Re-Election and Political Career Paths in the Uruguayan Congress, 1985–99
DAVID ALTMAN and DANIEL CHASQUETTI
Given the presumed marginal – or at best the ‘rubber-stamp legitimising’ – character of Latin American legislatures, they ‘have escaped careful scrutiny’. Even in cases where legislatures are supposed to play a much more significant role than the continental average, such as Chile, Costa Rica and Uruguay, knowledge of legislative politics is still far from conclusive. This article studies re-election patterns of legislators in Uruguay during the four post-authoritarian elections. During these elections in Uruguay, we observed a decreasing but still high rate of turnover of legislators. These high rates of legislative turnover are affected by a significant number of legislators who do not seek re-election. While inter-party electoral volatility strongly influences the rates of incumbent re-election, intra-party volatility does not seem to have an impact on this phenomenon. Lastly, the closed and blocked lists in conjunction with the Uruguayan multiple simultaneous vote, and the fact that a legislator belongs to the Senate, are additional institutional features that help to explain the turnover and incumbent re-election in the legislature.
URUGUAY IN COMPARATIVE PERSPECTIVE This paper analyses legislators’ re-election rates in Uruguay during the legislative elections of 1989, 1994 and 1999. Even in cases where legislatures are supposed to play a much more significant role than the continental average, such as Chile, Costa Rica and Uruguay, knowledge of legislative politics is still far from conclusive. During these elections in Uruguay, we observed a decreasing but still high rate of legislative turnover. These high rates of legislative turnover are affected by a significant number of legislators who do not seek re-election. While inter-party electoral volatility strongly influ- ences the rates of incumbent re-election, intra-party volatility does not seem
David Altman is Assistant Professor in the Instituto de Ciencia Polı´tica at Pontificia Universidad Catolica de Chile, Chile, while Daniel Chasquetti is Assistant Professor in the Instituto de Ciencia Polı´tica at the Universidad de la Repu´blica, Uruguay. We are grateful to Daniel Buquet, Rossana Castiglioni, Brian Crisp, Mark Jones, Paul Mueller, Rafael Pin˜eyro, Peter Siavelis, and the anonymous referees of the Journal of Legislative Studies for their helpful comments. We are extremely indebted to Santiago Lopez and Ximena Machado for their fine research assistance in building the database for this article. This research was funded by the KONCECYT Project No. 1040920.
The Journal of Legislative Studies, Vol.11, No.2, Summer 2005, pp.235–253 ISSN 1357-2334 print=1743-9337 online DOI: 10.1080=13572330500158656 # 2005 Taylor & Francis Group Ltd 236 THE JOURNAL OF LEGISLATIVE STUDIES to have an impact on this phenomenon. Lastly, closed and blocked lists in con- junction with the size of electoral districts are additional institutional features that help to explain the turnover and incumbents’ re-election in the legislature. From a comparative perspective, Uruguayan rates of legislative turnover fall somewhere in the middle (middle–low) of those of its neighbours. Latin American legislatures show a great variety in terms of incumbent re- election, and in Table 1 Uruguay is compared with other American countries. While Mexican and Costa Rican legislators cannot run for immediate re-election,1 their Chilean counterparts show a very high rate of incumbent re-election.2 Of course, even Chilean legislators fall far below the high regis- tries of the US legislators. Panamanian legislators are an interesting case given that they seek re-election almost at the same rate as US legislators but they barely succeed in this endeavour.3 Argentina shows exceptionally low re- election rates for incumbents.4 Although re-election is allowed in Argentina, its legislative patterns of re-election are more similar to those of Mexico and Costa Rica than to those of Chile. ‘Since 1983, the overall stability of member- ship in the Argentine Chamber of Deputies has been fairly low. The average legislator has served only one four-year term in office. The percentage of “newcomers” has always exceeded 40 per cent, while only 20 per cent of incumbents obtained immediate re-election.’5 In Colombia re-election rates have been reported as moderate. Archer and Shugart show that around 48 per cent of legislators in both chambers are re-elected,6 a very similar percentage to that of Uruguayan legislators
TABLE 1 RE-ELECTION RATES IN SOME COUNTRIES OF THE AMERICAS
Percentage Percentage Percentage Length of Seeking Winning Returning to Country Term Re-election (of those seeking) Office
United States (1996) 2 88 94 83 Panama (1999) 4 87.5 49.2 43 Chile (1993) 4 76 78 59 Colombia (1990) 4 n.a. n.a. 48 Uruguay (1999) 5 72.3 64.9 47 Brazil (1995) 4 70 62 43 Venezuela (1993) 5 n.a. n.a. 32 Bolivia (2002) 4 n.a n.a. 22 Argentina (1997) 4 26 67 17 Costa Rica (2001) 4 0 0 0 Mexico (1997) 3 0 0 0
Notes: Represents the average of both chambers (40.28 per cent Deputies, 24.20 per cent Senate). Sources: For Uruguay and Bolivia authors’ calculations, United States, Chile, Brazil and Argen- tina, Colombia, Venezuela, Panama. POLITICAL CAREER PATHS IN THE URUGUAYAN CONGRESS 237 in the elections of 1999. In Venezuela during the 1988–93 legislature only 38 per cent of legislators were freshmen.7 Legislative turnover in Brazil is also moderate or moderate–low. Samuels and Mainwaring note that political decisions account for about half of the turnover because a significant share of incumbents decide to run for executive offices, such as governor or vice-governor. Nonetheless, as in Uruguay, ‘the proportion of successful re- election-seeking legislators in Brazil has been increasing since 1986’.8 If these data are correct and one considers the percentage of those legis- lators returning to office, Uruguay falls exactly in the middle of these eight countries in the sample. More specifically, it falls between Colombia and Brazil. But of course, if we do not consider those countries where legislative re-election is not permitted, Uruguayan scores look quite different. Why is it so important to study the patterns of legislative turnover and re-election? It is well known that re-election rates have important conse- quences for the quality of the democracy. Very low re-election rates not only generate deficiencies in legislative expertise, but they can also increase the dedication costs of policy making and promote unwanted behaviour in the internal life of political parties. First, high legislative turnover may impede legislators from gaining expertise and seniority and therefore can reduce the quality of parliamentary work. During the last years, Latin American legislatures have dealt with complex reforms, such as the reform of the state, the opening of the economies and the processes of regional integration. The importance of those decisions requires individuals with high qualifications. Second, high legislative turnover rates can cause great instability in the trajectories of political leaders. Individ- uals who decide to enter the ‘legislative career’ must have, as the Constitution demands, a full-time dedication to that job. The risk of not being re-elected generates personal costs difficult to quantify. In addition, this phenomenon can lead to some sort of elitisation of political cadres where only those who have a guaranteed economic sustenance will predominate in the legislative arena. Third, high legislative turnover rates can generate detrimental behaviour for the parties’ internal life. As we will explain later, during the last 15 years some legislators from the very same sector, with the objective of guaranteeing their re-election, unleashed true fratricide-wars with the simple objective of improving their position on their ballots.9 Of course, this behaviour is contingent upon the type of relationship the legislator maintains with his or her leader, who is the person that holds the power to make the list, and is also dependent on the voting perspectives of his or her political sector. For the aforementioned reasons, it seems that a moderate rate of legislator turnover is the suitable dose for the health of Congress and political parties. In this way, it is possible to avoid the ‘petrifaction’ of political leaders. It also 238 THE JOURNAL OF LEGISLATIVE STUDIES allows the entrance of new representatives without sweeping out the main core of legislators. One of the simplest yet strongest assumptions within the literature that deals with legislatures and legislator behaviour is Mayhew’s assumption that, when permitted, Congressmen are interested in nothing but being re-elected, and it is this main and indisputable motivation which generates predictable behaviour patterns. As Navia argues, ‘seeking re-election does not automatically result in being re-elected, but being an incumbent helps a great deal to secure a seat in the Chamber of Deputies’.10 Nonetheless, in a brand new volume on Latin American legislatures, Morgenstern claims that ‘Latin American legislators (a) are not homogeneous re-election seekers, (b) follow different strategies into office, (c) are restrained by different constitutional arrangements, and (d) operate within a vastly different party alignment. As a result, the Latin American legislatures do not look or act like the U.S. Congress’.11 Uruguay defies, as do its fellow Latin American countries, the assumption that legislators are only re-election seekers. Given the great variation in the number of incumbents seeking re-election in Latin American countries, several scholars question Mayhew’s assumption regarding re-election. For instance, Mainwaring is reticent to extrapolate Mayhew’s assumption to the Brazilian Congress, claiming that it is too restrictive. He says ‘this assumption has the advantage of parsimony, and it is reasonable to assume that winning elections is a major motivating force for most politicians. Nevertheless, it is too restrictive for the Brazilian context, and it must be modified’.12 He is followed by Morgenstern, who says that ‘these data suggest an important impediment in the creation of a unified theory of legislative behavior. Since not all (or even almost all) legislators seek re-election, even within any given country it is incorrect to assume homogenous legislators all driven by a similar motivation’.13 In Uruguay, the legislative election creates a much smaller expectation than the presidential election. Given the characteristics of the institutional design, which establishes the direct election of the government head and the concurrency of presidential and legislative elections, political leaders, analysts and citizens in general show more concern for knowing who the President of the Republic will be than for knowing who the members of the future Congress will be. Phenomena like the election of minority governments and the formation of coalition governments have motivated the study of many aspects of the executive–legislative relationship. Along these lines, miscella- neous subjects such as cabinet formation, legislator productivity and disci- pline, and the influence of the electoral system in the configuration of the parliamentary maps have been analysed and evaluated.14 In all the cases, an important role in the equation of the presidential government is recognised POLITICAL CAREER PATHS IN THE URUGUAYAN CONGRESS 239 and attributed to the legislative branch of government. However, the rate of turnover in the Uruguayan Congress is an under-studied phenomenon. Only Bottinelli15 has analysed this subject for the post-authoritarian elections, and he shows that between the elections of 1989 and 1994 the House of Representatives had a turnover rate of nearly 70 per cent, which is remarkable compared to the rates of turnover in 1942–66, estimated at around 30 per cent.16 According to Bottinelli, what explains this phenomenon is the increase of electoral volatility within the parties during period 1971–94, a double effect of the tri-party competition in the countryside, and demographic changes that determine some sort of first-past-the-post competition within parties.17 We agree that relatively high rates of legislator turnover exist in Uruguay. Nevertheless, we present new analytical and empirical explanations of incumbent turnover. This article is divided into four sections. In the first section the different theoretical elements of legislator turnover are analysed. In the second section some methodological definitions for the research will be offered and the empirical results related to the rates of turnover and legislative re-election. The third section offers possible explanations for the causes of the high rates of turnover. The fourth starts elucidating the fate of incumbents who failed in their re-election attempts, taking as a reference the legislature of 1990–95. And finally, we try to systematise the main conclusions of the study.
INSTITUTIONAL FACTORS INFLUENCING LEGISLATIVE RE-ELECTION Electoral systems generate important incentives for legislators’ careers. The decision to seek re-election and the possibility of obtaining it are closely related to two central elements of the electoral system: the candidate selection process and the magnitude of the electoral districts.18 At the same time, re- election is also contingent upon the levels of electoral volatility that are present in the political system: the higher the volatility the lower the chances of re-election. First, the form of candidate selection can vary according to the character- istics of the system. The main forms of tailoring legislative lists are the follow- ing: (a) closed and blocked lists, (b) closed but not blocked lists, (c) open lists and, finally, (d) customised candidacies. Electoral systems with closed and blocked lists stimulate the existence of strong and institutionalised parties.19 In these cases, party leaders control candidate selection and establish their order on the list.20 Naturally, this favours party discipline during legislative mandates, since those legislators that do not respect party decisions can be punished in the ensuing list making. Closed but not blocked lists diminish the leaders’ capacity for control. This alternative allows the voter to modify 240 THE JOURNAL OF LEGISLATIVE STUDIES the order of legislators that the party presents and might thus increase the likelihood of electing legislators who are not among the party’s favourites. Obviously, this increases the autonomy of legislators during their parliamen- tary mandates and therefore it may diminish the degree of cohesion of parties. Nevertheless, in case a legislator seeks re-election this autonomy imposes certain restrictions over his or her behaviour given that in order to be included on the list, he or she has to maintain a stable bond with the party. Open lists contribute to weak and non-cohesive parties. This system not only allows the voter to modify the order of legislators in the list, but it also gives the citizen the chance of adding or deleting names from the list.21 If this is the case, the luck of the legislator does not depend on the will of the party alone, but on the relationship that is maintained with his or her constituencies. Open lists favour undisciplined behaviour from legislators with high levels of decisional autonomy.22 Finally, first-past-the-post electoral systems rely on uni-nominal electoral districts and in consequence on highly personalised candidatures.23 As in the case of the open lists, this modality favours legislator autonomy and might trigger undisciplined parties. Accordingly, legislators remain more attentive to the particular demands of their constituencies than to the party lines. Second, electoral systems can exhibit great differences of district magnitudes that directly affect the level of proportionality of the system and therefore the re-election opportunities.24 District magnitudes also influence the power of control among party leadership. If closed and blocked lists are in place, the larger the district magnitude the stronger the influence of party leadership. Other things being equal, the smaller the district magnitude the weaker the influence of party leaders and the stronger the local leadership. The Uruguayan electoral system presents closed and blocked lists within multi-member districts. As we have seen, closed and blocked lists favour the party leaders’ control, but in Uruguay the effects of this modality are modified by the presence of the multiple simultaneous vote (MSV). MSV allows the voter to elect the party but at the same time it allows the election of those legislators who will represent the party in Congress. Unlike systems in which each party presents a closed and blocked list, in Uruguay each party presents an important variety of closed and blocked lists that compete among themselves within the party. Intra-party competition softens the rigidity of closed and blocked lists and transforms the Uruguayan system into a kind of intrapartisan preferential vote. This also has a great impact on the possibility of legislative re-election and the level of party discipline. The relationship between the legislator and his or her fraction leader will not be determining, because this system allows the undisciplined legislator to open and run a new list.25 The Uruguayan electoral system contains an important variety of electoral districts in terms of magnitude. POLITICAL CAREER PATHS IN THE URUGUAYAN CONGRESS 241 Senators are elected in a 30 member-unique national district and deputies are elected in 19 multi-member districts of diverse magnitude.26 Taking into account all legislative positions to be filled in the election, Uruguay presents two big districts (the Senate and Montevideo), a medium-size district (Canelones), and 17 small districts that vary between two and four seats each. Third, the level of electoral volatility directly affects the possibility of legislator re-election. The positive or negative variation in the total share of the votes of a party increases or diminishes the chance of incumbent re-elec- tion. Electoral volatility can be calculated with the Pedersen Index, which measures the net change of votes of all the parties between two continuous elections.27 Since 1946 in only four of the 11 elections the electoral volatility surpassed ten per cent (1946, 1958, 1989 and 1994). In 1946, 13.1 per cent was caused by the strong electoral decline of the National Party and the significant growth of the Colorado Party. In 1958, 17.9 per cent volatility was caused by a similar phenomenon, the abrupt decline of the Colorado Party and spectacular growth of the National Party. In 1989 the high levels of volatility, 13.3 per cent, were caused by the appearance of the New Space Party. Finally, in 1994, volatility of 11.6 per cent was related to the virtual tie among the three largest parties and the permanence of a fourth relevant party.
INCUMBENT RE-ELECTION AND TURNOVER We will call rate of parliamentary turnover the variation in percentage terms of the composition of Congress between one election and the next. The starting point for the measurement of the rate of parliamentary turnover is the legislators elected in 1984. The criterion for this measurement will be the changes in the lists of legislators after each election (1989, 1994 and 1999). In other words, we take the list of the 130 legislators elected in an elec- tion and we contrast it with the lists of the elected legislators in the following election.28 We consider both chambers as a single legislative body.29 We call rate of re-election the variation in percentage terms of those legislators seeking re-election and those obtaining it.30 This index measures the success of re-election seekers and it has special significance given that it will give us some clues regarding citizens’ acceptance or rejection of individ- ual parliamentary handling. Several types of legislators are either excluded from or included in calculations: (a) substitute legislators who entered the leg- islature during the previous period and seek to be elected as legislators are excluded; (b) legislators who died during their mandate and legislators that decided to leave politics are considered non-re-elected; (c) legislators who are elected to executive positions (Presidency, municipal government) are considered re-elected; (d) legislators who were included as substitutes of frac- tion leaders are considered re-elected legislators; and (e) legislators who are 242 THE JOURNAL OF LEGISLATIVE STUDIES elected Vice-President of the Republic are considered re-elected because they occupy seat number 130. Finally, it must be said again, our measurements of rate of parliamentary turnover and rate of re-election take into account the Senate and the House of Representatives as a single legislative body, albeit we control for Senate in our models. Therefore, for our purposes, it does not matter whether a senator seeks re-election in the House of Representatives or a deputy does the same at the Senate level.
Turnover and Incumbent Re-election Table 2 shows that the rate of parliamentary turnover has been relatively high in Uruguay during the last three elections, albeit there is a tendency to reduction. In 1989 it reached 66.1 per cent, in 1994 it fell to 63.8 per cent and in 1999 to 53.8 per cent. The average rate of parliamentary turnover during this period is around 61 per cent and it widely surpasses the 30 per cent estimation of turnover for the period 1942–66.31 Such a high rate of parliamentary turnover does not imply an explicit rejection by the citizenship of more than half of the legislators who end their mandates. The electoral supply shows that not all legislators seek re- election. Although an ample majority of legislators do seek re-election, it is also true that some legislators pursue other positions (President of the Republic, Municipal Mayor, to name a few), others retire from public life, and others die during their mandates. Therefore the rate of parliamentary
TABLE 2 RATE OF PARLIAMENTARY RENOVATION AND INCUMBENT RE-ELECTION SEEKERS
New Legislators Success of (rate of Re-Election Seekers Total parliamentary Non-Seekers Seekers (Rate of Failure Election Legislators renovation %) (%) (%) Re-Election %) (%)
1989 130 86 (66.1) 40 (30.8) 90 (69.2) 44 (48.9) 46 (51.1) FA-NEa 14 (70.0) 6 (30.0) TPb 30 (42.8) 40 (57.2) 1994 130 83 (63.8) 43 (33.1) 87 (66.9) 47 (54.0) 40 (46.0) FA-NE 16 (61.5) 10 (38.5) TP 31 (50.8) 30 (49.2) 1999 130 69 (53.0) 36 (27.7) 94 (72.3) 61 (64.9) 33 (35.1) FA-NE 33 (73.3) 12 (26.7) TP 28 (57.1) 21 (42.8)
Notes: (a) FA-NE: Frente Amplio and Nuevo Espacio (b) TP: Tradicional Parties Source: Authors’ elaboration based on Albornoz and Corte Electoral. POLITICAL CAREER PATHS IN THE URUGUAYAN CONGRESS 243 turnover conceals the phenomenon of legislative desertion. During the elec- tions of 1989, 1994 and 1999, approximately 30 per cent of legislators did not pursue re-election. The next question is what happens to those legislators who do seek re-election. In 1989 the rate of re-election reaches 48.9 per cent, in 1994 it ascends to 54 per cent, and in 1999 to 64.9 per cent. In absolute numbers this indicates that in 1989, 44 of 90 legislators obtained their objective; in 1994, 47 of 87; and in 1999, 61 of 94. As the rate of parliamentary turnover tends to decrease, the rate of re-election tends to increase. In 1989, the rate of parliamentary turn- over was 66.1 and TR 48.9; in 1994, the rate of parliamentary turnover was 63.8 and rate of re-election 54.0; finally, in 1999, the tendencies crossed each other when the rate of parliamentary turnover fell to 53.0 and the rate of re-election rose to 64.9 per cent.
EXPLAINING THE HIGH RATES OF PARLIAMENTARY RENEWAL The high rate of parliamentary renovation and its tendency to decrease can fundamentally be explained by the variation in electoral volatility. The evi- dence suggests that the tendency of decreasing volatility might be associated with the fall in the rate of parliamentary turnover and the increase in the rate of re-election. The association of electoral volatility, rate of parliamentary turn- over, and rate of re-election also suggests that the political party to which legislators belong is a key element for their chances of re-election. The data show that legislators from the so-called ‘traditional parties’ (Colorado and National) face greater obstacles in achieving re-election than legislators from the relatively new parties (Frente Amplio and Nuevo Espacio). In other words, legislators of parties that decline in electoral terms have less probability of being re-elected than those from parties that show a sustainable growth. The re-election rates for the legislators from left-wing parties oscillate between 61.5 and 73.3 per cent, while for the traditional parties it fluctuates between 42.8 and 57.1 per cent (see Table 2). Bottinelli argues that a second explanation for the rates of incumbent success while seeking re-election could be internal volatility (at the fraction level).32 Applying Pedersen’s Index of Volatility at the fractional level within political parties we found that the Colorado Party is the one that dis- plays the lower volatility. The National Party and the Frente Amplio show quite similar levels of internal volatility for the elections of 1989 and 1994, but they diverge in 1999. In the elections of 1999 the National Party shows a drastic increase of internal volatility and the Frente Amplio a strong reduction. A contrast between the evolution of intra-party volatility and the rate of re-election of the legislators from each party shows the non-existence of an important bond between both phenomena. 244 THE JOURNAL OF LEGISLATIVE STUDIES Table 3 illustrates the levels of intra-party volatility and the rate of re- election for the three larger parties. In it, the re-election rate does not seem to be sensitive to the changes in intra-party volatility. Indeed, within the Colorado Party a phenomenon contrary to the expected appears, as both the internal volatility and the rate of re-election increase. Nonetheless, re-election rates increase to a much more pronounced degree. In the National Party the behaviour of both variables is still more irregular, because only in 1999 does it behave as expected. In this case, intra-party volatility increases and the rate of re-election minimally diminishes. It is only in the Frente Amplio where the expected behaviour is observed with a clear decrease in internal volatility and an increase of successful rates of re-election. Unlike the relationship between electoral volatility of the assembly taken as a whole and the rates of re-election, dissimilar records in all parties demonstrate the non-existence of an association between intra-party volatility and legislative re-election.33 At the beginning, this article claimed that the success or failure of those re- election seekers was related to certain elements of the electoral system such as closed and blocked lists and district magnitudes. The closed and blocked lists used by fractions in Uruguay grant a great deal of power to fraction leaders. Although some fractions use Conventions, Congresses or other executive bodies to approve their list preparation, in most cases this is nothing more than a mere formality.34 On the one hand, big electoral districts, like the Senate or the Department of Montevideo, make possible the election of legislators without their own electoral capital (votes). Usually, the lists of legislators are headed by figures of national prestige who obtain such an important volume of votes that they guarantee the entrance to the parliamentary arena of legislators barely known by the citizens. This phenomenon is even stronger in Montevideo than in the Senate of the Republic. In the latter, the citizenry knows most legislators. On the other hand, something very different may
TABLE 3 INTRA-PARTY ELECTORAL VOLATILITY AND RE-ELECTION RATES
Colorado Party National Party Frente Amplio
Rate of Rate of Rate of Volatility Re-election Volatility Re-Election Volatility Re-Election
1989 20.3 42.1 52.1 43.7 49.7 58.3 1994 30.8 48.3 47.5 53.1 50.2 68.7 1999 32.2 75.0 60.7 48.0 22.0 74.3
Source: Authors’ elaboration based on Albornoz and Corte Electoral. POLITICAL CAREER PATHS IN THE URUGUAYAN CONGRESS 245 happen in the small electoral districts of the countryside. Of these 17 electoral districts, approximately half choose only two deputies (which is the consti- tutional minimum, product of the demographic change). Under this scheme of competition, parties’ expectations are reduced to obtaining at least one seat in a district, which leads local leaders or deputies to try to head a list that allows them to obtain the intra-party triumph. But to do so, candidates must necessarily have their own electoral votes allowing them to negotiate with other local leaders and/or leaders of the national fractions of the party. This is not always the case and in many circumstances the loss of the fraction endorsement or the impossibility of finding an agreement with smaller caudillos entails an electoral defeat.35 We expect something in between in middle sized departments. Table 4 shows the performances of re-election seekers in small and large electoral districts. Data from Table 4 support the idea that the combination of big electoral districts with closed and blocked lists of fractions increases the chance of re-election for those legislators who seek it. Small electoral districts, on the other hand, offer fewer chances of re-election. Nonetheless, it is not clear whether the differences among large, medium and small departments are statistically significant. In order to test the statistical significance of all variables mentioned in the previous section and to take into account the binary character of our dependent variable (re-election success or failure), we use Binary Time Series Cross Sectional Analysis (BTSCS). We expect that the smaller the electoral district the lower the probability of incumbent re-election. Also, contrary to common knowledge, we believe that intra-party volatility has no effect on the prospects of incumbent re-election. Moreover, if a legislator belongs to a new party (Frente Amplio and Nuevo Espacio), she or he is more likely to be re-elected. Finally, we should also mention another factor that might have an impact on the rate of parliamentary renovation during the post-authoritarian
TABLE 4 FAILURE AND SUCCESS OF RE-ELECTION SEEKERS BASED ON THE SIZE OF ELECTORAL DISTRICTS
Re-election
Electoral Districts Failure (%) Success (%) Total
Senate 18 (29.5) 43 (70.5) 61 (100.0) Big (Montevideo) 51 (46.4) 59 (53.6) 110 (100.0) Medium (Canelones) 10 (45.5) 12 (54.5) 22 (100.0) Small (Rest) 40 (51.3) 38 (48.7) 78 (100.0) Total 119 (43.9) 152 (56.1) 271 (100.0)
Source: Authors’ elaboration. 246 THE JOURNAL OF LEGISLATIVE STUDIES re-election rates: time itself. It is possible that, as time goes by, legislators become more acquainted with legislative life and therefore the seniority of the whole legislature increases. However, there is a problem here. The constitutional reform of 1996 implied a new set of electoral laws.36 We considered including a dummy indicating that the elections of 1999 were held under a different electoral system, but we would then have the problem of auto-correlation (the Pearson’s correlation between both variables is 0.866 significant at the 0.001 level). Therefore, we simply tested whether time definitively affects incumbent re-election rates. Of course, in order to have a more conclusive opinion we should have more data generated through new elections under the same rules of the game. Table 5 reports the results of the models explaining incumbent re-election. The empirical findings fit our theoretical expectations. The overall model has a high statistical significance (p , 0.001) and performs satisfactorily, predicting over 60 per cent of the cases correctly. Three of the five independent variables (Senate, Left and Year) are significant and have the anticipated sign. Since size of electoral district and intra-party volatility are not statistically discernible from zero, we cannot reject the null hypotheses that either of these independent variables has an effect on incumbents’ re-election rates. In Model 2 we drop the variable Senate, given that part of this information is included in the variable Size-Electoral District. The model performs significantly worse than Model 1. In Model 3 we drop variable Size-Electoral District and the overall performance of the model improves. Model 3’s Chi
TABLE 5 LOGISTIC REGRESSION ANALYSIS
Model 1 Model 2 Model 3
Independent Coef (Std. Err.) Coef. (Std. Err.) Coef. (Std. Err.)
Left (þ) 0.708 (0.314) 0.651 (0.318) 0.712 (0.295) Year (þ) 0.065 (0.033) 0.060 (0.035) 0.064 (0.033) Size-Electoral 0.007 (0.168) 0.179 (0.185) – – District (þ) Intra-party 0.011 (0.010) 0.013 (0.010) 0.010 (0.010) Volatility (2) Senate (þ) 0.953 (0.370) – – 0.959 (0.337) Constant 2129.42 (66.394) 2121.15 (71.730) 2129.05 (65.87) Overall Correctly Classified: 63.4% 57.1% 63.4% Log likelihood 2162.593 2165.679 2162.59 Wald chi2(4) 20.25 12.57 20.24 Prob . chi2 0.0011 0.0136 0.0004