Legislative Voting in Argentina

According to our main argument, variation in legislators’ roll call votes—more specifically, in their recovered ideal points—should be systematically related, ceteris paribus, to their strategic electoral value as defectors. In addition to the scope conditions previously discussed, we also need to account in our empirical analysis for other contextual factors that affect legislative behavior in Argentina. Specifically, we need to consider the role of actors other than legislative party leaders, whose influence is likely to be reflected in patterns of legislative voting.

Earlier studies of voting behavior have demonstrated that the main dimension of conflict in the Argentine Congress is captured by the government-opposition divide (Jones and Hwang Reference Jones and Hwang2005; Jones et al. Reference Jones, Wonjae and Micozzi2009). There is ample evidence showing that this dimension properly accounts for the overwhelming majority of legislative votes; that is, a one-dimensional spatial model correctly distinguishes around 95 percent of roll call votes. This configuration is also consistent with the findings of Hix and Noury (2016), who examined 16 legislatures and found that government-opposition dynamics are the main drivers of voting behavior in most institutional contexts. Therefore, we focus on legislators’ placement along the latent government-versus-opposition dimension. Specifically, we examine whether the variables of interest identified in our argument move the predicted ideal points of government (or opposition) legislators toward (or against) the government’s position. Absent any relaxation of induced party discipline, a legislator’s alignment with the government or opposition parties should completely explain that legislator’s vote choices.

In addition to electoral incentives, the literature on Argentine legislative politics claims that governors are an additional source of influence over the behavior of national legislators (Gervasoni and Nazareno Reference Gervasoni and Nazareno2017). Governors in Argentina have control over party resources, including access to the ballot, which they could use to influence the composition of the province’s delegation in the national congress (De Luca et al. 2002). For example, according to Jones and Hwang (Reference Jones and Hwang2005, 272), given “superior resources and institutional advantages, a provincial boss who is a governor will on average be more powerful and thus have greater ability to affect the political future of deputies belonging to their provincial delegation, than a boss who does not control the governorship.” The evidence in Cherny et al. Reference Cherny, Figueroa and Scherlis2018 suggests that governors’ ability to nominate legislative candidates varies significantly by province and electoral cycle. Nevertheless, their study shows that most of the viable nominations belonging to the party controlling the provincial government correspond to candidates selected by governors.

Using data from an expert survey, Gervasoni and Nazareno (Reference Gervasoni and Nazareno2017) find support for the notion that Argentine governors influence the voting behavior of their party’s provincial delegation. However, empirical evidence derived from the actual behavior of deputies offers a different picture. For example, Jones and Hwang (Reference Jones and Hwang2005) examine roll call data from six legislative terms in the Argentine Chamber of Deputies and find not only a lack of provincially based effects on the main dimension of conflict but also no signs of gubernatorial effects on legislators’ voting behavior.

From our perspective, which centers on legislators’ opportunities to shirk, we expect that belonging to a provincial party delegation that has a copartisan in the governor’s office should move a legislator’s ideal point closer the federal government along the government-opposition dimension.^{Footnote 3} This expectation is grounded in studies on fiscal federalism and subnational politics in Argentina. According to this literature, governors are financially dependent on the largess of the federal government. As such, the power of the purse, which is in the hands of the federal government, can be used to pressure governors to side with the central government (Benton Reference Benton2009; González Reference González2012). The empirical record suggests that funds discretionally managed by the president are disbursed regularly in exchange for legislative support from provincial delegations (Gibson and Calvo Reference Gibson and Calvo2000).

Another reason that would make Argentine legislators, irrespective of political orientation and party, more likely to vote with the federal government is public employment. Compared to other Latin American countries, public employment in Argentina is high and is often used for electoral purposes (Calvo and Murillo Reference Calvo and Victoria Murillo2004; Gervasoni Reference Gervasoni2010). In several provinces, there are more public employees than private ones; and in most of these provinces, the salary of public employees is higher than the salary of employees in the private sector. Provincial and local governments depend heavily on the transfer of fiscal resources from the central government to sustain these expenditures. As a result, dependence on federal transfers is also likely to affect legislators’ incentives to side with the government in roll call votes.

These contextual factors notwithstanding, we also need to account for another potential determinant of legislators’ incentives to defect from the party line. As we have noted, in CLPR systems, the party leadership decides candidates’ position on the ballot before the elections. Such a placement is not trivial: it affects the electability of individual legislators in a district, the overall traits of the party’s congressional delegation, and the party’s electoral returns. Party leaders, however, care not only about winning elections but also about the cohesion of their legislative delegation and their ability to pass or block legislation.

In response to these tradeoffs, party leaders create lists with loyalists, experts, and sometimes popular outsiders (Serra Reference Serra2011; Galasso and Nannicini Reference Galasso and Tommaso2015). Buisseret et al. (Reference Buisseret, Folke, Prato and Rickne2019) model the optimal party list choice under CLPR, considering how the party leadership allocates candidates of heterogeneous quality across list ranks depending on the party’s electoral goals and competitive environment. If voters have partisan preferences but also prefer voting for more competent candidates, then parties should assign strong and popular candidates in their ballots’ topmost positions (Buisseret et al. Reference Buisseret, Folke, Prato and Rickne2019). This strategy comes with a nontrivial consequence: strong, popular, and competent candidates may also be less willing to be whipped by party leaders and more likely to want to retain an aura of autonomy. This effect and the others discussed in this section must be controlled for when testing the implications of our main argument.

Data and Estimation

We employ roll call data to examine legislators’ propensity to represent the views of their constituents on the plenary floor. To generate these estimates, we use the method based on Markov chain Monte Carlo (MCMC) simulation within a Bayesian framework proposed by Clinton et al. (Reference Clinton, Jackman and Rivers2004). This approach treats the unknown ideal points, bill parameters, and legislators’ utilities as random variables and conditions on the observed roll call data. The strategy is to impute values to these variables and estimate by regression the ideal points and bill parameters. The MCMC algorithm repeatedly performs these imputations and regressions and generates a large number of samples from the posterior density of the model parameters. Using this estimation procedure, we obtain summary statistics used for inference.

We consider voting decisions recorded in the Argentine Chamber of Deputies between February 28, 2008, and November 25, 2009, to match the coverage of the surveys used to place voters and legislators on a common ideological space.^{Footnote 7} A total of 251 recorded plenary votes were cast in this period. In most of these votes, the outcome was decided by a unanimous (87) or supermajority (159) decision. In the remaining 5 votes, the outcome was decided by a very thin majority.^{Footnote 8} To estimate legislators’ ideal points with and without defection opportunities, we consider two disjoint sets of roll call votes. The first one consists of the 159 nonunanimous lopsided votes. The second one includes the 5 closely contested votes.^{Footnote 9} Lopsided votes and short voting records pose no problems in the Bayesian approach (Clinton et al. Reference Clinton, Jackman and Rivers2004). For this reason, we use the MCMC scaling approach to analyze our roll call data.^{Footnote 10}

The left panel of figure 1 displays the ideal point of the average legislator in each of the main parties represented in the Argentine Chamber of Deputies, estimated using the sample of 159 lopsided votes. Dots are point estimates, and the spikes depict 95 percent confidence intervals. The dotted vertical line is centered at zero solely for illustrative purposes. From one extreme to the other, we can identify the government party and its main ally (the Frente para la Victoria and the Frente Cívico por Santiago); a couple of “swing” parties (Solidaridad e Igualdad SI-ARI, and the Partido Justicialista); and a bloc of opposition parties (conformed by the Socialist Party, PRO, the UCR, and the Coalición Cívica).

Figure 1.

Legislators’ Ideal Points and Voters’ Policy Preferences

This pattern confirms that the recovered ideal points correspond to legislators’ positions in regard to the government. The evidence in figure 1 also shows that for most parties (with the exception perhaps of the Socialist Party), the ideal points of their members are not identical. Instead, as the 95 percent confidence intervals indicate, intraparty differences in the estimated ideal points calculated using lopsided votes exist. With respect to voters’ ideology, the public opinion survey includes information about respondents’ partisanship. We use these data, in conjunction with our common-space scores, to estimate the ideological location of the average voter in each of the seven major parties with legislative representation.

The right panel of figure 1 shows the point estimates of the ideological location of each party’s representative voter (marked by dots) and their 95 percent confidence intervals (indicated by the spikes). The locations are arranged from left to right (with the Socialist Party on one end and the PRO on the opposite one). This alignment clearly differs from the configuration presented in the left panel of figure 1. The Coalición Cívica is a case in point. The ideological location of its representative voter is closer to the Frente para la Victoria than to the PRO; yet when it comes to casting votes in the legislature, its representative member tends to vote with the latter rather than the former.^{Footnote 11} More important, taken together, these different patterns provide prima facie evidence that Argentine legislators face a dilemma between voting along the government-opposition divide or according to their voters’ policy preferences.^{Footnote 12}

Moving on to our main variable of interest, Malapportionment, we use the distribution of seats per number of inhabitants to identify those electoral districts where voters are underor overrepresented. For example, based on the 2010 census data, the province of Tierra del Fuego had 131,661 inhabitants and 5 seats in the Chamber of Deputies. The population of the province of Salta was almost ten times larger (1,239,111 inhabitants), yet it only had 2 additional legislative seats. So there was one representative per 177,016 inhabitants in the latter province, whereas in the former, there was one representative per 26,332 inhabitants. Considering all of Argentina’s electoral districts, slightly more than half of the legislators in our sample represented 165,469 or fewer inhabitants, while the remaining 124 represented 171,468 inhabitants or more. We use this threshold to capture party leaders’ incentives to exploit malapportionment for electoral purposes. Specifically, the variable Overrepresented takes the value of 1when the population-to-seats ratio in a legislator’s district is 165,469 or less, and 0 otherwise.

In addition to these two measures, we consider other variables to control for the contextual factors that characterize the Argentine case. First, to capture legislators’ loyalty toward their party leadership, we differentiate legislators who belong to the federal government’s legislative party bloc from legislators who belong to various opposition blocs.^{Footnote 13} Given the government-opposition divide that characterizes Argentine legislative politics, we expect these allegiances to capture the degree of party influence on legislators’ voting behavior. More specifically, in the absence of any other sources of influence on legislators’ roll call voting behavior, this variable should capture all the variation in our estimated ideal points.

Regarding gubernatorial effects, most governors in the period studied were aligned with the federal government. As such, a significant number of government legislators (72 percent) were part of the incumbent party at both the national and provincial levels. The country, however, also had several opposition governors during this period (e.g., Neuquén’s Jorge Sapag, Santa Fe’s Hermes Binner, Buenos Aires City’s Mauricio Macri, and Tierra del Fuego’s Fabiana Ríos). In this case, roughly 38 percent of the opposition legislators were part of the incumbent party at the provincial level. We use this information to create the variable Governor. This indicator takes the value of 1 if a legislator’s party controls the provincial government and 0 otherwise.

We also consider another reason why Argentine legislators, regardless of their political orientation, may vote with the federal government: dependence on federal transfers. To capture public employment in a legislator’s home constituency, the variable Public Employment measures the total number of public employees as a proportion of the number of heads of household in the municipality where this legislator resides.^{Footnote 14}

A final variable is used to control for legislators’ positions on their party’s list. Argentine electoral districts exhibit large differences in their magnitudes. As a result, a third place in the province of Buenos Aires ballot (out of 35 places) is not the same as a third place in the province of Chaco (where 3 or 4 candidates are elected, depending on the electoral calendar). To account for these differences, we created the variable Relative Position. It is calculated as

$$\left[ {1 - \left( {{{Position} \over {1 + {\rm{Ma}}{{\rm{x}}_{Position}}}}} \right)} \right],$$

where Position is a legislator’s place on the party ballot and Max _Position is the maximum value of Position in each district. According to this metric, the third-place candidate in Chaco receives a value of

$$\left[ {1 - \left( {{3 \over {1 + 4}}} \right)} \right] = 0.4,$$

which is equivalent to a candidate ranked 16th in Buenos Aires.^{Footnote 15}

Main Results

Our sample consists of 256 rather than 257 deputies because the Argentine Chamber of Deputies president seldom casts any votes (and therefore we cannot estimate an ideal point). To capture the effect of electoral incentives on legislators’ behavior, we estimated a series of models by ordinary least squares (OLS).^{Footnote 16} In our baseline specification, the dependent variable is a legislator’s ideal point, estimated using the lopsided votes. Next, we estimated the same model, but using the ideal points generated using close votes as the dependent variable.^{Footnote 17}

Table 1 shows the results of our OLS models. Negative coefficients indicate a progovernment effect of the covariates, while positive coefficients indicate a proopposition effect. The first model presents a “null” specification using the ideal points estimated from roll call records on lopsided votes as the dependent variable and, as the main covariates, the contextual variables that, according to the existing literature, should explain legislative behavior in Argentina (i.e., our controls).

Table 1.

Determinants of Legislators’ Voting Behavior

*** p < 0.01, ** p < 0.05, * p < 0.1

Robust standard errors in parentheses

The estimated coefficient associated with the variable Government confirms that for the most part, the government-opposition divide defines legislators’ voting behavior. The estimated coefficient of the variable Governor reveals that having a provincial governor from the same party also influences legislators’ voting behavior.

Moving from 0 (being in the opposition at the provincial level) to 1 (being aligned with the provincial governor) shifts a legislator’s position closer to the federal government. The results also show that public employment in a legislator’s home constituency impacts the legislator’s ideal point. The effect is statistically significant but substantively quite small. A one-standard-deviation increase in public employment corresponds to only a –0.06 decrease in the average legislator’s ideal point. The coefficient associated with the variable Relative Position is statistically indistinguishable from zero. Its interpretation, however, is not straightforward, as the dependent variable is measured in the government-opposition dimension. A similar specification that also includes an interaction between Relative Position and Government (not shown) suggests that the effect of party list placement on legislators’ ideal points is null in the case of both government and opposition legislators.^{Footnote 18}

We explicitly test hypothesis 1 in the second model of table 1, where we also use the ideal points estimated from lopsided votes as the dependent variable. Here, we examine how legislators’ electoral incentives affect their voting behavior. If the contextual factors considered in model 1 are completely determinative of legislators’ vote choices, then the coefficient associated with our key indicator of legislators’ responsiveness—the interaction between voters’ ideology and malapportionment—should be statistically indistinguishable from zero.

The results from model 2 are consistent with hypothesis 1. The effect of voters’ policy preferences on a legislator’s ideal point is statistically indistinguishable from zero in underrepresented provinces but positive and statistically significant in over-represented ones. Recall from the right panel of figure 1 that negative (positive) values of voters’ ideological position correspond to those located to the left (right) of the political spectrum. Therefore, a rightward (leftward) movement in the ideological position of voters belonging to the legislator’s party is associated with a corresponding shift in the legislator’s ideal point toward the opposition (government) camp. The marginal effect of voters’ ideology on a legislator’s recovered position is positive and statistically significant (coefficient 0.205, Z-score 3.91) in overrepresented districts and statistically indistinguishable from zero (coefficient 0.043, Z-score 1.34) in underrepresented ones.

These results provide evidence in support of our argument that legislators from overrepresented districts are more likely to accommodate the views of their party’s voters in their voting decisions than legislators from other districts.^{Footnote 19} Moreover, these findings not only conform to our theoretical expectations, but they are also consistent with other evidence. For example, the responses to a set of questions included in a contemporaneous survey conducted by the Universidad de Salamanca’s Parliamentary Elites of Latin America project (PELA, see Alcántara Reference Alcántara2018) indicate that legislators from malapportioned districts express higher levels of responsiveness to their constituents than legislators from other districts.^{Footnote 20}

A critical test of our main argument involves a comparison between lopsided and close votes. We undertake this task in model 3, which is similar to model 2 but with ideal points constructed using close votes as our dependent variable. The statistical results are consistent with our expectations; we cannot reject hypothesis 2. In the case of votes decided by a thin margin, the only coefficient that is statistically indistinguishable from zero at the 95 percent confidence level is the one associated with the variable Government. Compared to the “null” specification (model 1) and the strategic defection (model 2), the results of model 3 indicate that the party’s alignment along the government-opposition dimension fully explains the variation in its members’ vote choices.

Further examination of the five close votes in our sample provides additional support for our main argument. Three of these votes correspond to an extremely salient and politically divisive bill involving a tax program for agricultural exports. Its passage in the Chamber of Deputies was a key event in a bitter confrontation between agricultural producers and President Cristina Fernández de Kirchner. The conflict originated when the Ministry of Finance issued a decree raising taxes on agricultural exports and linking the tax rates to changes in international prices. The export tax rate hike generated an immediate response by rural producers, who implemented a series of lockouts, protests, and roadblocks. To crush the revolt, President Kirchner, whose party controlled both chambers of Congress, introduced a bill seeking legislative ratification for the tax hike. Despite the comfortable majorities enjoyed by the ruling party in both houses of Congress, the bill barely passed the Chamber of Deputies and was defeated in the Senate.

The other two votes correspond to bills that were discussed after the tumultuous legislative treatment of the agricultural tax bill. The first one took place on October 16, 2008, and involved article 10 of the 2009 budget proposed by President Kirchner. The article was highly controversial, as it delegated significant budgetary authority to the executive branch (so-called superpowers). The second vote was held on March 18, 2009, and the issue at stake was the timing of the forthcoming midterm elections. Fearing a defeat, President Kirchner sought to get legislative approval to hold the elections earlier than initially scheduled. The opposition interpreted the move as an attempt to hold a snap election, intended to tilt the electoral balance in the government’s favor. Thus, this vote was also highly contentious and was decided solely on the basis of the government-versus-opposition division.

Investigating the Mechanism: Party-led Defection?

The results in table 1 lend support to our main argument. Leaders tolerate strategic deviations from the party line when such defections have the potential to enhance their parties’ electoral fortunes. Ideological shirking, however, could also be the result of legislators’ unilateral actions. For example, some legislators may want to enhance their reputation before pursuing an executive office at the provincial level or seeking a senate seat. Others may want to strengthen their support among voters to improve their chances of a hostile takeover of the party’s leadership or to build a reputation with constituents before creating a new party. Furthermore, some legislators may shirk simply because they plan to retire.

As we have noted, recorded votes do not constitute a random sample of the universe of legislative decisions. As a result, ideal points recovered using roll call votes may not accurately reveal legislators’ individual preferences. To determine if legislators’ personal motivations, rather than partisan influence, drive their legislative behavior, we need to consider so-called position-taking actions instead of recorded votes.

As Alemán et al. (Reference Alemán, Calvo, Jones and Noah2009) note, cosponsorship data can be used to obtain a sincere measure of Argentine legislators’ pre-floor preferences. Following both Alemán et al. (Reference Alemán, Calvo, Jones and Noah2009) and Calvo (Reference Calvo2014), we use contemporaneous cosponsorship data to measure legislators’ policy preferences. The location of each legislator, as recovered by this estimate, represents a relatively inexpensive signal to voters, with few costs to the party and potentially large individual reputational benefits. From the perspective of the party leadership, if these positions do not affect the success of their party’s legislative agenda, they should not be subject to party discipline.

If our theory is correct, a comparison between legislative behavior in plenary votes and cosponsorship activity should allow us to distinguish between party-led defection and differences motivated by legislators’ preferences. Existing studies suggest that partisanship (Micozzi Reference Micozzi2014) and belonging to the government or the opposition parties (Alemán and Calvo Reference Alemán and Calvo2013) are strong predictors of joint bill drafting in Argentina. As a result, legislators’ position-taking efforts should resemble voting behavior in lopsided votes rather than in closely disputed ones. However, our theory is “agnostic” with respect to the effect of malapportionment on legislators’ behavior in the case of cosponsorship.

Models 4 and 5 in table 1 correspond to models 1 and 2 but are estimated using as the dependent variable legislators’ policy positions recovered from cosponsorship data (rather than lopsided votes). As before, negative coefficients indicate a progovernment effect of the covariate, while positive coefficients indicate a pro-opposition effect. The “null” specification (model 4) reveals that the government-opposition divide affects legislators’ position-taking behavior, which is consistent with previous studies of cosponsorship.

Regarding legislators’ electoral incentives, model 5 shows that the marginal effect of voters’ ideology on a legislator’s recovered position is indistinguishable from zero in overrepresented districts and negative and statistically significant in underrepresented provinces. In other words, the results reveal that in their position-taking behavior (as measured by cosponsoring bills), legislators do not appear to follow their voters’ ideological views. In contrast, voting on the plenary floor directly affects each party’s legislative strategy. Only a select group of legislators cater to their electoral base through their voting behavior.