Local Economic Voting and the Commodities Boom

Economic voting has been widely accepted as a crucial factor explaining individual electoral behavior in comparative politics (Duch Reference Duch2001; Duch and Stevenson Reference Duch and Stevenson2008; Hellwig Reference Hellwig2010; Nadeau et al. Reference Nadeau, Lewis-Beck and Bélanger2013; Lewis-Beck and Stegmaier Reference Lewis-Beck and Ratto2013). In Latin America, individual-level studies have mostly associated electoral behavior with sociotropic evaluations of the national economy (Gélineau Reference Gélineau2007; Lewis-Beck and Ratto Reference Lewis-Beck and Ratto2013; Singer Reference Singer2013, Reference Singer2015).Footnote ¹ Similarly, studies focusing on aggregate data have looked only at national economic conditions to explain voting behavior across Latin American democracies (Murillo et al. Reference Murillo, Oliveros and Vaishnav2010; Remmer Reference Remmer2003, Murillo and Visconti Reference Murillo2017).

Less attention has been paid to local economic conditions and their marginal effect on electoral behavior, although the literature on political economy provides evidence of the geographic distribution of policy preferences as a result of local distributive outcomes (Ardanaz et al. Reference Ardanaz, Murillo and Pinto2013; Scheve and Slaughter Reference Scheve and Slaughter2001). As exceptions to this trend, Books and Prysby (Reference Books and Prysby1999) and Johnston et al. (Reference Johnston, Pattie, Dorling, MacAllister, Tunstall and Rossiter2000) focus on the impact of local socioeconomic variables on electoral behavior, using survey data. The former show that state-level unemployment rates have an effect on voting intentions in U.S. presidential elections through the evaluation of presidential economic performance. The latter find that both local- and ward-level unemployment rates, as well as evaluations of regional prosperity, shape voting decisions in parliamentary elections in England and Wales. Drawing on aggregate data in Argentina, Porto and Lodola (Reference Porto and Lodola2013) find that agribusiness’s contributions to the municipal GDP, in conjunction with federal fiscal policies, affect the electoral support for national legislators at the municipal level in the province of Buenos Aires.

Furthermore, the literature on economic voting assumes that voters are able to attribute the responsibility for economic performance to politicians, conditional on economic openness and institutional variables (Duch and Stevenson Reference Duch and Stevenson2008; Hellwig Reference Hellwig2014). Institutional factors facilitating the assignment of responsibilities include presidential regimes, a unified government, and presidential re-elections. All of these increase the incidence of economic voting and have been shown to influence national-level electoral behavior in the region (Hellwig and Samuels Reference Hellwig and Samuels2008; Gélineau Reference Gélineau2007; Ratto Reference Ratto2011; Tagina Reference Tagina2012).

However, the assumption about voters’ ability to discern politicians’ responsibility for economic outcomes is not widely accepted. Achen and Bartels (Reference Achen and Bartels2016) and Bermeo and Bartels (Reference Bermeo and Bartels2013) suggest that U.S. and European citizens punish governments for natural disasters and economic crises outside their control. Similarly, Campello and Zucco (Reference Campello and Zucco2016) argue that Latin American voters blame or reward incumbents for economic performance associated with the price of commodities and international interest rates, which are outside their control. They suggest, though, that information facilitates the attribution of responsibility over economic outcomes (Campello and Zucco Reference Campello and Zucco2017). This view is also supported by Rho and Tomz’s experimental evidence (Reference Rho and Tomz2017) showing that once trade distributive effects are explained, individual respondents are more likely to espouse self-serving material views on openness.

Whereas most analyses of policy responsibility focus on national economic performance, this study builds on the work of Porto and Lodola (Reference Porto and Lodola2013) on local economic conditions and federal policymaking. Like their study, this one focuses on a salient export tax with distinctive redistributive consequences for soybean-producing areas. That is, it compares the local wealth impact of increasing agricultural prices on electoral behavior before and after the countrywide protests that took place in 2008 against a hike in a federal export tax.

The interaction of local economic conditions with federal tax policies, we argue, affects the incumbent’s national electoral performance. We expect greater local wealth to have a positive marginal effect on the support for the incumbent, in accordance with economic voting. Yet salient national policies—especially tax policies—have a geographically differentiated impact, which tends to shape voting patterns in distinct ways across local jurisdictions. Voters come to realize the influence of these policies when information makes them aware of their redistributive consequences. Therefore, when these policies are widely perceived to erode the accumulation of local wealth, we expect a negative marginal effect on the vote for the national incumbent. Data on Argentine department-level soybean production at different electoral periods in time is used to identify these local economic effects.

The Agricultural Boom and Local Electoral Behavior in Argentina

Argentina is a suitable case for studying electoral behavior during the commodity boom because we can isolate the influence of local-level wealth while controlling for sudden shifts in national tax policies that gained salience during this period. By focusing on Argentina, we can map the varying geographic distribution of export wealth that mounting agricultural prices created across departments that are similar in numerous characteristics. Moreover, by examining this local wealth at different moments through time, we can capture changes in salience that featured an export tax designed to appropriate a share of this wealth.

Argentina is a major exporter of agricultural goods. At the onset of the new millennium, a threefold devaluation and an upward trend in agricultural prices improved the terms of trade, because two-thirds of country’s exports are primary products or manufactured goods of primary origin (INDEC n.d.). Argentina’s main export products were soybeans and soy byproducts. Soybean wealth spread in the traditional agricultural zones due to a capital-intensive model relying on the subcontracting of agriculture-related services, such as planting and harvesting, while increasing dependence on inputs like agrochemicals, machinery, and seeds (Barsky and Dávila Reference Barsky and Dávila2008; Bisang et al. Reference Bisang, Anlló and Campi2008; Gras Reference Gras2009). Given low labor mobility in Argentina, soybean wealth also boosted the demand for unrelated services in the towns serving these areas. These effects were heightened because they followed a period of economic malaise marked by low agricultural prices and an appreciated exchange rate in the 1990s. As Millán describes it in his ethnographic account of an Argentine rural town in 2007:

Therefore we expect the agricultural boom to generate local wealth effects that drove voters’ decisions beyond its broader effect through fiscal revenue or national macroeconomic growth, which was positive between 2003 and 2013 although almost null in 2009 and 2010 (World Bank, World Development Indicators).

In 2003, as soybean prices began to take off, Néstor Kirchner was elected president with 23 percent of the vote. Argentina had experienced a profound institutional and economic crisis in 2001–2, and Kirchner had run as one of three candidates of the Peronist Party. He was the runner-up, but the frontrunner withdrew, anticipating a defeat in the runoff. Once in power, Kirchner consolidated a Peronist-based coalition labeled Front for Victory (FPV), which won the 2005 midterm elections. In 2007, his wife, Cristina Fernández de Kirchner, was elected president with 45 percent of the vote. After her husband’s unexpected death in 2010 generated a massive wave of sympathy, she was re-elected the following year with 54 percent of the vote.

The Kirchners’ electoral success has been associated with the economic bonanza that soybeans triggered. The federal government collected taxes on exports, the revenues from which could be administered discretionally and were used to finance redistributive programs favoring core constituencies in urban areas and poor peripheral provinces (Richardson Reference Richardson2009; Mazzuca Reference Mazzuca2013). Export taxation became a vital source of fiscal revenue for the federal government and had a substantial impact on crops, especially soybeans (Freytes and O’Farrell Reference Freytes and O’Farrell2016; Richardson Reference Richardson2012).Footnote ² The highest tax rates were imposed on soybeans and their derivatives, which were not subject to export quotas because they were not consumed domestically (Barsky and Dávila Reference Barsky and Dávila2008; Richardson Reference Richardson2009). By contrast, meat, wheat, and maize were regulated by both price controls and export restrictions after 2005 in an effort to contain domestic food prices.

The soybean export tax rate, originally established in 2002, was raised under Néstor Kirchner in 2005 and in early 2007 and, after the election of Cristina Kirchner, again in late 2007. However, as growing prices accompanied these tax hikes, they did not seem to hinder the influx of wealth that export-oriented areas were accumulating—or at least its effect was not salient for local populations.

On March 11, 2008, at the beginning of the harvest season, the Ministry of Finance raised export tax rates and linked them to international prices on a sliding scale.Footnote ³ Agricultural producers reacted by launching massive protests, including lockouts and roadblocks, which caused food shortages. In response, progovernment labor unions and organized groups of the unemployed staged countermobilizations (Barsky and Dávila Reference Barsky and Dávila2008; Giarracca et al. Reference Giarracca, Teubal and Palmisano2008; Hora Reference Hora2010). These protests increased the public salience of the export tax and voters’ information about its net effects through both the actions of agricultural organizations and media coverage.

A coordinating committee organized these protests, bringing together the Rural Society of Argentina (SRA), which represents the largest and richest producers; the Rural Confederations of Argentina (CRA), which gathers midsized producers; the Agrarian Federation of Argentina (FAA), an association of small farmers; and the Confederation of Agricultural Cooperatives (CONINAGRO), representing the numerous cooperatives that dot the Argentine countryside, rather than producers themselves (Lattuada Reference Lattuada2006). As coordination overcame prior interorganizational conflicts, agricultural producers’ mobilization resulted in 329 departmental lockouts (Murillo and Mangonnet Reference Murillo and Mangonnet2016). The president of the Argentine Association of Research Regional Consortia (AACREA), the largest technical association in the sector, briefly explained the discontent of those days.

In describing their role as information providers to town dwellers, in the same interview, the Executive Committee of AACREA explained:

To mitigate the protests, President Cristina Kirchner, who controlled both chambers of Congress, introduced a bill seeking legislative ratification for the tax raise. The bill included promises of redistribution to poorer provinces of the hinterland. The bill passed the Chamber of Deputies despite defections from the government caucus but was defeated in the Senate by one vote, further raising the salience of the tax.

The media coverage amplified the salience of the tax’s redistributive consequences. For instance, the number of articles referring to export taxes obtained using the search engine of La Nación, a major national newspaper, peaked in 2008 at 1,440, when the annual average during Néstor and Cristina Kirchner’s administrations was 515 per year. In the 4 months between the announcement of the tax hike and the legislative vote, a third of total news coverage in the two largest newspapers, Clarín and La Nación, and in the progovernment Página 12 was dedicated to covering it—even when the former two expressed negative views of the export tax and the latter a positive evaluation (Zunino Reference Zunino2015, 229, 281). Indeed, the topic was on the front page of the three mentioned newspapers for all but five days of this period (Zunino Reference Zunino2015, 233).

Empirical Analysis

The analysis examines the economic determinants of the local context on presidential and legislative elections (only for the Chamber of Deputies) across Argentine departments for the 2007–15 period. The sample includes all departments of the country’s 23 provinces.Footnote ⁸ We exclude the City of Buenos Aires because it is a fully urbanized district with no agricultural production. Departments are the geographic units for computing most economic statistics in Argentina. They are also the lowest tier for which national-level electoral data are available.Footnote ⁹

Estimation

A visual inspection through exploratory spatial data analysis, parametric tests, and least squares diagnostics suggests the possibility of spatial dependence in our data.Footnote ¹⁸ Spatial dependence, or autocorrelation, is problematic, as it violates the underlying assumptions of linear regression. Spatial effects in the form of correlated residuals or high or low values of the dependent variable clustered in space (i.e., lack of independent observations) can yield biased and inefficient estimates (Anselin Reference Anselin2002; Anselin and Bera Reference Anselin and Bera1998). Consequently, we proceeded by estimating a spatial error model. Such a model incorporates spatial effects through an adjustment of the error term.Footnote ¹⁹ We estimated the following equation:Footnote ²⁰

${y_i} = \alpha + Harves{t_i}\beta + Lockout{s_i}\gamma + {X_i}\delta + {\varepsilon _i},$

where y_i is the vote share for the FPV or Cambiemos in a given department i; Harvest_i is the log of the number of soybean harvested hectares; Lockouts_i is the logged number of departmental lockouts occurring in 2008; X_i denotes a matrix of agricultural variables (agricultural capital and smallholding farms) and controls, which includes lagged vote shares and sociodemographic covariates; and ε_i is the error term.Footnote ²¹ More important, the error term has the following form:

${\varepsilon _i} = {W_i}{\varepsilon _i}\lambda + {\xi _{i,}}$

where ε_i W_i is the vector of error terms ε_i weighted by W_i, which is a matrix of spatial weights or “connectivity matrix” (Anselin and Bera 1988) specifying the degree of interdependence among observations; and ξ_i is the vector of uncorrelated, homoscedastic errors.Footnote ²²

Results

Tables 1, 2, and 3 present the results. All models test the effect of soybean harvests and the controls. We next proceed to add agricultural capital, small farms, and lockouts. Tables 1 and 2 display the models for the Chamber of Deputies.Footnote ²³ Table 3 shows the models for presidential elections.

Table 1

Legislative Vote and Local Wealth in Argentina, 2007–2009

^* p < 0.10

^** p < 0.05

^*** p < 0.01

Table 2

Legislative Vote and Local Wealth in Argentina, 2011–2015

^* p < 0.10

^** p < 0.05

^*** p < 0.01

Table 3

Presidential Vote and Local Wealth in Argentina, 2011–2015

^* p<0.10

^** p < 0.05

^*** p < 0.01

In accordance with hypothesis 1, harvested soybeans increased the support for the FPV in the 2007 legislative election, suggesting a wealth effect independent of tax policy. We focus on model 2 (table 1). A one-percent increase in the number of soybean-harvested hectares is correlated with an extremely small increase in the average support for the FPV, at nearly zero. However, there are substantive statistical effects as percentage changes in soybean harvests get larger. For example, a percentage change from the mean to its maximum value in the 2007 season—from Guatraché in La Pampa to General López in Santa Fe—is associated with a 1.5 percent increase in vote share for the FPV.

As expected, the relationship between soybean harvests and FPV’s electoral performance is negative and statistically significant at the conventional levels in most of our models on the 2009–2013 elections (tables 1, 2, and 3), as stated in hypothesis 2. We interpret the results of model 4 in table 1 for the 2009 midterm election. A percentage change from the mean to the maximum of the harvested soybeans in the 2009 season—from Capitán Sarmiento in Buenos Aires to General López in Santa Fe—is correlated with an average decrease of 2.13 percent in the share of valid votes for FPV candidates to the Chamber of Deputies, other things being equal. Similarly, a percentage change from the median—Patagones, in Buenos Aires—to the maximum value is correlated with an average decrease of 5.8 percent in the share of FPV’s valid votes.

The marginal effects of harvested soybeans on FPV’s vote shares in the 2011 and 2013 elections are smaller than those of the 2009 election. While in the expected direction, the coefficients that model 1 and 2 (table 2) and model 2 (table 3) yield for the 2011 legislative and presidential election, respectively, do not reach statistical significance. The exception is model 1 (table 3) on the presidential election, but without controlling for lockouts, capital, and property size. In the 2013 midterm election (model 4, table 2), a percentage change from the mean to the maximum of the 2013 harvest season is correlated with an average decrease of 1.5 percent in the share of valid votes for the FPV. As theorized, tax policy effects tended to decline in the post-2009 elections.

Though it is negative, we find no statistically significant relationship between the 2008 lockouts and FPV’s vote share in 2009 (model 4, table 1). Neither do we find significant effects in subsequent elections. We do find statistically significant effects with regard to agricultural capital and smallholding farms according to our expectations, though only in 2009 and 2011, respectively.

As predicted in hypothesis 3, there is a positive and statistically significant relationship (at the 1 percent level) between harvested soybeans and Cambiemos’s presidential vote share (models 5 and 6, table 3). We interpret the coefficient in model 6. A percentage change from the mean to the maximum of harvested soybeans in the 2015 season—from San Antonio de Areco in Buenos Aires to Río Cuarto in Córdoba—is correlated with an average increase of 1.5 percent in the share of valid votes for Mauricio Macri, other things being equal. Similarly, a change from a value around the median—Pocho, Córdoba—to the maximum level of harvested soybeans is correlated with a 2.7 percent average increase. These magnitudes should be viewed in the context of a highly competitive national election that ended up yielding a surprising, unforeseen result at the polls.

In the 2015 election, we also find a positive, significant effect at the .1 level for 2008 lockouts. A percent increase from its mean to maximum value—Puán, in Buenos Aires—is associated with an average increase of 4 percent in Cambiemos’s vote share. Substantively, this quantity shows that the resentment that the 2008 conflict induced in the sector might have contributed to prospective voting for Macri’s candidacy rather than retrospective voting against the FPV in the 2009–2013 elections. Control variables, such as agricultural capital and education, also have a positive and statistically significant effect on the vote share for Cambiemos. On the contrary, the coefficient for poverty is negative and significant.

In short, our findings seem to support our main three hypotheses regarding the effect of local wealth on electoral behavior. That is, soybean wealth had a positive marginal effect on incumbent support in 2007—unaffected by fiscal policymaking—and a negative effect as national export taxes were internalized in 2009 and subsequent elections. Furthermore, as expected, we find a substantive positive effect on support for Macri in 2015. Such electoral support had significant implications.

The Aftermath of Macri’s Election

Macri’s administration made a turnaround that favored the agriculture sector. On his third day in office, President Macri abolished trade restrictions and all export taxes except for soybeans, which were scheduled to decline 5 percent annually, thereby providing immediate incentives for agricultural investments (Infobae 2016). The decision was announced in a town considered to be the capital of soybeans, with a symbolic effect, as the CRA’s president noted: “The president, 72 hours after taking office, went to Pergamino, to a ‘corral,’ to announce the end of export taxes and restrictions. He fulfilled his campaign promise, and we put the items on the agenda” (Chiesa 2015).

Furthermore, Macri’s high-rank appointments in agriculture represented a dramatic shift from his predecessor. The Ministry of Agriculture was established in 2009, having previously operated as a secretariat in the Ministry of the Economy. Under the Kirchner administration, the appointees’ professional background was either political (with no connections to the sector whatsoever) or technical, especially those who came from the National Institute of Agricultural Research (INTA). By contrast, Macri filled these positions with former top officials from agricultural organizations, including SRA, CRA, and AACREA.

We analyze the appointments to minister, chief of staff, secretary, and undersecretary by tracking their previous careers.Footnote ²⁴ We grouped them into six categories.Footnote ²⁵ As figure 1 shows, only 1 of the 28 appointed officials during the Kirchner terms had previous experience in the aforementioned rural organizations. Conversely, 12 of 29 appointees under Macri had previously occupied important positions in at least one agricultural organization. Most remarkably, Macri appointed Ricardo Buryaile, a former CRA vice president and agrodiputado, as minister of agroindustry—as the ministry was renamed—and Ricardo Negri, from AACREA, as his secretary of agriculture (equivalent to deputy minister). In personal interviews, both these officials averred that Macri’s victory signaled a dramatic shift toward agricultural interests.

Figure 1

Officials Appointed to the Ministries of Agriculture and Agroindustry, 2009–2017

After a cabinet reshuffling in 2017, Macri replaced Buryaile with Luis Miguel Etchevehere, then president of the SRA, further confirming the tendency to appoint representatives of agricultural organizations to this ministry; other members of this organization were also appointed in conjunction with Etchevehere. Indeed, Canelo and Castellani (Reference Canelo and Castellani2017, 23–27) report that agroindustry is the ministry in Macri’s cabinet with the greatest number of high-ranking officials coming from business organizations and the private sector. Representatives of agricultural organizations welcomed this increase in their policymaking influence and recognized that it reduced their incentives to protest, just as the literature on labor unions’ political exchanges predicted (Pizzorno Reference Pizzorno1978). The following quotations from leaders of the SRA, CRA, and FAA illustrate this perception.

Local-level representatives have a similar perspective, as we heard when participating in an assembly of approximately one hundred delegates from local agricultural associations in 2016. There was a consensus among these delegates about demanding specific policies without resorting to protests that could hurt the government. Local-level representatives stressed the need to be cautious without neglecting the demands of producers suffering economic hardship, emphasizing the importance of the success of this strategy. This was a recurring theme. Many delegates stressed the need to abandon protests as a mechanism to draw the government’s attention, and reported local-level discussions in which producers asked their representatives for prudence in the relationship with the national government, given their recently gained access to policymaking.

Thus, not only did local economic conditions drive the FPV’s electoral performance from 2007 to 2015, but they also favored the victory of a conservative president, who subsequently increased the sector’s policy influence. Macri’s party, which was born in the City of Buenos Aires (Vommaro et al. Reference Vommaro, Morresi and Bellotti2015), took advantage of those local conditions to gain support in the agriculturally rich areas in 2015. Local economic conditions thereby contributed to party building. Once in office, Macri cemented this relationship with traditional agricultural regions while expanding it across the poorer hinterland (Murillo Reference Murillo2017).

The connection between economic interests and political organizations in Argentina, which favored the building of a successful center-right coalition, contrasts with the experience of Bolivia. Eaton (Reference Eaton2016, Reference Eaton2017) describes how opposition to Evo Morales’s populist presidency was built around the agricultural producers of Santa Cruz, including soybean farmers, who were afraid of land reform and resentful of export quotas. These economic actors financed rival political elites, leading to an increasingly polarized confrontation with the Morales government, also around 2008. However, Eaton argues that Morales’s concessions to these sectors were crucial in demobilizing them, thereby depriving political elites of their financial support for building a successful conservative party in Bolivia.

Conclusions

This research about the relationship between departmental economic conditions and electoral behavior in Argentina complements a literature that focuses on the impact of the commodities boom on incumbent support (Campello and Zucco Reference Campello and Zucco2016; Mazzuca Reference Mazzuca2013; Richardson Reference Richardson2009). By looking at this effect at different points in time, we are able to distinguish the marginal impact of agricultural wealth and the internalization of tax policies at the department level. The marginal effect of high agricultural prices favored the incumbent coalition in 2007. Yet it became negative once widespread protests raised the salience of redistribution through an export tax, even when the rate of that tax remained unchanged, thanks to the agricultural producers’ mobilization. These findings highlight both the effect of local conditions and the ability of voters to identify policy effects on their wellbeing. We expect future studies to analyze further the characteristics that define local contexts that could be relevant in explaining national voting patterns.

We have also shown the electoral effect of export-oriented agricultural areas on Macri’s support in 2015. In return, Macri fulfilled his campaign promise to end export restrictions and appointed representatives of agricultural interests. Indeed, when he could not deliver the scheduled 5 percent reduction in soybean taxes, due to a fiscal deficit in 2016, he found patience on their side.

Although it is too early to evaluate this strategy, Macri seems to be leading not just the first electorally successful center-right political party in Argentina but also the first one connected to agricultural interests.Footnote ²⁶ This modern rural vote is based on a model that expanded the demand for local services. As a result, it is associated not only with agricultural production and related services but more generally with nonagricultural services, since local populations indirectly reaped the benefits of agricultural growth. The relationship between modern agriculture and a democratic electoral right is a political novelty in Argentina. The durability of this alliance remains to be evaluated.