Marta Arretche

University of São Paulo (USP), Brazil

ABSTRACT

The article distinguishes federal states from bicameralism and mechanisms of territorial representation in order to examine the association of each with institutional change in 32 countries by using constitutional amendments as a proxy. It reveals that bicameralism tends to be a better predictor of constitutional stability than federalism. All of the bicameral cases that are associated with high rates of constitutional amendment are also federal states, including Brazil, India, Austria, and Malaysia. In order to explore the mechanisms explaining this unexpected outcome, the article also examines the voting behavior of Brazilian senators constitutional amendments proposals (CAPs). It shows that the Brazilian Senate is a partisan Chamber. The article concludes that regional influence over institutional change can be substantially reduced, even under symmetrical bicameralism in which the Senate acts as a second veto arena, when party discipline prevails over the cohesion of regional representation.

Keywords: Federalism; Bicameralism; Senate; Institutional change; Brazil.

Introduction

A well-established proposition in the institutional literature argues that federal states tend to take a slow reform path. Among other typical federal institutions, the second legislative body (the Senate) common to federal systems (Lijphart 1999; Stepan 1999) has been described as particularly important for creating opportunities to veto policy change (Tsebelis 1997). The Senate not only hampers institutional change by contributing an additional decision-making arena to democratic deliberations (Tsebelis 1997), but at the same time it adds a significant subnational veto to legislative processes (Rodden 2006; Stepan 1999). The Senate is described more often than not as the legislative body where subnational elites are entitled to protect constituent unit rights and block changes that may impinge upon subnational rights. Because of the supposed influence of subnational elites on senators' behavior, as well as the supposed association between federal states and bicameralism, an influential proposition associated with federalism argues that proposals for institutional reform will be more difficult to approve: federalism equals a strong Senate, which equals subnational veto power, which leads to the preservation of the status quo.

Unpacking these assumptions, however, leads us to interesting findings. This article distinguishes federal states from bicameralism and mechanisms of territorial representation in order to examine the association of each with institutional change in 32 countries. I use constitutional amendments as a proxy. The study reveals that only bicameralism appears to systematically reduce the rate of constitutional amendment. The evidence clearly illustrates that bicameralism tends to be a better predictor of constitutional stability than federalism. Further evidence that federalism does not necessarily act as a constraint on institutional change is the finding that all of the bicameral cases that are associated with high rates of constitutional amendment are also federal states, including Brazil, India, Austria, and Malaysia.

In order to explore the mechanisms explaining this unexpected outcome, the second part of the article examines the voting behavior of Brazilian senators constitutional amendments proposals (CAPs). I measure the voting cohesion of same-state Senators affiliated with national parties from January 1989 to December 2009.

I find that the Brazilian Senate is a partisan Chamber. Senators elected by the same state show low voting cohesion, even when the proposed amendments will negatively affect the interests of their respective states. I also find that federal issues have a marginal effect on the cohesion of Senate voting, both at the level of regional legislative caucuses and of national party-based caucuses.

Based on the evidence of the Brazilian case, the article concludes that regional influence over institutional change can be substantially reduced, even under symmetrical bicameralism in which the Senate acts as a second veto arena, when the second Chamber functions in a partisan manner; which is to say, if party discipline prevails over the cohesion of regional representation. Under these conditions neither federalism nor bicameralism slows reform.

Federalism and Bicameralism

The association between federalism and bicameralism is founded on solid historical evidence. A second Chamber, based on territorial considerations, was created upon the founding of the U.S. federation and the system was replicated in virtually all subsequent federations. The European confederations, such as Switzerland and Germany, also adopted an Upper House based on territorial considerations. Bicameralism, understood as a second Chamber representing constituent units, is largely predominant among federal states (Massicote 2001; Patterson and Mughan 2001; Neiva 2004). By contrast, the Upper House disappeared (or was strongly weakened) in various unitary European countries¹ due to the fact that the Senate served as a refuge for the nobility before universal suffrage took full effect. As a result, "legitimacy of upper houses in class-based systems was repudiated, while the legitimacy in territorially-based systems was maintained." (Tsebelis and Money 1997, 32-3).

The association between federalism, bicameralism, and their constraining effect on institutional change requires further specification though. Under asymmetric bicameralism, in which the Upper House has limited attributions and the Lower House retains legislative supremacy, the Senate veto power will be limited only to those policy areas under its legislative jurisdiction. Indeed, for a sample of 51 countries, Neiva (2006) convincingly demonstrates that the existence of a second Chamber based on territorial considerations does not necessarily mean that this legislative Chamber has real power. Case studies point in the same direction. As surprising as it seems, "the Belgian's Senate power was severely reduced when it was elected in its new formula in 1995." (Lipjhart 1999, 206).

Hence, only under symmetrical bicameralism should there be an association between the existence of the Senate and a constraint on institutional change. Even so, veto power theory argues that the probability of institutional change is associated with the number of political actors required to approve a given institutional change (Tsebelis 1997). Electoral results can reduce the number of veto actors (Stepan 1999; 2004). Hence, even under symmetrical bicameralism, it is possible that the two Houses operate as one veto player if both Houses have a similar partisan composition. In this case, they can be counted as only one actor, even if the Upper House is entitled to veto legislative change. This hypothesis, however, raises the issue of the Senate's composition and the parliamentary behavior of its members.

The argument that the Upper Chamber acts as an additional veto point is based on the condition that its composition and behavior are different from that of the Lower Chamber. Additionally, it assumes that Senators represent territorial affiliation, and for this reason Senators vote in accordance with the preferences of governors, regional elites, or even the public opinion of their districts of origin. Hence, the argument supposes that Senators' parliamentary behavior is potentially non-partisan, because federal states may be predisposed to weaken national party organizations (Desposato 2001; 2004). Federalism should thus have the effect of weakening national party affiliations while at once strengthening territorial affiliations.

Historical evidence casts doubt upon this assumption, however. The consolidation of national parties suggests that representatives may act more responsively to their party leaders than the above logic suggests. In this sense, partisan affiliations can divide the allegiances of legislators within the same subnational district. Chibber and Kollman (2004) argue that the formation of national party systems, even in federal states, resulted in both the weakening of subnational parties and the influence of local elites. As a result, partisan discipline has substituted independent judgment as the voting standard in national legislatures. In India, the U.S., and Canada, national parties have replaced local representation as the clearest predictor of parliamentary behavior. In Australia, the independence of local notables vis à vis their parties gradually diminished and was ultimately substituted by a strong partisan discipline in both Houses of the legislature (Castles and Uhr 2005, 57; Patterson and Mughan 200, 43). In Austria, the Länders are "under-represented in the federal decision-making arena and have no effective veto powers in the Federal Council that would allow them sustained influence over the federal policy-making process" [and] "provincial governors are frequently bound to the intentions of their headquarters, so that party-political loyalties may overshadow Lander interests" (Obinger 2005, 197-8). In Canada, "MPs from a province do not speak for their territorial governments" (Banting 2005, 94). Even in Germany, with its powerful Bundesrat - the federal council representing the 16 Länders -, veto strategy dynamics are interpreted in accordance with the partisan composition of the legislative chambers (Horeth 2008). A series of case studies therefore illustrate that partisan affiliations divide local loyalties, even in federal states.

In sum, although the original federal formula was to constrain institutional change by creating a legislative Chamber that would represent territorial interests (the Senate), in many countries this design operates de jure, rather than de facto. The relationships - among federalism = Upper Chamber = subnational veto power - can be indeed weak.

Institutional Change

How best to measure institutional change is a subject that reliably results in considerable controversy. The concept of institutional change is encompassing, rendering it difficult to find measures to compare different contexts. The same reform can have distinct impacts depending on the nature of preexisting policies. Furthermore, policy change can occur in variegated ways, varying from incremental to endogenous change (Mahoney and Thelen 2010) and transformations of paradigm (Hall 1993). Changes can involve the renegotiation of some element while leaving others unaltered (Thelen 2003, 225). Institutional change may also involve the addition of new attributes while maintaining characteristics of the previous policy intact (Pierson 2004, 137). Change can also be the result of converting existing institutions to fit new goals without undertaking formal reforms (Immergut 1992).

To address the problem of measurement, this paper adopts the annual rate of constitutional amendments as an indicator of institutional change. Although this measure only captures a fraction of the institutional changes experienced by any given country, constitutional amendments typically reflect important alterations of the status quo. This is particularly so in federal states, where major decisions over policy and the distribution of authority must often be articulated with subnational units (Keating 2005, 19-20). Furthermore, in most countries constitutional amendments represent among the most difficult legislative initiatives. They are expected to be especially rare in federal states. Constitutional amendments therefore provide more than an adequate proxy for assessing the extent to which an institutional configuration results in changes to the status quo. Focusing on amendments implies excluding important political changes, but because the units of analysis are uniform, this method possesses the advantage of generalized comparability. The data at the heart of this paper's analysis were borrowed from Lutz (1994)².

Table 1 presents a sample of 32 countries and their respective institutional characteristics. Based on Lutz's data (1994) on the annual rate of constitutional amendments, I evaluated the institutional attributes of each country for the period covered by the dependent variable³.

Federalism and Constitutional Stability

Is it a generalized fact that federalism tends to produce policy stability? Are federal states systematically associated with institutional stability, independent of particular institutions that might lend support to this result?

Figure 1 depicts the annual rate of constitutional amendments differentiating between federal and unitary countries within the sample⁴. It is clear that the distribution of values is very asymmetrical. There is no evidence of a significant relationship between the type of state and constitutional stability. The percentage of cases with high institutional stability (in which the approval of constitutional amendments approximates zero) is the same between the two groups of countries at 30 percent.

The annual average rate of constitutional amendments in federal states is 2.5 and 2.3 in unitary states. If we exclude New Zealand (clearly an outlier), the difference increases; the average approval of constitutional amendments in unitary states falls to 1.8 per year. However, given the dispersion of values, the average tells us very little. Lutz (1994) considers rates above 1.24 to be elevated in comparative terms. The percentage of countries that fall within this parameter in the two groups of countries is exactly the same: 50 percent.

Although merely descriptive, Figure 1 provides evidence to suggest that the federal states are in fact more conducive to constitutional change, which runs contrary to what theory should predict. This inference can be explained by endogenous⁵ factors, however. The adoption of a federal system is an inherent expression of conflicts over national decisions. It is plausible to assume that such contexts render constitutions lengthier, more complex, and will ultimately lead to a higher probability of legislative activity focusing on the Constitution. Hence changes to the rights or attributions of subnational units will entail a constitutional revision.

However, the endogeneity that gives origin to processes of change is, as a condition, insufficient to explain the apparent facility with which constitutional amendments are enacted. Certainly, lengthier constitutions tend to emerge from more conflictive contexts. These bigger constitutions, in turn, also tend to require a greater number of amendments because they typically regulate an elevated number of issues⁶. However, the theory stipulating that federal states should operate as a constraint on institutional change would lead us to believe that the rate of constitutional amendments ought to be smaller, in spite of endogenous pressures for amendments. In other words, Figure 1 does not support the notion that, given constitutions of the same size, equal endogenous pressures ought to produce smaller amendment rates in federal as opposed to unitary states.

Bicameralism and Constitutional Stability

It is possible that, by creating additional veto arenas, bicameralism is more effective at promoting policy stability than federalism alone. Given that the legislative powers of Upper Chambers vary substantially (Lijphart 1999; Stepan 1999; Neiva 2004; 2006), the sample cases distinguished between unicameral and different types of bicameral legislatures, based on the proposition of Tsebelis and Money (1997), which stipulates that variations in senatorial powers include attributions to initiate, alter, or block legislation. The configuration of these variegated powers will, in turn influence the Senate's interactions with the Lower House in myriad ways. Within the aforementioned scheme, bicameral systems are thus categorized into three types: (i) the Senate possesses powers that are equal to those of the Lower House or wields exclusive powers to propose constitutional amendments (strong Senate); (ii) the Senate is authorized to propose constitutional amendments, but a joint session with the Lower House is required to approve amendments (Senate with veto power); and (iii) cases in which the Lower House is decisive in initiating and approving constitutional amendments (weak Senate). We can therefore expect that only cases (i) and (ii) - a strong Senate or one with veto power - can have a negative effect on the annual rate of constitutional amendments⁷.

The relationship between constitutional stability and legislative power is illustrated by Figure 2. The first comparison provides an informative inference on how unicameral legislatures perform in relation to strong Senates; it contrasts polities with only one veto arena against those with a second arena that has the authority to initiate, modify, or block legislation. Confirming what is to be expected of veto player theory (Tsebelis 1997; Tsebelis and Money 1997), strong Senates appear to reduce the probability that constitutional amendments will be approved. The average rate of constitutional amendments in unicameral countries is 3.5, whereas it is 1.8 in polities with a strong Senate. Once again, however, given the asymmetry of the dispersion, the average for each group cannot be considered an accurate or reliable indicator. Another way of comparing the two country groups is to employ Lutz's value of 1.24 as a controlling value (1994). Seventy percent of unicameral countries in the sample (9 out of 13) show annual rates of constitutional amendments higher than 1.24, whereas this number is 45 percent in countries with a strong Senate (5 out of 11).

An analysis of the distribution of values confirms this result. If we exclude New Zealand, both the groups have almost the same number of cases: 11 polities with a strong Senate and 12 unicameral countries. In countries with unicameral systems, six illustrates rates close to 1.0; in three cases this rate is around 2.0 and 3.0. By contrast, four cases show that the number of constitutional changes can be considered very high; that is to say, a rate that is higher than four amendments per year. In countries with a strong Senate, in turn, six illustrate high constitutional stability with rates of less than 1.0; four countries have a moderate rate of change (around 2.0 and 3.0 amendments approved per year). In contrast, only one case (India) exhibits a very elevated rate (above 4.0). The average and the distribution of values therefore show that Senates have a slight advantage in maintaining greater constitutional stability.

The group whose Senate is characterized by a Senate with veto powers is too small, including only two countries (Venezuela and Argentina). Although it confirms the proposition that Upper Houses possessing veto power (but without power to initiate legislation) tends to produce low rates of constitutional change, the inference should be taken with caution due to the group size.

Finally, the association between bicameralism and constitutional stability is partly confirmed by the performance of countries with weak Senates. This group illustrates curious results, as five cases exhibit high constitutional stability (that is to say, a rate of approximately zero) and two cases show elevated rates of change (above five amendments approved per year). As a group, the average would obviously not be an adequate measure, especially given the bimodality of the distribution and the small number of cases. However, the majority of the cases illustrate high rates of constitutional stability.

The relationships depicted in Figure 2 therefore indicate that the presence of a second veto arena tends to inhibit constitutional change independent of the Senate's legislative powers. In short, bicameralism tends to be associated with diminished rates of constitutional amendments. This result confirms the proposition that additional veto arenas have a higher probability of preserving the status quo. Evidence indicates that bicameralism therefore tends to be a better predictor of constitutional stability than federalism.

However, high rates of constitutional amendments are associated with bicameralism in certain cases, independent of the Senate's legislative powers. Despite its strong Upper House, India approved an average of seven constitutional amendments per year between 1950 and 1992, and Brazil approved almost three amendments per year between 1989 and 2006. Although their Senates are regarded as comparatively weak, Austria approved an average of more than six constitutional amendments between 1975 and 1992, and Malaysia approved an average of five amendments per year between 1957 and 1992.

It is worth noting that all cases in which bicameralism is associated with high rates of constitutional amendments are also federal. This evidence confirms last section's results: federalism does not necessarily act as a constraint on institutional change.

Therefore, although a second Chamber increases the likelihood that the status quo will be preserved, even when combined with federalism this outcome is far from assured. Moreover, this inference cannot be treated as a residual phenomenon that is irrelevant to the analysis. It indicates that bicameralism - even when symmetric - is not a sufficient condition to prevent constitutional change. With a second Chamber that wields veto power it is possible to approve changes to the status quo, therefore devising democratic solutions for national policy conflicts. There are factors that facilitate constitutional change, even under symmetrical bicameralism, and the following analysis of the Brazilian case aims to explore these factors.

Electoral Rules and Constitutional Stability

The Senate is an institution associated with a country's constituent states or provinces, and is typically attributed the role of an encumbrance on institutional change. Territorial representation would imply that a second Chamber should have a distinct composition from the Lower Chamber, providing regional interests with a veto. Under these conditions, populational minorities would have exponential opportunities to wield a veto (Rodden 2006; Stepan 1999). However, an alternative proposition holds that if a second legislative House has the same composition as the first, and therefore similar majoritarian preferences, it cannot be regarded as a second veto (Tsebelis 1997; Hiroi 2008). This second hypothesis, however, requires the admission that "parties matter."

The proposition of Chibber and Kollman (2004) holds that parties become important in central parliamentary arenas under two conditions. First, when authority over policy is centralized; and, second, when parties control parliamentary behavior. The previously examined constitutional amendments fulfill the first condition, because their dependent variable is legislation considered in central decision-making arenas. On the second count, the election rule for Senators is frequently employed as an indicator of whether regional actors possess influence over the parliamentary careers of Senators (indirect choice), or alternatively, if parties are represented in the Higher Chamber (direct choice).

Figure 3 illustrates the rules for Senator selection and the annual rates of constitutional amendments. The sample obviously excludes unicameral systems. Taken as an isolated explanatory factor, the rules for selecting Senators appears to have little if any influence on constitutional stability. The behavior of the two polar groups - with rules that are both direct and indirect - illustrates similar behavior. The average rate of constitutional amendments of the first group is 1.3, whereas in the second it is 1.8. The percentage of cases that enjoy constitutional stability is the same: 70 percent.

This result, however, must be qualified. The bimodality of distribution, the small number of cases, and the dispersion of values all cast doubt on the definitive nature of these results. They serve merely as an indication that the proposition requires greater empirical evidence. In other words, whether or not the rule for selecting senators affects institutional stability or not still requires more rigorous confirmation.

Furthermore, although the association between electoral rules and parliamentary behavior has been amply explored in the literature, case studies suggest that just because senators are selected by an indirect rule does not necessarily imply that parties matter does not in central parliamentary arenas.

In Canada, "senators are hand-picked up by the [provincial] prime minister, who may adjust the Senate's membership in harmony with the government party majority in the House of Commom" (Patterson and Mughan 2001, 47). In the German case, where senators also are appointed by governors, the "the Bundesrat can be exploited as a partisan veto instrument under divided government" (Horeth 2008, 418).

In other words, it is possible that parties control the parliamentary behavior of senators, even under the indirect rule of selection.

In summary, neither federalism nor regional control over the naming of senators appears to have an important influence on constitutional stability. For its part, bicameralism seems to have a larger impact in preserving the status quo by creating a second opportunity for a veto.

However, even under bicameralism there are cases of considerable constitutional change. These cases suggest that bicameralism alone is not a sufficient condition to impede institutional change. The period subsequent to the unveiling of Brazil's 1988 Constitution represents one striking case of symmetrical bicameralism, with a strong Senate and at once an elevated rate of constitutional amendments. Brazil therefore provides a particularly revealing case for analysis.

Constitutional Stability in Brazil

Until now, this paper has examined relations between the institutions of federal states, on the one hand, and the preservation of the status quo, on the other. It is possible to argue that in isolation these institutions have a systematic impact on institutional stability; but it is equally possible to argue that this result occurs only when they are combined (Pierson 1995; Ragin 1987).

Additionally, rates of constitutional stability examined in previous sections failed to identify the direction of change, mainly due to comparability issues. That is to say that the relations reflected the annual rate of constitutional amendments, independent of the legislation in question and its impact on the status quo. In Canada, for example, significant institutional changes took place from the 1970s onwards. But these reforms implied a strengthening of provincial governments in both their policy competences and their budgetary authority (Banting 1995; Chibber and Kollman 2004; Keating 1999). It cannot be expected that territorially-based interests will mobilize to veto legislation if it fails to negatively affect their interests (Desposato 2004). Similarly, it is not plausible to assume that vetoes will be deployed when a decision-making process does not involve politically divisive issues, even if their content is federal (Horeth 2008). Accordingly, it is plausible to assume that the German rate of constitutional amendments, which is higher than other averages and elevated by Lutz's (1994) standards, can be explained by the approval of constitutional amendments whose content was strategically designed to avoid the qualifying demands of the Bundesrat⁸. Or, alternatively, high rates of constitutional amendments can be the result of previous commitments for small changes in the status quo. In this case they might reflect the anticipated veto power of opponents (Immergut 1992). Therefore, without controlling for the content of constitutional change the relations between the institutions examined in previous sections and institutional stability remain indeterminate. In short, more precise evidence of change is required to explore this issue.

Finally, evidence presented in Figure 3 is based on the hypothesis that the direct election of senators regiments party control into the electoral process and, as a consequence, increases partisan discipline. However, this issue is highly debated in the literature, especially when applied to federal states. In theory, regional control over party lists and over resources for parliamentary survival weakens national parties, even under direct elections (Samuels 2003).

The Brazilian case can help us better explore these questions. Brazil is a federal state with a strong Senate (Neiva 2004; 2006; Stepan 1999). The election of the Upper Chamber is direct and a partisan affiliation is a necessary condition for candidacy. For the approval of constitutional amendments, Brazilian bicameralism is rigorously symmetrical. A proposal for such an amendment is governed by the same rules in both Houses and must navigate between both until an identical text can be agreed upon and approved.

While the literature tends to view symmetrical bicameralism as a necessary condition for constraining the approval of constitutional amendments, the Brazilian experience provides a clear disconfirming case. Between 1992 and 2009, more than 62 constitutional amendments were approved in the National Legislature, and the rate of amendments tallied 3.6 per year. Of these amendments, 28 dealt with issues of direct relevance to federal interests. The rate of 3.6 is higher than the average encountered in all the groups examined in previous sections and ranks as "elevated" according to Lutz's criteria (1994). If we measure only the issues of exclusive federal interest, the annual rate would be 1.7, even higher than Lutz's index (1994) for measuring elevated rates of constitutional change (1,24).

Most important for the current analysis, a large part of approved constitutional amendments imply losses for subnational governments and, in particular, the states. A cursory examination reveals that these amendments negatively affected state and municipal revenues, subnational governments' ability to exercise authority over the collection of taxes, the implementation of policies, and the allocation of their own revenues. Rights conferred to subnational governments by the 1988 Constitution were suppressed by constitutional amendments approved in the Senate between 1989 and 2006. Given the content of these proposals, one would be led to believe that subnational governments mobilized their vetoes. However, if such mobilizations occurred, they were obviously unsuccessful.

Thus, we have a case in which federalism = strong Senate ≠ subnational veto power ≠ preservation of the status quo. One way of exploring this parliamentary behavior is to assess roll-call voting patterns with regards to constitutional amendments.

Between January 1989 and December 2008, the Brazilian Senate submitted 100 constitutional amendment proposals (CAPs) to roll-call votes. Of these, 60 had a direct impact on subnational issues and 23 of these were approved. Of the 60 federal CAPs submitted to a roll-call vote, 13 were rejected. The remaining 19 were either filled, affected by the previous approval of another constitutional amendment, or postponed.

Of the 13 unapproved CAPs, five had to do with subnational electoral legislation, and senators rejected them unanimously. Their rejection, therefore, can be explained by the calculus of parties. Except for this one issue area, electoral reform, the remaining CAPs that had been rejected by the Senate consisted of proposals that had already been approved by the Chamber of Deputies. The Senate rejection means that a new text should be voted by the Lower House. In these cases, legislative rejection reflects the Upper Chamber's strategic use of its institutional resource - the veto - as a means of ensuring that the final text elaborated by the two Houses brings about the desired compromise.

The examination of approved constitutional amendments reveals that the majority of reforms imposed losses on the revenues of subnational governments, as well as reduced their authority over taxes, the implementation of their own policies, and the allocation of their own revenues. How can these results be explained?

Of the 60 legislative issues submitted, 99 roll-call votes were considered for the analysis, once unanimous and invalid votes were excluded. The cohesion of senators from the same states in these votes is represented by Figure 4. For each unit of the federation, total votes are distributed among (i) the absolute number of roll-call votes in which three senators were absent or abstained from voting; (ii) the absolute number of roll-call votes in which senators voted differently, whether two or three were present, and, (iii) the absolute number of roll-call votes in which two or three senators voted together⁹.

As can be observed, senators from the same state did not vote cohesively. Of the votes cast for the 60 federal CAPs, many of which negatively affected subnational interests, the systematic behavior of senators from the same state was to vote divisively. The cohesion between state-based caucuses - even if we consider only the two senators present - never exceeded 70 percent of the votes. For state-based caucuses of senators to vote together is the exception rather than the rule. Only 10 state-based caucuses - of 27 - voted the same in more than 50 percent of the votes. In few states (Amapá, Sergipe and São Paulo), this rate is smaller or equal to 30 percent of the votes. In sum, state-based voting cohesion is not the regular behavior of the Senate. Brazilian senators did not harbor similar preferences when it came time to vote on issues that affected the interests of their respective states and districts of origin.

One way of exploring the effect of federalism on the behavior of state-based caucuses is to compare the difference between the rates of voting cohesion among senators from each state on different issues. Figure 5 illustrates the results of taking the percentage of roll-call votes in which two or three senators voted together on i) matters that affected subnational interests; ii) matters that did not affect subnational interests, and (iii) all roll-call votes. The left column presents the percentage of roll-call votes for each state in which senators voted in the same way on the 99 roll-call votes on the 60 "federal" CAPs selected for analysis. The column in the middle represents the same indicator for roll-call votes that did not affect subnational interests. Finally, the right column represents the same indicator for all the roll-call votes of the CAPs during the same period: from January of 1989 to December of 2008.

The data indicate that in 18 of the 27 states, matters affecting subnational interests have the effect of increasing voting cohesion of state-based legislative caucuses. This is not a significant effect, however. Rarely does this effect elevate the cohesion of senators in 20 points (Bahia, Ceará, Paraíba and Santa Catarina).

The mechanisms typically cited in the literature as explaining the voting behavior of senators - such as the pressure of governors, regional public opinion, and economic, political, and state elites - cannot explain the Brazilian case.

How do the Brazilian senators behave, then, in partisan terms? Figure 6 presents party discipline rates in the Senate for the seven main national parties on roll-call - valid or non-unanimous - voting of all CAPs from January 1989 until December 2009, distinguishing once more by the issue at stake. Discipline was measured in relation to the vote of the leader of each party for each one of the roll-call votes. In this way, the values of the y-axis refer to the average percentage of senator discipline in relation to votes of the respective party leader. The column on the left presents results for the roll-call votes on the "federal" CAPs, the column in the center refers to the non-federal CAPs, whereas the column on the right provides the same indicator for all the roll-call votes of the CAPs during the period.

As can be observed, party discipline in the Senate is systematically elevated for all of the seven major national parties. The rates of party discipline exceed 75 percent for all parties. Partisan discipline in the Senate follows a similar pattern as that of the Chamber of Deputies (Figueiredo and Limongi 1999). This pattern extends to issues of federal interest (Cheibub, Figueiredo and Limongi 2009; Arretche 2007; 2009). In short, the Senate is also a House that operates on partisan terms.

Second, the comparison among the columns on the left and the center for each party show once more that votes on issues of federal import reduces party discipline in the Senate. This reduction, however, is very small.

These results confirm the findings obtained for rates of cohesion in state-based senatorial caucuses. That is to say that the effect of federalism on the Senate - or the organized regional pressures affecting the parliamentary behavior of senators - tends to exercise little influence. These results confirm those obtained by Desposato (2001; 2004) with respect to the limited effect of federalism on the cohesion of national parties in central parliamentary arenas.

Final Considerations

In federal states proposals for institutional change that threaten the rights of the federation's constituent units will likely encounter greater legislative resistance in obtaining approval. However, there is no systematic relationship between federalism, per se, and constitutional stability.

By contrast, stability is more likely to be found in a political institution often encountered in federal states: bicameralism. A second veto arena - particularly when it can rely on strong legislative powers - tends to produce greater constitutional stability, especially in matters of amending the Constitution - the most difficult of institutional reforms.

A strong Senate is not, however, a sufficient condition to inhibit constitutional change. Brazil is an interesting case to study because it exhibits high rates of constitutional amendments under symmetrical bicameralism. Moreover, the content of approved constitutional amendments renders it even more interesting, as the majority of these reforms negatively affected the interests of subnational governments. From 1989 to 2008, the approval of 23 constitutional amendments generally resulted in the suppression of subnational revenues in favor of the Union as well as augmenting the regulatory authority of the federal government over the collection of subnational taxes, the implementation of subnational policies, and the allocation of subnational revenues. The mechanisms typically viewed as exerting pressure on the federal legislature - such as the pressure of governors, economic and regional elites, or public opinion - do not appear to have affected the legislative behavior of senators, otherwise the constitutional amendments would not have passed.

By contrast, the voting cohesion of state-based senatorial caucuses was systematically low. Partisan cohesion characterizes the parliamentary behavior of the Brazilian Senate. The expected effect of federalism did not significantly influence the cohesion of national partisan caucuses. Indeed, it was a marginal effect. The low cohesion of state-based caucuses is explained by partisan affiliation. Senators from the same state voted divisively because they are disciplined by their own party. This signifies that the approval or rejection of legislative issues in the Senate can be explained by the veto power of the parties and not regional interests. In this case, it is not the states that are represented by the Senate. Instead, it is the parties who are represented through state districts.

Regional influence over institutional change can be substantially reduced, even under symmetrical bicameralism in which the Senate acts as a second veto arena, when the second Chamber functions in a partisan manner; which is to say, if party discipline prevails over the cohesion of regional representation.

Notes

1 The United Kingdom is perhaps the best known example of a gradual weakening of the Upper Chamber.

"Schemes to abolish or reform the Upper House began circulating in the 1880s and culminated in the Parliament Act of 1911, which placed the ultimate power of decision in the Lower House. The Lord's powers were further reduced in 1949 legislation that limited their delaying power over nonfinancial legislation from three years to one. The Lord's power over financial legislation is even more reduced." (Tsebelis and Money 1997, 34)

2 "Toward a Theory of Constitutional Amendment" is a pioneering study on the requirements to amend constitutions and the effects of these requirements. The study also examines annual rates of amendment in all the U.S. states and a sample of 32 countries.

3 The sample contains 33 observations, because, in the United States, the approval of constitutional amendment 17 in 1913 changed the election of senators from indirect to direct.

4 A classification based on the availability of data from the Forum of Federations. In the cases in which this information was not available for the period examined, additional consultations with historical sources were necessary.

5 I would like to thank Fernando Limongi and Pedro Neiva for their suggestion with respect to this problem.

6 In fact, the theory of Lutz (1994) holds that the size of the Constitution combined with strategies for the approval of constitutional amendments is the most important explanatory factor in accounting for rates of amendments. Lutz demonstrated empirically that there is a high correlation among these variables for a sample of 34 countries. He also provided evidence to support the notion that the rules for approving constitutional amendments are more important then the size of the Constitution.

7 The data were obtained from Mastias and Grange (1987), Tsebelis and Money (1997), Lijphart (1999), Patterson and Mughan (2001), Massicote (2001) and Neiva (2004; 2006). In cases in which information emited by authors conflicted, databases were consulted and experts on the country in question were contacted.

8 For an analysis of these strategies in the German case, see Horeth (2008).

9 This method of measuring cohesion provides advantages over those typically employed, especially over the Rice Index. Given the reduced size of caucuses (three senators) and the fact that only two senators showed up to vote in a significant number of roll-call votes, the Rice Index tends to obtain extreme results, rendering the averages of dubious credibility.

Moreover, it avoids the methodological decision of attributing one meaning to a legislator's strategic decision to abstain or simply not show up, since it critically depends on the issue under examination. However, it makes it more difficult to establish a general rule for comparing roll-call votes.

To resolve these problems, the method chosen in the figure simply separates the valid votes and the non-unanimous roll-call votes in which the three senators from each state absented themselves or abstained from a vote. Additionally, this method considers equally valid a scenario in which only two senators or the three senators from each state vote.

Therefore, states units or parties in which one senator voted are only excluded because of the logical impossibility of measuring the cohesion between these cases.

Bibliographical References

Arretche, Marta. 2007. The veto power of territorial governments in Brazil: Political institutions and representatives voting behavior. Brazilian Political Science Review 1 (2):40-73.

Arretche, Marta. 2009. Continuidades e descontinuidades da federação brasileira: de como 1988 facilitou 1995. Dados 52(2): 377-423.

Banting, Keith. 2005. "Canada. Nation-building in a federal Welfare State". In Federalism and the Welfare State. New world and European experiences, ed. Herbert Obinger, Stephan Leibfried and Francis Castles. Cambridge: Cambridge University Press.

Banting, Keith G. 1995. "The Welfare State as statecraft: Territorial politics and Canadian social policy". In European social policy: Between fragmentation and integration, ed. Stepahn Leibfried and Paul Pierson. Washington, CD: Brookings Institution Press.

Castles, Francis G., and John Uhr. 2005. "Australia: Federal constraints and institutional innovations". In Federalism and the Welfare State. New world and European experiences, ed. H. Obinger, S. Leibfried and Francis G. Castles. Cambridge: Cambridge University Press.

Cheibub, J. A., A. C. Figueiredo, and F. Limongi. 2009. Political parties and governors as determinants of legislative behavior in brazil's Chamber of Deputies, 1988-2006. Latin American Politics and Society 51: 1-30.

Chibber, P. K., and Ken Kollman. 2004. The formation of national party systems. Princeton/Oxford: Princeton University Press.

Desposato, Scott. 2004. The impact of federalism on national party cohesion in Brazil. Legislative Studies Quarterly XXIX (2): 259-85.

Desposato, Scott. 2001. Institutional theories, societal realities, and party politics in Brazil. Ph.D. thesis, University of California, Los Angeles.

Figueiredo, Argelina, and Fernando Limongi. 1999. Executivo e Legislativo na nova ordem constitucional. Rio de Janeiro/São Paulo: FGV/ANPOCS.

Hall, Peter. 1993. Policy paradigm, social learning, and the State: The case of economic policy making in Britain. Comparative Politics 25 (3): 275-96.

Hiroi, Taeko. 2008. The dynamics of lawmaking in a bicameral legislature: The case of Brazil. Comparative Political Studies 41 (12): 1583-606.

Horeth, Marcus. 2008. A successful failure: The contested implications of Germany's federal reforms. German Politics 17 (4): 408-23.

Immergut, Ellen M. 1992. "The rules of the game: The logic of health policy-making in France, Switzerland, and Sweden". In Structuring politics. Historical institutionalism in comparative analysis, Sven Steinmo, Kathleen Thelen and Frank Longstreth. Cambbridge: Cambridge University Press.

Keating, Michael. 1999. "Managing the multinational State: Constitutional settlement in the United Kingdom". In The dynamics of decentralization. Canadian federalism and British devolution, Trevor C. Salmon and Michael Keating. Montreal: McGill-Queen's University Press.

Keating, Michael. 2005. The government of Scotland. Public policy making after devolution. Edinburgh: Edinburgh University Press.

Lijphart, Arend. 1999. Patterns of democracy. New Haven: Yale University Press.

Lutz, Donald S. 1994. Toward a theory of constitutional amendment. The American Political Science Review 88 (2): 355-70.

Mahoney, James, and Kathleen Thelen. 2010. "A theory of gradual institutional change". In Explaining institutional change. Ambiguity, agency, and power, James Mahoney and Kathleen Thelen. Cambridge: Cambridge University Press.

Massicote, Louis. 2001. Legislative unicameralism: A global survey and a few case studies. The Journal of Legislative Studies 7(1): 151-70.

Mastias, Jean, and Jean Grange. 1987. Les seconds chambres du Parlement en Europe Occidentale. Paris: Economica.

Neiva, Pedro. 2006. Os determinantes da existência e dos poderes das câmaras altas: federalismo ou presidencialismo? Dados. Revista de Ciências Sociais 49(2): 269-99.

Neiva, Pedro. 2004. Estudo comparativo de câmaras altas: o poder e o papel do Senado nos sistemas presidencialistas. Ph.D. diss., Instituto Universitário de Pesquisas do Rio de Janeiro.

Obinger, Herbert. 2005. "Austria: Strong parties in a weak federal polity". In Federalism and the Welfare State. New world and European experiences, ed. Herbert Obinger, Stephan Leibfried and Francis Castles. Cambridge: Cambridge University Press.

Patterson, Samuel C., and Anthony Mughan. 2001. Fundamentals of institutional design: The functions and powers of parliamentary second Chambers. The Journal of Legislative Studies 7 (1): 39-60.

Pierson, Paul. 1995. "Fragmented Welfare States: Federal institutions and the development of social policies". Governance: An International Journal of Policy and Administration 8(4): 449-78.

Pierson, Paul. 2004. Politics in time. Princeton: Princeton University Press.

Ragin, Charles. 1987. The comparative method. Moving beyond qualitative and quantitative strategies. Berkeley: University of California Press.

Rodden, Jonathan. 2006. Hamilton's Paradox. Cambridge: Cambridge University Press.

Samuels, David. 2003. Ambition, federalism, and legislative politics in Brazil. Cambridge: Cambridge University Press.

Stepan, Alfred. 2004. "Electorally generated veto players in unitary and federal systems". In Federalism and democracy in Latin America, ed. Edward Gibson. Baltimore/London: Johns Hopkins University Press.

Stepan, Alfred. 1999. "Para uma nova análise comparativa do federalismo e da democracia: federações que restringem ou ampliam o poder da demos". Dados 42 (2): 197-251.

Thelen, Kathleen. 2003. "How institutions evolve: Insights from comparative historical analysis". In Comparative historical analysis in the social sciences, ed. James Mahoney and Diestrich Rueschemeyer. Cambridge: Cambridge University Press.

Tsebelis, George. 1997. Processo decisório em sistemas políticos: Veto players no presidencialismo, parlamentarismo, multicameralismo e multipartidarismo. Revista Brasileira de Ciências Sociais 12 (34): 89-118.

Tsebelis, George, and Jeannette Money. 1997. Bicameralism. Cambridge: Cambridge University Press.