ELITE CONFLICT AND STATE CAPACITY

In this section, I present a theory that explains the pervasiveness of low state capacity and its possible development as the outcome of a conflictive relationship between a ruler and a nonruling economic elite.

The argument centers on two actors. A political ruler, whose power emerges from military dominance, decides whether to expropriate an enemy economic elite and whether to invest to expand future capacity. The ruler strives to maximize the spoils from office, which can be achieved by holding on to power for as long as possible and by increasing the state’s ability to tax. Thus, state capacity, conceptualized as the state’s ability to implement policy, is realized here as its ability to raise taxes. The ruler expands capacity through costly investments, and decides whether to invest based on the private returns to those investments. The nonruling economic elite, in turn, control some productive asset and can use part of their income to resist the ruler and potentially gain power. The elite seek to maximize profits from their economic assets, which entails avoiding expropriation as well as minimizing their future tax burden.

These opposing distributive preferences between the ruler and economic elite result in a situation of latent conflict.Footnote ⁴ Such distributive conflict erupts in the absence of institutions that can generate credible bargains among them, such as institutions of limited government, and when no actor is overpowering—if the latter were the case, fighting would be futile, and the weaker faction would prefer not to incur its cost.

The argument takes as its starting point a common phenomenon throughout history: the low-capacity trap. When the ruler coexists with a strong economic elite in the conditions outlined above, he has few incentives to undertake the necessary investments to increase state capacity. This is because he fears that these investments generate backlash by prompting the economic elite to attempt to seize power and then use the enhanced capacity to their own advantage and against the ruler. It therefore makes little sense for the ruler to invest in enhancing capacity—doing so would reduce the likelihood that he survives in power to reap the benefits of the investment. The ruler can attempt to deplete the economic elite’s own source of power by expropriating their assets, but a strong elite can successfully fight back, wrestle power away from the ruler, and reverse the expropriation.

Thus, a sufficiently strong economic elite, by credibly threatening to seize political power, can deter both the expropriation attempts and the investments in capacity by the ruler, leading to a low-capacity equilibrium. Furthermore, marginal changes to the economic elite’s resources are not enough to escape this situation. This is because the elite can compensate for a small decline in their overall resources by increasing the fraction they dedicate to trying to seize power.

This characterization of low-capacity equilibria helps explain why the emergence of capable states is a relatively rare occurrence. In contrast to stationary bandits that effectively monopolize theft in their domains and have incentives to invest in the future, rulers who face continuous challenges from domestic elite groups have to spend their energy and resources trying to hold on to power. Features of this low-capacity equilibrium can be found in late-medieval and early-modern European states, such as Britain under the first Tudors; in nineteenth-century and early twentieth-century Latin America, such as Chile and the Dominican Republic, and in some present-day states, such as Afghanistan, Haiti, and the Democratic Republic of Congo (North, Wallis, and Weingast Reference North, Wallis and Weingast2009; Pincus and Robinson Reference Pincus and Robinson2011; North et al. Reference North, Wallis, Webb and Weingast2012).

An escape from this low-capacity equilibrium can be achieved by a large disruption to the existing balance of power: for example, by a negative shock that temporarily weakens the economic elite relative to the ruler and clears the way for political consolidation and investments in state capacity. The shock reduces the elite’s ability to resist the ruler, who, undeterred, uses this window of opportunity to permanently destroy the elite’s source of power through the expropriation of their productive asset. By facilitating the political consolidation of the ruler and lengthening his time horizon, temporary ruler strength enables investments in state capacity that would have been prohibitively costly otherwise.Footnote ⁵

Temporary ruler strength can emerge from many sources, and shocks that weaken the nonruling elite are one example. More generally, a way out of low-capacity equilibria can be achieved by substantial, temporary changes to the relative balance of power that favor the ruler. Positive shocks to the ruler include, for example, asymmetric access to a new military technology or support from an external power. Negative shocks to the elite include episodes that reduce their ability to generate profits, such as the introduction of new technologies that make existing systems of production obsolete or adverse weather events.

In this article, I focus on economic shocks that severely impact the economic elite. These can come in the form of a negative price shock, which asymmetrically affects the economic elite when the ruler can still call on other resources—such as mobilizing landless peasants vying for the elite’s land—to support his cause.

In short, when facing low enough prices that temporarily weaken a nonruling economic elite, the ruler has an incentive to expropriate them and consolidate his power. Investments in future capacity and, with them, the foundations of a capable state, are more likely to take place.

Thus, under the scope conditions of the theory—evenly matched elite factions with competing distributive preferences interacting in the absence of institutions that can sustain a credible bargain between them, and asset expropriation as a way in which rulers can weaken their opponents—two key hypotheses that can be brought to the data emerge:

1. 1. When the nonruling economic elite faces low enough prices, rulers are more likely to expropriate the elite’s assets.

2. 2. When the nonruling economic elite faces low enough prices, rulers are also more likely to invest in expanding state capacity.

BACKGROUND

Ideally, these hypotheses would be evaluated by randomly assigning prices to a sample of ruler-elite pairs in conflict. As an approximation to this ideal design, I use a large economic shock—the Great Depression—as an exogenous change in agricultural commodity prices in Mexico, and take local political rulers and the local economic (landed) elite as the relevant actors. Mexico in the aftermath of the revolution—which lasted from 1910 to roughly 1919—is an attractive case to examine; it satisfies the scope conditions of the theory and, because of its political situation, offers a sizeable number of subnational units, which allows for meaningful quantitative analysis.

Local hacendados—the owners of large estates—had developed credible rent-sharing arrangements with the political elite during the pre-revolutionary period (e.g., Haber, Maurer, and Razo Reference Haber, Maurer and Razo2003). However, after the revolution, the Mexican state collapsed. The incipient national institutions of the old regime disappeared during the turmoil of the civil war, and the power vacuum was filled at the local level by a confederation of regional warlords, as well as other caciques (political bosses), with local power bases (e.g., Brading Reference Brading1980; Knight Reference Knight2005).

The old pre-revolutionary bosses, members of large landowning families, were completely discredited politically after the revolution and stopped holding public office. They nonetheless “wielded unseen but surely felt power, [and] had the means of coercion or could easily purchase them” (Wasserman Reference Wasserman1993, 128).Footnote ⁶ In some places, the old landed elite staged a comeback after several years of political retreat, while in others they lost influence definitively.

Local politics during the period proceeded largely independent of national politics, as municipal factions wrestled control of the available economic resources: land and taxes. During the period, municipios relied on their own resources to run the local government, with little support from state and federal resources. They retained the prerogative of taxing land, services, and the production and sale of some commodities. The local bureaucracy staffed tax collection offices, public safety forces (e.g., municipal guards), and sometimes provided basic public services, such as water provision, public sanitation, and the administration of local public markets, slaughterhouses, and burial grounds (Wasserman Reference Wasserman1993; Aboites Reference Aboites2003).

The conflict over local power—and its associated rents—often turned violent as the pre-revolutionary elites struggled to preserve their position and the new bosses sought to establish themselves locally. The outcome of these conflicts hinged on the resources that these groups were able to mobilize in their favor; while the old elite could turn to their estates to support their bid for power, the rising notables could rally land-hungry peasants (e.g., Wasserman Reference Wasserman1993; Knight Reference Knight2005).Footnote ⁷

Despite their retreat from local political life after the outbreak of the revolution, hacendados were largely able to secure their properties during the years of civil war (1910–19), even in the face of limited land redistribution implemented by the various armed factions as part of their recruiting strategies. In the 1920s, with agrarian legislation already in place, land reform was used as a political instrument to demobilize peasants in conflictive regions, but it was largely limited to claims of past dispossession and by villages outside of haciendas (e.g., Sanderson Reference Sanderson1984; Gaona Reference Gaona1991; Dell Reference Dell2012; Saffon Reference Saffon2015).

Thus, by the 1930s, local landowners still retained the main source of wealth in rural municipios. The period of mass land redistribution only came, as Figure 1 shows, after the Great Depression, and particularly after 1934. The revival of the land reform program was supported at the national level by President Lázaro Cárdenas (1936–1940), but its implementation had a very local dynamic, and could only happen with support from local authorities.Footnote ⁸ Local political leaders could (and did) encourage land redistribution by coordinating land invasions that effectively expropriated the hacendados’ estates and increased the likelihood of a petition’s formal approval. At the same time, they were also able to stop the process at various stages, if it was in their interest.

FIGURE 1.

Land Redistribution in Mexico (1916–1960)

Modern ejidos, the communities formed from redistributed land, are exempt from property taxes. However, during the period of analysis, and well into the 1950s, redistributed land was taxable, and was one of the most important sources of local revenue (Aboites Reference Aboites2003).Footnote ⁹ Local rulers could therefore benefit from an increased capacity to tax land. Whether the struggle between hacendados and the new local political leaders determined the expansion of local state capacity, perhaps mediated by land reform, remains an open question in the literature.

In the Mexican post-revolutionary context, the argument developed above suggests that, in the areas that were badly hit by a negative price shock and landowners’ resources were diminished, the new political bosses should have been more likely to expropriate land (Hypothesis 1), and to invest to increase state capacity (Hypothesis 2). Furthermore, these effects should arise specifically where a competing elite—the landowners—were present.

Illustrative cases. A number of cases exemplify this pattern of landlord domination that persisted through the revolution, followed by agrarian unrest that culminated in land reform, and the final consolidation of the new local political leadership—often military caudillos or local caciques—that instigated expropriation (e.g., Singer Reference Singer1988; Harper Reference Harper2009). The available record also suggests that, in these cases, political consolidation was also accompanied by an expansion in local state capacity. To illustrate the theory, in this section I contrast two badly hit regions that experienced political consolidation, land expropriation, and rapid growth of capacity (Cañada and San Felipe), with a region that was only mildly exposed to the price shock (Naolinco), and where political instability persisted, land expropriation was limited, and state capacity increased at a much slower pace.

The local ruler and cacique Ernesto Prado from Michoacán’s region Cañada de los Once Pueblos, which was badly hit by the Great Depression, provides a good example of a path that was conducive to state capacity development:

San Felipe del Progreso, in the state of Mexico, also suffered a negative price shock in the aftermath of the Great Depression. In her ethnographic study of the municipio, Margolies (Reference Margolies1975) describes a similar process to the one in the Cañada. A handful of hacienda owners prospered in the late nineteenth century by expanding both their export-oriented production and, given the proximity to mines and its workers, domestic-oriented grain production. They dominated local political life—at least one of them directly served as municipal president during the pre-revolutionary period—and had a considerable degree of authority not only within their estates but in the surrounding towns. “In short, [the hacendado] sought a comfortably mutualistic relationship with townsfolk and a toadyish receptivity which permitted him to do whatever he wanted in any contingency. [He could] satisfy both insignificant whims [. . .] and regulatory demands” (Margolies Reference Margolies1975, 32).

Hacendados, the economic elite, were able to successfully resist the revolutionary turmoil (1910–19) through self-financed armed defense groups, and the newer local revolutionary political leadership, the local ruler, saw them as a threat. By the 1930s, these local authorities reacted with a “deliberately calculated” indifference to illegal land invasions on the haciendas, and “repeatedly denounced the landlords as ‘provocators’ and ‘subverters’ whose only aim was to create problems for the government” (Margolies Reference Margolies1975, 43). Macario Durán, a local agrarian leader from the 1930s that rose to the municipal presidency in the early 1940s, “exercised absolute power from 1940 to 1957” (Torres-Mazuera Reference Torres-Mazuera2012, 51), which allowed him to profit from lands formally held in common property and monopolize local trade.

In both cases, the available data, described in more detail below, suggest that the price shock hit these regions negatively, and that, as the theory implies, both expropriation of assets from the landed elite and one measure of local state capacity, the number of bureaucrats, increased substantially. The Cañada region roughly corresponds to the municipio of Chilchota, which saw its commodity potential—the gross income that could be potentially generated given the local agro-climatic conditions and commodity prices—fall by 30% from 1930 to 1940. Over the same period, land was redistributed for the first time, and the total number of bureaucrats rose from four to 35. In San Felipe, commodity potential fell by 20%. Land redistribution, though already underway by 1930, increased by a factor of eight over the next decade, while the total number of bureaucrats tripled from 15 to 45. Relative to national trends in the number of bureaucrats, which on average increased by a factor of 1.5 during the period (or ten bureaucrats per municipio, in absolute terms), these were large increases.Footnote ¹⁰

Naolinco, in the central region of the state of Veracruz, provides an example of failed expropriation and continued conflict. In the aftermath of the revolution, a radical agrarista movement, led by governor Adalberto Tejada, took momentum regionally. Armed peasants organized into rural defense forces and successfully took over local governments across the state (Fowler-Salamini Reference Fowler-Salamini1978). While a sizeable number of provisional land redistributive grants were approved, in places like Naolinco, hacendados were able to continuously avoid their implementation through violence, by employing (costly) hired guns to intimidate peasants and the new local rulers. The most prominent armed group that sold protection was the Mano Negra, led by the hacendado and representative of the local economic elite Manuel Parra, who took over local administrations in the region and reportedly killed thousands of peasants in the region during the period (Santoyo Reference Santoyo1995). After Parra’s death in 1943, agrarian unrest continued. His hacienda was eventually expropriated and redistributed in the 1950s under the leadership of Guillermo Cedeño, who nonetheless had to flee the municipio himself after several attempts on his life.

Contrary to the previous cases, in Naolinco, commodity potential only fell by 10% (close to the mildest possible decline across the country), which could help explain the landowners’ ability to resist far-reaching land reform and challenge the agrarian leadership. In the 1930s, there was only one definitive redistributive grant with expropriated high-quality land, which amounted to less than 1% of the municipio’s surface area. While the number of bureaucrats increased—following a nationwide upward trend—it did so at a slower rate than the other cases, from 23 to 30 public officials (a 30% increase).Footnote ¹¹

RESEARCH DESIGN

These cases highlight patterns that are consistent with the observable implications of the argument. However, opportunities for a definitive consolidation of the local political leaders that emerged from the revolution, as well as the hypothesized expansion of state capacity, likely varied from region to region in unobserved ways. This makes a systematic evaluation of the proposed theory challenging. To overcome these difficulties, I rely on one source of exogenous variation to the landed elite’s available resources, and thus to their ability to resist and challenge the rising local political leaders. Price shocks induced by the Great Depression affected crop values differentially and were largely unexpected (e.g., Hamilton Reference Hamilton1992). They can thus provide credibly exogenous variation in the economic resources available to the landed elite between regions suitable to produce different crops.

As the theory suggests, a large negative shock to the economic power of elites can increase investments in state capacity—potentially with long-term consequences. In Mexico, these price shocks came during a period of political reorganization following the turmoil of the revolution. If local leaders successfully consolidated power after a large shock that depleted the landed elite’s resources, and their regional political machines were in turn integrated into the national political system under the Partido Revolucionario Institucional (PRI) regime, these initial increases in local capacity could have persisted.

Figure 2 illustrates this component of the identification strategy with two selected commodities. After 1929, some internationally traded agricultural commodities, such as coffee, display a sharp decline in their prices, while others remain relatively stable (rice) or even slightly increase (bananas).Footnote ¹²

FIGURE 2.

Commodity Prices Before and After the Great Depression

Given the availability of historical data, I implement two estimation strategies: a difference-in-differences approach that uses two periods (1930 and 1940) and compares differential changes in state capacity and land redistribution, and a cross-sectional estimation that compares these outcomes' levels after the shock was realized.

Difference-in-differences design. The main results are obtained through a difference-in-differences approach that compares changes in outcome y _it, a municipal-level measure of local state capacity or land redistribution, in differentially shocked places (e.g., Topalova Reference Topalova2010; Dube and Vargas Reference Dube and Vargas2013):

(1) $$\begin{equation} lny_{it}=\alpha + \beta _1 ln\bar{V}_{it} + \lambda _t \times X_{i,1930} + \lambda _t + \gamma _i + \varepsilon _{it,} \end{equation}$$

where $\bar{V}_{it}$ is a measure of commodity potential in time t for municipio i, X _{i, 1930} is a vector of time invariant pre-shock controls that are interacted with the time fixed effect λ_t, γ_i are municipio fixed effects, and ε_it is an error term. Commodity potential, the main variable of interest, aggregates the value of the potential production of a municipio’s crops at a given point in time, relative to the rest of the country. It is defined as

$$\begin{equation*} \bar{V}_{it}=\sum _{g=1}^G \frac{\bar{P}_{gt} \times Suitability_{ig}}{Avg.\;Suitability_{g}}, \end{equation*}$$

where $\bar{P}_{gt}$ is the average price of crop g in time t ∈ {1920s, 1930s}, Suitability _ig is a municipio-specific crop suitability measure (in metric tonnes) determined by agroclimatic conditions, and $Avg.\;Suitability_{g}=\frac{1}{N}\sum _{i=1}^N Suitability_{ig}$ is a national average. $\bar{V}_{it}$ captures the relative availability of resources for the landed elite, who produce commodities for the market.Footnote ¹³ Parameter β₁ in the equation can quantify the effect of the price shock through the channel that the theory suggests: the temporary weakening of the landed elite, associated with a decline in commodity prices, leads to the expropriation of land, along with an increase in the incentives to invest in local capacity by local rulers.

With exogenous controls, the key identification assumption is $E(\varepsilon _{it} | \lambda _t, \gamma _i, ln\bar{V}_{it})=0$. That is, a shocked municipio should have to maintain the same difference to an unexposed control municipio had it not been shocked by the Great Depression. Figure 4 and Table C.5 (in Section C.3 of the Online Appendix) present evidence that supports the plausibility of this assumption, using pre-Depression data on state capacity and land redistribution.

Cross-sectional design. For a cross-section one decade after the shock, I estimate the following equation:Footnote ¹⁴

(2) $$\begin{equation} y_{i,1940}=\alpha +\beta _0 ln\bar{V}^{1920s}_i + \beta _1 S^{1920s-30s}_i + \delta X_i + \varepsilon _i, \end{equation}$$

where y _{i, 1940} is a municipio-level outcome of local state capacity or land redistribution for 1940. $\bar{V}^{1920s}_i$ is the initial commodity potential (using the 1920–1929 price average) prior to the price shocks, S ^{1920s − 30s}_i is the percentage shock to the commodity potential attributable to the Great Depression, X _i is a vector of covariates, and ε_i is an error term.

The percentage shock to commodity potential is given by

$$\begin{equation*} S^{1920s-30s}_i =\frac{\bar{V}^{1930s}_i - \bar{V}^{1920s}_i}{\bar{V}^{1920s}_i} \times 100. \end{equation*}$$

Here, the identification assumption is $E(\epsilon _i| \bar{V}^{1920s}_i, S^{1920s-30s}_i,X_i)=0$, which is a relatively stringent condition. It requires, for instance, that unobservables related to state capacity/land redistribution not be correlated with the initial crop suitability in each municipio, nor with its change induced by international price fluctuations. However, to the extent that one can regard these variables as exogenous—they are constructed from geographic features, along with internationally determined prices—the estimate for β₁ should be similar to the difference-in-differences approach.

For long-term outcomes, I estimate a similar version to Equation (2):

(3) $$\begin{equation} y_{i,2000}=\alpha +\beta _1 ln\bar{V}^{1920s}_i + \beta _2 S^{1920s-30s}_i + \delta X_i + \varepsilon _i, \end{equation}$$

where y _{i, 2000} is a present-day outcome.

MEASURES AND DATA

To characterize capacity, I draw from the notion of infrastructural power—“the capacity to actually penetrate society and to implement logistically political decisions” (Mann Reference Mann1986, 170). I operationalize this concept by assessing the available resources of the state to implement policy, focusing on a key component: the presence of government officials in municipios (e.g., Acemoglu, Garcia-Jimeno and Robinson Reference Acemoglu, Garcia-Jimeno and Robinson2015; Soifer Reference Soifer2008). Almost any governmental action requires implementing agents. Bureaucrats gather information and enforce decisions, and keeping them on the payroll requires resources. For this reason, the number of bureaucrats at any given time (absent outside funds) reflects the realization of costly investments in expanding the capacity of the local government.

Adopting this resource-based approach to capacity, however, requires distinguishing between the ability of state actors to implement political decisions—such as extracting information and resources or maintaining order—and their incentives to do so. In the context of post-revolutionary Mexico, where constant government presence was concentrated in a few places and nonexistent in most of the territory, this challenge can be partially addressed. The decision to set up a minimum number of bureaucrats is a precondition to implementing subsequent policies; it is necessary for the operation of local government.

Archival evidence suggests that bureaucrats during the period indeed filled essential roles in local governments, and were unlikely appointed solely for patronage purposes.Footnote ¹⁵ Local expenditure budgets from the period indicate that policing and tax collecting positions were, respectively, the most common functions; together they account for almost half of all bureaucrats in the municipios with available information. The positions that follow are basic administration (mostly city hall members), the local judiciary, and municipio representatives in smaller towns. Other functions, including local prison keeping, sanitation, and market and cemetery administration, employ a negligible number of people. Together, these secondary functions account on average for less than 10% of public employees.Footnote ¹⁶

The total number of bureaucrats is available by municipio from the 1930 and 1940 population censuses, as reported by the respondent’s main occupation, and was entered from historical census reports for this project.Footnote ¹⁷ For 1940, but not 1930, these data are disaggregated in a way that allows to distinguish between federal, state, and municipal bureaucrats.

As a measure of the second main dependent variable, asset expropriation, I rely on land redistribution microdata at the agricultural unit (ejido) level from Sanderson (Reference Sanderson1984), which includes the outcome of petitions at the national level (including denied grants), date of redistribution, and basic characteristics of the landholding. I focus on redistributions that include land with access to water, the most valuable landholding type, and the productive core of large agricultural estates. Expropriation of marginal land would not have substantially affected hacendados’ resources, and thus their ability to resist local political leaders.Footnote ¹⁸

For long-term outcomes, a similar measure of capacity—number of bureaucrats—is available from the 2000 population census. I use this outcome to keep a consistent measure over time, though there are reasons to think that it might not be an ideal measure of capacity in present-day Mexico.Footnote ¹⁹ To complement this outcome, I also use local taxes as a proportion of local GDP.Footnote ²⁰ This measure partially reflects the realized extractive capacity of local governments, given past investments in capacity.

Crop suitability is available as total production capacity (ton/ha) for low-input-level rainfed crops (1961–1990), from FAO’s Global Agro-Ecological Zones. The suitability measure is calculated using information about local climate, soil types, slope, and rainfall. These data are spatially merged with municipios to obtain a local-level suitability measure.Footnote ²¹ Present-day municipio maps were individually modified for this project to follow 1940 borders when possible, using georeferenced period maps from the 1940 population census. Price data for a number of internationally traded crops come from the Global Financial Data repository.Footnote ²²

The theory presented above suggests that negative economic shocks affect state capacity through the weakening of a nonruling economic elite. As a measure of the presence of a nonruling economic elite—in this case, the older landowning elite—I use the existence of large estates in a municipio. I turn to the classification of settlements in the 1930 population census and restrict the analysis to places with at least one ranch, hacienda, or finca (estate).Footnote ²³ Unsurprisingly, large estates are present in a large majority of municipios in 1930. Of course, powerful landowners in post-revolutionary Mexico were not restricted to traditional hacendados, and during the revolutionary period some post-revolutionary local leaders became hacienda owners themselves;Footnote ²⁴ however, given available sources, hacienda presence meaningfully captures the presence of a traditional elite whose wealth was based primarily on the exploitation of agricultural commodities.

Additional covariates used in the analysis include total population, the proportion of people working in agriculture, the proportion of people living in cities, and dispersion of settlements (number of localities per hectare), from the 1930 and 1940 population censuses.