Item-Vetoes and Override Attempts in Multiparty Legislatures

Transcription

1 Item-Vetoes and Override Attempts in Multiparty Legislatures Valeria Palanza Pontificia Universidad Católica de Chile Gisela Sin University of Illinois, Urbana-Champaign Abstract This paper analyzes the dynamics behind vetoes and veto overrides in the context of a multiparty legislature. The paper argues that a key difference introduced by item vetoes is that, used strategically, they allow the President to break up the coalition that enacted the law, making overrides more difficult. By deleting articles selectively, the President leaves enough distributive goods in the bill so that support for an override is not available. We find that total vetoes, which affect all legislators equally, are more likely to be overridden than partial vetoes. Second, and against the received wisdom that override attempts are more likely under divided Congress, we find that in multiparty legislatures the likelihood of override attempts is higher under plurality government. We analyze an original dataset on vetoes and overrides in Argentina that comprises the period We also analyze cases that illustrate the main arguments developed in the paper. Paper prepared for presentation at the 2012 Meeting of the International Political Science Association, Madrid. Earlier versions of this paper have been presented at the Primer Encuentro Internacional Legislativos en América Latina: Mirada Crítica y Agendas Pendientes, October 2011, Belo Horizonte, Brasil and at the 2012 Meeting of the MPSA in Chicago, USA, and the 2012 Meeting of the LASA in San Francisco, USA. We thank Jose Cheibub, Christian Grose, Andrés Mejía Acosta, Aníbal Pérez Liñán, Anthony Pezzola and Erica Moreno for extensive comments on previous versions of this paper. We also thank Ernesto Calvo, Daniel Chasquetti, Constanza Figueroa-Schibber, Eric Magar, and Juan Moraes for exchanging data and ideas. Bárbara Castillo, Nicolás Díaz, Felipe Monestier, Cristóbal Sandoval, Jamie Scalera, Esteban Torre, and Giancarlo Visconti provided valuable research assistance at different stages of this project. Valeria Palanza gratefully acknowledges funding by FONDECYT # and by Iniciativa Científica Milenio Proyecto NS100014, Min. de Economía, Fomento y Turismo de Chile. 1

2 I. Introduction On October 2001, two months before President Fernando De la Rúa resigned the presidency in the midst of public unrest, Congress enacted a law (law 25500) that modified a previously approved Budget Law so that each Senator and each Representative could grant fifty pensions (pensiones graciables) per year without restrictions on who could receive them. The bill was very simple, stating that each legislator, regardless of political party affiliation, would be able to give away the same number of pensions every year for the next ten years. Although originally this type of benefit had been instituted for individuals that had served the country in extraordinary circumstances, throughout time the use of the pensiones graciables by legislators had become highly arbitrary. The President, who argued that the pensions were used with clientelistic purposes, vetoed the law in full. However, a month later, legislators unanimously overrode the presidential veto in both chambers. On December 1991, Congress passed a bill (law 24057) that aimed to promote non-profit organizations that studied and helped citizens in the use of urban space, rural land, housing, and transportation. When the bill reached the President, Carlos Menem, he imposed an item veto that removed all the prerogatives Congress had granted and that established the organizations right to act in those issues not being attended to by the federal government, to receive preferential treatment when soliciting loans or subsidies, and to work as consultant for the federal government when the latter made decisions on the subject. The language left in the bill was vague, stating that the federal government would promote the creation of such organizations, but without any specifics as to how. Indeed, the partial veto was substantively a total veto, although legally it was not: the rest of the bill became law. In this case Congress did not override the veto. Understanding why the President chose a full veto in one case and a partial veto in the 2

3 other is part of an interesting puzzle that we claim also affects the likelihood of congressional overrides. Why did Congress override the President in the first example but not in the second? Under which conditions do legislators decide to override the President? In this paper we focus on the dynamics of veto overrides and its connections to total and line item vetoes. If presidential vetoes are, as suggested by Cameron (2000), rare events, successful overrides led by Congress are even more unusual. We pay attention to these events as they provide insight to the bargaining process that takes place between the branches during the legislative process. In this paper we approach the topic from a comparative perspective, that is, we take into account variations in political environments along with override arrangements that differ from those predominantly analyzed in the literature. As Alemán & Schwartz (2006) have shown, the veto prerogative varies considerably within Latin American separation-of-powers systems. We draw on empirical evidence from Argentina not only to understand the nature of veto bargaining in this specific case, but also to enable further theorization, fundamentally regarding the implications of (i) override attempts and (ii) the success of such attempts. Argentina s separation-of-powers system is composed of an executive that has a fouryear mandate and a bicameral legislature with staggered elections. Lower chamber representatives are elected through proportional representation under closed party lists of variable magnitude (from five seats to over thirty seats), which results in a multi-partisan distribution of seats in Congress. Each province elects three Senators through direct elections, 3

4 and seats are allocated by a 2/3 rule: the party with the most votes wins two seats, while the party with the second largest number of votes wins the third seat. 1 As in other Latin American countries, in Argentina the executive enjoys ample legislative prerogatives. Perhaps most importantly, the President is entrusted the constitutional prerogative to enact legislation through non-statutory means, i.e., decree authority. Other prerogatives include exclusive proposal power on key legislative areas such as the budget. In addition to these capacities, traditional veto power, as it is known in the United States, is extended to include the ability to veto parts of legislation approved by the legislature while the remaining content is enacted. Veto overrides need the approval of two-thirds of the House and the Senate membership, and are taken by a roll call (Schinelli 1996:413). It is puzzling that despite the concentration of prerogatives in executive hands, the high levels of party discipline, and the overall status that Presidents enjoy as leaders of their parties; vetoes occur, and furthermore, they are overridden. An important element in this puzzle is the distinction between total and item vetoes. 2 First, we find that Congress is more likely to override total vetoes. Given a veto, the fact that it nullified a bill in full is an important factor leading to an override. In this article we suggest that the line-item veto prerogative has the capacity to undo the coalition that initially supported the bill in each chamber. We claim that by using the line item veto, the President not only eliminates the items that are far removed from his own preferences, but also keeps enough 1 Before the Constitutional reform of 1994, Senators were elected by the provincial legislature, two per district. 2 We refer to item vetoes alternately and equivalently as item vetoes and partial vetoes. Te former terminology derives from US politics, whereas the latter is more commonly used in comparative politics, especially in Latin America. See Alemán and Schwartz (2006) for more variations in terminology. 4

5 particularistic goods in the bill that a sufficient group of legislators will prefer not to override the presidential veto. 3 In this article we also suggest that, contrary to findings regarding veto enactment, varieties of divided government matter for successful overrides. 4 Periods during which the President s party enjoys but a plurality in Congress are a strong predictor of successful overrides. We argue that during these periods, when the governing party does not have a majority in at least one of the chambers (despite being the most numerous party), it uses overrides as an instrument to gain credibility in its commitments with other parties. Because the governing party knows that in the future they will need the support of additional political parties to reach a majority, overriding presidential vetoes helps the party maintain credibility within the working coalition. For issues that are of great importance to the extra legislators, the plurality party in Congress is better off by overriding the presidential veto. We analyze an original dataset that comprises every law enacted in Argentina from December 1983 until December 2007 and present quantitative results. Additionally, in support of the causal mechanism identified, we present four case studies, which suggest that, indeed, partial vetoes break up coalitions by differentiating particularistic goods. The paper proceeds as follows: in the next section we introduce the state of the literature. In Section III we provide evidence in support of our argument, specify and explain several decisions that we made regarding our variables of analysis, and present the patterns that emerge from the Argentine setting. In Section IV we test a few preliminary hypotheses and reach conclusions, which we summarize in the concluding section. 3 In particular, at least one third of the membership of one of the chambers will not have incentives to override the presidential veto. 4 We follow Calvo (2011) in using two categories of divided government; we explain this further in section III. 5

6 II. Vetoes and Bargaining Between the Branches Strategic approaches to the study of vetoes, developed with the US case in mind, have produced a substantial understanding of the role of vetoes in separation-of-powers systems. However, most of these contributions are rendered null when brought to the test in institutional contexts that differ from the case of the United States. In what follows we review the contributions and introduce some of the complexities brought in by analyzing the effects of vetoes in comparative perspective. Early models that were used to interpret vetoes assumed complete information (Romer and Rosenthal 1978), and predicted that in equilibrium vetoes (and overrides, for that matter) should never occur. Later developments relaxed the complete information assumption by introducing uncertainty about the President s preferences, i.e., which kind of bills she would accept or reject. Those models found that when the preferences of Congress and the President are sufficiently divergent, and as a result Congress is uncertain over what the President will find acceptable, vetoes do occur. Cameron s (2000) sequential veto bargaining model, in which Congress updates its beliefs about the President s preferences after the first veto-bargaining episode, places emphasis on veto chains and the policy concessions attained through chains. Cameron finds that the President may veto a bill she prefers to the status quo because it helps her build a reputation that will extract concessions from Congress in a second iteration of the process. Matthews (1989) focused on the effect of veto threats, arguing that the President uses them as a tool to reveal information about her preferences. He finds that a President may issue a veto threat to obtain a broader concession from Congress even though she would accept Congress's ideal point. Groseclose and McCarty (2000) develop a blame game in which they 6

7 include an electorate that is uncertain about the President s preferences. With the electorate as an audience, Congress may present the President with a bill he will necessarily veto, and this reveals to the electorate the true preferences of the President. One of their main findings is that vetoes tend to decrease the popularity of the President. McCarty (1997) examines the effects of presidential honeymoon periods, and argues that the President has strong incentives to veto bills during the initial months of his term in order to establish a reputation as an extreme. Because Congress anticipates this possibility, it passes bills closer to the President s ideal point during the honeymoon period. As a result, McCarty claims that over the course of a presidential term, the accommodation of Congress to the President s preferences should decline. These contributions permeate this article, and we acknowledge them here. Once we abandon the well-known features of the US setting, institutional variations among cases make analysis more complex, and predictions taken from the US setting provide little guidance. Among comparativists, Shugart and Carey (1992); Carey and Shugart (1998); and Mainwaring and Shugart (1997) pointed out the relevance of the veto prerogative, especially given the leverage it provides presidents in the legislative process. Yet while the work of Carey and Shugart set the research agenda in many aspects of the institutional life of new democracies, students of separation of powers systems delayed focusing on the analysis of vetoes. Perhaps the most expansive analysis of the veto prerogative in different settings is that by Alemán and Schwartz (2006), which undertakes the analysis of veto prerogatives in comparative perspective. Presidential veto powers are, in their words, richer, more varied, and more regionally distinctive than hitherto appreciated. They classify the patterns observed in eighteen Latin American countries into four game forms while distinguishing consequential details. Of those game forms, several represent contexts where presidents have line-item or item-reduction 7

8 vetoes, which expand presidential prerogatives by allowing the president to revise versions passed by Congress. This is the case in fifteen countries in Latin America: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela. 5 This argument is also central in Tsebelis and Alemán (2005) who claim that item vetoes and the possibility to make amendatory observations empower presidents by enabling them to choose the final form of any bill. Recent work by Indridason (2011), who also places emphasis on the line-item veto, does a fine job of bringing together important issues in legislative politics to assess commonly held expectations regarding vetoes. He advances a model of veto bargaining over public goods and pork barrel spending in the presence of credit claiming incentives. Indridason s theoretical findings, i.e., (a) that the item veto will reduce pork barrel spending, and (b) that the gains attained from the reduction of pork may be offset by greater incentives to attach riders to legislative proposals, provide theoretical guidance even as they call for empirical testing. A related literature has focused on the incidence of vetoes in specific Latin American countries: Uruguay (Buquet, Chasquetti and Moraes 1998; Magar and Moraes 2003, 2011), Argentina (Figueroa Schibber 2005; Magar 2001; Mustapic & Ferretti 1995), Chile (Aristegui 2010; Bravo & Castillo 2010; Magar 2001), and Brazil (Hidalgo 2010). These works provide a valuable comparative tool and show us how the veto prerogative is used in different institutional settings, and under what conditions vetoes seem to prosper. On average, presidents produce four vetoes per year in Chile, approximately four and a half vetoes in Uruguay, thirteen vetoes per year in Brazil and fourteen in Argentina. A common feature of these studies is their focus on 5 A fundamental difference between these fifteen countries is that while in Argentina, Brazil, Ecuador, and Uruguay presidential revisions of bills (by using the item veto) are not required to undergo a send-back vote in Congress, all other cases do have this requirement. 8

9 partisan affinities of the President and Congress and levels of significance of laws as explanatory variables. These studies provide a wealth of hypotheses, which we try to address in this article. Given our interest in multiparty presidential systems, we focus on the case of Argentina using a dataset that includes all bills passed by Congress and all presidential vetoes that took place in the last 25 years. For this reason prior work on this country s legislative process is of particular interest here. Where the literature on the Argentine legislative process has made most progress is possibly in regards to legislator preferences (Alemán, Calvo, Jones and Kaplan 2009; Calvo 2011; Jones and Hwang 2005), and the internal organization of Congress as it affects lawmaking (Jones, Saiegh, Spiller and Tommasi 2002; Calvo 2011). This paper draws from insights presented by these studies, while undertaking the topic of vetoes, which these studies have surprisingly ignored. In Argentina, work by Figueroa-Schibber (2005) focuses on how an institutional innovation, such as partial overrides, contributes to the success of bargaining between Congress and the President. By analyzing overrides during Menem s presidency, she finds that the number of overrides is related to the strength of the President: the stronger the President, the fewer the number of overrides. Earlier contributions include Molinelli (1985), a broad study on veto overrides in Argentina between , and Mustapic and Ferretti (1995), focused on the period immediately following the transition to democracy. Where the literature on Argentina has made most progress is perhaps in the analysis of legislator preferences (Jones and Hwang 2005, Alemán et al 2009), and the internal organization of Congress and the legislative process (Jones et al 2002, Calvo 2011). This paper draws importantly from insights presented by these studies. III. Vetoes & Overrides in Multi-Party Presidential Systems: The Case of Argentina 9

10 During the 24-year period between the return to democracy in Argentina, on December 1983, and the end of Nestor Kirchner s presidential term in December 2007, Congress enacted 3,258 laws. Of these laws, 362 (above 10%) faced presidential vetoes, and Congress overrode 42 of those vetoes. What explains these numbers? Under what conditions do congressional overrides occur? And under what conditions do congressional preferences prevail? The Argentine case presents a particularly interesting puzzle given the extent of legislative prerogatives in hands of the President. The President initiates legislation, and certain legislation can, in fact, only be proposed by the President. This ensures that a substantial portion of proposals will represent presidential preferences. Additionally, because the President can enact highly significant legislation by decree without ever consulting Congress, the existence of vetoes is further puzzling. During the time period analyzed in this paper Argentine Presidents enacted 694 decrees (DNUs) legislation which is equivalent to congressional statutes except for legislators say in the process. Due to the institutional layout and the nature of the interactions that take place during the legislative process in Argentina, one would expect that whenever legislation goes through Congress, the two branches have reached an agreement that will enable approval in relative proximity to presidential preferences. Why else would the President forego legislating unilaterally, by decree? Yet this does not seem to be necessarily the case. Palanza (2009) argues that Argentine Presidents legislate via Congress when they have no alternative, that is, when legislating via decree is not acceptable to legislators. 6 In other words, Presidents resort to 6 The reasons why it may be unacceptable are numerous, from constitutional limitations to pressure imposed by constituents and interest groups. 10

11 legislating through the congressional path when the President s and legislators preferences are such that Congress would see it in its interest to undo an executive decree. In sum, we expect confrontation between the branches in the legislative process, and we see vetoes and overrides as indicators of that tension. The case of Argentina provides an interesting vantage point given the presence of the item veto prerogative in addition to the total veto. Furthermore, the override procedure is also complex, as Congress can override parts of vetoes, proceeding to override vetoed articles one by one, overriding vetoed articles in full as well as only in part. In what follows we take a closer look at these interactions. III. 1. A Tale of Four Vetoes Our argument indicates that there are dynamics enabled by the partial veto that alter the nature of the negotiation that takes place between the President and Congress. In particular, we have claimed that by deleting a selection of articles in a law the President gains the ability to break up the enacting coalition, which we suspect was put together in part through selective incentives. We suspect that if it were not the case that item vetoes break up the enacting coalition, making overrides harder to achieve, we should observe that attempts to override total and item vetoes are equally distributed, which they are not. Interestingly, although the number of total and partial vetoes is equivalent (182 and 180 respectively), the number of override attempts for bills that were vetoed in full is higher. In what follows we present four cases that illustrate our argument. A. Coalition Cohesion around Broad Benefits The two examples we present next illustrate how some vetoes do not alter the enacting coalition, as they affect issues that are broadly and evenly spread in society. No specific set of legislators 11

12 that supported the bill during initial approval should a priori be compelled to change their position with regards to the bill after the veto. This implies that override attempts are more likely to occur than if a set of legislators changed their preferences regarding the bill. Case 1: Law 24305, National Program Against Foot and Mouth Disease At the end of 1994, Congress approved an important measure that aimed to eradicate the foot and mouth disease, as constant outbreaks of the disease had been problematical for the commercialization and access to export markets of Argentinean cattle. The law charged the federal agency in control of animal health with implementing a plan to stop the spreading of the disease through control measures such as quarantine of animals infected. Most importantly, it charged the agency with the responsibility to instrument a national vaccination plan aimed at eradicating the disease. Toward this objective, the law exempted from all taxes the import, manufacturing, and marketing of vaccines, as well as the actual vaccination work. This was the most important segment of the law, as the high price of vaccines was the main impediment to exterminate the disease. In January 1995, the President item vetoed the tax break. Because his veto affected the enacting coalition equally, Congress was able to override the presidential decision. Case 2: Law 25019, National Program for Wind and Solar Energy On September 1998, Congress enacted a law that promoted the development of wind farms by exempting them from any taxes during their first fifteen years of development. Congress sought to create monetary incentives that would promote investment on alternative sources of energy. However, when the law reached the President, item vetoed a fundamental portion of the bill: tax benefits. Without the tax benefits, it was highly unlikely that wind farms could prosper in their initial phases. The President argued that the tax cuts would have a negative 12

13 impact on federal government income, as well as on the energy market. Again, because the law was very specific, and the veto did not affect groups of legislators selectively, it was overridden by unanimity in the Senate and by a super-majority in the House. B. Coalition Splits around Selective Benefits Selective incentives, which reward a set of legislators while not affecting (or perhaps punishing) other legislators, have the power to break up enacting coalitions. The two cases analyzed next illustrate how legislators react to Presidential line item vetoes, attempting to override the President in those items that affect articles of broad impact and not attempting to override the veto affecting articles that provide selective incentives. Case 3: Law 24191, General Budget for 1993 President Menem presented this budget proposal to Congress in September After passing through the Budget Committee, it reached the House floor in December, where a few amendments were added. The Executive would veto four of those amendments: (i) Article 21, which granted funds from the Ministerio de Desarrollo Social (Ministry of Social Development) to the Ministerio de Obras Públicas, Salud, y Economía (Ministry of Public Works, Health and Economics). The funds were aimed at five social programs, and Congress specified that a Bicameral Congressional Committee would be in charge of overseeing the use of those funds; (ii) Article 26, which established payments of 5 million pesos to proprietors whose land was to be expropriated to allow for the construction of a highway along Northern access to Buenos Aires; (iii) Article 39, which granted 14 million pesos to citizen security projects with resources from the Fondo de Desarrollo Regional (Regional Development Fund); and (iv) Article 41, which established that the budget for the two chambers of Congress would be distributed proportionally to the number of seats in each. Not without debate, the Senate passed the Lower 13

14 Chamber s version later that month, and the President vetoed these four articles on December 30, The options facing partial vetoes are such that Congress might have chosen to do nothing, or to try to override the veto either entirely or in part. In March of 1993, the Lower Chamber tried to override the President s decision in only two of the four vetoed articles. The articles that faced an override attempt were articles 21 and 41 (which arguably affect benefits that are evenly spread in society), and those that did not were articles 26 (which in one case benefitted a small group of citizens in Buenos Aires), and 39 (which drew resources away from specific regions). This override attempt was successful, and although the Senate did not bring the override to a vote, we believe it provides evidence that specific changes introduced through partial vetoes effectively break up the enacting coalition. In this case, while the 2/3 required for an override came together in support of the two articles that affected them evenly, they did not in the case of the two articles that provided selective incentives, which quite possible could be the two articles that enabled initial approval of the bill. Case 4: Law 24938, General Budget for 1998 The approval of this bill provides evidence that Congress can garner support for overrides more easily when vetoes affect measures that benefit a large number of legislators. In the 1998 budget law, the President vetoed (i) a salary increase for elementary and high school teachers, (ii) the creation of a new government agency to support the development of small and medium size companies, (iii) granting funds to a hospital that treats children recovering from severe burns, (iv) funds for the recovery of the area surrounding the rivers Atuel and Diamante, (v) subsidies for a transportation system in Patagonia, (vi) funds for the National Institute for Water and the Environment, (vii) the extension of subsidies received by three provinces (Fondo de 14

15 Reparación Histórica) and the incorporation of eight additional provinces as recipients of such funds (the provinces already receiving the funds were Catamarca, San Juan and San Luis, the new ones were Santiago del Estero, Salta, Jujuy, Tucumán, Chaco, Misiones, Mendoza and Córdoba), and (viii) funds for the recovery and refilling of digging sites in the greater Buenos Aires area and for the Merchant Marine Academy. Out of all these measures vetoed by the President, Congress insisted on the one that extended and expanded the subsidies for eleven provinces. A month after this first override, Congress also overrode the veto concerning the funds for recovery of digging sites in Buenos Aires and the Marine Merchant Academy. While this latter override is harder to interpret, and is possibly part of a side negotiation (especially because it occurs later in time) the first override illustrates that the one aspect that legislators were able to override was precisely one in which benefits were widely spread. III.2. Hypotheses and Variables In this section we present our variables of analysis and a set of descriptive hypothesis that are in line with common hypotheses in the literature on vetoes and veto bargaining. The section is organized into three parts; in each we present independent variables. Before advancing, however, we present our dependent variable, congressional overrides. We operationalize overrides in two instances: (1) as override attempts, and (2) in terms of their success given that an attempt to override a veto was made. In the case of override attempts, we seek to uncover the conditions under which legislators try to override presidential vetoes, regardless of their success in the attempt. We believe there are systematic differences between vetoed laws that legislators try to override and those others, the majority, where no effort is made. In this case, our variable is a dichotomous variable that takes the value of 1 when there is 15

16 an override attempt. Regarding the success of override attempts, we seek to understand the conditions leading to success, that is, those that enable undoing presidential decisions. This is also a dichotomous variable where 1 represents a successful override. Over a total of 3,258 laws enacted during the period, only 362 were vetoed, which is the number of cases we observe in order to evaluate the conditions on the likelihood of override attempts. Likewise, because we can only evaluate the likelihood of successful overrides conditional on attempts to override, the number of observations in that case is 62. A. Identifying and Classifying-- the Bills We built a dataset including all the bills that became law between the return to democracy in 1983 and the end of Néstor Kirchner s presidency in 2007, and we identified all bills that received a veto. Additionally, we coded whether each veto was a total veto or an item veto, generating an indicator variable, Item Veto that takes the value 0 when a veto is total and 1 when it is partial. In many presidential systems, Presidents can propose bills directly, that is, without needing to go through a legislator as in the United States. In Argentina, the President can send bills directly to Congress. Therefore, one of our independent variables identifies this difference. Presidential Proposal is a dichotomous variable that takes the value 1 when the bill was proposed by the President and 0 when Congress proposed it. We expect Congress to be inclined to override vetoed laws that were initially proposed by Congress. B. Partisan Support in Congress In the identification of divided and unified government, we acknowledge the distinct characteristics of divided government in multi-party legislatures and follow Calvo (2011), who 16

17 distinguishes among three categories. The first, Unified Government, includes the periods in which the President s party has more than fifty percent of the seats in the House and the Senate. The second category, Plurality Government, identifies the periods in which the President s party is the largest party in the House and/or the Senate, but falls short of a majority in at least one chamber. 7 The third category, Divided Congress identifies the circumstances in which the largest parties in each chamber are different. In general, these variables seek to represent the distance between Congress and the President. Following the established literature on vetoes and overrides one expects that overrides would be more likely when the distance between the President and Congress is greatest, i.e. under divided government. Previous studies argue that honeymoon periods following the beginning of a new Administration should produce a smaller probability of vetoes and therefore, overrides (e.g. McCarty 1997). We created a dichotomous variable Honeymoon, which takes the value 1 during the initial 3 months of a President s term, the period typically considered the honeymoon phase to assess these hypotheses. C. Scoring Legislative Significance Scoring legislative significance is a complex and difficult task. Mayhew s (1991) outstanding effort, later extended by others (e.g., Cameron 2000, Clinton & Lapinski 2006), capitalizes the existence of sophisticated analyses and publications that are regrettably not available in the Latin American setting. Any student of Latin American legislative politics that has faced the necessity of scoring legislative significance has taken on this task aware of the 7 It could be that the party has a plurality in both chambers, or that it has a plurality in one chamber and a majority in the other. Plurality corresponds to the periods , , , and

18 many limitations one faces. Work by Monestier (2010) reviews the multiplicity of criteria used in recent studies for the purpose of establishing different levels of significance of legislation. While regretting the discordance of approaches, he suggests that much would be gained by a unifying criterion that would enable comparisons. Despite the difficulty, we agree with Cameron (2000) that legislative significance is hard to define but easy to recognize. In this paper we extend previous work in Molinelli et al (1999) by maintaining three levels of significance. This classification has a profound conceptual similarity with the one developed by Mayhew (1991), although the depth of methodological rigor in Mayhew is unfortunately not feasible in this context, for the reasons presented above. In Molinelli et al. s (1999) categorization, the three levels are: (1) landmark legislation, (2) important legislation, and (3) minor legislation. 8 We maintain the criterion established there and extend its application beyond 1998 to all laws enacted until December In what follows we describe each category. Table 1 provides examples of the laws included in each category, whereas Table 2 illustrates vetoes within each category. Level 1 Landmark legislation: This category includes only major pieces of legislation that establish or profoundly reform areas of legislation that have broad effects across the nation. Only extensive and profound reforms to these laws enter this category (narrow reforms are placed in the next category). A few treaties belong in this category given the scope and extension of their consequences, as do all budgetary laws (see Table 1 for more examples). Level 2 Important legislation: This category includes legislation that admittedly has varying degrees of importance, yet ultimately falls short of producing the broad effects of landmark 8 We acknowledge that there are gains to derive from an approach that allows for four categories, such as the one proposed by Cameron While we plan to advance in that direction, the analysis included here maintains the three categories in Molinelli et al (1999). 18

19 legislation. It includes legislation that reforms aspects of landmark legislation, and legislation that regulates important issues but is constrained in terms of scope. Table 1 provides examples. Level 3 Minor legislation: This category includes pieces of legislation that are mostly symbolic or that establish issues that are important only for a very small group of individuals, with minor consequences for the broader population. Laws creating monuments, transferring buildings between jurisdictions, deciding where international organizations will locate, enabling citizens to accept honors, among others, are typical examples of laws in this category. [ Table 1 about here ] [ Table 2 about here] Palanza and Sin (2011) find that landmark legislation is more likely to be vetoed, and in particular, it is likely that the veto will be partial as opposed to total. We expect levels of legislative significance to have a positive effect on the likelihood of override attempts, as legislative significance should motivate legislators to undo presidential vetoes. Regarding the effect of legislative significance on the success of override attempts, we expect it may have a marginal effect, yet the main effect should come from legislators ability to rebuild a coalition, considering most vetoes of landmark legislation are partial. IV. Describing the Patterns In this section we present the patterns uncovered by our large-n statistical analysis, as well as illustrate aspects of our argument and the causal mechanism we propose through case analyses. Yet before making sense of overrides, we first summarize the conditions under which vetoes occur. IV.1. The initial probability of a veto 19

20 Table 3 shows that of the statutes initially passed in Argentina during the period , only 362, or 11%, were vetoed. The number of bills vetoed, as a percentage of bills enacted during the period, seems large compared to US figures, where between 1945 and 1992 a total of 17,428 bills were enacted and only 434 of them were vetoed equivalent to 2.3% (Cameron 2000). [Table 3 about here] Palanza and Sin (2011) show that the level of significance of legislation, and whether the bill originates in Congress or the Presidency are forceful factors determining the likelihood that the President will veto legislation. More specifically, when considering Landmark legislation, the likelihood of vetoes increases considerably. Furthermore, when the President s party in Congress sustains merely a plurality of seats (as opposed to a majority), and when the piece of legislation is in the class of Landmark legislation, the President is consistently more likely to veto. Importantly, we also found that the President is much less likely to veto legislation she proposed. This suggests that the bargaining that takes place when enacting legislation is not limited to partisan dynamics, but holds instead an important institutional component: which institutional actor proposes the bill affects the likelihood of vetoes. We also uncover that when the President s party has merely a Plurality in Congress, the likelihood of vetoes increases. Plurality is the only variable that captures the distance in preferences between Congress and the President that consistently yields statistically significant results. These results are contrary to a long-held belief that divided Congress, that is, the situation in which the President holds less than a plurality of seats in at least one chamber, should affect the likelihood of vetoes positively. 20

21 Furthermore, the distinction between item and total veto reveals a much more complex situation. One of the most striking effects when we examine this difference is the likelihood that a piece of legislation receives an item veto relative to a total veto when the President introduces said legislation. We found that if the President introduces legislation and said legislation is vetoed, it will surely receive an item veto instead of a total veto. With respect to the three partisan configurations of government, we observe that the probability of total vetoes under each circumstance is almost the same. Regarding item vetoes, we note that even though the differences between the effects of the three scenarios are more noticeable, they are still small. These results are in line with what we find regarding the probabilities of vetoes in general. The variables that intend to capture the distance between President and Congress do not have a strong effect when predicting the probabilities of total and item vetoes. Another important result in Palanza and Sin (2011) is related to the different levels of legislative significance. The likelihood of line item vetoes vis à vis total vetoes increases significantly when legislation is landmark. We conjecture that the President vetoes items of landmark legislation that were part of the agreement in Congress among different party factions, which enabled passage, but with which the President disagrees. Indeed, the higher probability to item veto landmark legislation that the President herself proposes suggests higher levels of disagreement between the branches. IV.3. The initial probability of overrides We know from the literature that if vetoes are rare, overrides are rarer still. But although vetoes are rare, they are less so in Argentina. However, as shown in table 4, in Argentina, there were merely 62 override attempts during the 24-year period under study. This implies that Congress took steps to sustain initially 21

22 passed legislation in 17% of the cases. Forty-two overrides were successful during this period, which implies that 11.6% of all vetoes were successfully overridden, or more impressively, that 68% of all attempts succeeded. [Table 4 about here] As a benchmark, we know that during the period analyzed by Cameron, 50% of all vetoes faced override attempts in the United States. In comparison, Argentina s 17% seems meager. We claim that this difference stems from the existence of line item vetoes. As mentioned above, we argue that this happens precisely because the line item veto breaks up the coalition that supported the enactment of the bill. Congress attempted to override about 20% of all the bills that were vetoed in full, and succeeded most of the time, as 14% of those bills suffering total vetoes were successfully overridden. However, override attempts are considerably fewer when bills are item vetoed: Congress tried to override only 14% of executive decisions in the case of item vetoes. 9 Furthermore, in only 9% of bills vetoed partially, Congress overrode successfully at least a portion of the items comprised by the presidential veto. It is striking to consider that of the 362 vetoes, 320 were not overridden; yet this does not imply that all 320 laws died. 10 In fact, all partially vetoed laws that were not overridden were 9 Because the use of asymptotic analysis may be inappropriate, we also conducted Fisher's exact test to analyze the association between two events. With Fisher's exact test the significance of the deviation from a null hypothesis can be calculated exactly, rather than relying on an approximation that only becomes exact at the limit as the sample size grows to infinity. The drawback is that this test only tells us whether two events are related, but not how they are related. The p-value of the Fisher s exact test for the independent variable, total and item veto, is.06 when the treatment variable is a dichotomous variable where 1 means that the override attempt had received a total veto. 10 We acknowledge that congressional response is not limited to override attempts, and there is evidence that suggests that veto chains do occur in Argentina. Table 2 above includes two examples of total vetoes that arguably are part of a chain (although they were passed under 22

23 enacted as they were following the partial veto. This emphasizes why the existence of both total and item vetoes is such an important characteristic of the institutional setting in Argentina, one holding many lessons and implications. Out of the 320 laws that were vetoed and not overridden, only 49% had faced total vetoes (that is, 157 laws). The remaining 163 laws, having received vetoes affecting only parts of the legislative text, were enacted without further ado. Partial vetoes provide additional complexity. When the President can veto portions of laws, he may choose to veto every article of relevance, leaving just a few insignificant items in the statute. Partial vetoes may vary tremendously in terms of how much is deleted from the bill. For instance, law 24057, which we discussed in the introduction. A law that is stripped of most of its content certainly ceases to fulfill its purpose, and one is left to wonder why adopt this route instead of a total veto. We claim that it is because the few items left in the statute are perks intended to deter a number of legislators from supporting an override to the bill. This is how enacting coalitions are broken up by item vetoes. In cases where the President can only impose total vetoes, as in the United States, an override will at least receive support from the majority that passed the bill initially, and support from additional legislators may be needed depending on just how much support it received in that instance. In the United States support from the required 2/3 occurs in 50% of the cases. If we were to assume, in the case of Argentina, that the President does not pull apart the enacting coalition by issuing partial vetoes, we should expect the same amount of override attempts for total and line item vetoes (given that the number of total vetoes and line item vetoes is the same). This is not the case; Figure 1 shows that we find that the probability of attempts to different presidents, which according to Cameron 2000 would preclude the possibility): laws (on accidents at the workplace), and (on risks at the workplace). In this version of the paper, however, we limit our analysis to this single possibility of death absent overrides. 23

24 override bills that were vetoed in full is more than double that of attempts to override bills vetoed partially. [Figure 1 about here] This suggests that when the President partially vetoes bills, she is picking and choosing articles in order to dissolve the coalition that initially supported the bill. What affects, therefore, the chances that given a veto, Congress will attempt to revert the President s decision? Logit results in Table 5 confirm most of the findings in Cameron 2000: (a) The level of legislative significance affects those odds: in particular, belonging to the category of landmark legislation increases the odds that Congress will attempt to override vetoes; (b) Partisan effects are present: the likelihood of attempts decreases when the President s party holds less than a plurality in the lower chamber. The latter of these findings is perhaps the most striking, given how counter-intuitive it is: upon first consideration, one expects that less support for the President in Congress causes overrides to be more, not less, likely. [Table 5 about here] A novel finding, linked to the particular institutional setting of Argentina, stems from the distinction between total and item vetoes. Table 5 shows that across specifications, facing an item veto consistently decreases the odds of override attempts, i.e., congress is more likely to attempt to undo total vetoes. Figure 1 illustrates the magnitude of this effect: the odds of an override attempt when changing from an item to a total veto are more than doubled. As suggested, this is precisely because the enacting coalition is broken through partial vetoes (Figure 1). [Figure 2 about here] [Figure 3 about here] 24

25 Likewise, the odds of an override attempt are 2.6 times greater when facing landmark as opposed to non-landmark legislation (Figure 2 shows predicted probabilities). When analyzing the likelihood of attempts given types of vetoes and levels of significance, Figure 3 shows that override attempts are more likely to affect landmark legislation in general, but even more so if the President total vetoes the legislation. On Successful Attempts: Observed Overrides Are the determinants of override attempts related to those of success? We have already noted the high success rate of attempts (.68). Are there specific political scenarios, or characteristics of the legislation that provide greater insight as to when we should expect success? Table 6 provides results from the analysis conducted to tackle this question. Across specifications, we find that the single significant effect on the likelihood of success is that of plurality: when the President s party in Congress holds a plurality (no more, no less), override attempts are more likely to succeed. [Table 6 about here] We conjecture that if legislation, especially under plurality government, is the result of a compromise among different groups to reach a majority, successful overrides might be one of the mechanisms used by the plurality party leadership to show its level of commitment to the groups that assisted in reaching a majority. The fact that the likelihood of override success is so high speaks of a Congress that is able to coordinate, and avoid attempts that are deemed to fail. But most importantly, these results suggest that Congress might also act through re-passage of vetoed legislation, an analysis that is beyond our goals in this article. 25

26 [Figure 4 about here] Figure 4 shows that the greatest effect on the likelihood of overrides is that of plurality, 72%. This effect is not only large in magnitude it is also the only statistically significant effect on the likelihood of success. Again, we argue that in environments in which the governing party needs the support of legislators outside their own party in order to reach a majority, overrides are a matter of commitment and credibility. So as to maintain the support of those legislators in future instances of the game, the governing party provides the legislators needed to reach the requirement of two-thirds. The case of the Budget Law for 1998 presented above shows precisely this situation: the override of articles affecting members of the coalition that helps the governing party reach its majority. V. Conclusion This paper analyzes the dynamics behind vetoes and veto overrides in the context of a multiparty legislature. In particular, it focuses on the distinction between total and item vetoes, suggesting that the latter introduce complexity to the legislative bargaining process and alter the strategic game entirely. The paper argues that a key difference introduced by item vetoes is that used strategically, they allow the President to break up the enacting coalition and make overrides more difficult. The President achieves this by deleting specific articles from a piece of legislation that is nonetheless enacted. The deleted articles, typically targeted benefits placed in the bill to gather support from different groups of legislators, do not take anything away from enough legislators that putting together an override coalition is not possible. This explains why we find that total vetoes, which affect all legislators equally, are more likely to be overridden than partial 26