Party Capacity in New Democracies: How Regime Type Affects Executive Recruitment

Transcription

1 Party Capacity in New Democracies: How Regime Type Affects Executive Recruitment David J. Samuels and Matthew S. Shugart Prepared for presentation at Ruling Politics: The Formal and Informal Foundations of Power in New Democracies, Harvard University, November

2 1) Introduction When Portuguese radio began playing Grandola Vila Morena on April 25, 1974, it was not possible to imagine what world historical significance the first step of the Revolution of the Red Carnations would have. It was the beginning of what Huntington (1991) famously would call the Third Wave of democratization. In the subsequent thirty years, about sixty new democracies would be born. To be sure, Huntington captured an important moment in world history. Yet scholars quickly noticed limits to global democratization, and turned their attention to failures of democratic consolidation (e.g. O Donnell 1996, Schedler 2001), and to the emergence of illiberal democracies (Zakaria 1997) or competitive authoritarian regimes (Levitsky and Way 2002). In particular, scholars have lamented the failure of political parties to adequately represent citizens interests in many new democracies (e.g. Mainwaring and Scully 1995, Moser 2001, Mainwaring 1999, Levitsky and Cameron 2003; Hicken 2009). Many see political parties as the weakest link in new democracies (e.g. Randall and Svasand 2002, Carothers 2006; Mainwaring and Torcal 2006; Finkel et al, 2007), for example by frequently violating their stated policy platforms (Stokes 2001). This focus on parties shortcomings reflects E.E. Schattschneider s famous dictum which has become conventional political science wisdom-- that democracy is unthinkable in the absence of political parties. If parties cannot become more faithful agents of democratic representation and accountability, or so the thinking goes, then new democracies are in peril of illiberal backsliding. One cannot put a finger directly on any particular element of state-society relations. As with concepts such as legitimacy, democratic representation and accountability are extraordinarily difficult to measure accurately. Broadly considered, capacity refers to parties

3 ability to maintain a consistent role as collective actors in democratic politics nominating candidates, aggregating voters interests by proposing platforms and running competitive national campaigns, and managing government power. 1 Somewhat more precisely and using a different term ( institutionalization ), Mainwaring and Torcal define parties capacity to act as agents of democratic representation as 1) party organizations have a life of their own and are not subordinated to personalities; 2) parties have strong roots in society; 3) voters see parties as legitimate. 2 In this paper, we focus on the first element of party capacity. All activities parties undertake as agents of democratic representation likely flow from their ability to nominate candidates for executive office who embody the organization s collective goals. To our knowledge no research has considered the possibility that party capacity is a function of how they recruit executive talent. Yet the logic is straightforward: if parties lack capacity to keep their agents in check, they will also lack the ability to act as collective agents of representation and accountability. That is, they will fail to become responsible parties--organizations capable of programmatically linking broad groups of voters to government policy-making (APSA 1950; Manin et al. 1999; Kitschelt et al. 2010). To speak of parties capacity to represent citizens interests is to speak of an informal political relationship and variation in executive recruitment offers a useful window into the sources of party capacity. What accounts for variation in the degree to which party organizations subordinate or are dominated by the candidates they nominate for executive office? No party can be assured of fidelity, ex post. Like any collective actor, parties suffer potential agency losses upon nominating 1 Other scholars of parties have used the term, capacity, to refer to such concepts as strategic adaptation (Baker and Scheiner 2004), establishing effective control of government (Lapalombara and Anderson 1992), or mustering winning coalitions on the floor of a legislative chamber (Binder 1997). 2 They also add a fourth element, stable patterns of inter-party competition exist, since they are discussing both party and party-system institutionalization. 3

4 someone to serve as potential national executive. Agency losses can derive from the problem of adverse selection the fact that ambitious politicians have personal goals that may or may not perfectly overlap with the organization s goals---or from moral hazard the fact that any contract is difficult to enforce after it is signed. Mechanisms to guard against agency losses are the essence of what we mean by capacity building: parties with higher capacity are those that structure the incentives of politicians who seek higher office. The key way parties vet and structure potential future leaders incentives is by requiring them to work with their colleagues in the national legislature. Leaders with legislative experience have been socialized to work collaboratively to achieve the parties collective goals; parties know that leaders with such experience are relatively more likely than an outsider to continue to work collectively after winning national executive office. In short, leaders prior legislative experience serves as the key operational indicator of parties organizational capacity. Party organizational capacity may be an amorphous notion. And we recognize that having a life of its own is also an informal notion as far as social-science concepts go. Yet this is because intra-party politics is itself an informal game in which strategizing behind closed doors rather than the party statutes often determine winners and losers. Nonetheless, in this paper we suggest that the broad contours of democratic constitutional design the formal rules of representative democracy play a critical role in shaping parties informal relationship with society. This is because variation in the structure of executive-legislative relations across democratic regimes shapes parties capacity to recruit executives from within their ranks. In earlier work (Samuels and Shugart 2010), using a global sample of all democraticallyelected national executives from , we showed that parties in parliamentary systems have greater capacity to recruit national executives from within the ranks of legislators and that 4

5 such recruitment patterns have powerful consequences for the extent of intra-party cohesion and consistent policy representation. However, because we pooled all democracies, it remains unclear whether our finding was driven by patterns of executive recruitment in older democracies. On the one hand, because most older democracies are in Western Europe, perhaps parties capacity to recruit executives from within the legislature is a regional phenomenon--a product of the particular political context present in earlier waves of democratization (Lipset and Rokkan 1967). If this is the case, our institutional argument is flawed, because parliamentarism may itself be a function of this path-dependent historical context. On the other hand, it is also possible that institutions shape party capacity in both older and newer democracies. To the extent that this is the case, then there is nothing particularly different about the socio-political context of 3 rd Wave democracies. Instead, what matters is the fact that the pool of older democracies is dominated by parliamentary systems, while newer democracies are far more likely to have directly-elected presidents. It remains unclear how parties in newer democracies stack up against those in older democracies. Given the scholarly hand-wringing about party capacity in newer democracies, it is certainly possible that parties in all 3 rd -wave democracies lack the capacity to recruit executives with significant legislative experience. After all, parties in any new democracy--regardless of regime-type--face a difficult challenge in terms of developing capacity simply because a legislature--at least a democratic one--has not been in place for very long. Rather obviously, in their first years, parties in new democracies therefore lack the capacity to recruit national leaders who possess significant legislative experience. However, as time passes, parties in any democratic regime could develop such capacity. The question for the study of representation and 5

6 accountability in new democracies is whether parties operating in different formal democratic institutional contexts develop such informal capacity at different rates, and to different degrees. Perhaps not surprisingly given the gist of our previous work, in what follows we show that in newer democracies, the formal institutional structure of executive-legislative relations significantly shapes parties capacity to recruit executives who are likely to be good agents. As 3 rd wave parliamentary democracies mature, we have already seen a steady increase in the frequency with which their parties recruit prime ministers from within the legislature. However, we observe no such trend for directly-elected executives in either pure or semi-presidential systems. And as for premiers in 3 rd -wave semi-presidential systems, they tend to fall in between pure parliamentary premiers and presidents in terms of their prior legislative experience. Finally, we find no difference in the rate at which parties in parliamentary systems develop this capacity in 3 rd wave cases versus countries that democratized in earlier eras. In short, contrary to Mainwaring and Zoco (2007) for example, it is the type of democracy--and not when it was born--that most shapes politicians incentives and capacity to invest in party-building efforts. These findings undermine conventional generalizations about party weakness in newer democracies: in terms of their informal capacity to recruit insider leaders, parties in newer democracies are not weaker than their counterparts in older democracies--as long as the form of democracy is parliamentary. Party weakness in 3 rd wave democracies is a function of the formal institutional structure. That is, the true source of party weakness in contemporary democracies appears to be the separation of powers. In the next section, we situate the process of executive recruitment theoretically, arguing that democratic regime-type offers important clues to parties capacity to carry out some of the most fundamental roles assigned to them in democracy. We then provide empirical support for this claim. 6

7 2) Parties and Executive Recruitment: Theoretical Framework In this section we explain why differences in constitutional structure impact the way parties select candidates for executive office. Political parties fulfill all the key functions of democratic governance: nominating candidates, coordinating election campaigns, aggregating interests, formulating and implementing policy proposals, and managing government power. However, the manner in which parties fulfill these functions varies considerably according to constitutional structure of executive-legislative relations--specifically whether the system is parliamentary, presidential, or semi-presidential (Samuels and Shugart 2010). Nearly every democracy fits into one of three basic types: parliamentary, presidential or the semi-presidential hybrid. In a parliamentary system, the executive originates from and survives at the pleasure of the legislative majority. Voters vote for only a legislative candidate or party list, and thus may not vote separately for an executive candidate. The prime minister and cabinet collectively depend on the confidence of parliament, which also makes them direct agents of their own parties. This relationship differs fundamentally from what we find in separation-of-powers systems. In pure presidential systems the executive is elected directly and separately from the legislature, and survives independently of the legislative majority s desires, for a fixed term. And under semi-presidentialism, the president is also directly elected for a fixed term, but there is also a prime minister and cabinet dependent on parliamentary confidence. What difference does variation in the organization of executive-legislative relations make for the way parties recruit national executives? It is useful to think of representative democracy as a chain of delegation between principals and agents. The first link in the chain connects voters to elected officials, who are accountable through the process of repeated elections. The 7

8 key difference in the chain of delegation between pure types of executive structure parliamentary and presidential derives from the relationship between voters and the executive branch. In the former type, voters elect a legislature and a legislative majority then invests a government with power, and holds it accountable. Executive authority--vested in a cabinet headed by a prime minister--derives from the results of parliamentary elections, and only indirectly from the voice of the people. The contrast with pure presidential systems is straightforward: in these systems voters directly elect an executive, and the cabinet is an agent of the president, not the legislature. The difference between direct and indirect election of a national executive has important implications for the types of candidates parties select. To win even a slice of political power, parties must pursue multiple and often competing goals: votes, office, and policy (Strøm 1990). Once politicians have decided to join forces, the key problem they confront is one of delegation to a leader: to help make strategic decisions about which goals to emphasize at which times, all parties delegate some degree of authority to a single person who will stand as the party s candidate for national executive office. Whether or not this presidential or prime-ministerial candidate is also de jure the head of the party organization, parties entrust prospective presidents and PMs with their collective reputation. This process of delegation, while offering parties the advantage of a national leader who articulates the party vision in the campaign, also entails a potentially great danger. Principalagent theory tells us that all organizations face the problem of adverse selection, the risk that the group may nominate for the top spot an individual who is a poor leader, or who only partially shares the organization s collective interests. Our fundamental claim is the following: the formal constitutional structure shapes the nature of this contracting problem because it powerfully 8

9 structures the choices parties must make when nominating someone. Specifically, the danger of adverse selection is minimized in parliamentary systems but is greatest in pure presidential systems. Under parliamentarism a single electoral goal drives the resolution of any and all collective action problems for the party: winning legislative seats. All collective decisions regarding delegation to a leader focus on what the party needs to accomplish in legislative elections and in the legislature in order to maintain its brand name. Thus a party s candidate for the top office acquires popular legitimacy only indirectly, through the party s own internal selection mechanism and the parties collective performance in legislative elections. By contrast, when the executive branch is constituted via a direct presidential election, the candidate selected by the party must win an entirely separate contest. The strategies that the presidential candidate follows to win a nationwide plurality or majority of the votes may not fully overlap with those that the party chooses in its quest for legislative seats. In fact, because voters vote separately for president and legislature, the voting constituency of the president may differ quite substantially from that of the legislative party. Precisely because of the separate election, parties may even nominate a presidential candidate who only partially shares the party s goals--because the type of politician who makes a good servant of the party s interests may be poorly correlated with the type who has wide electoral appeal. What s more, once in office, the president s term is fixed, implying that the legislative party and the party organization lose the ability to control the executive s actions in the same way that a parliamentary party can, through the ultimate threat of removal before the next election. The implication for leadership selection is clear: because executive and legislative origin is fused in parliamentary systems, parties can informally vet potential executive candidates by requiring them to serve in legislative institutions for a period of time before being nominated for 9

10 the top job. Through service in the legislature, ambitious politicians can be screened and monitored over a period of time. In any case, parties know that any future leader will be responsible to the parliamentary caucus. This gives us good reason to expect prime ministers in parliamentary democracies to have completed a period in the legislature prior to entering office. We expect parliamentary PMs to have served in the legislature, and to have served longer, than directly-elected presidents, due to the incentive for parties in separation-of-power systems to look beyond their ranks of legislators for candidates who possess broad enough electoral appeal to win a direct popular election. This argument should ring familiar, as it is a more precise elaboration of Linz s (1994) claim that presidents tend to be outsiders while prime ministers tend to be insiders. Although Linz did not make this point, we extend his logic to suggest that there is no reason to expect third wave systems to differ from older democracies in these terms. Again, the formal institutional structure of democracy rather than the era in which it emerged--that most shapes party capacity to act as agents of democratic representation. 3) Institutions, Age of Democracy, and Party Capacity: Hypotheses Our argument focuses on how formal institutions channel parties leadership selection decisions and how these decisions embody a core element of party capacity. It is possible, however, that non-institutional factors might account for deficient party capacity in newer democracies. Mainwaring and Torcal s argument, for example, is rooted in a historical sequences approach pioneered by Lipset and Rokkan (1967): parties in older democracies, the argument goes, emerged as relatively capable agents of democratic representation and accountability due to particular historical patterns of modernization, industrialization, Church- 10

11 State conflict and other (unrepeatable) dynamics that are absent or far weaker in 3 rd wave democracies. This fundamentally sociological approach suggests that by implication, given the absence of such factors, parties in newer democracies are necessarily weaker agents of democratic political representation. In contrast, our institutional hypothesis suggests that socio-economic factors and particular historical sequences are not the key factors shaping party capacity. Instead, weak party capacity is rooted in constitutional structure: while parliamentarism can ameliorate the dilemmas parties face in recruiting executive candidates, parties in presidential systems confront greater challenges of finding candidates who can win a presidential election while also being a good party agent. Let us elaborate specific hypotheses at this point. Most basically, we expect executive recruitment patterns to differ across democratic regimes as they age. If a new democracy is parliamentary, as it matures parties will have more opportunities to use the legislature as a proving ground to monitor potential future prime ministers. That is, in parliamentary systems PMs years of prior service as a legislator should increase as the democracy ages. This should be the case regardless of when the democracy emerged. In contrast, if Mainwaring and Torcal are correct, then we should see a difference in party recruitment capacity even in parliamentary systems between older and newer democracies. In contrast, we expect party recruitment capacity to be lower in presidential democracies, because of the leadership-selection dilemmas posed by the regime type. As noted above, the set of skills needed to win a separate and direct contest for a single national office may correlate only loosely with the skills needed to be a good party servant. Spending significant time in the legislature--with all the necessary compromises legislators must make both with party colleagues 11

12 and with other parties--might even be a liability in a campaign for executive office. Under the separation of powers, presidents are less likely to have significant legislative experience in new democracies. Again, this should be the case regardless of when the democracy emerged. In the case of semi-presidential systems, capacity building should be limited by some of the same tensions inherent in pure presidentialism. On the one hand, parties need executive candidates who can win a popular election. On the other hand, the mixed nature of these regimes--specifically that in addition to the president there is a premier, who depends on the confidence of parliament--suggests that patterns of capacity building could be likewise mixed. Still, partisan control over the premiership in all semi-presidential systems is attenuated by presidents influence, because presidents normally have authority to nominate (and sometimes outright to appoint and fire) the premier, breaking parties hold over executive recruitment that applies in pure parliamentary systems. 3 However, assemblies in semi-presidential systems retain some influence, because the premier does depend on legislative confidence. Thus, in terms of whether executives legislative experience will increase with democracy s age, we expect an intermediate level of party capacity in the case of premiers in semi-presidential systems. We further expect the impact of age of democracy on premiers prior legislative experience to vary according to the two subtypes of semi-presidential democracy first noted by Shugart and Carey (1992). In other words, the formal distinction between the weak presidency of premier-residential regimes and the strong presidency of president-parliamentarism should also impact party capacity. In the premier-presidential subtype, the prime minister s and cabinet s survival in office depend exclusively on the confidence of the legislative majority, implying that the president has relatively limited bargaining power to impose his or her choice of 3 See Shugart (2005) for details on the specific powers of presidents in semi-presidential systems. 12

13 premier. Yet in the president-parliamentary subtype, the president has formal authority to dismiss the premier, making the latter an agent of the president as well as of the legislative majority. Given this formal institutional difference, premiers in the first subtype should resemble their counterparts in pure parliamentary systems: in both systems, PMs average prior legislative experience should increase with democracy s age. However, presidents in premierpresidentialism should resemble other directly-elected executives in terms of prior legislative experience. As for president-parliamentary systems, given the directly-elected executive s formal powers, we have no reason to expect either premiers or presidents to be notably different from presidents in pure presidential systems. In sum, parties informal capacity as agents of popular representation is not shaped primarily by geographic region or the era in which democracy emerged, Parties should develop informal capacity as voters agents as democracy ages, yet the formal institutional context should powerfully shape such capacity. Parties in parliamentary democracies should develop capacity quickly as the system ages; this will show up in an increased tendency for their prime ministers to have prior legislative experience. In contrast, we expect significantly lower capacity to recruit insider executives in pure presidential systems, regardless of their age or when they were born. Parties in semi-presidential systems should be relatively presidentialized, due the popular election of a key executive politician. However, parties in premier-presidential systems can be expected to exhibit an increased tendency to draw their premiers from the ranks of legislators, but still at a lesser rate than for parliamentary parties. 4) Data and Measures 13

14 We draw on our own database of national executives, which contains information on the career paths of over 800 executives (prime ministers and directly-elected presidents) from all democracies from (Samuels and Shugart 2010). 4 These data provide information on our dependent variable--whether and how long a national executive served in the legislature prior to assuming the top office. These same data provide information on whether a regime was presidential, parliamentary, or semi-presidential (and which subtype). Our main independent variable, is age in years of Democracy, where Democracy is defined as scoring five or better on the Polity IV Democracy indicator for at least five consecutive years. A total of 59 countries qualify; Table 1 summarizes the data. Of primary interest are the third wave democracies those that have undergone democratization since This group provides about half of our sample - a set of 405 executives--267 prime ministers and 138 elected presidents. Of the prime ministers in this sub-sample, 111 are from pure parliamentary systems, and 156 are premiers in semi-presidential systems. For each country we know the year democracy began, and thus can calculate the age of democracy at the time any prime minister or president ascended to the position. --Table 1 Here-- 5) Comparing Regime Types on Party Capacity: Analysis Because our sample covers a relatively long time-period, we can determine whether newer democracies whatever the constitutional format--differ from older democracies in terms of building capacity, as measured by their tendency to recruit executives from within the legislature. Indeed, because the 3 rd wave started right in the middle of our 60-year time-period, 4 All data are available at 14

15 we can also compare the first 30 years of older democracies against the experience of third-wave democracies thus far. Executive experience by regime type and time period: Aggregate data Let us first consider the overall prior legislative experience of presidents and prime ministers, using difference-of-means tests to see if broad patterns match our expectations. This test only includes politicians who obtained legislative experience (if any) under democratic auspices. (We consider those whose gained experience under prior non-democratic regimes below.) Table 2 shows that about one-third of chief executives in third wave pure presidential systems have prior legislative experience, compared to nearly 9 in 10 parliamentary prime ministers. Including presidents in semi-presidential systems barely alters the picture. --Table 2 Here-- Table 2 also reveals that executives in older presidential systems have somewhat more legislative experience (56.5%) than those in 3 rd wave pure presidential democracies (33.8%). 5 This difference in prior legislative experience between newer and older pure presidential democracies is statistically significant. However, no significant difference appears between older and newer pure parliamentary systems (89% vs. 93.9%). 6 However, party capacity in both older and newer presidential systems is much lower than in both older and newer pure parliamentary systems. In Table 3 we expand our consideration of the legislative experience of prime ministers to include semi-presidential systems, including their subtypes. These comparisons include only the 5 This difference of means between newer and older (pure and semi-) presidential democracies is likewise significant. 6 There are too few examples of semi-presidential systems among the older democracies to include them in the analysis. 15

16 third wave democracies, because too few examples of this regime-type exist prior to We expect semi-presidential premiers to have significantly less legislative experience than parliamentary PMs, and we that those operating the premier-presidentialism will have greater experience than their counterparts in president-parliamentary systems. Difference-of-means tests provide strong support for each hypothesis: Nearly two thirds of premiers under premierpresidentialism have prior legislative experience, against less than half of those under presidentparliamentarism. Despite this tendency, they are still significantly less likely to have legislative experience than are PMs in pure parliamentary systems. --Table 3 Here-- These findings suggest that the party capacity in semi-presidential regimes may be enhanced by adoption of the premier-presidential subtype--the one in which the formal powers of the presidency do not include dismissal of a premier or the cabinet. Nonetheless, parties in such systems still do not possess the same sort of capacity as their pure parliamentary counterparts. In short, the presence or absence of a directly elected presidency matters for recruitment of executive talent from within the legislature--and so does the formal power of the president if there is one. Presidents have less legislative experience than parliamentary prime ministers, and premiers in semi-presidential systems have less experience than their counterparts in pure parliamentary systems. All of these findings support our argument that constitutional structure shapes parties capacity to minimize adverse selection problems arising from executive recruitment. Moreover, we also confirm that parties in 3 rd wave parliamentary systems exhibit the same level of capacity to recruit premiers with legislative experience as parties in older parliamentary democracies. 16

17 Age of democracy and years of experience: Statistical analysis of 3 rd wave democracies So far we have looked only at the aggregate levels of legislative experience of executives in different democratic regimes. We now proceed to a statistical test of our hypothesis about the impact of the maturation process of different types of 3 rd wave democracies. We expect executives number of years of prior legislative experience to increase with age of democracy. If the new democracy is parliamentary, partisan capacity should emerge early in the regime s life, and PM s experience level should increase as democracy matures. However, we expect lower rates of capacity building in presidential systems, with semi-presidential parties recruitment patterns for premiers falling in between. We report regression results in which executives years of legislative experience, denoted as L, is the dependent variable. The primary independent variables of interest are age of democracy, denoted D, and regime type. The nature of our data suggests a logarithmic transformation, because a linear model run on raw quantities would predict (at some point) unrealistically long-lived politicians. Moreover, we expect a one-year increase in the age of democracy to have less impact on executive s prior legislative experience at high values than at low ones. That is, the difference between having a 30 year old democracy and one that is 29 years old on party capacity is likely less than that between democracies that are five and four years old. A further important methodological point is in order: because L can take the value of zero when an executive has never served in the legislature, we would lose all such observations if we took the logarithm, because the log of 0 is not definable. Thus we actually take the log of L+1, adding one year to each executive s actual number of years of legislative service. The regression format thus will be as follows: log 10 (L+1) = log 10 a + k log 10 D, 17

18 excluding control variables (discussed below). This equation can then transformed into a fixed exponent format by taking the antilog of L, D, and a. Thus, solving for L, we have L = ad k 1. This format simply reflects the equation of a straight line between two axes, both of which are logarithmic. (See Taagepera 2008 for an elaboration of these methodological points). As for the main independent variable, the key to the equation is not D simply the number of years of experience as a democracy but rather k, the exponential term on D (which is the same as the coefficient on the log of D in the regression equation). If all executive candidates were outsiders (i.e. they had zero prior legislative experience), logically the exponent k on D would be zero. And if all executives had served in the legislature as long as the regime had been democratic--a possibility that defies human mortality at some point in the life of a democracy--then logically k=1. For regimes that provide their parties with the incentive and the means to vet their executive candidates within the legislature--that is, parliamentary systems--we thus expect k to be considerably greater than zero, but also less than 1. For elected presidents, on the other hand, we expect k to be less than it is for parliamentary systems, and we expect the difference in slopes to be significant. We run two basic regressions using OLS with clustering of standard errors by country: one including only data from pure presidential and parliamentary systems to test for differences in party recruitment capacity between the pure-type regimes, and a second that incorporates 18

19 semi-presidential systems. 7 Because of the typically short tenure of premiers in semi-presidential systems, the full model has almost twice as many observations as the pure-type model. 8 To test for differences across democratic regimes, in both models we add a dummy variable for prime minister and an interaction of this variable with the logarithm of D. The comprehensive model also includes a dummy for semi-presidential premier and an interaction of this with the logarithm of D. These variables test each of our core hypotheses: that the increase in L as democracy ages will be greatest for parliamentary prime ministers and least for presidents (of any type), and in between for semi-presidential premiers. The regressions also include control variables for executives who had experience in a prior regime, either authoritarian or an earlier democracy. These are included because if we are interested in the extent to which parties build capacity to recruit from within the legislature in a new democracy, we need to control for executives experience under different institutions, particularly in light of the fact that some third wave democracies emerged from authoritarian regimes that nonetheless offered politicians opportunities to gain legislative experience. Table 4 reports results: Model 1 for pure-type executives and Model 2 for all regimetypes. Transforming these various estimates into their fixed-exponent formats and setting the prior-regime dummy variables to zero, 9 we get: (1) L= 1.5D.46 1 for parliamentary systems and 7 The reason for clustering is that observations within any one country are unlikely to be independent of one another. Clustering avoids the danger of under-estimating standard errors on our coefficients. 8 It is not simply the case that semi-presidential systems double the total because there are always two executives in such systems. In fact, our data include almost three times as many premiers (156) as presidents (56) in semipresidential systems. This is because prime ministers come and go more frequently than presidents in such systems. 9 The transformation is done as follows, using Model 1 as a guide: The exponent for presidents is simply the coefficient on log(age of democracy), while we must take the antilog of the constant. For parliamentary PMs, we need the sum of the coefficient on log(age of democracy) and the coefficient on the interaction term. This sum is our exponent, while our constant term is the antilog of the regression s constant, plus the coefficient on the PM dummy. 19

20 (2) L= 1.5D.17 1 for pure presidential systems. --Table 4 Here-- Equations (1) and (2) confirm our central expectations: executives in 3 rd -wave parliamentary systems gain increasing prior legislative experience as democracy ages, and at a greater rate than is the case for executives in pure presidential systems. 10 The difference between the exponents on D for both systems (that is, the coefficients on logd shown in Table 4) is statistically significant, with that for presidents in the pure type (.17) being just over a third of that for parliamentary prime ministers (.46). In fact, the ratio of the two estimated exponents (.17/.46=.37) almost exactly matches the actual ratio of the percentages of executives with any prior legislative experience in these two regime-types (.34/.89=.38). Thus even limiting our sample to just newer democracies, parliamentary prime ministers are more than three times as likely as presidents to have prior legislative experience, and the rate of increase in recruiting experienced executives for parliamentary parties is likewise more than three times that for presidentialized parties. Taken together, these aggregate percentages and estimated exponents provide strong support for our hypotheses about the impact of formal regime structure on parties capacity building as democracy matures. To illustrate these results further, Figure 1 plots the results of a Clarify simulation of the results from Model 1 in Table 4. First, note that the 95% confidence intervals on the estimates for pure parliamentary and presidential systems are separate from one another by the fourth year 10 Why do the equations, with their constant terms greater than 1.0, predict more than one year s worth of experience for the executive in a democracy s first year, even when we have controlled for those who obtained legislative experience in a previous regime? The reason is that many transitions begin with a constituent assembly. In many cases Polity does not score these interregnum years as fully democratic because the transition remains incomplete. For our purposes, we consider service in such an assembly to be equivalent to legislative experience, thus allowing many politicians to have a year or so of experience before ascending to the new democratic regime s executive. 20

21 of democracy. Second, after about the seventh year the lower bound of the 95% confidence interval for presidents is almost flat, highlighting the fact that parties in pure presidential systems have very limited capacity to recruit insider candidates even in relatively mature systems of this type, parties nominate many successful presidential candidates with no legislative experience at all. In contrast, for parliamentary systems we see a clear and continuing increase in L up to the thirty-year maximum allowed by our data. --Figure One Here-- In short, given that the line L=D is an upper bound on those who started their careers within the current regime, there is a large space below that line in which the predicted values could have appeared. Yet the empirical results for parliamentary and presidential executives closely match our expected differences between the principal regime types. As for hybrid systems, Model 2 in Table 4 confirms that the premiers of semi-presidential systems have increased legislative experience at a rate that is in between those for the pure-type executives: (3) L= 1.8D.22 1 A Clarify simulation (not shown) confirms a statistically significant difference between the legislative experience of semi-presidential and parliamentary prime ministers but only for regimes older than about 12 years. The greater variability for premiers in semi-presidential systems probably results from the fact that presidents have considerable discretion in the selection of a premier, except in cases of cohabitation, when the opposition controls parliament (Samuels and Shugart 2010). Some presidents choose as premier a politician with welldeveloped links to legislative parties, while others appoint personal cronies with no legislative experience. The former type of premier will resemble their counterparts in parliamentary systems, while the latter will resemble presidents in their degree of prior legislative experience. 21

22 Table 5 reports results that further probe semi-presidential systems. The model for all semi-presidential premiers simply confirms Equation 3 above, while the other two models offer separate estimates for each regime subtype. It is immediately obvious that all the explanatory power of age of democracy comes from the premier-presidential subtype, as no relationship exists for president-parliamentary premiers. Thus, just as we as expected, there is more evidence of party capacity-building in 3 rd wave premier-presidential systems than in presidentparliamentary regimes, at least as concerns their premiers. However, the overall poor fit for this model makes us cautious in drawing as firm a conclusion as compared against the clear contrast between pure presidentialism and parliamentarism. --Table 5 Here-- We can get a more detailed picture by inspecting plots by executive type. Figure 2 graphs results for 3 rd wave parliamentary prime ministers (upper panel) and for presidents (lower panel). Each panel includes a diagonal line at L=D. Any points well above this line are executives who gained much or all of their legislative experience under a prior regime. 11 If the prior regime was authoritarian, the symbol for that executive is shaded; otherwise, experience was gained in a previous democracy. The bottom panel also distinguishes presidents in pure and semipresidential systems with different symbols. --Figure 2 Here-- The figures reveal that plotted points for parliamentary prime ministers mostly follow the regression line, even though some appear well above it, close to the L=D line. These prime ministers had served in the legislature almost as long as their country had been a democracy (in our data, Spain s José Luis Rodríguez Zapatero is the champion in this respect). Yet several 11 As discussed in a previous note, it is possible for an executive to have a legislative experience and yet have L>D, for instance in the case of service in a transitional constituent assembly. 22

23 had little or no legislative experience before assuming the top office, as was the case for one Czech and two Hungarian PMs who entered office after their countries had been democratic for twelve years. However, thus far no prime minister in any third wave parliamentary democracy older than fifteen years has come to the position without prior legislative experience. Of course, many countries had less than fifteen years of democracy at the endpoint of our dataset, so we can t rule the possibility that a parliamentary PM with no legislative experience will come to power in a much older democracy at some point in the future. 12 In the lower panel of Figure 2, we see a very different pattern. Far more presidents had no legislative experience before being elected. Examples in democracies at least twelve years old at the time of the president s election include Valdas Adamkus in Lithuania, Yayi Boni in Benin, and Michelle Bachelet in Chile. We find such examples all the way out to the maximum observed age of democracy, thirty years. Still, not all presidents are outsiders. Even as there are many presidents with no prior legislative experience, some have served sufficient time to place between the regression estimate and the L=D line. 13 Consider the case of Portugal, a semipresidential system and the original third-wave democracy: In 1996, when Portuguese democracy was twenty years old, Jorge Sampaio was elected following sixteen years of legislative experience. Yet ten years later Portugal swung to the other end of the range: its next president was Aníbal Cavaco Silva, who had no legislative experience whatsoever. 12 The following parliamentary democracies in our sample are 15 years old or older as of the date of the most recent executive to enter our dataset: Estonia, Greece, Hungary, Papua New Guinea, Solomon Islands, Spain, and Turkey. 13 If the regression that is reported in Table 4 is repeated only on executives whose years as a legislator were greater than zero, the resulting equation is L= 3D For parliamentary prime ministers with nonzero legislative experience it would be L= 2D The difference in exponents is significant, but the predictions of these equations overlap considerably, owing to the different intercepts. Of course, we can t just ignore the many executives with zero experience, but the point of the exercise is that when parties in presidential or semi-presidential systems do turn to their legislative ranks to recruit a (winning) presidential candidate, they tend to choose a fairly experienced legislator, and these presidents length of prior legislative service increases as democracy matures. Yet even mature democracies continue to have presidents with no such experience. 23

24 It is important to note that our results on the impact of democratic regime type hold up even when controlling for legislative experience gained under either a prior non-democratic or democratic regime. In Figure 2, the points well above the L=D line represent such executives; quite simply, the panels suggest that many more executives in pure or semi-presidential democracies obtained legislative experience in prior democratic or authoritarian regimes, while few PMs in parliamentary systems did. In fact, about 12% of all presidents had some legislative experience under a non-democratic regime, while less than 4% of parliamentary prime ministers did, a statistically significant difference. In addition, far more presidents in pure presidential systems had legislative experience under a previous democratic regime (14%) than did parliamentary prime ministers (8%). 14 Does this finding on service in prior regimes enlighten us further about party capacity under parliamentary and presidential (pure or semi-) regimes? We can only speculate, but we suspect it does. It could be that the lower capacity-building observed in presidential systems makes parties (and voters) in these regimes more prone to select personalities who were already prominent before the current regime was founded. This explanation would be consistent with our theoretical claim that parties in parliamentary systems prefer candidates they know from close observation as legislative colleagues in the current democratic regime. In addition, parliamentary parties need to rely relatively less than their presidential counterparts on candidates who have spent years cultivating personal popularity directly among voters. In presidential and semipresidential systems, by contrast, parties have an incentive to nominate candidates with a personal appeal, and often, especially in the earlier years of a new democracy, such candidates will be those who already had a political career under a former regime. Thus difference in priorregime experience tends to support our claim that presidentialism limits party capacity-building. 14 The significance level of this difference is just over

25 Are Third Wave Democracies Distinct? Do 3 rd wave democracies develop party capacity at a lower rate than older democracies did? In particular, do third wave parliamentary democracies develop party capacity at a lower rate than older parliamentary democracies? Recent scholarly claims, which focus mainly on other aspects of party capacity, imply a positive answer to these questions (e.g. Mainwaring and Torcal 2006; Mainwaring and Zoco 2007). To assess this argument, we repeat our analysis of the impact of the age of different types of democracy on parties capacity to recruit executives with legislative experience, but this time compare party capacity across waves of democracies. Several caveats regarding comparability of older and newer democracies are in order. First, older democracies are not perfectly comparable to third wave cases because many earlier transitions did not replace authoritarian regimes but reestablished democracy following WWII s relatively brief but painful hiatus. Additionally, while nearly all of the parliamentary democracies that emerged or were reestablished in the decade and a half following the Second World War have survived to the present day, many of the presidential democracies did not. In fact, this trend of breakdowns of presidential democracy in the 1960s and 1970s is what originally inspired Linz (1994) to advance his influential argument that presidentialism per se made democratic survival difficult. Table 6 reports results. In Model 1 we consider only the first 30 years of older parliamentary democracies; Model 2 includes all parliamentary democracies, of any age; and Model 3 is for the first 30 years of the older presidential democracies. Model 1 shows almost the same rate of growth in legislative experience of prime ministers as we saw in the third-wave parliamentary democracies above. Model 1 results in an equation of L = 1.8D.47 1, while the 25

26 result for 3 rd wave was L= 1.5D.46 1 (Equation 1, above). Yet surprisingly, when we include all years of the older parliamentary democracies, we actually find that the exponent is lower than it is for newer democracies (.38 vs..46)! This finding--that third wave parliamentary democracies exhibit no less, and perhaps even greater capacity to recruit from within their ranks than older democracies is certainly inconsistent with hand-wringing that parties in all third wave democracies are relatively weak. It also supports our contention that the key factor limiting party capacity is institutional rather than a function of historical sequences or sociological factors. --Table 6 Here-- As for older presidential democracies, 15 Model 3 in Table 6 results in an equation of L= 1.6D This shows a somewhat greater rate of increase in legislative experience for older presidential democracies than we found in the third wave, where Equation 2 was L= 1.5D However, the coefficient for the older presidential democracies fails to meet statistical significance, in part because we are dealing with just 20 cases. Third wave presidential systems could be building capacity more slowly than their earlier counterparts, but results are inconclusive. 16 We just don t have enough older presidential democracies to know. In the final analysis, we find very clear evidence that the capacity building of parties in parliamentary democracies is much greater as democracy matures than is the case in presidential democracies. This finding holds whether or not we pool data across all cases or break the data 15 We do not show a regression on the older presidential democracies of any age, which results in no relationship. This is due to the inclusion of the US, where very few presidents have had legislative experience in recent decades. 16 A separate regression (not shown) likewise failed to show significant difference across eras for presidents. However, we can again compare the exponents of the two pure types among the older democracies to the ratios of the aggregate percentages of executives with prior legislative experience. Table 2 shows these latter percentages to be 93.9% for prime ministers and 56.5% for presidents, a ratio of If we confine the analysis to the first thirty years only, as we should, the percentages drop to 90.4% and 56.3%, a ratio of This is almost precisely the ratio of the exponents for the two pure-type executives in the first thirty years of the second wave: 1.62 (.47/.29). Thus the relationship of what we have termed capacity building in the two types bears the same relationship to the aggregate patterns in the second wave as we found in the third wave. So capacity building in third wave presidential systems could be lower than in the second wave, the small number of second wave presidents and the insignificant coefficient in Model 3 of Table 12 notwithstanding. 26

27 down into older and newer democracies. When we attempt to compare cases either within or across era, we confirm that parliamentary democracies develop capacity relatively quickly regardless of when they emerge, while pure and semi-presidential 3 rd -wave democracies do not. 4) Conclusions and Implications Political scientists traditionally have held that political parties are essential to democracy. In theory, parties perform all of the basic functions necessary for a political system to maintain a connection between voters and those who govern them: for example, recruiting candidates, developing collective policy programs, and organizing government business. In the real world there is often a great disjuncture between theory and practice. In particular, scholars have recently expressed concern that parties in many third wave democracies lag behind their counterparts in older systems in terms of their capacity to fulfill the tasks democratic theory assigns to them. We question whether party capacity is a function of the era in which democracy was established or the institutions under which democracy operates. We analyzed one aspect of party capacity to serve as a link between voters and policy-makers: the process by which they recruit leaders who will later serve as national executives. To avoid adverse selection problems, parties have powerful incentives to recruit leaders who will remain faithful to the group s principles and enact the group s goals. To achieve this outcome, parties have a more specific incentive: to rely on insiders, politicians who have worked with their party colleagues in the legislature, where they can observe the potential executive in action and structure his or her incentives. Early in any democracy s life the pool of experienced legislators is limited. As a result, parties may be forced to rely on outsiders who lack legislative experience, or on politicians 27

28 who gained legislative experience under a prior regime. Yet as any democracy matures, the pool of experienced insiders increases. Thus it is possible that parties will recruit executives with more legislative experience as democracies age. Our key claim, which builds on a broader argument we develop elsewhere (Samuels and Shugart 2010) is that the formal institutional structure of legislative-executive relations shapes the extent to which parties will tend to recruit insiders, regardless of when or where the democracy emerged. We showed that parties capacity to recruit insiders is very much in evidence in parliamentary democracies, where the executive originates from within the legislature, and survives at the pleasure of the legislative majority. Parties in parliamentary democracy can ensure the selection of executives who will serve their interests. In contrast, when the executive emerges from an electoral process that is separate from that for the legislature, parties have far less control over the types of executives they recruit. Due to the separate election, they must ensure that the candidates they nominate are capable of appealing directly to voters. Thus parties in pure presidential systems exhibit far lower capacity to recruit from within as their parliamentary counterpart. Our analysis also showed that parties in semi-presidential systems mimic the regime s hybrid format and exhibit intermediate capacity building. The premier-presidential subtype exhibits increased experience over time, particularly for its premiers, consistent the formal balance of inter-branch power in this subtype, which is tilted in favor of the legislature and thus the parties. Nonetheless, capacity building remains relatively lower in these regimes than in parliamentary systems, where there is no separate presidency to break the chain of delegation running from voters through legislative parties to the executive. 28

29 In short, democratic regime-types--presidential, parliamentary or semi-presidential-- shape politicians and parties strategies, which in turn shape the party-voter linkage. The literature on parties and political representation in democracy assumes it is normatively desirable for parties to recruit executives from within the ranks of legislators. To the extent that this is true, our results suggest that parliamentarism strengthens programmatic representational linkages, while presidentialism weakens them. Parliamentary systems appear to offer the most suitable conditions for strong parties to develop, regardless of the era in which democracy emerges. 17 Moreover, premier-presidentialism could be preferable to pure presidentialism or the presidentparliamentary subtype. Premier-presidentialism allows considerable flexibility--some presidents are insiders while others are outsiders--a feature that may be desirable for party-building in contexts where such capacity might otherwise be quite limited. Our findings clarify the discussion about the nature of party weakness in 3 rd wave democracies. The source of such weakness is institutional: the fact that most recent transitions to have created democracies with a directly-elected executive. Where parliamentarism emerges, parties are likely to quickly develop the capacity to overcome adverse selection problems by recruiting prime ministers from their legislative ranks. Scholars should pay closer attention to the role of the constitutional form of executive-legislative relations than to the distinction between third-wave and older democracies when it comes to this important dimension of party capacity. In the end, it is the type of democracy--and not when it was born--that most shapes politicians incentives and capacity to invest in party-building efforts. 17 However, it is not true that capacity is conditioned only by democratic regime-type, because democratic institutions may be endogenous to socio-economic conditions: separation of powers systems are more likely to be adopted where party development in highly unequal societies (Shugart 1999). 29

30 References American Political Science Association Toward a More Responsible Two-party System: A Report of the Committee on Political Parties. American Political Science Review 44(3): 1 14 (supplement, part 2). Baker, Andy and Ethan Scheiner (2004). Adaptive Parties: Party Strategic Capacity under Japanese SNTV. Electoral Studies 23(2): Binder, Sarah Minority Rights, Majority Rule: Partisanship and The Development of Congress, New York: Cambridge University Press. Carothers, Thomas The Backlash Against Democracy Promotion. Foreign Affairs 85(2): Finkel, Steven, et al The Effects of U.S. Foreign Assistance on Democracy Building, , World Politics 59(3), April Hicken, Allen Building Party Systems in Developing Democracies. New York: Cambridge University Press. Huntington, Samuel The Third Wave: Democratization in the Late 20th Century. Norman: University of Oklahoma Press. Kitschelt, Herbert et al Latin American Party Systems. New York: Cambridge University Press. Lapalombara, Joseph and Jeffrey Anderson Political Parties. " Routledge Encyclopaedia of Government and Politics. London: Routledge, Levitsky, Steven and Maxwell Cameron Democracy without Parties? Political Parties and Regime Change in Fujimori s Peru, Latin American Politics and Society, 45(3):

31 Levitsky, Steven, and Lucan Way The Rise of Competitive Authoritarianism. Journal of Democracy 13(2): Linz, Juan J Presidential or Parliamentary Democracy: Does It Make a Difference? In The Failure of Presidential Democracy, Volume 2: The Case of Latin America, ed. J. Linz and Arturo Valenzuela. Baltimore: Johns Hopkins University Press, Lipset, Seymour Martin, and Stein Rokkan Party Systems and Voter Alignments: Cross- National Perspectives. New York: Free Press. Mainwaring, Scott P Rethinking Party Systems in the Third Wave of Democratization: The Case of Brazil. Stanford, CA: Stanford University Press. Mainwaring, Scott, and Timothy Scully Introduction: Parties and Party Systems in Latin America. In Building Democratic Institutions: Party Systems in Latin America, ed. Scott Mainwaring and Timothy Scully. Stanford, CA: Stanford University Press, Mainwaring, Scott, and Mariano Torcal Party System Institutionalization and Party System Theory after the Third Wave of Democratization. In Handbook of Party Politics, ed. R. Katz and William Crotty. London: Sage, Mainwaring, Scott and Edurne Zoco Political Sequences and the Stabilization of Interparty Competition Electoral Volatility in Old and New Democracies. Party Politics 13(2): Manin, Bernard, Adam Przeworski, and Susan Stokes Elections and Representation. In Democracy, Accountability, and Representation, ed. A. Przeworski, S. C. Stokes, and B. Manin. New York: Cambridge University Press, Moser, Robert Unexpected Outcomes: Electoral Systems, Political Parties and Representation in Russia. Pittsburgh: University of Pittsburgh Press. 31

32 O Donnell, Guillermo Illusions about Consolidation. Journal of Democracy 7: Randall, Vicky and Lars Svasand Party Institutionalization in New Democracies. Party Politics 8(1): Samuels, David and Matthew Shugart Presidents, Parties, and Prime Ministers. New York: Cambridge University Press. Schedler, Andreas Measuring Democratic Consolidation, Studies in Comparative and International Development 36/1 (Spring 2001): Shugart, Matthew Presidentialism, Parliamentarism and the Provision of Collective Goods in Less-Developed Countries. Constitutional Political Economy 10(1): Shugart, Matthew S., and John M. Carey Presidents and Assemblies: Constitutional Design and Electoral Dynamics. New York: Cambridge University Press. Stokes, Susan C Mandates and Democracy: Neoliberalism by Surprise in Latin America. New York: Cambridge University Press. Strøm, Kaare A Behavioral Theory of Competitive Political Parties. American Journal of Political Science 34(2): Taagepera, Rein Making Social Science More Scientific: The Need for Predictive Models. New York: Oxford University Press. Zakaria, Fareed The Rise of Illiberal Democracy. Foreign Affairs 76(6):

33 Table 1. Countries by regime type and number of observations. Country Democratic type Year democracy established President s in sample PMs in sample Years of prior democrac y Albania Parliamentary Argentina Presidential Armenia SP (Pres-parl) a Bangladesh Parliamentary Benin Presidential Bolivia Presidential Brazil Presidential Bulgaria SP (Premier-pres) Chile Presidential Croatia SP (Premier-pres) Czech Rep. Parliamentary Dominican Rep. Presidential Ecuador Presidential El_Salvador Presidential Estonia Parliamentary Georgia Presidential Ghana Presidential Greece Parliamentary Guatemala Presidential Honduras Presidential Hungary Parliamentary Indonesia Presidential Korea Presidential Latvia Parliamentary Lithuania SP (Premier-pres) Macedonia SP (Premier-pres) Madagascar SP (Premier-pres) Malawi Presidential Mali SP (Premier-pres) Mexico Presidential Moldova SP (Premier-pres) b Mongolia SP (Premier-pres) Mozambique SP (Pres-parl) Namibia SP (Pres-parl) Nepal Parliamentary Nicaragua Presidential Pakistan Parliamentary Panama Presidential Papua New Guinea Parliamentary

34 Paraguay Presidential Peru 1 SP (Pres-parl) Peru 2 SP (Pres-parl) Philippines Presidential Poland SP (Premier-pres) Portugal SP (Pres-parl) Romania SP (Premier-pres) Russia SP (Pres-parl) Senegal SP (Pres-parl) Slovakia SP (Premier-pres) c Slovenia SP (Premier-pres) Solomon Islands Parliamentary South Africa Parliamentary Spain Parliamentary Taiwan SP (Pres-parl) Thailand Parliamentary Turkey Parliamentary Ukraine SP (Pres-parl) Uruguay Presidential Zambia Presidential Pre-democratic type according to Gandhi, except for Dominican Republic and Guatemala and South Africa (all coded by Gandhi as democratic for some years prior to becoming democratic by our criteria). Notes a. Premier-presidential from b. Parliamentary from c. Parliamentary before

35 Table 2. Percentages of prime ministers and presidents with prior service as legislator, by regime type and third wave vs. older democracy Older Democracies Third Wave Democracies Prime ministers Presidents Difference of means (PMs vs. Presidents) Pure types only 93.9 ( ) n= ( ) n=46 Pure types only 89.0 ( ) n= ( ) n=65 Including presidents in semi-presidential systems 89.0 ( ) n= ( ) n= *** 55.2*** 46.0*** *** Significant at.01 level

36 Table 3. Percentages of prime ministers with prior service as legislator: semi-presidential and parliamentary democracies of the third wave (1) Parliamentary prime ministers (2) Semi-presidential premiers (2a) Subtype: Premierpresidential (2b) Subtype: President-parliamentary Difference of means: % w/prior service as legislator 89.0 ( ) n= ( ) n= ( ) n= ( ) n=47 1 vs *** 1 vs. 2a 23.3*** 1 vs. 2b 44.3*** 2a vs. 2b 21.0*** *** Significant at.01 level ** Significant at.05 level

37 Table 4. Regression models for age of democracy and years of legislative experience Dependent variable: log(l+1), where L is years as legislator Variable Model 1: Puretype executives only Age democracy, D, logged.17*** (.06) Model 2: Including semi-presidential systems.19*** (.06) logd X Prime minister.29*** (.08).27*** (.07) logd X Semi-Presidential PMs ** (.11) Legislative career began in authoritarian regime.51*** (.12).50*** (.07) Legislative career began previous democracy.56*** (.07).53*** (.05) Prime minister.05 (.08).02 (.07) Semi-presidential PM (.09) Constant.14** (.06).17*** (.05) Number of observations Prob > F R-squared In both models, the coefficient for Age democracy refers to presidents. In Model 2, an executive who is a semi-presidential premier takes on a value of 1 on both the Semi-presidential premier and Prime minister variables. Robust standard errors in parentheses; clustered on Country *** Significant at.01 level ** Significant at.05 level

38 Figure 1. Relationship between age of democracy and years of prior legislative service, parliamentary prime ministers vs. presidents (simulated values)

39 Table 5. Regression models for age of democracy and years of legislative experience, semipresidential systems Dependent variable: Years as legislator, logged (L) Variable Semi-Pres PMs Age democracy, logged (D).22** (.10) Premier-Pres PMs.29** (.14) Pres-Parl PMs -.01 (.17) Legislative career began in authoritarian regime.52*** (.10).38*** (.11).80*** (.10) Legislative career began previous democracy.55*** (.06) --.63*** (.08) Constant.24*** (.08).22 (.13).32** (.12) Number of observations Prob > F R-squared Robust standard errors in parentheses; clustered by country *** Significant at.01 level ** Significant at.05 level

40 Figure 2. Prior legislative experience of executives, by regime type