Abstract
What forces shape the behavior of incumbent legislators in a federation? Do subnational political elites (particularly governors) influence legislators’ decisions on national policies? Do legislators’ own political backgrounds and office ambitions motivate their actions in the chamber? We address these questions by estimating the causal effect of gubernatorial influence and individual political careerism on the voting behavior of regionally based legislators (i.e., senators) in Argentina, where electoral and candidate nomination rules provide little room for individualistic behavior in Parliament. Taking advantage of roll call voting data, we calculate the distance between each senator and her or his national party leader in the chamber. We document evidence that, on average, senators from governors’ parties systematically break party unity. This effect grows as gubernatorial power increases. Moreover, we find that senators with successful local level political careers are also more likely to behave autonomously from their national party leaders. Finally, we show that the impact of senators’ office ambitions on their voting patterns is surprisingly small.
Introduction
What forces shape the behavior of incumbent legislators in a federation? Do subnational (state) factors influence legislators’ decisions about national policies? Do legislators’ own political backgrounds and office ambitions motivate their actions in the chamber? We address these questions by estimating the causal effect of gubernatorial influence and individual political careerism in shaping the behavior of senators in Argentina – a country where considerable authority resides in subnational territorial units (provinces), and both party-centered electoral rules and candidate nomination mechanisms provide little room for individualistic behavior in Parliament. 1
We thank Claudia Avellaneda, Alejandro Bonvecchi, Carlos Gervasoni, Juan Negri, Patricia Otero Felipe, Aníbal Pérez-Liñán, and Javier Zelaznik for their helpful suggestions and comments.
There is a growing interest among comparative analysts of federal regimes in understanding whether powerful subnational political actors (particularly state governors) are able to affect the voting behavior of national legislators (Ames 2001; Desposato 2003, 2004; Carey and Reinhart 2004; Jones and Huang 2005; Cheibub, Figuereido, and Limongi 2009; Rosas and Langston 2011; Cantú and Desposato 2012). Because governors normally enjoy prominent political positions and control substantial economic resources, it is reasoned that federal representatives may be keen to follow governors’ bidding in the chamber and thus behave autonomously from their national party leaders, especially when their interests are in conflict. The underlying assumption of these studies is that gubernatorial effects are more likely to be found in countries where the electoral system and party nomination procedures place political careers in the hands of state party leaders rather than the national leadership and individual candidates.
Another strand of research on legislative behavior focuses on legislators’ individual-level attributes, such as prior political experience and office ambition. For example, research shows that legislators seeking reelection to the US Congress tend to behave in a more disciplined fashion – and so vote along national party lines – than those who aspire to higher elective office (Van Der Silk and Pernacciaro 1979; Hibbing 1986; Herrick and Moore 1993; Francis and Kenny 1996; Herrick 2001). Moreover, comparative studies that focus on legislators’ political backgrounds indicate that representatives with personal vote-earning attributes (e.g., strong local ties and individual support bases) are less dependent on national parties for career progression. Therefore, these legislators are more likely to be individualistic and thus to break party unity in floor votes (Carey 1997; Mainwaring and Pérez-Liñán 1997; Owens 2003; Morgenstern 2004; Sieberer 2006; Tavits 2010a, 2010b). No systematic research, however, has yet integrated these different contextual and individual sources of legislative behavior into an empirically testable model.
This article is designed to fill this gap in the literature. We study the potential impact of gubernatorial pressures and individual political careers on the voting behavior of regionally based legislators (i.e., upper house representatives) by calculating their ideal points on a single dimension. Based on information from roll call votes, these ideal points provide an indicator of the distance between each senator and her or his national party leader in the chamber. A senator that frequently votes with or against the party leader is considered to be “closer to” or “further from,” respectively, the party leader in that dimension. In examining the influence of governors over senatorial voting behavior, we assess the contextual power of subnational interests over national policy making – a relevant topic in the study of federalism. In analyzing the impact of political careerism, we evaluate the role of individual motivations and the extent of personal-versus-party voting – a critical issue that has been raised in the literature on electoral systems and representational styles.
Contrary to the growing and sophisticated body of literature on the lower chambers in Latin America, research on the upper chambers is underdeveloped. 2 This absence is striking for several reasons. First, bicameralism has a long political tradition in the region and exists in almost half the continent's current presidential democracies. Second, Latin American senates are far from mere revisionary chambers as they typically enjoy extensive constitutional prerogatives and exclusive lawmaking powers in certain areas. Third, an institutional mixture of presidentialism, federalism, strong bicameralism, and high-level legislative careerism is only found in the United States. This feature makes the upper chambers in Latin America a promising laboratory to examine the logic of party unity and defections in contexts of multiple principals (i.e., the country's president and state governors), where interests may at times come into conflict (Carey 2009).
Notable exceptions are Londregan (2000), Ingall and Crisp (2001), Crisp and Ingall (2002), Crisp and Desposato (2004), Langston (2006), Llanos and Sánchez (2006), and Hiroi and Neiva (2013).
Our analysis focuses on Argentina, which is an ideal case for examining the role of gubernatorial effects and political careerism on national legislative activity due to its particular institutions. The highly decentralized nature of the country's federalism makes governors influential actors that dominate provincial politics and can block significant public policies at the national level. Political career opportunities, moreover, are essentially decided at the provincial level given Argentina's party-centered electoral rules and provincial party organizations’ control of candidate nomination mechanisms. It is thus reasonable to expect that governors, who are frequently the de jure or de facto presidents of their provincial parties, will influence the voting behavior of their copartisan legislators. In addition, incumbent senators in the period we studied varied largely in terms of political background and office ambition. This variation allowed us to explore whether different types of political careers lead to patterns of distinctive behavior in the chamber by affecting the degree to which senators are loyal to their governors or national party leaders. In the field of comparative politics, most institution-centered studies on legislative behavior have exclusively focused on examining whether legislators elected under different electoral systems show different voting patterns. Our research contributes to this debate by emphasizing the possibility that legislators behave rather differently under the same institutional structure (Mainwaring and Pérez-Liñán 1997).
Our empirical analysis indicates, first, that average governors systematically affect the behavior of their copartisan senators in the chamber. Furthermore, we find evidence that the likelihood of senators voting against their national party leaders in the chamber (i.e., voting further from national party leaders’ ideal points) increases as the power of their copartisan governors increases. Second, senators’ political backgrounds also affect their likelihood of voting along national party lines, regardless of their party affiliations. The statistical results show that senators with personal vote-earning attributes are more prone to break party unity during floor votes than are those without such attributes. Third, roll call voting estimations are less conclusive regarding the effect of political ambition on senatorial voting behavior. Contrary to our theoretical expectations, we find no empirical evidence that senators who aspire to higher office are more likely to break with their congressional party leaders. However, we document that senators with static ambition who belong to the provincial opposition tend to vote along national party lines, while statically ambitious senators who are affiliated with the governor's party did not exhibit a systematic pattern of voting behavior. We attribute this phenomenon to a competing-principals dilemma, whereby statically ambitious senators need to maintain good relations with both their copartisan governors and their national party leaders.
The rest of the article proceeds as follows: The second section expands upon the theoretical arguments that underlie our hypotheses that gubernatorial influence and political careerism affect senatorial behavior in Argentina. The third section describes our data and discusses the operationalization of relevant variables. The fourth section tests our model of senatorial voting behavior, which is an ideal-point model based on all published roll call votes covering three successive Senates between 2001 and 2007. This strategy yields individual-level ideal points that indicate how close each senator's voting behavior is to his or her respective national party leader's position in the upper chamber. The last section considers the potential generalizability of our results and how they contribute to the current debates on legislative behavior.
In this section we outline Argentina's decentralized political system and offer a series of testable hypotheses on how gubernatorial influence and individual careerism could potentially affect senatorial voting behavior.
There has been a recent growth of literature on the power and influence of state governors on Latin American legislatures. The empirical verification of gubernatorial effects on national legislative activity in the region, however, remains largely elusive. Previous analyses of roll call voting in the Brazilian Chamber of Deputies, for example, have uncovered no systematic evidence that governors affect federal legislators’ propensity to vote with their party's majority (Ames 2001). Likewise, Carey and Reinhardt (2004) documented no direct effect of alliance with governors on cohort voting unity, while Desposato (2003, 2004) and Cheibub, Figuereido and Limongi (2009) detected only weak gubernatorial influences on the behavior of both individual legislators and state delegations. Similarly, Jones and Hwang (2005) found that Argentine governors exercise no influence over the homogeneity of their party cohorts’ voting behavior in the Chamber of Deputies, a phenomenon the authors attribute to the willingness of provincial bosses to delegate control of their deputies to the national party leadership in exchange for future fiscal benefits. Nonetheless, Rosas and Langston (2011) found that Mexican governors whose terms end after those of their state copartisan legislators in the national assembly are able to substantially increase their voting unity. Using a different method and research design, Cantú and Desposato (2012) also revealed a substantial incidence of same-party governors on congruent voting behavior within party delegations in the Mexican legislature.
The central argument of this line of research is that gubernatorial influence over national legislative politics is ultimately possible to the extent that governors have some measure of control over legislators’ career advancement. This form of gubernatorial control is more likely to be found in countries where the electoral system allows voters to only vote for a party list and permits state party leaders to nominate candidates and manage campaign resources coveted by politicians for career progression.
In Argentina members of the Chamber of Deputies and the Senate are elected according to party-centered rules that give party leaderships control over politicians’ career paths. 3 All else being equal, this peculiarity should encourage federal representatives to adhere to the national party line in Parliament (Stratmann and Baur 2002; Morgenstern 2004; Sieberer 2006; Carey 2007). 4 However, candidate selection and the formation of party lists (i.e., who runs and in what position) are carried out at the provincial level by party delegations according to their own statutes. 5 The highly decentralized nature of Argentina's candidate nomination procedures thus magnifies the control that subnational – not national – party leaders exert over career-oriented politicians (Jones 1997, 2001, 2008; De Luca, Jones and Tula 2002; Jones, Saiegh, Spiller and Tommasi 2002; Jones and Hwang 2005; Lodola 2009). As almost every elective position (and an overwhelming majority of appointed posts) is determined within the provinces, legislators fearing the possibility of not being (re)nominated should be responsive to their provincial patrons when voting in Parliament.
Senators are elected for staggered six-year terms (with one-third of the seats renewed every two years) in provincial districts by a fixed majority-minority formula whereby two seats are allocated to the plurality party and one seat to the first runner-up. Federal deputies are elected for four-year staggered terms (half the chamber is renewed every two years) under a closed-list proportional representation system in province-wide multimember districts with relatively low magnitudes. In both cases, there are no legal restrictions on the number of consecutive terms one can serve.
In contrast, electoral systems that encourage candidates to cultivate a personal vote make parties less relevant for a legislator's political advancement, thus creating incentives for individualism during floor votes (Carey and Shugart 1995; Ames 2001).
Admittedly, formal candidate selection procedures vary across the major parties. Despite such variation, governors had the power to decide candidacies for both national and subnational legislatures in the period under study (Lodola 2010: 183–185).
Provincial governors are also able to affect the political careers of federal representatives – and, consequently, their voting behavior – because they retain control over campaign financing and enjoy political discretion over a vast amount of resources transferred from the central government. Although political parties receive public funds for campaigning, extant legislation requires 80 percent of all public funding to be transferred directly to provincial party delegations. The remaining money is retained by national parties to spend on their operative activities and national media appeals, not to support individual campaigns. Certainly, the public funds legally allocated to provincial party chapters represent a modest fraction of the money that individual candidates need to run competitive campaigns. 6 But incumbent governors have access to additional economic resources that can help politicians get elected. This is due to the country's federal fiscal system, which prioritizes gubernatorial – rather than presidential – discretion over the use of intergovernmental transfers (Wibbels 2005; Bonvecchi and Lodola 2011).
In Argentina the election of a senator generally entails more risk and more expense than that of a federal deputy. Because of higher district magnitudes, a federal deputy may win election to the lower chamber by concentrating on winning votes in a municipality or provincial region, whereas a senatorial candidate will usually need to campaign throughout the whole province.
In effect, most Argentine provinces only raise extremely low proportions of their budgets and depend heavily on revenues from the federal government to cover their annual outlays. The bulk of this money comes directly from a revenue-sharing mechanism and is transferred by statute with no strings attached. These resources are therefore legally considered part of the provinces’ budgets and are audited by the provincial legislatures, which are usually politically controlled by governors (Jones, Sanguinetti, and Tommassi 2000; González 2010). Other smaller transfers have specific purposes (typically for public infrastructure), but in practice the central government finds it very difficult to monitor or sanction the misuse of these funds (Bonvecchi and Lodola 2011). Finally, in addition to their own tax incomes and federal transfers, some provinces also collect sizeable royalties for the extraction of natural resources (e.g., oil, gas, and minerals) in their territories. In this context where a large quantity of resources is not subject to local taxation and there is a lack of accountability in how resources are spent, governors are able to allocate money and manpower to copartisans’ election campaigns, making up the difference between national party disbursements and the amount individual candidates need to successfully vie for election.
Because legislators are career-oriented and governors have the ability to influence their career paths, we should expect senators to feel more independent of their national party leaders but less so of their provincial (state) bosses. Even in the Argentine context of relatively disciplined parties (Mustapic 2000; Jones 2002), such independence may be reflected in legislators defecting from the party-line when voting (in this article, voting further from congressional party leaders’ ideal points).
This argument is grounded on the assumption that all governors are equally capable of successfully lobbying legislators from their provinces and so create divisions in party unity, especially when provincial interests are at odds with the interests of national party leaders. However, we argue that gubernatorial influences vary to a large degree across provinces for two reasons. First, the presence of a copartisan governor in a legislator's home province primarily affects the possibility of observing gubernatorial effects in voting behavior. Whereas a same-party governor is likely to support a legislator's career aspirations, a governor from a different party is extremely unlikely to do so. Second, as we discuss in the next section, the relative strengths of Argentine governors vary across provinces and over time according to their institutional, electoral, and partisan resources. Therefore, powerful governors should be more influential than weak ones. This discussion leads us to our first hypothesis:
Hypothesis 1: Senators from provinces ruled by copartisan governors are more likely to cast dissident votes against their national party leaders in the chamber than are other senators. This effect increases as gubernatorial power increases.
Besides the influence of provincial governors, legislators’ personal attributes may condition their parliamentary voting behavior and the incentives to defect on party line votes. Our other hypotheses suggest that different types of political careers lead to different incentives to vote along national party lines. This is because political careers distinctively affect the degree to which senators owe loyalty to their provincial and national leaders. One former Argentine senator with a long legislative career had the following to say:
I was a candidate on six occasions […]. Of course, I am indebted to the governor who did the list. But he also looked at the pools to see who the best candidates were. And my votes counted. So, I am not going to do stupid things here because someone tells me what to do [me lleva de las narices]. Before doing that, I am going to sit down and discuss it with the governor, the president or whoever else. 7
Interview with Oscar Lamberto, Buenos Aires, 8 November 2005.
“Political careerism,” as used in contemporary political science, is a slippery term. It typically encompasses the analysis of political recruitment (or how candidates are attracted by established elites or parties to compete for public office), pathways to power, and office ambition. In this article we refer to political careerism as separately denoting (a) the career paths that get politicians into the Senate (i.e., political experience) and (b) politicians’ office goals at the end of their senatorial terms (i.e., political ambition).
The link between political experience and legislative behavior has not received much scholarly attention. There are a few studies in US politics that associate legislators’ prior political experience with congressional organization and activity. For example, Matthews (1960) and Canon and Stewart (2009) found that former congressmen socialize more rapidly (i.e., receive more committee assignments, have more chances to reach leadership positions, and show a more active committee performance) than do former state governors. In his classic work, Fenno (1986) claimed that freshmen senators with prior legislative experience have shorter adjustment periods in the chamber than those without such experience. But the literature on legislative behavior in the US Congress focuses more on the ideological positioning of legislators than on their political backgrounds (McCarty, Poole, and Rosenthal 2001).
Better guidance for studying the role of political experience (our first indicator of careerism) on voting behavior is provided by a large and growing strand of comparative politics research that explores the effects of personal reputations on legislative party unity (Mainwaring and Pérez-Liñán 1997; Owens 2003; Morgenstern 2004; Sieberer 2006; Carey 2007; Tavits 2010a, 2010b). These studies contend that personal vote-earning attributes – such as strong local ties to a community or region and a successful local-level political career, especially in executive office – make legislators more independent from their congressional party leaders. We refer to them here as “local political bosses.” 8 Being powerful locally is a potential source of independence within the party, which can promote individualistic attitudes in Parliament for three interrelated reasons (Tavits 2010b). First, local notables from any political party have an independent (typically, both partisan and non-partisan) electoral support base. Because party leaders know that locally vocal candidates attract votes that benefit the party in general, they have few incentives to punish those who vote against the party line. Second, local bosses have significant influence (if not direct control) over the execution of public funds and tentacle-like political machines, which allows them to build relations with a variety of organized groups and distribute material benefits and privileges to their supporters. These broad personal networks naturally reduce dependence on party patronage for campaigning and winning votes. Third, political bosses are less dependent on party leadership for nomination and have more career options (political or otherwise) outside their parties than do politicians without local ties. For both reasons, local notables face much lower career costs in the event that they defect on party line votes.
Jones and Hwang (2005) also regard a small group of politicians or a single individual who dominates parties at the provincial level as a provincial party boss. According to them, Argentine deputies are the bosses’ subordinates, because the bosses may control the deputies’ political careers.
Therefore, our second hypothesis posits the following:
Hypothesis 2: Local political bosses occupying a seat in the Senate are more likely to vote against their national party leaders in the upper chamber than are other senators.
Although prior political experience in highly ranked local office may be a source of independent activity in Parliament, most theories on legislative behavior identify legislators’ office ambitions (our second indicator of political careerism) as the key drivers of their policy preferences and consequent voting patterns. A relevant theoretical question thus arises: what types of political ambitions generate incentives for more individualistic behavior during floor votes? We focus on two types of ambition (Schlesinger 1966): static (the desire to retain current office) and progressive (the pursuit of higher office).
Scholarly literature on American politics contends that statically ambitious legislators have stronger incentives to behave in accordance with their national party leaders because their power-building activities continue to be played out under constraints imposed by their comrades in the assembly. In contrast, legislators who aspire to higher office have weaker incentives to behave in a disciplined fashion because their political networks of activity are defined outside the legislature. In exploring roll call scores for US senators, Van Der Silk and Pernacciaro (1979) reported that senators seeking an elective position in the Senate (e.g., chairperson, floor leader, or whip) change their voting behavior to approximate the decisions of their party colleagues, while seekers of higher office move toward the electorate median. A joint study by Herrick and Moore (1993; see also Hibbing 1986) found that members of the US House of Representatives with ambitions beyond the chamber do not adhere to the voting behavior of their parties, unlike legislators with static and “intra-institutional” (i.e., leadership positions in the House) ambition, who adopt the typical voting patterns of their copartisans. Additionally, Herrick, Moore, and Hibbing (1994) and Rothenberg and Sanders (2000) revealed that departing legislators tend to change their ideological (partisan) position substantially more than their continuing colleagues, while Francis and Kenny (1996) showed that House of Representatives members running for higher office shift their legislative voting record away from their national parties (and closer to their state parties) as the election approaches.
This discussion suggests that progressively ambitious legislators should be more prone to behave autonomously from their national party leaders, which is our third hypothesis:
Hypothesis 3: Senators with progressive ambition are more likely to vote against their national party leaders in the upper chamber than are other senators.
It is less evident, however, whether Argentine senators with static ambition would behave as predicted by the aforementioned studies – that is, to vote along party lines. This is because US-based research on the ambition-behavior link draws heavily upon the assumption that constituencies control political careers (Miller and Stokes 1963). Yet, as we have discussed in detail, in Argentina provincial party leaders rather than individual candidates and voters manage the contours of political careerism. Incumbent legislators seeking reelection, therefore, face a competingprincipals dilemma. On the one hand, they need to maintain good relations with their party leaders in the chamber, who are chief arbitrators in the battle for legislative resources (i.e., committee, chairmanship, and party directorate assignments) and enjoy agenda control over bills (Jones 2002). 9 On the other hand, they also seek renomination, which largely depends on behaving in a manner showing commitment to their provincial party leaders. 10 Therefore, statically ambitious senators will only adhere to the national party line if the condition of guaranteed renomination is met. Unfortunately, we are unable to determine with precision when this occurs for each senator. We may expect, however, that senators with static ambition who are affiliated to the provincial opposition will be less exposed to pressures from competing principals than other senators. Because there are no incentives for provincial opposition senators to demonstrate loyalty to their governors (e.g., by reserving the right to dissent with the national party when governors and national leaders are in disagreement), we would expect to observe higher levels of national party unity among senators from the provincial opposition.
In the Argentine Senate, the allocation of committees and leadership positions is decided by the president (i.e., the vice-president of the country) in consultation with party leaders and based on the percentage of seats held by parties. Unlike in the United States, there is no seniority system.
Arguably, senators aspiring to higher office (i.e., the presidency or a governorship) do not face such a dilemma because they have an independent power base among the electorate that frees them from party nomination (De Luca 2008).
Hypothesis 4: Provincial opposition senators with static ambition are less likely to vote against their national party leaders in the upper chamber than are other senators.
In sum, our theory of legislative behavior builds on existing institutional explanations that emphasize the role of electoral systems and party nomination strategies, and on research on legislative political careerism. We argue that variations in federal senators’ voting behavior (i.e., their defection on national party line votes) may be explained by both contextual factors (i.e., the influence of powerful provincial governors) and individual factors (i.e., legislators’ prior political experience and office ambitions). Although other important work has looked at related questions in Argentina, we are the first to empirically examine state-level determinants and individual-level motivations of legislative behavior together. This set of factors better captures the complexity and causal heterogeneity of legislators’ decisions to vote in line with or against their national parties in Parliament.
Our empirical analysis is based on 1,020 roll call votes held in the Argentine Senate between 10 December 2001 and 5 December 2007. These data represent three full two-year legislative periods (2001–2003, 2003–2005, and 2005–2007). The periods under analysis cover the last year of President Fernando de la Rúa's (Alianza, 1999–2001) tenure, who resigned in the midst of a profound economic and political crisis; the tenure of interim President Eduardo Duhalde (Partido Justicialista [PJ] 2002–2003); and the tenure of President Néstor Kirchner (PJ, 2003–2007). Table 1 provides information on the number of roll call votes held in each legislative year (running from March to February). Following a change in Senate rules regarding the use of roll call votes in December 2002 and the introduction of the electronic voting system in February 2004, the number of recorded votes drastically increased. Previously, senators only had to cast roll call votes for the election of authorities, veto overrides, and impeachments. Most of the bills discussed on the floor were approved by a show of hands. Under the new rules, the floor has recorded votes on almost all bills with a few rare exceptions.
Number of Roll Call Votes in the Argentine Senate, 2001–2007
Number of Roll Call Votes in the Argentine Senate, 2001–2007
Source: Cámara de Senadores (various years).
We estimated ideal points of senators based on roll call data, which served as a summary of their voting behavior. The core assumption of ideal-point estimations is that a legislator's preference (i.e., ideal point) can be represented in a low-dimensional Euclidean space, and that the legislator's utility declines as the distance between her or his ideal point and a reference point increases. The use of ideal points to examine congressional activity is the norm in the study of the US legislature (e.g., Aldrich, Berger, and Rohde 2002; Poole and Rosenthal 2007) and has become increasingly popular in the study of Latin American legislatures (e.g., Morgenstern 2004; Jones and Hwang 2005; Alemán and Saiegh 2007; Cheibub, Figuereido and Limongi 2009; Rosas and Langston 2011; Cantú and Desposato 2012).
Our point of departure is a one-dimensional version of the Bayesian estimation procedure originally developed by Jackman and his colleagues (Jackman 2001; Clinton, Jackman, and Rivers 2004). 11 The decision to estimate a one dimensional model was based on the consensus that Argentine politics is strongly patterned by a government-opposition cleavage (Jones and Hwang 2005; Jones, Hwang, and Micozzi 2009). Under this interpretation, legislative voting separates senators into government and opposition camps based on their party affiliations. One end of that dimension is occupied by the PJ and the other by the Unión Cívica Radical (UCR). 12
Among the multiplicity of estimation methods typically used in political science (see Clinton, Jackman, and Rivers 2004), we rely on a Bayesian estimation model because our data comprise a relatively limited number of senators and roll call votes. This model uses Markov chain Monte Carlo (MCMC) methods, which generate large number of samples for inference.
During the period under analysis, these two parties achieved a noteworthy level of seats in the Senate. The PJ averaged 59.9 percent of the seats, while the UCR averaged 27.5 percent. The rest of the seats belonged to several provincial parties (9.5 percent) and small national parties (3.1 percent).
With a one-dimensional model, the utility function of each senator is expressed as:
where y* ij is a choice between a “yea” position (ζ j ) and a “nay” position (ψ j ) for each senator i on each bill j. 13 In line with the assumption of utility maximization, y ij = 1 if y* ij > 0, y ij = 0 otherwise. If we assume that the function is negative quadratic (i.e., a senator's utility is maximized when her or his ideal point xi completely overlaps either a “yea” or a “nay” position), utilities are expressed as:
Ideal-point estimations require assuming every vote as “yea” or “nay.” Other attitudes such as abstentions should be treated as missing. Because abstentions under the old Senate rules were considered de facto “nay” votes (see Cámara de Senadores, 2 October 1996: 356), we coded one abstention held in 2002 as such. According to the new internal rules, abstentions do not count for quorum purposes (Article 212). We therefore regarded 163 abstentions held between 2003 and 2007 as missing (Kikuchi 2012).
where η ij and υ Ij are errors. Assuming that η ij and υ Ij have a joint normal distribution, equation (1) for the utility differential y* ij can be expressed as a linear regression with the unobserved ideal points x i and unknown bill specific parameters β j and α j :
We assume ε ij ~ N (0, 1). x i is a (n × 1) matrix of ideal points, β j is a (1 × m) matrix of discrimination parameters, and α j is an m-vector of intercepts. Using a hierarchical probit model estimated by a Bayesian simulation, we obtain the parameters β j , α j , and x i .
Another assumption in this ideal-point estimation is that senators who serve for multiple terms have different ideal points in each biennial period. However, because ideal points denote the relative position of each legislator in a given legislative period, intertemporal comparisons of ideal points (e.g., comparing the ideal point of a legislator in 2001 with that of a legislator in 2007) are problematic if they are estimated separately. Analysts have developed two different techniques for intertemporal models: identifying “similar bills” and fixing the ideal points of “reference legislators.” The former procedure utilizes information about vote cut points of similar bills across time, assuming that intertemporal comparisons are possible if legislators in different periods vote on similar bills (Bailey 2007). The latter instead fixes some legislators’ ideal points for all the legislative periods so that they can be used as reference points (Treier 2011). We relied on this procedure because there are no “similar bills” in our roll call data. We thus created two imaginary senators named PJ Loyalist and UCR Loyalist – who always vote in line with their party leaders (or party majority in the leader's absence from a session) – and included them into the estimation.
We fixed the position of UCR Loyalist at −1.0 and PJ Loyalist at 1.0 to identify the one-dimensional model. 14 We also constrained the ideal points to have mean zero and standard deviation one across senators as a prior restriction (Clinton, Jackman, and Rivers 2004). 15 We then used Gibbs sampling to generate 300,000 samples from the joint posterior density of the parameters and obtained the summary statistics used for inference. The first 5,000 iterations were discarded, and then every thousandth sample was saved. The information on 788 roll call votes was omitted from the estimation because they were either unanimous or lopsided with fewer than 5 percent of the senators supporting the losing side.
We imposed this restriction at the final stage of estimation by “postprocessing” (Rivers 2003).
We used the pscl package developed by Jackman (2011) for this estimation.
After obtaining discrimination parameters for each legislative period, we were able to determine how many of them are distinguishable from zero. The discrimination parameters β j indicate “how change in xi translates into support for bill j” (Jackman 2001: 229). If a large number of these parameters in a one-dimensional model are distinguishable from zero, there may not be higher dimensions in the data. Our model consistently fits the data because the median percentage of roll call votes that simultaneously achieved discrimination is 88.4 percent. 16 This result indicates that the party dimension mainly underlay the observed policy space in the Argentine Senate between 2001 and 2007. Figure 1 displays senators’ mean ideal points in each legislative period and their 95 percent posterior confidence intervals for that single dimension.
Ideal Points for Argentine Senators, 2001–2007
The percentage of votes that achieved discrimination in each legislative period is 88.9 (16/18) in 2001–2003, 86 (129/150) in 2003–2005, and 93.8 (60/64) in 2005–2007.
Estimating ideal points based on senators’ roll call votes was the first step of our empirical analysis. We then computed our dependent variable, which is the distance (or absolute difference) between a senator's ideal point and that of her or his national party leader in each legislative period. Our assumption is that the ideal point of a congressional party leader represents the position held by her or his national party. Party leaders were excluded from the analysis because the distance between their position and that of their national parties is naturally zero. This exclusion reduced the number of senators to 196. 17 The value of the dependent variable (DISTANCE) ranges from 0 (eight senators from seven different provinces) to 1.70 (one senator from the Tierra del Fuego province), with a mean of 0.35 and a standard deviation of 0.40.
A total of 26 party leaders (8 in 2001–2003, and 9 in both 2003–2005 and 2005–2007) were excluded from the estimation. Although the PJ switched its congressional leadership from José Luis Gioja to Miguel Pichetto in December 2002, we used Pichetto as the PJ leader to calculate our dependent variable because he always voted along national party lines under Gioja's leadership.
The first independent variable (GOVERNOR) serves to examine whether gubernatorial pressures systematically lead copartisan senators who hail from the governor's province to behave autonomously from their national party leaders in the chamber (Hypothesis 1). We drew on a modified version of González's (2013) composite index of gubernatorial power. The index encompasses governors’ institutional, electoral, and partisan resources. 18 Institutional resources refer to gubernatorial tenure potential – that is, whether the incumbent governor is constitutionally allowed to run for reelection, and whether the governor's term outlasts those of her or his copartisan representatives in the Senate.
The original index does not include gubernatorial institutional resources.
Naturally, a governor who can be reelected and whose political life in the provincial executive office exceeds the term of a copartisan senator is more influential than her or his lame duck counterparts. This is simply because the governor can reward loyalty by securing jobs or (re)nomination for those copartisans who must leave their posts before the end of the governor's executive term (Rosas and Langston 2011). For each senator, we created two dummy variables: one indicating whether a senator came from a province where the governor was allowed to compete for reelection, 19 and another revealing whether a senator's legislative term was outlasted by the governor's. 20
Gubernatorial reelection rules have varied across Argentine provinces and within provinces over time (Calvo and Micozzi 2005; Lucardi and Almaraz 2013). Between 2001 and 2007, 20 out of 24 provinces allowed for gubernatorial reelection, with 16 of them limiting the incumbent governor to serve for two consecutive terms.
To compute differences in the expected length of the political careers of governors and senators, we considered the contrasting political time-horizons of the former (who are elected for four-year periods) and the latter (who are elected for six-year periods). Because the 1994 constitutional reform mandated the Senate to be completely renewed in 2001 thus introducing a new staggered calendar, only one-third of the senators elected that year served for their full six-year terms, while one-third served for two years (2001–2003); the remaining one-third served for four years (2001–2005).
Gubernatorial electoral resources are measured as the share of votes received by the governor's party in the last provincial executive election. Finally, partisan resources refer to both governors’ capacity to influence policy making in their districts (i.e., the share of seats held by the governors’ parties in their provincial legislatures, and whether they held majority control of their assemblies) and vertical copartisanship (i.e., whether a governor and the president belong to the same political party). The index of gubernatorial power is a composite measure of these shares and dummy variables, with the latter contributing 0.5 points to the index in order to balance the effect of each measure. The index thus ranges from a high value of 4 to a low of 0, with higher numbers implying greater gubernatorial power. In our data, the GOVERNOR variable ranges from 0.75 (Mendoza, 2003–2005) to 3.66 (San Luis, 2003–2005) with a mean of 2.23 and a standard deviation of 0.59.
The second set of independent variables serves to test whether senators’ personal attributes (i.e., their prior political experience and office ambitions) led to individualistic behavior in floor votes. Data collection on the background characteristics of senators and their observable office goals drew upon multiple sources – such as senators’ online biographies; rosters from the national government, provincial and municipal governments, and legislative bodies; official data on candidate nominations; and the Directorio Legislativo (CIPPEC 2002, 2009).
Hypothesis 2 implies that senators with strong local ties and autonomous electoral support bases should be more independent of their national party leadership in the chamber (and so vote further from their ideal points) than senators without such personal attributes. Because the Argentine Senate has a number of institutional and material advantages over the Chamber of Deputies, prominent politicians (e.g., former presidents and governors) seek to obtain senatorial seats. 21 In particular, the Senate offers longer terms, greater individual influence (veto power) over the legislative process due to its comparatively smaller size, and more economic resources to develop politically rewarding activities outside the chamber. 22 To test this hypothesis, we generated a dummy variable (POLITICAL BOSS) to ascertain whether a senator was the president or a provincial governor before becoming a member of the upper chamber (1 if yes, 0 otherwise). Of the senators in our sample, 10.7 percent possess such political experience. 23
Indeed, according to Lodola and Almaraz (2013), 23 percent of all individuals who occupied a governorship between 1983 and 2011 ran for the Senate at end of their executive mandates. Almost 94 percent of them did it successfully.
The Senate also enjoys exclusive competence to confirm or deny all major administrative and judicial (including members to the Supreme Court of Justice) appointments.
Arguably, former mayors can also be considered political bosses. Yet it is worth noting that contrary to other federal countries in the region, such as Brazil, Argentina's institutions make local governments politically dependent and financially weak actors (Lodola 2010: 92–97). Because provinces are constitutionally granted the exclusive right to determine both the scope and content of municipalities’ autonomy, only mayors from a few highly populated districts in the Buenos Aires metropolitan area are able to build and maintain their own political networks of electoral support. None of them, however, occupied a senatorial seat in the period under analysis.
Hypotheses 3 and 4 refer to whether senators’ progressive and static ambitions influence their voting behavior. The systematic examination of political ambition may be problematic because ambition is a psychological predisposition rather than an observable behavior (Hibbing 1986). To address this, we used a senator's decision to pursue a given office as a surrogate for her or his psychological predisposition, which is assumed to exist prior to the senator's legislative activity (Herrick and Moore 1993: 772). This allowed us to infer senators’ office ambitions from the posts (both nominated and elected) they actually occupied or vied for at the end of their terms (Herrick and Moore 1993). We coded senators who held or sought election to a higher office as having progressive ambition and those running for reelection as having static ambition. Certainly, this is an impure measure because we infer pure preferences from observable choices. 24 It may be the case, for example, that a senator desired to become governor but chose to compete for reelection knowing that her or his chances of winning the governorship were low. Alternatively, perhaps a senator wanted to remain in the chamber but was unable to be renominated. Hence, our approach is likely to underestimate the effect of political ambition as the senators in our sample may not have had the opportunity to fulfill their actual career goals. Nonetheless, we acknowledge the limitations of our coding scheme and have interpreted the statistical results accordingly.
Measuring ambition as “expressed desires” requires survey data that are not available.
To capture senators’ career aspirations, we thus created two dummy variables: the STATIC variable identifies whether senators ran for reelection at the end of their terms (1 if yes, 0 otherwise); the PROGRESSIVE variable, those who were vying for either the presidency or a governorship (1 if yes, 0 otherwise). Nearly 28 and 14 percent of all senators included in our sample pursued static and progressive ambitions respectively. 25
Moreover, almost 39 percent of the senators either decided to run for a lower office (federal deputy, mayor, provincial legislator, or municipal councilor) or to assume a nominated position in the executive branch of their home provinces. An additional 7 percent were nominated in a national bureaucratic post, and 3 percent returned to their provinces to assume a position linked to their party machines. Finally, close to 9 percent of the senators in our sample quit politics either through retirement or death.
For theoretical reasons explained in the previous section, under some circumstances we expect the effect of provincial governors and individual political careerism to be conditional on whether senators belong to the governor's party or to the local opposition. To test these propositions (Hypothesis 1 and Hypothesis 4), we included several multiplicative interaction terms with the COPARTISAN variable. This allowed us to determine whether a senator was a member of the governor's party (1 if yes, 0 otherwise). In our data set, 67.35 percent of the senators have a copartisan governor.
The model also incorporates some additional variables. First, we included a dummy variable to test for significant differences in voting behavior between senators representing metropolitan provinces (i.e., Ciudad de Buenos Aires, Buenos Aires, Córdoba, and Santa Fe) and those representing peripheral provinces. Given the increasing role that the latter have played in the maintenance of presidential legislative coalitions in Argentina (Gibson 1997; Gibson and Calvo 2000; Zelaznik 2011), we expect senators from peripheral provinces to be more cohesive – and therefore to vote closer to their national party leaders – than those from metropolitan districts. Second, party dummy variables were included because party-specific factors may influence the likelihood of defections on national party line votes. The UCR, SMALL NATIONAL PARTIES, and PROVINCIAL PARTIES variables indicate whether (1 if yes, 0 otherwise) a senator is affiliated with the UCR; small national parties such as Frente País Solidario, Partido Socialista, and Recrear; or provincial parties such as Movimiento Popular Neuquino, Partido Liberal, Renovador de Salta, Vecinalista Partido Nuevo, Cruzada Renovadora de San Juan, Fuerza Republicana, and Partido Nuevo. Senators who belong to the opposition may be more prone than pro-government senators to break party unity given that they have lower stakes in the policy-making process (Owens 2003; Tavits 2010a). Similarly, parties with more centralized nomination rules should be more coherent, while small parties with no real chance of winning due to defections should have smaller incentives to preserve unity in floor votes (Bowler, Farrell, and Katz 1999).
Table 2 displays the OLS estimates of gubernatorial influence and political career effects on the distance between senators’ ideal points and those of their national party leaders in the Argentine Senate between 2001 and 2007. Overall, these estimates provide strong support for the claim that both gubernatorial pressures and personal vote-earning attributes generate incentives to behave autonomously from congressional party leadership and, in turn, to break party unity in the chamber. The causal impact of political ambition, however, is somewhat elusive.
Determinants of Voting Distance between Senators and Their National Party Leaders
Determinants of Voting Distance between Senators and Their National Party Leaders
Note: Robust standard errors are in parentheses.
p < 0.05.
Source: Authors’ own compilation.
First, in accordance with Hypothesis 1, we find that, on average, governors influence their copartisan senators so as to defect on national party line votes, and that this effect increases as gubernatorial power increases. Note that the statistical model includes the GOVERNOR variable, the COPARTISAN variable, and the interaction term between them. As expected, the coefficient for GOVERNOR is not statistically significant at any reasonable level of confidence. This result indicates the lack of gubernatorial influence over senatorial behavior when COPARTISAN equals zero.
In other words, regardless of their institutional, electoral, and/or partisan resources, governors are not able to influence the voting behavior of opposition senators from their provinces. To discern whether gubernatorial power systematically affects party unity in the upper chamber by pressuring copartisan senators to break with their national party leaders, we calculated conditional coefficients and standard errors (see Table 3). 26 If senators and governors share party affiliation, the conditional coefficient for GOVERNOR (which measures the marginal effect of gubernatorial power over voting behavior) is positively signed – thus indicating a tendency of these senators to vote further from their congressional party leaders as gubernatorial power increases – and statistically significant at the 0.05 level; this is consistent with our theoretical expectations. 27 Specifically, a one point increase in the index of gubernatorial power when copartisanship equals one is associated with a 0.188 unit increase in the distance between a senator's ideal point and that of the senator's party leader in the chamber – almost one-half the standard deviation of our dependent variable.
Conditional Coefficients of Model 1
Note: Robust standard errors are in parentheses.
p<.05.
Source: Authors’ own compilation.
Notice that the coefficient for COPARTISAN captures the effect of extremely weak governors (i.e., gubernatorial power index or GOVERNOR equals zero) on the voting decisions of their copartisan senators. It has a positive sign but does not reach statistical significance.
The calculation of the conditional coefficient for GOVERNOR when copartisanship equals one is (-.049 + .237).
Second, our results also lend credence to Hypothesis 2, which states that regardless of party affiliation, local political bosses occupying a senatorial seat are consistently linked to defections from the party line. Indeed, the coefficient for the POLITICAL BOSS variable has the correct positive sign and is statistically significant at the 0.05 level of confidence. Those who have served as president or as a provincial governor have had the opportunity to build and maintain a strong territorial political structure. In turn, this makes these political bosses unfearful of breaking with their national party leaders. This is because national party leaders exercise little, if any, influence over political bosses’ career advancement. In fact, in most provinces politics is built around personalities, not around national level party politics. As a former PJ governor crudely explained: “The national party does not exist here. […] We are autonomous politicians and do politics with our own mechanisms and tools.” 28 Thus, provincial bosses have no strong incentives to abandon their personal interests for those of their national parties.
Interview with Vicente Joga, Formosa, 10 April 2006.
Third, the results are rather mixed with regard to our second indicator of individual political careerism – namely, senators’ political ambitions. For instance, Hypothesis 3 (progressive ambition makes incumbent senators more independent in Parliament) is not supported in the case of Argentina. Although the coefficient for the PROGRESSIVE AMBITION variable has the correct positive sign – thus suggesting that seeking higher office seems to increase the likelihood of party defections – it is statistically insignificant. It is possible that measuring political ambitions based on observable choices made by senators at the end of their mandates, as we do in this article, rather than as expressed desires is not a good indicator of careerism. Furthermore, it is also possible that some of the senators who aspire to become governors (especially those who belong to the provincial opposition) may choose a different strategy – for example, maintaining good relations with their national party leaders (by voting close to leaders’ ideal points) in order to get resources that would allow them to successfully challenge the gubernatorial incumbent.
We did, however, find empirical support for Hypotheses 4 (statically ambitious senators affiliated with the provincial opposition are less likely to break party unity than are other senators). Indeed, the negative and statistically significant coefficient for the STATIC AMBITION variable indicates that opposition senators (i.e., COPARTISAN equals zero) show a systematic tendency to vote in line with their national party leaders’ recommendations. Moreover, as shown in Table 3, the conditional coefficient that measures the marginal effect of static ambition when COPARTISAN equals 1 is negatively signed and statistically insignificant. Our explanation for this result refers to the competing-principals dilemma discussed above. Statically ambitious senators from the governor's party need to demonstrate a certain commitment to their national party leaders so as to receive resources and positions in the chamber. But they also need to show loyalty to their provincial party leaders (especially governors, as stated by Hypothesis 1) in order to obtain renomination. The need to please both their comrades in the chamber and their bosses at home lead them to develop an “erratic” pattern of voting behavior.
Finally, with regards to the effects of control variables, we found no systematic difference in voting behavior between senators who hail from metropolitan areas and those from peripheral provinces as literature on Argentina's political federalism would suggest. We did find, however, that senators from the UCR tend to vote in a less disciplined fashion than those from the PJ, which is our benchmark category. This may be related to the denationalization of the Argentine party system, which has particularly affected the UCR over the last decade (Torre 2003).
There is a growing interest in the comparative study of legislative behavior with regard to the issue of national representation versus local representation. Whereas prior research has mostly considered institutional effects, this article looks at both contextual and individual-level determinants of defecting from the national party line. We developed an idealpoint model that examines whether gubernatorial pressures and political careerism effects lead regionally based legislators (senators) to break with their national party leaders in the upper chamber. We empirically tested our model in the case of Argentina, where the institutional setting offers few incentives to develop individualistic behavior in Parliament. Drawing on roll call data for the 2001–2007 period, our analysis provides systematic evidence of the powerful role that provincial party elites (governors) exercise over the legislative process. In the direction predicted by our hypotheses, the results indicate that, on average, senators from the same party as a governor systematically break party unity by voting further from their congressional leaders, and that this effect increases as gubernatorial power increases. Moreover, provincial political bosses sitting in the Senate are more likely to vote against their congressional leaders. We found less conclusive evidence regarding the effect of political ambition on senatorial voting behavior. Contrary to previous studies on the US Congress, we found no empirical evidence that senators with progressive ambition are more likely to behave independently and so vote against the party line in Parliament. We did find, however, that opposition senators with static ambition tend to vote along party lines, while statically ambitious senators from a governor's party do not show a clear voting pattern – a phenomenon we attribute to competing pressures from national and provincial party leaders.
Our findings provide an important addition to the institution-centered approaches that are used to study party unity in comparative politics. These studies assume that a given institutional arrangement provides each legislator with the same structure of incentives and overlook individual-level variance in the incentives and abilities to defect on national party line votes. Our study also has implications for the literature on representational styles as it shows that incentives for individualism in Parliament do exist in countries with party-centered electoral institutions and candidate nomination mechanisms.
This article leaves a number of significant questions open for further research. First, with regard to the nature of legislators’ political autonomy from national and subnational party leaders, analysts should consider legislators’ autonomy not only as a function of the institutional setting but also as the result of politicians’ electoral security – an issue briefly mentioned in this article. Second, future research should consider other factors that might influence incentives to vote with or against national party leaders, such as committee membership (Sieberer 2006) and affiliation to political and social organizations. Third, researchers should consider the importance of non-institutional variables – critically, the characteristics of a bill – to investigate voting unity and the impact of subnational interests (Owens 2003; Cheibub, Figuereido, and Limongi 2009). Naturally, not all issues on which senators vote are of equal importance for provincial and national leaders. Ideally, one would use data on presidential and gubernatorial positions on each vote. But this information is simply unavailable. An indirect strategy would be to identify issues that directly affect provincial interests and/or where governors’ positions are likely to be at odds with that of the national government – tax and fiscal legislation are obvious candidates.