Using the Subnational Comparative Method to Study Latin American Politics

As Giovanni Sartori (Reference Sartori1970, 1035) reminded us long ago, to compare is to control. Herein lies the chief appeal of the subnational comparative method. By definition, subnational jurisdictions in the same country have in common their shared status as subnational jurisdictions; no subnational unit is more or less a constituent unit in the country of which it forms a part. This means that analysts can control for many factors that vary from country to country but that do not vary within a country, including such variables as whether the national government uses presidential or parliamentary forms, whether legal institutions reflect civil or common law traditions, and whether or not the country belongs to any given international organization.Footnote ⁴

According to Snyder (Reference Snyder2001b, 96), “subnational units within a single country can often be more easily matched on cultural, historical, ecological, and socioeconomic dimensions than can national units,” though he is careful to emphasize that “withinnation comparisons do not necessarily improve our ability to hold these dimensions constant” (italics in original). Because subnational regions in country X can be less similar to other subnational regions in country X than they are to subnational regions just across an international border in country Y, Snyder’s central recommendation is that scholars analyze contiguous subnational units across national borders. Surprisingly few scholars have heeded this call to simultaneously conduct within-nation and between-nation comparisons of contiguous units, even though Snyder’s article on the subnational comparative method is by far the most influential and widely cited for those who have adopted it in their research. Far more common are studies that explicitly compare subnational units within the same country while controlling for the national government, as discussed in the following section.

Why is the potential autonomy of subnational jurisdictions so important for the subnational comparative method? After all, there are good reasons to suspect that subnational governments are rarely fully autonomous, or entirely able to disregard the preferences of powerful national actors. Unlike the nation-state, which often enjoys significant autonomy relative to supranational actors, thanks to the sovereignty norms that govern the international system, subnational governments face multiple limitations on their autonomy; the relationship between subnational and national governments is simply not analogous to the relationship between states and supranational actors (Gibson Reference Gibson2013). In the context of Latin America’s “centralist traditions” (Veliz Reference Veliz1980), the “potential autonomy of subnational governments” can be understood to refer to their ability to adopt and implement policies that diverge from the preferences of the national government in a relevant policy area. In other words, can elected subnational officials ignore or evade the various inducements and constraints coming from national governments in the attempt to dictate subnational outcomes?

Here it is important to clarify that my argument is not that constraints on autonomy invalidate the use of the subnational comparative method, but instead that the degree to which subnational governments are constrained and the source of these constraints should inform theory building in general and the plausibility of holding constant the role of the national government in particular. As I will argue below, the less the autonomy of subnational governments, the greater the need to fold the national government into theory building, as opposed to discounting it theoretically. More generally, the source of the threat to autonomy—whether it is coming vertically from national or transnational actors or horizontally from other subnational governments—should be incorporated into theories that try to account for variation in outcomes across subnational units.

While the subnational comparative method is by no means limited to Latin America, one can argue that it is especially indicated there for a number of reasons, some of which have to do with how decentralization was designed in the region and some of which have to do with broader political dynamics. First is the predominance of symmetrical rather than asymmetrical approaches to administrative decentralization. When Latin American countries decided to devolve responsibilities for expenditures like education, health care, or infrastructure, they rarely excluded individual subnational units de jure from receiving these responsibilities (regardless of concerns about de facto unevenness in capacity), just as they rarely gave individual units special responsibilities not enjoyed by other units at their same level. Even as asymmetrical reforms have become more prominent elsewhere, often in response to geographically concentrated ethnic groups (Hooghe and Marks Reference Hooghe and Marks2016), the Latin American approach (outside of Bolivia, where the 2009 Constitution enables the formation of indigenous municipalities) has been decidedly symmetrical, which probably aids in the use of the subnational comparative method by presumably making it safer to assume that subnational units have been endowed with similar prerogatives by the national government.

Second is the significant extent of fiscal decentralization in the region (Escobar Lemmon Reference Escobar-Lemmon2001; Fukasaku and Hausmann Reference Fukasaku and Hausmann1999; Wibbels Reference Wibbels2005). In general, national governments transferred not just policy responsibilities but the resources necessary to make these responsibilities actionable and therefore meaningful (Garman et al. Reference Garman, Haggard and Willis2001). Fiscal decentralization thus increased the possibility of comparing whether and how units responded differently to their new formal responsibilities. Furthermore, to increase the fiscal resources under the control of subnational governments, those who designed decentralization relied heavily on automatic revenue sharing via transfers from the national government and largely eschewed the devolution of new tax bases to subnational units.Footnote ⁵ Given pronounced unevenness in the ability of subnational governments to collect their own taxes in most countries, this design feature further increased the applicability of the subnational comparative method by making it possible to compare how different subnational units used resources delivered from above.

Third, despite important differences in sequencing decisions (Falleti Reference Falleti2010), administrative and fiscal decentralization often came coupled with political decentralization. The introduction of elections to select subnational chief executives has made it more plausible to treat the units these individuals govern as substantially more independent than they were when their leaders were appointed by national governments, whether democratic or authoritarian. Meanwhile, other developments probably reinforced the ability of newly elected subnational governors and mayors to act much more independently than they had in the past. For example, the concomitant collapse of party systems that many Latin American countries experienced removed some of the partisan controls through which national party leaders formerly kept in line the subnational party leaders who governed in subnational spaces (Cyr Reference Cyr2017; Willis et al. Reference Willis, Garman and Haggard1999). At the same time, persistent deficiencies in the capacity of the central state, especially in the form of national governments that were unable to enforce legislative and policy controls across the national territory, can be thought of as a permissive condition that probably enhanced the de facto political autonomy of subnational governments.

Vertical Challenge 1: Accommodating the “National” in the Subnational Comparative Method

The first vertical challenge facing scholars who use the subnational comparative method has to do with how they should account for the role of the national government at the theoretical level. As the literature on decentralization has shown, the national government does not just go away after decentralization is adopted, but instead typically continues to alternately engage, favor, and punish different subnational governments in ways that can be difficult to measure and document.

The continued but uneven salience of the national government in most policy realms creates an important challenge for the subnational comparative method. Here I distinguish between three main responses to this challenge by scholars; while most omit national causes of subnational variation, others have taken a very different approach by exclusively positing national causes of subnational variation.Footnote ⁶ Still others have articulated hybrid theories that blend attention to national and subnational causes of subnational outcomes. Each approach to theory building has distinctive strengths, but the hybrid approach holds special appeal for those policy fields in which decentralization has resulted in shared authority rather than its neat reassignment from top to bottom. In other words, the content of the policy field itself will hold clues for the form that theory building should take. According to the articles in this special issue, it may be more difficult for subnational units to defend their status as autonomous units and make policy choices that deviate from what the national government prefers in some policy areas, like property rights (SánchezTalanquer this issue), than in others, like health care (Otero-Bahamón this issue).

Among work that theoretically brackets the national government, the most common approach to case selection has been to use the Most Similar Systems Design (MSSD) rather than the Most Different Systems Design (MDSD). This preference is possibly due to the very structure of the subnational comparative method, which encourages us to see subnational units as similar rather than different, and despite the fact that it can be productive to explore why subnational units that are different from one another in most relevant respects nevertheless share a common outcome of interest.Footnote ⁷

A pioneering MSSD study is Snyder’s book on reregulation in four coffee-producing Mexican states (2001a), where gubernatorial decisions are explained as the result of their interactions with state-specific societal groups (in both their oligarchic and grassroots forms), and where governors are conceptualized as equally upwardly accountable and equally bound by the dual imperative of securing votes for the governing party (PRI) and maintaining stability (2001a, 27). Snyder illuminates an intriguing paradox, whereby an avowedly neoliberal president (Carlos Salinas) sanctioned illiberal behaviors by governors in order to produce PRI majorities. As Snyder goes on to note, Salinas “did not merely tolerate subnational deviations from the neoliberal line, he selectively promoted them” (2001a, 28, italics mine), and promotion by Salinas seems especially important in the case of one of Snyder’s four states (Oaxaca), whose governor enjoyed a close personal relationship with the president. But if Salinas encouraged reregulation in some states more than others, perhaps in response to variation across states in the nature or depth of the electoral threat facing the PRI, this uneven behavior on the part of the president could have shaped variation in cross-state regulatory outcomes in ways that are not specified in the theory.

Another classic example of work that holds the national government constant theoretically, although by combining MSSD with MDSD, is Alfred Montero’s Reference Montero2001 study of industrial policymaking. Montero argues that subnational governments will successfully implement industrial policy only when delegative government and horizontal embeddedness are present at the subnational level; both conditions are necessary and cumulatively sufficient. Minas Gerais and Rio de Janeiro are compared as similar secondary industrial states in Brazil that produced different outcomes (industrial policy only in the former case), and the positive outcome in Minas is then compared with a similarly positive outcome in a very different case in another country, that of Asturias in Spain.

Like Snyder’s, Montero’s theoretical framework stands out for the richness of the combined attention it pays to state and societal actors at the subnational level, and also for the theoretical absence of the national government. Brasília is absent from the theory, but it does tend to shape outcomes in various ways in the cases, as in the argument that “intergovernmental conflicts with other states and the federal government encouraged Minas’s leaders to construct an elaborate economic technocracy designed to attract new investments to the state” (2001, 26). Not only does the center appear to play a motivating role as a trigger of what happened in Minas, but the vertical circulation of elites between Minas and the national government during military rule (65) could have shaped the logics of horizontal embeddedness among Minas elites in the 1990s that Montero emphasizes.

To be fair, Montero does begin to theorize the role of the national government when he discusses how decentralized cases like Brazil and Spain, where the national government is hypothesized to play no causal role, might differ from more centralized cases like Mexico, where democratization in the 1990s (a national-level event) explains the emergence of new industrial policy practices in the Mexican states (2001, 12–15).

A powerful, more recent example of the decision to treat the role of the national government as a theoretical constant can be seen in the study of policy diffusion across Brazilian municipalities by Natasha Borges Sugiyama (Reference Sugiyama2013), who explicitly (and laudably) clarifies that she wants to go beyond “a simple tale of vertical diffusion” (2013, 126).Footnote ⁸ For Sugiyama, municipal policy diffusion takes place in a fashion rather isolated from national-level dynamics; municipal officials are very aware of what’s going on in other municipalities, but national officials are mostly absent from their decisions about whether to emulate those municipalities or not. Perhaps because of her focus on larger cities, Sugiyama’s interviews with mayors suggest that federal transfers were too insignificant to shape municipal decisionmaking.Footnote ⁹ One could argue, however, that the role of the federal government in municipal policy adoption is undertheorized here, important in the empirical material but missing from the theory, which exclusively emphasizes municipal-level variables. The data show that Brazil experienced a dramatic increase in the adoption of its Programa Saúde da Familia (PSF) after 1998, which is the year that the federal government introduced greater funding to incentivize its adoption by municipalities (2013, 70, 73).

While the temptation to hold constant the role of the national government, and to privilege only subnational causes of subnational variation, is understandable, this treatment of the national government can also be questioned.Footnote ¹⁰ Decentralization helped to create “potentially autonomous policy jurisdictions” in Latin America, but not unambiguously. This is because, although the design elements mentioned above, such as administrative symmetry, automatic revenue sharing, and political decentralization, all helped justify the use of the subnational comparative method, other features of decentralization and its aftermath have worked in the opposite direction. For example, in the rapid push to decentralize, often the relevant legislation either failed to articulate a precise division of labor between levels of government in different policy fields, or specifically opted for a system of shared rather than exclusive competencies.

It is also the case that the same national politicians who signed off on the decentralization of authority then often doubled down on deconcentration, deploying (and funding) agents of the national government to compete with decentralized authorities in (some) subnational territories. Not only did these duplicative efforts undermine the rationality of public budgeting, but they also served to puncture “autonomous policy jurisdictions” at the subnational level. In other cases, national officials, or what Jessica Rich (Reference Rich2013) calls “activist bureaucrats,” have sought to mobilize civil society actors against subnational politicians in ways that may advance national reform objectives, but by eroding the autonomy of subnational governments.

Furthermore, decentralization often triggered attempts to recentralize authority and resources, sometimes across the board and sometimes in ways that targeted particular subnational units, but all of which have made the bracketing of the national government in “post-decentralization” Latin America a lot riskier as a research strategy (Dickovick and Eaton Reference Dickovick and Eaton2013; González Reference González2016). This is especially so in the context of judicial systems that are often too weak to punish national governments for unconstitutional breaches of subnational autonomy.

If the dominant approach has been to exclude the national government from causal explanations of subnational variation, some more recent work using the subnational comparative method has moved far in the opposite direction, not only theorizing the role of national-level variables but emphasizing them exclusively over more purely subnational variables. Consider the explanations of subnational regime type by Carlos Gervasoni (Reference Gervasoni2010, 2018) and André Borges (Reference Borges2016). In Gervasoni’s work, Argentina’s national-level revenue-sharing system, which the federal government uses to automatically transfer the taxes it collects to provincial governments, is the sole causal variable that explains the uneven democratic performance of provinces. The greater the share of these transfers in total provincial revenues, the lower the level of democracy in the province. While the causal mechanisms Gervasoni considers play out within the provinces, the theory itself only has space for national institutional rules and cannot explain subnational regime change in the absence of any change in the content of the rules.Footnote ¹¹

With respect to the same dependent variable, Borges looks to the party system rather than to the fiscal system (and to Brazil rather than Argentina) but develops an argument that similarly emphasizes national-level dynamics. Borges explains subnational democratization as a response to the nationalization of the party system: states with more nationalized legislative elections are more democratic, and “national party dynamics have gradually undermined subnational incumbents’ power over state institutions” (Reference Borges2016, 171).

Between the extremes of these two options, which have led scholars either to control entirely for national causes or to endow them exclusively with explanatory power, scholars have successfully deployed a third, “hybrid” option. In this line of work, scholars develop theoretical explanations for subnational-level variation that identify both subnational-level and national-level causal variables. Hybrid theories are appealing because they avoid the dangers of overstating the autonomy of subnational units, which exclusively subnational theories tend to do, while also avoiding the opposite tendency, in exclusively national theories, to underestimate the potential autonomy of these units.

Two representative examples of hybrid approaches, representing the earlier wave of work on economic policy and the more recent wave of work on regime type, are those of Erik Wibbels (Reference Wibbels2005) and Agustina Giraudy (Reference Giraudy2015). Wibbels, for example, explains variation in the policy preferences of subnational officials (and the associated implications for national-level macroeconomic performance) as a function of four independent variables: one that is subnational (the level of political competition within the province), two that are national (the rules that govern the federal fiscal system and the representation of provinces in national legislatures), and one that straddles the two levels (the relationship between national and provincial leaders within parties). Turning to the persistence of subnational undemocratic regimes, Giraudy’s explanation similarly combines national-level variables like the president’s fiscal and partisan powers with subnational variables, such as the degree to which governors can depend on elite cohesion or mass support within their provinces.

A third representative example, though at the municipal rather than provincial level, is Donna Lee Van Cott’s Reference Van Cott2008 study of indigenous politics in ten small and medium-sized cities in Bolivia and Ecuador. After theorizing about the impact of Ecuador’s more flexible approach to decentralization (a national-level feature), she then attributes reform outcomes to municipal-level differences.

Considering these three scholars, one can identify a critical trade-off here, in the sense that hybrid theoretical frameworks blending causal attention to national and subnational variables are likely to come at the cost of greater complexity (and less parsimony), to the extent that this is seen as a negative.

Recent research shows that theoretically hybrid approaches can also be used to accommodate the vertical relationship between intermediate-level units and the local governments within those units.Footnote ¹² Scholars who are chiefly interested in using the subnational comparative method to compare intermediate-level governments have begun to factor in and theorize the various roles played by municipal governments, as seen in the “authoritarian province, plural cities” dynamic (Fenwick Reference Fenwick2016; Gibson Reference Gibson2013, 2015; Behrend and Whitehead 2016; Niedzwiecki Reference Niedzwiecki2018; Souza 2016).Footnote ¹³ Likewise, scholars whose main research site is municipal governments have worked in the opposite direction by considering more deeply how municipalities are affected “from above” by intermediate-level governments (Grindle Reference Grindle2007; Alves and Gibson Reference Alves and Gibson2019). Consider, for example, Veronica Herrera’s Reference Herrera2017 study of water politics in Mexican municipalities, which attributes the marketization of water policy to industrial elites and middle-class constituencies. While her focus is on municipal-level causes and outcomes, Herrera simultaneously factors in the role that intermediate-level governments played, which she conceptualizes as the level of institutional support for municipal water reform that state governments can provide or fail to provide (Reference Herrera2017, 82).

By paying theoretical attention to more than one level of government (i.e., national and provincial, or provincial and municipal), hybrid theories offer a more realistic way of theorizing the complexity inherent in multilevel political systems, along with the multiple threats to autonomy that subnational governments face.

Vertical Challenge 2: Relationships Between Subnational Governments and Transnational Actors

According to the discussion so far, scholars have paid a lot of attention to the vertical relationships between subnational and national governments, and they have differed greatly in the decisions they have made about how to treat the national government theoretically in research that seeks to explain subnational variation. But another important vertical challenge also affects the use of the subnational comparative method, though this second challenge has received far less explicit scholarly attention. Whether subnational units function as potentially autonomous policy spaces is also shaped by the vertical relationships that increasingly tie them to international or transnational actors, who can be conceptualized as operating above the national level of government.

Even though we live in an era of globalization, comparativists who study intergovernmental relations in Latin America have tended not to pay much attention to supranational dynamics. Perhaps this is because struggles between national, provincial, and municipal governments over which level gets to do what and with whose money often seem to play out as quintessentially domestic conflicts, safely protected by sovereignty norms and securely under the control of ministries of the interior and of finance. For whatever reason, and to the best of my knowledge, scholars who have used the subnational comparative method have mostly overlooked the question of how external actors might shape the autonomy of subnational jurisdictions (even though these actors did receive a fair amount of attention in the literature on why decentralization happened). But attention to these external actors is likely to be critical in order to defend the claim that subnational units represent “potentially autonomous policy jurisdictions.”

Even without much literature to draw on, it is nonetheless possible to argue that the subnational transnational challenge differs from the subnational national challenge in two key respects, which scholars who use the subnational comparative method should keep in mind. The first is that it refers to a much more heterogeneous set of actors, including bilateral aid organizations, international financial institutions, globalized NGOs, transnational advocacy networks, multinational corporations, and even individuals, in the form of transnational migrants. A second difference, related to this remarkable breadth in the identity of the external actors who might impact the degree of subnational autonomy, is the reality that this impact might be positive or negative, in the sense that disparate goals and preferences might drive these external actors either to favor or to oppose autonomy for subnational units. This variety stands in stark contrast to the first vertical challenge, in which it can more safely be assumed (and documented) that the national government mostly tries to restrict subnational autonomy, even when it has passed decentralizing legislation.

Beyond taking stock of the variety of external actors and the multiplicity of their preferences, scholars can usefully distinguish between two different kinds of relevant scenarios (“bottom up” vs. “top down”) that increasingly link subnational governments with extranational actors and forces. According to the first scenario, the leaders of subnational governments themselves have reached out and sought to activate possible allies or defenders outside their countries when their autonomy is threatened by political opponents at the national level. We know from the sizable literature on the “Europe of the Regions” (Hooghe and Marks Reference Hooghe and Marks2001) that the effect of the creation and strengthening of European Union institutions has been to bolster the position of subnational regions in regard to national governments; by the year 2000, regions had already staffed and funded more than 150 independent offices in Brussels. In Latin America, regional integration has afforded nothing of the kind to subnational governments, even though these governments have indeed been able to defend provincial interests indirectly through the lobbying of national governments over Mercosur’s Common External Tariff (Pezzola Reference Pezzola2018).

Drawing on the concept of “paradiplomacy,” scholars have begun to examine how subnational governments in Latin America directly engage in international relations, not as a way to defend their autonomy but instead as a means to encourage foreign investment in their regions or to promote regional exports (Maira 2010). For example, Jorge Schiavon (Reference Schiavon2006) finds that opposition-governed states in Mexico are not more likely to engage in international relations, and that wealth and proximity to the United States are, instead, the factors that explain their pursuit of paradiplomatic efforts abroad.

Rather than the Europe of the Regions or paradiplomacy, the deployment of external strategies by subnational governments in Latin America more closely recalls Keck and Sikkink’s “boomerang effect,” through which domestic actors partner with external allies to target national governments (Reference Keck and Sikkink1998). Subnational governments should probably be added to the long list of actors enumerated by Keck and Sikkink who may deploy the boomerang, including local social movements, foundations, and the media.

One can identify several salient examples of subnational elected officials appealing to international bodies to defend subnational autonomy in the face of what they frame as illegal intrusions by the national government. Consider the case of Gustavo Petro, a former guerrilla leader of the M-19 in Colombia, who was elected to a four-year term as mayor of Bogotá (2012–15). After pursuing a series of left-leaning reforms in service provision, including a controversial attempt to reverse the privatization of trash collection, Petro was suspended from office by Inspector General Alejandro Ordoñez (Eaton Reference Eaton2020). Arguing that his suspension was a breach of Bogotá’s 1993 autonomy statute, Petro took his case to the Inter-American Court of Human Rights, which ordered President Juan Manuel Santos to reinstate the mayor (who was able to finish his term). Another example of a mayor who pursued a similar course of action but to lesser effect was Caracas mayor Antonio Ledezma, who responded to what he considered illegal recentralizing actions by President Hugo Chávez by lobbying the Spanish Parliament and by engaging in a hunger strike on the steps of the Caracas headquarters of the Organization of American States.

If the first scenario involves subnational officials reaching out to extranational allies in the struggle to defend their autonomy from the center, the second, “topdown” scenario unfolds in the opposite direction, as external actors reach down into subnational spaces. The broader point is that when decentralization expanded subnational autonomy, some transnational actors benefited from this change while others were threatened by it. Thus, for the sake of simplicity, one could argue that external influence can take one of two forms: attempts by transnational actors to restrict the autonomy of subnational governments or to defend that autonomy.

Perhaps the most visible actors to take stock of are the same bodies—bilateral aid agencies and multilateral financial institutions—that promoted the very decision to decentralize to begin with and that, subsequent to decentralization, stayed on the scene, in the form of capacity-building programs for newly empowered subnational governments (Dickovick Reference Dickovick2014). Even here, however, one finds evidence of rival views about the merits of subnational autonomy. Although most of the prominent bilateral development organizations have continued to promote development, other actors, in the World Bank and the International Monetary Fund, voiced early support for recentralizing measures in Argentina, Brazil, and Colombia (Dillinger and Webb Reference Dillinger and Webb1999; Restrepo Reference Restrepo2015).

In addition to governmental organizations, scholars should canvas three other kinds of external actors—multinational corporations, transnational NGOs, and migrant associations—to determine whether and how their behavior might shape subnational autonomy. With respect to corporations, examples of how foreign companies can benefit from the autonomy of subnational units can be found in the literature on the fiscal wars that took place between Brazilian states in the 1990s (Rodríguez-Pose and Arbix Reference Rodríguez-Pose and Arbix2001). Governors’ ability to offer exemptions from the important tax on goods and services led to bidding wars, which probably resulted in less tax liability for automobile manufacturers.

But foreign companies may also have cause to fear subnational autonomy and to use the kinds of rules that were introduced by Chapter 11 of NAFTA, which expanded protections for foreign investors in relation to subnational governments. A case in point is the municipality of Guadalcázar, which sought to deny a land use permit to a U.S. company, Metalclad, which wanted to open a toxic waste facility in the municipality (Warner and Gerbasi Reference Warner and Gerbasi2004). In 1996, Metalclad secured a $16.7 million award from the court of international arbitration established by NAFTA to compensate for the negative financial impact of losing the permit. Warner and Gerbasi propose that cases like this have a galvanizing effect on foreign investors seeking to question almost any form of regulation by subnational officials, and a chilling effect on municipalities seeking to make use of their legal authority.Footnote ¹⁴

Like bilateral aid agencies, transnational NGOs are especially relevant as external actors that—for their own issueor policy-specific reasons—may bring to bear important assistance in the construction of subnational capacity, a critical asset, the absence of which may doom efforts by subnational governments to act autonomously. Peru is a case in point. External environmental NGOs, in partnership with national NGOs, have sponsored capacity-building activities in regional governments to enable them to effectively make use of formally devolved zoning and planning authorities (zonificación ecológica económica y ordenamiento territorial, or ZEE-OT). As Maria-Therese Gustafsson (Reference Gustafsson2016) shows in her comparison of outcomes across regions, some regional governments have been far more able than others to carry out and conclude this technical exercise in ways that impose limits on where mining can take place within the subnational territory—in direct opposition to the wishes of the national government in Lima.

In this sense, the case of resource extraction in Peru showcases a particularly interesting dynamic, whereby different kinds of transnational actors (e.g., environmental NGOs and transnational corporations) seek to bolster rival sides in the struggles that have unfolded between subnational and national governments over the scope of autonomy for the former. Transnational mining companies question the autonomy of local authorities to regulate extraction at the same time that transnational NGOs are engaged in capacity-building exercises to render their autonomy meaningful.

In addition, a vibrant literature on the impact of remittances on country-of-origin dynamics points to the importance of a relatively new kind of “external” actor: migrants and the hometown associations (HTAs) they have built to facilitate and coordinate the flow of resources back home. To date, scholars have largely asked about the impact of remittances on democracy (Meseguer and Aparicio Reference Meseguer and Aparicio2012) and development (Duquette-Rury Reference Duquette-Rury2014) rather than about how they may impact the potential autonomy of subnational units. The design of Mexico’s 3×1 program, whereby funds from municipal, state, and federal governments triple the amount of money sent by HTAs, points toward at least two possible rival hypotheses. On the one hand, remittances may increase the fiscal capacity of municipalities that otherwise simply do not enjoy sufficient resources to make use of their formal policy prerogatives. On the other hand, the required participation of higher levels of government, in the form of matching transfers, can link municipal governments to the preferences of those higher levels even more tightly, with negative consequences for their potential autonomy.

The Horizontal Challenge: Relationships Between Subnational Governments at the Same Level

In addition to challenges from actors who occupy higher rungs in vertical hierarchies, the potential autonomy of subnational jurisdictions can also be challenged horizontally. Here, too, we see significant variation in the literature in terms of how, and how much, scholars have emphasized these relationships between subnational units at the same level of government (intermediate or local).

Much of the literature tends not to focus on these relationships at all, or to privilege the vertical relationship with the national government to the exclusion of horizontal dynamics.Footnote ¹⁵ In the literature on subnational regime type, for example, interactions between provinces are unimportant for both Gervasoni and Giraudy, who both see struggles between democratic presidents and autocratic governors playing out in an almost exclusively bilateral fashion. Horizontal dynamics likewise seem to be theoretically irrelevant in the major policy-centered works discussed above, including research on regulatory, industrial, and fiscal policymaking by intermediate-level governments (Snyder Reference Snyder2001a; Montero Reference Montero2001; Wibbels Reference Wibbels2005), as well as work that explains municipal-level variation (Grindle Reference Grindle2007; Moncada Reference Moncada2016; Wampler Reference Wampler2007). Though horizontal dynamics have not featured prominently in research that deploys the subnational comparative method, I argue that relationships between subnational units can have the effect of either increasing or decreasing the autonomy of those units, and that they can be conceptualized as taking three main forms: cooperation, conflict, and contagion and diffusion.

Cooperation has received the most attention. Most generally stated, the asymmetrical power dynamics that characterize the relationship between the national government and even the most powerful of subnational units put the onus on cooperative behaviors between subnational governments to defend their collective prerogatives (Stoner-Weiss Reference Stoner-Weiss1997). In her sequential theory of decentralization, for example, Tulia Falleti considers the creation of subnational associations of provinces or municipalities as a “strong indication of the increased levels of autonomy of subnational officials” (Reference Falleti2010, 64). For Colombia, Falleti shows how the Colombian Federation of Municipalities lobbied successfully for additional fiscal resources and for the recognition of municipal autonomy in the 1991 Constitution (134).

Laura Flamand (Reference Flamand2009) and Alberto Pacheco (Reference Pacheco, Puga and Luna2008) likewise emphasize collective efforts on the part of Mexican governors, who succeeded in creating a powerful new lobbying organization (the National Conference of Governors) under the Vicente Fox administration. Cooperative horizontal relationships are also critical for Edward Gibson, who argues that the subnational hegemonic parties that sustain subnational authoritarianism and enforce “boundary control” are linked horizontally to a network of copartisan (and not necessarily authoritarian) provincial parties. As he notes in the Argentine case, “Peronist hegemonic provincial parties were … minority members of a chain of provincial parties dominated by parties from large competitive provincial party systems, [and] benefited from the reflected legitimacy of their national coalition” (2013, 156; see also Cruz Reference OlmedaOlmeda forthcoming).

I also argue (Eaton 2017) that lateral coalitions between subnational officials can serve to defend the autonomy of the units they govern, and that ideologically deviant governors and mayors are likely to lose if they confront the center on a bilateral basis. Whereas a coalition of governors in eastern Bolivia coordinated actions (e.g., work stoppages, strikes) against the national government, which was forced to accept those departments’ more liberal policies, regional presidents in Peru who tried to design policies at odds with national preferences were isolated from one another, unable to find common cause, and ultimately powerless to resist Lima’s divide-and-conquer tactics.

At first blush, conflict between subnational governments would appear to hinder the ability to act collectively in defense of their autonomy as subnational units. However, scholars have identified pathways through which conflictual rather than cooperative relations between subnational units can lead to greater autonomy. For example, Soifer (Reference Soifer2016) argues that while conflict between subnational regions in Ecuador and Colombia had the effect of depressing overall levels of public goods provisioning by the central state, it reinforced, at the same time, the ability of regional elites to pursue their distinct preferences in regard to public goods in a more autonomous fashion. In Brazil, conflict between the wealthier states of the South and the poorer states of the North and Northeast led to an institutional solution in the guise of the 1988 Constitution, which enhanced the autonomy of both through different accommodations: the devolution of important tax bases like the VAT, which pleased the former; and generous provisions for the automatic sharing of revenues from centrally collected taxes, which assuaged the concerns of the latter (Souza Reference Souza1997).

Moreover, while subnational units may clash or coalesce in quite visible patterns, they also influence each other horizontally through subtler and more indirect mechanisms of contagion and diffusion. While most of the diffusion literature focuses on the nation-state as the unit of analysis, recent scholarship suggests that subnational units are powerfully shaped by what happens in neighboring jurisdictions. Consider Sugiyama’s research on social policy diffusion across Brazilian municipalities. She found that social networks and shared professional norms between health bureaucrats in different municipalities facilitated the spread of common policy approaches. Thus, while her first causal variable (the ideological orientation of the mayor) is one that can be thought of as mostly independent from other districts, the second variable taps into the importance of professional relationships that transcend jurisdictional boundaries. Municipal bureaucrats network in professional forums, the effect of which is, de facto, to limit the autonomy of municipalities as separate policy jurisdictions. In other policy-centered research, Catalina Smulovitz (Reference Smulovitz2015) has shown that factors like the strength of civil society explain cross-provincial variation in the adoption of domestic violence laws, but that more protective laws have diffused across provinces over time.Footnote ¹⁶