autonomy; and last, they receive administrative responsibilities. The first two moves in this sequence should allow subnational authorities to build strongholds of supporters (because they have the resources to do so) and to win elections. Once this happens, they gain greater autonomy from the national executive, as illustrated in the bottom line of Table 1.

The assumptions thus far have been that the three types of decentralization take place and that a sequence among them can be established. Moreover, I have only taken into account the first successfully implemented policy within each type of decentralization and the first cycle of decentralization, which ends once the three types of reforms have occurred. Decentralization processes, however, could evolve differently in reality. Only one or two types of reforms could occur, the timing of policies could overlap, and successive reforms within each layer could affect those that follow. Some of these complexities will be revealed in the analysis of the cases in the following sections. Nonetheless, as long as at least two types of devolution of authority and two implementation moments can be identified, the proposed sequential reasoning could be modified accordingly and applied to cases and sequences that follow different patterns.

Colombia: The Subnational Path to Decentralization

In 1986, by initiative of Conservative President Belisario Betancur (1982–1986), the younger and less entrenched factions of the two traditional parties in congress (Liberal and Conservative) passed a constitutional amendment for the popular election of mayors. This law changed one hundred years of intergovernmental relations. Since 1886, the president had appointed the governors, who in turn appointed the mayors. President Betancur explained in the following terms his support for this measure:

However, the decision to popularly elect the mayors did not result solely from the president's political convictions. According to O'Neill (1999, 2003), presidents are more likely to implement political decentralization when the prospects of their parties winning future national elections are bleak, while at the same time strong pockets of support exist throughout the country that would win them elected positions at the subnational level. Although O'Neill's (2003) theory is compelling, she also notes “it would be absurd to ignore the importance of context-specific factors that affected decisions to decentralize” (1070). My contention is that when other types of decentralization are considered, those context-specific factors help to account not only for the timing of decentralization but also, and more importantly, for the type and content of the policy first implemented. In the case of Colombia, the social mobilizations against the shortcomings of the developmental state help to explain why and how decentralization came about. They reveal the presence of territorial subnational interests in the coalition that pushed decentralization forward, a presence that has been largely overlooked.

During the 1960s and 1970s, the planning and implementation of developmental policies had been transferred to parastatal institutions, relatively autonomous agencies attached to central offices and ministries. They were equipped with significant financial resources and were designed to operate in a cost-recovery basis and on a nationwide scale. These agencies supplanted the role of local government in areas such as urban planning, housing, health, education, and the provision of services such as electricity, water, and sewage. The coverage was not uniform, however. Large municipalities kept the management of more responsibilities, and peripheral, poorer regions were left largely unattended. The parastatal agencies tended to focus more heavily on the regions prone to private investment, which created profound regional inequalities (Collins 1988, 426–27; Maldonado 2000, 72). Moreover, local government expenditures had dropped from 18% of total expenditures in 1967 to 14% in 1978 and were concentrated in the largest cities. In 1979, the three largest municipalities (Bogotá, Medellín, and Cali) absorbed 72% of the total local government expenditures, even though they accounted for 26% of the population. After the rest of the departmental capitals were considered, only 13% was left to be spent in more than nine hundred remaining municipalities, where over 35% of the population lived (Collins 1988, 426; DNP and PNUD 1998, 39; Nickson 1995, 146). This created ample discontent among the inhabitants of the poorer regions.

Between 1971 and 1985, over two hundred civic strikes (paros cívicos) took place. These strikes “involved the total or partial paralysis of social and economic activity in urban centers and/or regions as a means of pressing the state to accede to demands” (Collins 1988, 425). Sixty percent of the strikes were related to problems in the delivery of electricity, water, and sewage; 9%, to problems with roads; 6%, to problems in education; and 5%, to ecological problems (Velásquez 1995, 246). The majority of these strikes occurred in midsized municipalities (with 10 to 50 thousand people) in the country's peripheral regions, particularly in the departments of the Atlantic coast (Maldonado 2000, 73). Broad sectors of the population participated in these strikes, voicing the territorial interests of the underdeveloped regions. As Jaime Castro, former mayor and member of the 1991 constitutional convention, said:

The civic strikes brought local government to the center of the political scene in several ways. First, they pointed out the deficiencies of the parastatal agencies and the local administrations in delivering public services. The national executive paid close attention to this problem. In 1980, a team of economic experts was formed to study how to improve the system of intergovernmental finances. Richard Bird led this team, whose findings and recommendations were published a year later (Misión de Finanzas Intergubernamentales 1981). When the next president, Belisario Betancur, was confronted with increasing economic problems and steadily declining municipal and departmental revenues, he passed an emergency plan that included some of these recommendations. Law 14 of 1983 sought to strengthen the collection of taxes in departments and municipalities. Departments were given a new tax on automobiles and the authority to update and simplify their existing taxes, whereas municipalities could modernize their tax bases—important for property taxes—and determine within certain parameters their own level of industry and commerce tax (Ocampo Gaviria and Perry Rubio 1983). This fiscal measure halted the trend of declining municipal and departmental revenues and, although its overall impact on the distribution of resources among levels of government was negligible (Wiesner Durán 1992, 117–29), it revealed the importance of subnational pressures. Second, the civic strikes were signs that the old system of handpicked mayors was coming to an end. Local bosses and traditional clientelist practices had proved inadequate in alleviating popular discontent. The political appointment of mayors had led to a system in which mayors were dependent on the legislator, the governor, or the president—whoever was politically responsible for their appointment—and only accountable to them. There were frequent changes of local administrations and corruption was pervasive (Gaitán Pavía and Moreno Ospina 1992, 150–51). Very often mayors were not native to the town they governed. A number of these became known as “professional mayors,” who “would travel around all the municipalities of one department until they were discredited in all of them” (Osorio, Luis Camilo, interview by author, Bogotá, July 30, 1998). Finally, the strikes showed that there were locally based citizens who were demanding accountability and better services in their municipalities. These were broad nonpartisan civic coalitions that helped to put municipal democratization on the agenda. Decentralization in Colombia was thereby initiated from below. It was fueled by the protests of the local communities. When the president proposed and the national legislators passed the political decentralization reform of 1986, they were responding to those subnational demands and interests voiced in the civic strikes.

What were the consequences of the direct election of mayors? The immediate result was a decline in the number of civic strikes. There were 51 strikes in 1987, 35 in 1988, and only 19 in 1989 (Correa Henao 1994, 48–54). Former guerrilla members were incorporated into the legal political system. In some cities and regions, the grip of traditional caciques and local bosses loosened, and competition for public office presented them with new challenges they had never had to face in the past (Angell, Lowden, and Thorp 2001; Velásquez 1995). The direct election of mayors also produced two major policy ratchet effects: (1) incrementalism in the political sphere and (2) coordination among subnational authorities.

The 1986 decentralization reform created an impulse to further develop political decentralization, and this impulse would prove difficult to reverse. At the beginning of 1991, a constitutional assembly convened in Bogotá. The assembly, in session from February to July 1991, was organized into five committees. The second committee was responsible for territorial organization. Two of the main issues discussed in this committee were the popular election of governors and the degree of autonomy to be conferred to the intermediate level of government. The assembly was split between the so-called departamentalistas, who were in favor of the popular election of governors, and the municipalistas, who opposed it. However, against a backdrop of popularly elected mayors, the election of governors came to be seen as an inevitable next reform, even by the municipalistas. As one of them said, “The popular election of governors appeared to be a complement to the popular election of mayors. It was the next step” (Castro, Jaime, interview by author, Bogotá, March 29, 2001).

Political decentralization in 1986 also fostered coordination among the beneficiaries of the reform. It created a group of followers interested in deepening decentralization. The clearest manifestation of such an effect was the creation of an association of mayors. In 1988, the first cohort of elected mayors created the Colombian Federation of Municipalities (Federación Colombiana de Municipios, or FCM). As expressed in its statutes, the mission of the association is: “[T]o represent the collective interests of the municipalities, to lead and support the development of the municipal management, and to promote the deepening of decentralization” (FCM 1991). In 1991, FCM was very active in lobbying conventionalists for the extension of the mayors' tenure from 2 to 3 years, for the recognition of municipal autonomy in the constitution, and for the transfer of more fiscal resources to municipalities (El Tiempo, Bogotá, 23 February and 23 March 1991). Despite the reluctance of the national executive, all these reforms were approved and political and fiscal decentralization were deepened as a result.

Although previous measures in the direction of transfer of revenues and expenditures to subnational governments had been taken, their impact in the distribution of resources between levels of government was negligible.¹¹

However, after the first round of political decentralization and the creation of FCM, a major fiscal decentralization reform was incorporated in the 1991 constitution. Article 357 of the new constitution established that the transfers to municipalities would increase from a level of 14% of the current national income in 1993 to 22% in 2002. This reform expanded not only the rate but also the base of the automatic transfers, which included thereafter both tax and nontax revenues. As a consequence, the total transfers to subnational governments (both departments and municipalities) passed from 38% to 52% of the current national income between 1991 and 1997 (Vargas González and Sarmiento Gómez 1997, 33).

The administrative counterpart to fiscal decentralization came about in 1993. The initial impetus to pass this reform came from the national executive, which was eager to establish a new distribution of responsibilities among levels of government as a means to cut the double spending and the deficit that fiscal decentralization had introduced in 1991. The national executive sent the administrative decentralization bill proposal to congress in mid-1992. It took 1 year from the presentation of the bill proposal until the final approval of Law 60 in August 1993. Law 60 became to be known as the “framework law” of administrative decentralization. It ruled on the distribution of responsibilities among levels of government regarding education, health, housing, and water and sewage. It was the result of compromises made by the national executive, the representatives of states and municipalities, and the national teachers' union. The national minister of education mediated between the interests of the ministry of economy and the department of national planning, who wanted to take decentralization of education to the municipal level, and those of the union, which was opposed to decentralization, particularly toward the municipal level. With the agreement of subnational representatives, the compromise reached between the union and the national government was that decentralization of education would take place toward the intermediate level of government, with funds guaranteed from the national level (Angell, Lowden, and Thorp 2001, 178). The departments thereby became responsible for paying and training teachers. They could also give vouchers to students with special needs. The municipalities were responsible for investing in the construction and maintenance of school buildings. Together, departments and municipalities were responsible for managing the educational services of preschool, primary school, secondary school, and high school. The national level retained jurisdiction over curricula and general educational guidelines, and the three levels shared responsibility for the evaluation of the educational system. Apart from the distribution of responsibilities among levels of government, the law also established the distribution of resources among the subnational units and the creation of committees (comisiones veedoras) at both the departmental and the municipal levels to ensure that the transfers were properly allocated according to the law. It also granted FCM 0.01% of the total transfers to the municipalities “for the promotion and representation of all its members…the districts and municipalities” (Article 37, Law 60). Administrative decentralization was thus favorable to subnational authorities, regarding both policymaking and fiscal capacities. This was largely due to the fact that political and fiscal decentralization had already taken place and subnational interests were effectively represented by the time administrative decentralization came about.

The process of decentralization in Colombia followed a sequence of reforms that conformed to the preferences of subnational actors. Political autonomy was devolved first, followed by resources, and finally by responsibilities. The decision to popularly elect mayors in Colombia had self-reinforcing effects on the next rounds of political, fiscal, and administrative reforms. It produced coordination among subnational authorities that led to fiscal decentralization and deepened political decentralization through the extension of the mayor's mandate and the recognition of municipal autonomy in the national constitution. It also produced a sense of incrementalism in the political elite that allowed for the approval of the popular election of governors. Administrative decentralization was the last, almost residual, type of reform. It was pushed through by the national executive. However, owing to the sequence of previous decentralization reforms, subnational actors and the teachers' union were able to get the guarantee that the fiscal resources necessary to afford the costs of the transferred services would also be transferred. As a result, this measure did not have a negative effect on the degree of autonomy of subnational executives with regard to the national government. As is evident in Table 2, this first cycle of political, fiscal, and administrative decentralization in Colombia empowered subnational executives.

Argentina: The National Path to Decentralization

Unlike the case of Colombia, Argentina's path of decentralization conformed to the preferences of the national executive. After the move away from developmentalism, the process of decentralization started with an administrative reform in 1978. It was followed by fiscal decentralization in 1988, and finally by political decentralization in 1994.¹²

On June 5, 1978, the national military junta passed two decrees transferring all national preschools and primary schools to the provinces, the city of Buenos Aires, and the territory of Tierra del Fuego. Retroactive to January 1, approximately 6,500 schools, 65,000 public employees, and 900,000 students (about one third of the primary public education system) were transferred to the provincial administrations. No revenues or fiscal capacities were transferred with the schools, and yet the transfer had a cost of 207 billion pesos—equivalent to 20% of the total national transfers (FIEL 1993, 148).

National interests prevailed in this first round of decentralization. In the context of an authoritarian regime, the national executive was able to impose on the provinces its most preferred outcome: administrative decentralization. The central government was interested in administrative decentralization for several reasons. First, they saw the provinces as enclaves of conservatism, in which future right wing political parties could develop. Second, the central government was interested in cutting the size of the federal bureaucracy and the national deficit, in the spirit of a neoliberal program of government (Novick de Senén González 1995, 138). Third, an increase in provincial revenues—which rose from 0.88% in 1976 to 1.56% of the GDP in 1977 (Kisilevsky 1998, 55)—established a favorable environment to transfer expenditures without resources. A report by the national ministry of education gave the following account of conditions before the 1978 transfer:

Despite the authoritarian regime, the governors voiced their concerns. Among others, the governor of Salta wrote to the minister of interior in November 1977: “by no means is the provincial treasury in a situation to afford the total costs of the services to be transferred” (Kisilevsky 1990, 20). At this time, however, the military's grip on power was at its strongest, and the unfunded transfer was imposed from above. The administrative decentralization of 1978 had disastrous fiscal consequences for the provinces. The allocation of provincial resources for education had to increase from 14% in 1977 to almost 20% in 1982 (IMF 1985), at the same time that automatic transfers to the provinces decreased from 48.5% to 29% of all shared revenues (FIEL 1993, 151). Thirteen percent of the primary schools (about 3,400 schools) closed down prior to 1980, and governors were forced to beg for discretionary transfers from the national executive to avoid further closures.

Unfunded administrative decentralization had four important policy effects: (1) it reshaped the preferences of governors toward political and fiscal decentralization; (2) it contributed to the reproduction of power of the national executive; (3) it produced a demonstration effect by providing an example that future policymakers could follow; and (4) it produced incrementalism within the educational sector toward further decentralization of responsibilities.

During the electoral campaign of 1983, at least six political parties (including the two main parties: Unión Cívica Radical, UCR; and Partido Justicialista, PJ) advocated for a constitutional reform (Leiva and Abásalo 2000). The common concern was to strengthen political institutions and to avoid future disruptions to democratic rule. Several proposals to reform the constitution were introduced in congress in the first 2 years of the democratic transition. At the end of 1985, President Raúl Alfonsín (1983–1989) ordered the creation of a council to study the matter. The council's recommendation for a constitutional reform included the creation of a mixed presidential system (with a prime minister), the strengthening of federalism, decentralization of the state, municipal autonomy, provincial control over natural resources, and limits on the president's authority to intervene in the affairs of the provinces (Consejo para la Consolidación de la Democracia 1986). The council's proposal was highly decentralizing, from both political and fiscal perspectives. Had it been implemented, it would have granted mayors constitutional autonomy, a prerogative they lack to this day. Governors would have had total control over natural resources (including oil) and more autonomy from the national executive in situations leading to federal interventions. Had this reform materialized, its political effects would have likely been similar to those of Brazil's 1988 constitution (on such effects, see Stepan 2000). Interestingly, the debate over the constitutional reform in Argentina became structured along partisan (rather than territorial) lines (Botana and Mustapic 1991; Smulovitz 1987), and the governors did not endorse this political decentralization reform. Instead, with the return to democracy, governors focused on a fiscal reform, exhibiting a shift in their expected preferences.

Given the design of the prior round of administrative decentralization, governors were eager to negotiate an increase in fiscal transfers. When the revenue-sharing law of 1973 expired at the end of 1984, governors pushed to have a new revenue-sharing law in place. Carlos Menem, who at the time was the governor of La Rioja, proposed that the interior provinces rebel and cut the supply of energy to the city of Buenos Aires until an agreement on fiscal transfers was reached with the president (Pírez 1986, 68). But President Alfonsín controlled the timing of the reform and was successful in delaying its approval. Meanwhile, he used discretionary transfers to buy the political support of opposition governors. Discretionary transfers amounted to 59% of the total transfers in 1985 and 54% in 1986 (Ministerio de Economía 1989, 177–79). Thus, from 1984 to 1987, Alfonsín gained bargaining power vis-à-vis the governors by using the fiscal transfers to the provinces—which they desperately needed after unfunded administrative decentralization—in exchange for political support (mainly in the Senate).

Only after the 1987 midterm elections, when the ruling party lost its majority in the House (passing from 51% to 46% of the seats) and five governorships to the PJ, President Alfonsín agreed to the governors' demand for redistribution of revenue-shared taxes. On January 7, 1988, congress passed a new revenue-sharing law (Ley de Coparticipación, or Law 23,548) by which the provinces were granted 57.66% and the national government 42.34% of all revenue-shared taxes, and the discretionary transfers were cut to 1% of the shared taxes. By all accounts, this fiscal decentralization law was a victory for the governors, which came about when an exogenous change (the midterm elections of 1987) altered the balance of power between the president and the governors inherited from the first round of reforms. But the reform was also instrumental to the national executive. By that point, mounting economic problems and adverse midterm electoral results had made it clear that the ruling party would not retain the presidency after 1989. If the PJ were to win the 1989 presidential election, the new co-participation law would guarantee resources to UCR governors.¹³

The provincial fiscal recovery did not last long, however. Soon after the new revenue-sharing law was passed, the national executive (now in the hands of the PJ) was able to push forward a second round of unfunded administrative decentralization, which neutralized the effects of fiscal decentralization. On December 6, 1991, the Argentine congress passed Law 24,049 according to which the administration of all national secondary and adult schools and the supervision of private schools were transferred to the provinces and the city of Buenos Aires. Two food programs and the few remaining national hospitals were also transferred. The estimated cost of the transfer was 1.2 billion dollars per year, the equivalent of almost 10% of the total provincial expenditures and 15% of the total national transfers. Over 2,000 national schools, 72,000 teachers, and 700,000 students were incorporated into the provincial systems of education, which also had to supervise more than 2,500 private schools. Article 14 of the law established that the cost of the transferred services would be paid with provincial resources, whereas Article 15 stated that whenever the revenues collected in a given month were below the average of the April–December 1991 period, the national government would transfer 1.2 billion pesos or the difference required to match that amount. Government documents and interviews with national and subnational officials suggest that such guarantee was not enacted and the transfer of responsibilities was largely unfunded (see Falleti 2003, 136–55).

The first round of administrative decentralization of 1978 had a demonstration effect for the second round of administrative decentralization. In 1991, as a result of the convertibility law, the absolute amount of revenues in the provinces had doubled—the automatic transfers passed from 4,810 million dollars in 1990 to 8,846 million in 1992 (Subsecretaría de Relaciones Fiscales y Económicas con las Provincias 1994, 15). In this context, as in 1978, it was easier to pass an unfunded administrative decentralization reform. Minister of Economy Domingo Cavallo appealed to the same arguments used in 1978 by Minister of Economy Martínez de Hoz to justify the transfer of responsibilities. In meetings with the governors, Cavallo argued that the increase in revenues would allow the provinces to afford the expenditures generated by the transfer of social services.

Finally, the 1978 decentralization reform also produced incrementalism. Although the national secondary schools were administered de jure by the national government until 1992, a process of decentralization of responsibilities was already under way. In the words of the governor of Mendoza from 1987 to 1991:

National officials also recognized this situation. Secretary of education Luis A. Barry said: Or as a member of the ministry of economy put it: “only in their plates were the schools national” (Pezoa, Juan Carlos, interview by author, Buenos Aires, February 13, 2001). Under these conditions, the governors were more inclined to accept a transfer of schools, even if it was to be funded primarily with provincial resources. The 1978 round of administrative decentralization enabled the national executive to pass a similar policy reform, albeit in a democratic context, 13 years later.

Political decentralization came last in the first cycle of market-oriented decentralization reforms in Argentina. It occurred in 1994, when President Menem (1989–1995 and 1995–1999) exchanged constitutional reforms as a bargaining chip for his reelection. Political autonomy was granted to the city of Buenos Aires, but various decentralization reforms proposed in the constitutional assembly by provincial representatives (and also included in the 1986 report of the Council for Democratic Consolidation) failed to pass. Reforms such as a higher share of subnational revenues, provincial control of natural resources, and constitutionally guaranteed municipal autonomy were all proposed in the constituent assembly; but due to the political pressure of the national executive all these fiscal and political decentralization proposals did not pass. In other words, the national executive was able to control the timing as well as the main contents of the political decentralization reform of 1994.

In sum, as a consequence of the first round of administrative decentralization, the preferences of the governors were reshaped. Because the 1978 transfer of schools was unfunded, governors were more concerned, after the return to democracy, with revenues than with a constitutional reform that would have granted them more political autonomy (e.g., by protecting them against federal interventions or granting them control of natural resources). Arguably, governors could have pursued both types of reforms at the same time, but they did not. Instead, between 1984 and 1987, they focused on fiscal decentralization and did not endorse the project of political decentralization. The 1978 reform also had demonstration and incremental effects in that additional unfunded administrative decentralization measures were made possible. Finally, the first round of administrative decentralization initiated a reproduction of the bargaining power of presidents, who were then able to control not only the timing of fiscal and political decentralization but also the contents and extent of these reforms.

The sequence of administrative, fiscal, and political reforms followed by Argentina resulted in a small change in the relative power of the governors and mayors. The share of expenditures increased, but by a lower amount than the changes experienced by Colombia, Mexico, or Brazil. The subnational share of revenues decreased. This was in spite of the fact that, from 1978, the Argentine provinces were allocated responsibilities whose cost amounted to approximately 35% of the total transfers they received from the center. Regarding policy-making authority in the educational sector, it remained unchanged until 1993, when the new federal law of education was passed. The appointment of subnational officials remained the same with the exception of the mayor of Buenos Aires, who became popularly elected in 1996. Finally, the territorial representation of interests in congress stayed more or less constant throughout the period. Despite the introduction of decentralization policies that transferred responsibilities, resources, and authority to subnational governments, the sequence in which the reforms took place meant that the intergovernmental balance of power remained unchanged in Argentina. Compared to their situation prior to 1976, governors had acquired more responsibilities and fewer fiscal resources, with no change in their political authority.