The degree to which voters can hold elected officials to accounts remains a central concern of political science. Most research has explored established democracies, which all happen to be parliamentary systems, with the exception of the United States. Democracy's spread around the world has increased the number of presidential systems, which generates an important new set of questions about the relationship between economics and elections, and thus about well democracy can work: To what extent do the institutions of presidentialism allow voters to hold the government accountable? To what extent do we see party or government accountability in separation of powers systems? That is, do voters reward or punish incumbent presidents and legislators, and to the same degree? Do particular partisan dynamics or institutional formats of presidentialism dim or enhance the prospects for accountability?
Scholars have yet to provide answers to questions such as these, although substantial debate persists about the presidentialism's relative merits. Some scholars disdain presidentialism and suggest that the separation of powers hinders accountability. For example, Linz (1990, 1994) argued that presidentialism's “dual democratic legitimacies” confuses voters and that presidential autonomy inhibits “party government” accountability.1
See, e.g., Lijphart 1992, Manin, Przeworski, and Stokes 1999, and O'Donnell 1994.
Others question Linz's claim and suggest that presidentialism provides advantages for accountability by enhancing identifiability and allowing voters to hold presidents and legislators accountable for different things, which may be normatively a good thing (Perrson, Roland, and Tabellini
1997; Samuels and Shugart
2003; Shugart and Carey
1992).
I seek to discover the ways and extent to which we observe accountability for the economy under presidentialism. Hundreds of books and articles have explored the relationship between economics and elections,2
For a recent review of this literature see Lewis-Beck and Stegmeier 2000.
and scholars have begun to reveal the ways in which political institutions and party-system configurations mediate this connection.
3 See, e.g., Anderson (1995, 2000) and Powell and Whitten (1993).
Yet how accountability might work within a presidential system remains largely unexplored, both empirically and theoretically.
4 Cheibub and Przeworski (1999) provide a partial exception, although they define accountability differently, focusing on incumbents' “survival in office.”
Can presidential systems live up to their democratic promise, or does something about the separation of powers inhibit accountability? Exploring this question could provide substantial insight into ongoing debates about the advantages or disadvantages of different institutional formats for democratic performance.
ACCOUNTABILITY AND CLARITY OF RESPONSIBILITY
I define accountability as the electorate's capacity to reward or sanction incumbent politicians (Manin, Przeworski, and Stokes 1999, 40).5
I do not include the second clause of Manin, Przeworski, and Stokes's definition, “… so that those incumbents who act in the best interest of citizens win reelection and those who do not lose them.” Scholars typically find that economics determines about one-third of the vote (Dorussen and Palmer 2002, 4). Thus asking whether economic swings are sufficient to “kick the bums out” is asking too much. We expect a correlation between the economy and the vote, but a correlation between the economy and the removal of incumbents may not appear. Still, this does not mean that incumbents are not responsive to voter sanction and do not try to change their policies to avoid additional future sanctions.
This paper thus follows what Susan Stokes calls the “normal economic voting” research program. In this approach, “voters use the past performance of the government to predict future performance and see the government as responsible for that performance” (Stokes
2001, 13). Scholars believe that voters' capacity to reward or sanction elected officials declines when they cannot discern responsibility for government performance. Powell and Whitten (
1993) thus argued that voters are more likely to punish or reward incumbents for the state of the economy when “clarity of responsibility” for outcomes is high. Given the separation of powers under presidentialism, a pertinent question is therefore whether clarity of responsibility affects voters' capacity to reward or sanction presidential governments, in both executive and legislative elections.
The theory of retrospective voting (Fiorina 1981; Lewis-Beck 1988) suggests that electoral accountability occurs because voters retrospectively judge whether governments have acted in their best interests and then reward or sanction them appropriately. Yet several factors might hinder voters' ability to sanction elected officials. First, voters might not be able to punish incumbents if no viable alternative exists (Anderson 2000; Manin et al. 1999). This problem is not limited to presidentialism, and in the analysis I include only competitive elections. Second, incumbent politicians might strategically conceal information about policy responsibility. This is also a problem everywhere, and we can reasonably assume that voters in presidential systems can potentially sanction governments, at least as much as they can anywhere else. In any case, it is unreasonable to assume that incumbents facing competitive elections conceal all information; instead, we should test to see whether partisan and/or institutional factors mediate the connection between economics and elections.
Third, citizens will be able to sanction governments only if politicians desire (and are eligible for) reelection or care about who succeeds them to office. Manin, Przeworski, and Stokes (1999, 48) suggest that limits on presidential reelection thus restrict voters' capacity to hold governments accountable. Limits on reelection do impede voters from rewarding or punishing a particular individual incumbent president (or legislator), but they do not impede voters from sanctioning the incumbent's party. In presidential systems, the vote for or against the party of a term-limited president or legislator is necessarily a judgment of both the individual incumbent and the incumbent's party. Manin and coworkers' suggestion assumes that incumbent presidents do not care about their legacy, have no interest in the continuity of their policies, and do not care about their party's success once they have left office. Yet last-term presidents care about their place in history and, thus, care about their performance in office. Moreover, since their legacy is profoundly shaped by who succeeds them to office, they also care about who that person turns out to be. This suggests that (to some degree at the least) they ought to attempt to convince voters that they share a personal and political affinity with their successor-candidate.
Voters can reward a successful president's party by electing his or her party's successor, or they can punish the incumbent's party by electing a rival party's candidate. There is therefore no reason why voters could not hold political parties as a whole and presidential governments as a whole—even those populated by term-limited politicians—accountable in a retrospective voting fashion. Thus while many problems inhere in the accountability relationship between citizens and governments, these problems are not limited to presidential systems, nor should we simply assume that presidential governments are necessarily worse than other forms of government in terms of accountability.
Aggregate economic voting studies typically assume that some voters consider national economic conditions before the election as part of their vote decision. Voters attribute reward or blame to the incumbent government for those conditions, and as a result the vote shifts. Yet electoral accountability declines when voters cannot discern responsibility for government policy, and voters are more likely to punish or reward the incumbent government for the state of the economy if clarity of responsibility is high, and not if clarity of responsibility is low (Powell and Whitten 1993).
Two related reasons explain this dynamic. The first highlights incumbents' strategic actions. All incumbent governments tend to lose votes (Nannestad and Paldam 2002). Powell and Whitten surmised that incumbents know this and thus prefer to diffuse political responsibility, for example, by blaming those with whom they might have shared control of government. A country's institutional and/or partisan makeup can help politicians diffuse responsibility, or it can complicate such activity. The second factor emphasizes voters' ability or willingness to gather information: As institutional or partisan complexity increases, voters will have greater difficulty discerning actual responsibility for government output, independently of whether politicians attempt to diffuse responsibility or not. Regardless of which factor weighs more heavily, lower clarity of responsibility should insulate incumbents from the factors that cause them to lose votes.
PRESIDENTIALISM AND CLARITY OF RESPONSIBILITY
Perhaps clarity of responsibility and thus voters' capacity for sanctioning in presidential systems depends on factors that scholars have identified for parliamentary systems.6
Powell and Whitten (1993) argued that incumbent party or coalition cohesion and the type of committee system are important for clarity of responsibility. Information on those variables is unavailable for most of the countries in this study.
Or presidentialism could affect how voters hold governments accountable in particular ways. For example, Manin, Przeworski, and Stokes (
1999, 47) imply that presidentialism obscures government responsibility generally and that coalition and divided governments under presidentialism are particularly bad for accountability. In contrast, Persson, Roland, and Tabellini (
1997) suggest that the separation of powers institutionalizes conflicts of interest between branches of government and thus encourages relatively greater information revelation than parliamentary systems, which should aid voters' efforts to discern policy responsibility.
A few single-country studies support the notion that differences exist in terms of voters' ability and/or willingness to sanction members of the same party who occupy different branches of government, such as Lewis-Beck (1997) on France and Erickson (1989, 1990) on the United States, but no cross-national research has yet confirmed or refuted these findings.7
Remmer (1991, 2001) is the only scholar who has explored the links between economics and elections in a cross section of presidential systems. However, Remmer was not interested in the impact of political institutions and explored economic voting in Latin America generally. Moreover, she did not explore the impact of economics on legislative elections.
How do voters in presidential systems think about blame and reward? Do they focus exclusively and always on the president, regardless of who controls the legislature? Are legislators in some partisan or institutional environments more or less likely to feel the brunt of the voters' wrath?
Several factors might sharpen or blunt clarity of responsibility and thus voters' propensity to reward or punish incumbents in presidential systems. The most important factor is unique to presidentialism and is institutional: the electoral cycle, or the concurrence or nonconcurrence of elections. Others (see especially Shugart 1995) have noted that the electoral cycle plays a key role in executive–legislative relations. Here I suggest that it also mediates the accountability relationship between voters and elected officials.
Concurrent elections under presidentialism most resemble elections under parliamentarism. When executive and legislative elections are held simultaneously, candidates for executive and legislative elections have strong incentives to coordinate election campaigns, in order to benefit from campaign economies of scale, coattail effects, and organizational advantages. That is, concurrent elections tend to encourage a team effort, whereby politicians employ their national party organizations and national partisan messages to get out the vote. This associates copartisans from both branches of government in voters' minds and encourages voters to hold them accountable for the same things. Consequently, in concurrent elections voters have relatively greater incentives to hold both branches coresponsible for the economy, all else equal.
In contrast, when either executive or legislative elections are held nonconcurrently, national policy issues such as the economy should matter relatively less. For legislative elections, following the literature on U.S. congressional elections (e.g., Erickson 1990), nonconcurrence should attenuate the impact of national factors on the vote because (all else equal) it focuses voters' and candidates' energies on local factors and candidates' personal qualities, including but not limited to any potential incumbency advantages due to constituency service. In nonconcurrent legislative elections we also tend to observe relatively less national party organization involvement and relatively more influence of local or regional political messages.
A similar logic should hold for executive elections. Nonconcurrence frees presidential candidates to campaign on entirely individualistic bases, meaning that the election result will depend relatively more on voters' perceptions of the candidates' personal qualities and relatively less on their partisan connections. That is, we have reason to believe “It might not be the economy, stupid” in nonconcurrent presidential elections, because national policy factors ought to matter relatively less.
Another institutional factor that might mediate clarity of responsibility, in legislative elections, is the electoral rule. Under presidentialism voters cast separate ballots for legislative and executive elections. Unlike in parliamentary systems, this offers the opportunity for substantially different dynamics to drive executive and legislative elections. The primary factor is whether the electoral system encourages or discourages localism and/or individualism, giving legislators the choice whether to run “with” or “against” their president and/or party (Carey and Shugart 1995; Shugart 1995). When the electoral rule for legislative elections encourages localism, the state of the economy should matter relatively less, all else equal. In contrast, when the rules nationalize legislative campaigns, the economy should matter more clearly and consistently.
Two other variables that we also find in parliamentary systems merit discussion. The first is coalition government. Coalitions tend to obscure any one party's responsibility for government output (Anderson 1995, 2000). As in parliamentary systems, governing in coalition might insulate incumbent presidents or legislative parties from the factors that might otherwise affect vote swings for or against them. A contrast of vote swings lends credence to this notion: Presidents who governed in coalition were relatively more insulated from losses, losing on average 4.68% (N=21), compared to average losses of 6.31% (N=54) for presidents who did not govern in a coalition. A similar result obtains for the legislative elections: Parties in coalition lost an average of 0.92% (N=32), while parties that governed alone lost an average of 3.59% (N=71).
Second, minority government might insulate presidents and legislative parties from vote swings due to economic performance. In unicameral presidential systems, minority government occurs when the president's party or coalition does not control more than 50% of the seats in the single legislative chamber. In bicameral systems, minority government occurs when the president's party or coalition does not control more than 50% of the seats in both chambers. When parties divide control over the branches of government, the incumbent president can claim that all parties must share responsibility for poor performance. Several studies confirm that minority parliamentary governments suffer smaller losses than majority governments (e.g., Powell and Whitten 1993 and Strøm 1990). As with coalition governments, minority presidential governments also appear to insulate incumbents from losses. Presidential parties running a candidate following a minority presidency lost on average 5.00% (N=51), compared to average losses of 7.68% (N=24) for candidates following a majority president. However, it remains to be seen whether the economy drives these differences.
The factors explored thus far permit us to place presidential systems on a continuum. At one end elections are nationalized and always concurrent, and at the other end legislative elections are localized and often not concurrent with the executive election. In the real world, Costa Rica might fall on one end of the continuum while Colombia fell at the other end, and the United States would fall somewhere in the middle. The question remains whether such variables affect clarity of responsibility for incumbents in either branch of government.
CLARITY OF RESPONSIBILITY AND ECONOMIC VOTING: ANALYSIS
I test the impact of clarity of responsibility and economic conditions on elections in 23 presidential democracies, defined as those in which both branches of government are directly elected for fixed terms and in which the head of government is not accountable to the legislature (Siaroff 2003).8
I include cases from Siaroff's Category 2 as well as Category 6, which “functions essentially like” Category 2. As Elgie (1999, 11) notes, regime classification “varies from one writer to another,” but the only case on which I differ from Siaroff is Bolivia, which Siaroff classifies together with Switzerland. Despite a quirk in Bolivia's constitution, it functions as a pure presidential system. See Inter-American Development Bank 2002; personal communication with John Carey and Octavio Amorim Neto.
I employ aggregate economic and electoral data. I included a case if the country had had at least two consecutive democratic elections—and if political and economic data were available. This resulted in a sample of 75 executive elections and 103 legislative elections.
9 Appendix 1 contains a list of the countries and elections. I explore differences across regime types, incorporating pure and semipresidential systems as well as parliamentary systems, in other work.
Scholars of economic voting typically employ one of the following three variables as the indicator of economic performance: GDP growth, inflation, or unemployment. I employ the percentage change in real per capita GDP growth (World Bank 2003).10
Unemployment information is unavailable for most of the countries in the sample. As for inflation, the expectations about the impact of inflation are far less clear than for that of GDP growth, supporting the use of the latter over the former, because it is unclear whether voters respond to the level or direction of inflation.
Following Pacek and Radcliff (
1995), I use the change in GDP growth in year
t–1 if the election was held in the first six months of the year, and the change in GDP growth in year
t if the election was held in the last six months of the year.
11 Quarterly figures are unavailable for most countries in the sample. Other operationalizations could provide more nuance, but data limitations restrict my ability to conduct such tests. For example, because the time series is relatively short for most of these countries, I cannot explore the impact of “shocks” relative to long-term trends (cf. Palmer and Whitten 1999).
I call this variable
Economy. If the economy matters generally in presidential systems, this variable should have a positive and significant coefficient.
If clarity of responsibility diffuses or accentuates economic voting effects, we should observe variables in multiple regression analysis that substitute for, decrease, or wipe out the impact of the economy on vote swings. The logic of the argument suggests that the effect of economics on elections is interactive with clarity of responsibility variables (Powell and Whitten 1993). Thus if the economy matters in general, we expect a positive and significant coefficient on Economy (all else equal), and if another variable blurs the impact of economic fluctuations on the vote, we would expect a negative and significant coefficient on the interacted term Economy * OtherVariable when that variable is added to the equation. The impact of the economy on the vote would then be the sum of the coefficient on Economy plus the coefficient on the interacted variable.
As per the discussion in the previous section, I explore the impact of the following independent variables: Concurrence, Electoral Rules (for legislative elections only), Coalitions, and Minority President, all operationalized as dummy variables. I enter each variable directly and in an interacted term (with Economy). The dependent variable in all regressions is the absolute percentage change in vote for the incumbent president's party from the previous election. In the models for executive elections, the dependent variable is the vote change in the executive election, and in the models for legislative elections, the dependent variable is the vote change in the legislative election. In all models, I employ ordinary least squares regression with robust standard errors.12
I do not employ models with country dummies because of the numerous cross sections relative to the overall sample size in the data set (with 23 cross sections I have an average of three executive elections and four legislative elections per country). I ran a series of regressions with country dummies, and although some of the dummies were significant, their inclusion did not substantially alter the results.
In all regressions I also included two control variables: the incumbent party's proportion of the vote in the previous election (Previous Vote), which should have a negative impact because parties tend to lose votes after being in office; and a dummy variable indicating whether or not the incumbent president is running for reelection (Reelection). This variable should return a positive coefficient because incumbents running for reelection have advantages in terms of name recognition and organization relative to other candidates. Consequently, they (or their party, in the case of legislative elections, due to potential coattail effects) ought to receive a higher vote percentage, all else equal.
Given my hypothesis about the importance of the electoral cycle, I first separate out concurrent from nonconcurrent elections.13
Analysis of a correlation matrix for executive elections revealed that the Economy * Concurrence variable was correlated at .91 with the Economy variable (no other variables were correlated at higher than .79). Correlation at this level is problematic and makes the inclusion of both variables in one model unhelpful. I also attempted to test for the impact of concurrence using a Proximity variable, which measures the proportion of the presidential (legislative) term that has elapsed relative to the legislative (presidential) term. Yet this variable, when interacted with the Economy variable, was similarly highly correlated with the Economy variable, at. 92.
provides results for a comparison of the effects of the economy on the vote in both concurrent and nonconcurrent
executive elections.
The first model tests for the impact of the Economy, Previous Vote, and Reelection in concurrent executive elections. The results conform to my expectations: The vote in the current period is correlated with the vote in the previous period, presidents running for reelection can expect a higher vote share than candidates running for their first term, and the state of the economy is strongly correlated with the vote for the incumbent party. These results are robust to the exclusion of any particular case.
The results for nonconcurrent presidential elections also conform to my expectations. In contrast to the results for concurrent elections, neither the previous vote nor the state of the economy affects the vote swing for the incumbent party's candidate. This result may sound foreign to scholars and residents of the United States, where presidential elections are always concurrent with legislative elections, and I acknowledge that the results rest on relatively few cases. However, they are fairly robust: Note that the standard error on Economy is three times the size of the coefficient. Future work should return to this issue as more data become available.
Table 1 reveals that voters only sanction presidents and their parties' candidate when executive and legislative elections are concurrent. I suggested above that in nonconcurrent elections voters may focus on the candidates as individuals and not as members of a party running candidates for a set of offices. As a result they may not connect either the performance of the incumbent president or the performance of the economy to their vote. For those who believe that executive elections should flow from the state of the national economy, the results for nonconcurrent elections are, at a minimum, discouraging, even if they are not conclusive.
A key question remains about the potential impact of incumbent presidents who are running for reelection. Cheibub and Przeworski (1999) and Manin, Przeworski, and Stokes (1999) suggest that voters cannot engage in retrospective economic voting when they literally cannot hold the person who has been responsible for managing the economy over the past few years (in theory at least) to accounts. Does incumbency enhance accountability for the economy, and does the absence of an incumbent running for reelection weaken accountability?
In 11 of the 61 concurrent elections an incumbent president ran for reelection. (No incumbents ran for reelection in nonconcurrent executive elections.) Given the results for concurrent elections in Table 1, we know that incumbency does matter on its own. In Table 2, I interact the Economy and Reelection variables to discover whether incumbency has both an independent and a joint effect. If incumbency drives the accountability relationship, this variable should return a positive and significant coefficient. In addition, the Economy variable should lose significance, if the hypothesis that the absence of an incumbent running for reelection attenuates accountability is correct.
The results do not support the extension of Cheibub and Przeworski's and Manin and coworkers' argument. Incumbency does not make for a tighter accountability relationship, at least for the economy. Voters hold incumbent parties to accounts, even in executive elections and even if the incumbent is not running for reelection, as long as elections are concurrent. Linking partisan ties across branches of government, rather than the presence or absence of an individual candidate, is relatively more important for overall accountability in presidential systems.
The next question is whether these results change when we add any of the other clarity of responsibility variables. Table 3 presents several models for concurrent executive elections, adding in each variable and its interaction term.14
In this and the subsequent series of regressions, I could not run fully specified models because adding in the interacted variables results in problematic degrees of multicollinearity (among the interacted variables).
For nonconcurrent elections, I have too few cases to conduct similar and reliable analysis— the results would be dependent on the influence of the existence of one case or another of (for example) minority coalition government.
Model 1 tests for any change when the president's party does not control a majority of the seats in the legislature, independently of whether the president ultimately obtains a majority through coalition government. This specifically tests whether voters distinguish situations when the president's team does not unilaterally comprise a legislative majority relative to all other situations. This situation arose in 41 of the 61 concurrent cases. The null cases are thus those in which the president's party controls more than 50% of the seats. This hypothesis suggests that shared responsibility for government output might dampen voters' propensity to hold the incumbent president's party's candidate responsible for the state of the economy. Thus, the coefficient on the interacted variable ought to be negative, offsetting the positive expected coefficient on Economy. The results indicate that although Previous Vote, Economy, and Reelection all continue to be strongly correlated with the vote, the relationship between the economy and the vote does not change for incumbent-party presidential candidates whose parties do not control a legislative majority.
Model 2 tests for any change when the president governs in coalition. This hypothesis tests whether voters distinguish situations when the president agrees to share power from situations where the president attempts to govern without other parties' assistance, regardless of whether that coalition is majority or minority. This situation arose in 15 of the 61 concurrent cases. The null cases here include both single-party minority and single-party majority presidents. As with Minority President, the introduction of this variable does not alter the results.
The results in Tables 1 through 3 reveal that in executive elections the only variable affecting the clarity of responsibility and thus the relationship between the economy and the vote is whether the presidential and legislative elections are held concurrently. When this occurs, the state of the economy strongly shapes presidential election results: In each model, an increase of approximately 1% in per capita GDP prior to the election results in an increase of a bit more than 1% in the incumbent party's vote. The introduction of partisan clarity of responsibility variables does not affect this relationship—that is, voter reward or sanction is not shaped by the presence of minority government or coalition government.
It is perhaps paradoxical that nonconcurrent executive elections do not clearly enhance clarity of responsibility for the economy. Nonconcurrent presidential elections certainly concentrate attention on the presidential campaign. Yet while we might suppose that nonconcurrent elections would concentrate voters' goodwill or ire about the economy onto the presidential race, the results here suggest that “solo” presidential elections tend to eliminate the connection between the state of the economy and the vote. The findings here, although based on relatively few cases, tend to qualify the notion that voters always hold presidents to accounts for the state of the economy: “It's the economy, stupid” only holds clearly when executive and legislative elections are held concurrently. Nonconcurrent presidential elections may turn on something else—candidates' personal characteristics or other policy outputs. This issue merits further exploration.15
With the addition of more cases and/or countries, I could test for differences both across and within groups of countries (i.e., countries that only have concurrent elections, countries that have both concurrent and nonconcurrent elections, and countries that have only nonconcurrent elections), to further explore the impact of concurrence and incumbency. However, in this sample, only four countries have executive elections that are sometimes concurrent and sometimes nonconcurrent (N=12 total, 6 nonconcurrent), and only two countries' presidential elections are always nonconcurrent (N=8).
I now turn to legislative elections. To parallel the analysis for executive elections, Table 4 displays the baseline results for concurrent and nonconcurrent elections.
The results again support my argument. In concurrent legislative elections both the economy and the party's previous vote are significantly correlated with the vote in the current election. In contrast, Previous Vote matters less and Economy is not significant in nonconcurrent elections, although it is in the predicted direction.16
For concurrent elections, if we exclude the 1990 Dominican Republic election, the Reelection variable becomes statistically significant. No other case affects the results. For the nonconcurrent elections, if we exclude the 1994 Ecuador election, the coefficient on Previous Vote becomes significant. No other case alters the results.
This implies (but does not confirm, since I have no variable testing for “Local Conditions” in both models) that local conditions matter relatively more in nonconcurrent elections and that the holding of both executive and legislative elections simultaneously emphasizes national policy matters in voters' minds.
As with executive elections, the next question is whether these results change when we add clarity of responsibility variables. I explore concurrent elections first, in Table 5.
In Model 1, I test for any change when the president's party does not control a majority of the seats (Minority President). This hypothesis is the same as in Table 3. This situation arose in 42 cases, and the null cases are thus those in which the president's party controls more than 50% of the seats. The hypothesis suggests that shared responsibility for government output dampens voters' propensity to hold the incumbent president's party responsible for the state of the economy in legislative elections. Thus, the coefficient on the interacted variable ought to be negative, offsetting the expected positive coefficient on Economy.
The results indicate that the Previous Vote, Economy, and Reelection variables remain strongly correlated with the vote and that the interaction of Minority President with the Economy variable is in the predicted direction and is significant. To illustrate the finding, suppose that the economy increased by 1%. If the president's party did not control a majority, then the president's party would gain 1.211-0.562=0.649%, about half of what it would gain if the president's party controlled a majority. This supports the clarity of responsibility hypothesis that voters reward or punish the president's party relatively less when that party does not control a legislative majority.17
This model excludes the 1995 election in Argentina as an outlier. Exclusion of no other case affects the results. When this case is included, the Previous Vote and Economy variables remain significant, while the Reelection variable is not significant, as is the Minority President * Economy variable, although it is still in the predicted direction. Excluding no other case alters the results.
Model 2 tests for any effect in legislative elections when the president governs in coalition (Coalition), regardless of whether the coalition is majority or minority. This situation arose in 18 elections. Similar to Model 1, this hypothesis suggests that shared responsibility for government output dampens voters' propensity to hold the incumbent president's party responsible for the state of the economy. Thus, the coefficient on the interacted variable again ought to be negative, offsetting the positive expected coefficient on Economy. The result confirms the hypothesis. When the president governs in coalition, the economy still affects the vote, but if the economy increased by 1% and the president governed in coalition, the president's party would only gain 0.952-0.807=0.147%, even less than if the president lacks a majority legislative contingent.18
This model excludes the 1999 Panamanian election as an outlier. When this case is included, the Previous Vote and Economy variables remain significant, while the Reelection variable is not significant, as is the Coalition * Economy variable. The exclusion of one other case affects the results, in a similar way: the 1998 Venezuelan election.
Finally, Model 3 tests for the impact of the Electoral Rules. Electoral systems that encourage localism or personalistic campaigning may tend to obscure the clarity of responsibility for the state of the national economy, while “nationalizing” electoral systems might maintain the clarity of responsibility for the economy. I created a dummy variable designed to test whether the most extremely “localizing” and/or “personalizing” electoral systems differ in a substantial way from other electoral systems. These include the systems of Brazil, pre-1973 Chile, Colombia, Cyprus, Peru, Taiwan, and Uruguay (Samuels and Shugart 2003).19
My operationalization collapses Samuels and Shugart's (2003) tripartite coding, because I found that no difference exists between the cases they coded with a value of one and those they coded with two but that a difference exists between the cases they coded as zero and all the other cases.
For concurrent elections, only six cases took a value of one. As with the other variables, we expect the coefficient on the interacted variable to be negative.
The results again reveal the robustness of the first two variables in the regression to the introduction of this variable. The interacted Rules*Economy variable is also significant, indicating that localizing electoral rules diffuse clarity of responsibility for the economy. However, I should acknowledge that this result merits additional exploration, because this model excludes one of the six “localizing” cases, the 1990 Peruvian election, as an outlier. When this case is included, there is no relationship between this variable and the dependent variable. However, the other variables in the regression do not change. In short, although this test suggests that localizing electoral rules affect clarity of responsibility in concurrent elections, the result is inconclusive.
In sum, for concurrent legislative elections, I obtain robust results that the economy is related to voters' propensity to sanction incumbents and that the party's vote in the previous election is related to its vote in the current election. In addition, partisan variables such as coalition and minority government diffuse the relationship between the economy and the vote. These findings, combined with the findings in Tables 1 through 3, suggest that in countries with a single-party majority legislative contingent and concurrent elections, voters reward or sanction executives and legislators for economic performance similarly. However, party-system fragmentation, minority government, and potentially the electoral rules can mediate this relationship and obscure the clarity of responsibility for economic outcomes in legislative elections.
I now conduct similar tests for the nonconcurrent legislative elections. The results appear in Table 6. Recall that I hypothesized that in general, the accountability relationship between economics and elections should weaken in nonconcurrent legislative elections, because national political factors matter relatively less in nonconcurrent elections. All else equal, nonconcurrent elections focus voters' and candidates' attention on local factors and candidates' personal qualities and constituency service. Moreover, national party organizations play less of a role in nonconcurrent legislative elections, whereas local or regional factors may weigh more. This does not mean that national factors are (or should be) irrelevant in nonconcurrent elections, it is only a statement of their relative importance. Specifically, the hypothesis suggests that the coefficients on all of the variables in the regressions should be weakly significant or not significant. For example, the performance of the national party in the previous election should matter less if the current election is nonconcurrent, weakening the coefficient on Previous Vote. Likewise, the Economy ought to matter less or not at all. Moreover, none of the interacted variables should be significant, because we don't expect voters to take national partisan factors into consideration in nonconcurrent elections.
Let me first comment on the results for the Previous Vote variable. Recall that in concurrent elections, this variable was always significant, and in the baseline model for nonconcurrent elections (Table 4), it remained significant, although the coefficient was much weaker than that for concurrent elections. Here, the Previous Vote variable reaches significance in two of three models and its coefficient remains smaller than for concurrent elections, a plausible finding given our intuition that nonconcurrent elections tend to concentrate campaigns and voters' minds on local or candidate-based factors rather than national partisan factors.
Next I explore the impact of the Economy variable. In concurrent elections this variable was robust to all alternative specifications. Here, Economy is significant in only one model, and it is always smaller in nonconcurrent elections than in any of the models for concurrent elections. These results thus again conform to the expectations generated by Table 4, which contrasted concurrent and nonconcurrent legislative elections and found that the relationship between the economy and the vote was significant in both types but both weaker and less robust in nonconcurrent elections.
Finally, I examine the impact of the other clarity of responsibility variables in Table 6. Recall that I argued that nonconcurrence tends to reduce the impact of national party organization and campaign messages. Thus the influence of all national-level partisan factors ought to be relatively diminished in nonconcurrent elections. The results are clear: In contrast to the results in Table 5, none of the interacted variables are significant in Table 6, indicating that no other clarity of responsibility variables mediate the impact of the national economy beyond the electoral cycle.20
The negative and significant result on Electoral Rules indicates that candidates from the president's party obtain a higher percentage of the vote when the electoral rules “localize” competition. This occurs because localizing electoral rules allows candidates to insulate themselves from national trends and base their campaigns on the “personal vote” and the incumbency advantage.
To sum up, the results for legislative elections in Tables 4–6 indicate that the economy matters in both concurrent and nonconcurrent legislative elections, although it has a greater impact and matters more clearly and consistently in concurrent elections. Moreover, the results indicate that partisan and institutional clarity of responsibility affects the relationship between the economy and the vote in concurrent elections but that the evidence for such a connection is much weaker in nonconcurrent legislative elections. In addition, when we compare the impact of the economy in executive and legislative elections (compare against Tables 1 and 3, e.g.) we can infer that voters reward or sanction presidents more than legislators for economic performance.21
Regression analysis that pooled all elections, however, did not statistically distinguish between the effect of the economy in executive and that in legislative elections, although the sign of the dummy variable distinguishing executive from legislative elections was in the predicted direction.
These results leave us with the following: In concurrent executive elections, the economy affects the vote, regardless of the clarity of responsibility. Yet in nonconcurrent executive elections, the economy never matters. In concurrent legislative elections, the economy affects the vote, but partisan and institutional variables mediate that connection. The economy also affects the vote in nonconcurrent legislative elections, but relatively less so than in concurrent legislative elections, yet neither partisan nor institutional variables mediate that link. These findings tend to support my argument about the particular way in which the electoral cycle affects clarity of responsibility and thus accountability in presidential systems.
CONCLUSION
This paper elucidates the conditions under which electoral sanctioning is more or less likely to occur under presidentialism. The results confirm neither the most optimistic views, expressed most clearly (if only at an abstract level) by Persson, Roland, and Tabellini (1997), nor the most pessimistic, perhaps expressed by Linz (1990, 1994). In many ways the findings echo Shugart and Carey's (1992) suggestion that the separation of powers can engender different forms of representation and accountability than parliamentary systems.
Clarity of responsibility and thus voters' capacity to sanction elected officials work in particular ways under presidentialism. When elections are nonconcurrent, a situation that cannot occur under parliamentarism, sanctioning for the state of the economy is weak. In executive elections this lack of a connection between the economy and the vote came somewhat as a surprise, but the relatively weak relationship in legislative elections followed theoretical expectations and research findings on U.S. elections. Because about one in five executive elections and two in five legislative elections around the world are nonconcurrent, these results support some observers' concerns about overall accountability under presidentialism.
In contrast, concurrence enhances voters' capacity to reward or sanction government officials for the state of the economy in presidential systems. These results qualify Linz's critique of presidentialism based on his notion of dual democratic legitimacies. Despite the formal separation of powers, institutions that promote close electoral linkage between the executive and legislative branches can generate “unified democratic legitimacy.” Ceteris paribus, when elections are concurrent voters treat the incumbent executive and his or her legislative supporters as a team and judge them as such.
However, the all-else-equal clause is essential, because in concurrent elections legislative fragmentation and “localizing” electoral rules can affect the degree to which voters hold both branches accountable for the economy. This implies the following: Under single-party majority government with concurrent elections, voters in presidential and parliamentary systems will not differ much in their capacity to sanction or reward governments. However, even under concurrent elections, when we see multiparty coalitions (for example), clarity of responsibility and thus voters' ability to sanction incumbents function differently under presidentialism and parliamentarism. Under parliamentarism, as Anderson (1995, 2000), Powell and Whitten (1993), and others have argued, partisan and institutional variables tend to diffuse clarity of responsibility. Under presidentialism, the same phenomenon occurs in concurrent legislative elections but not in executive elections. This separates the degree to which voters sanction executives and legislators and means that executives in many presidential systems are subject to relatively greater sanctioning or rewards than legislators, a situation that cannot by definition arise under parliamentarism.
Presidentialism generates distinct forms of accountability. The electoral cycle most clearly affects this dynamic, but electoral rules and party system features also mediate voters' ability to sanction incumbents. Under a limited set of conditions presidentialism can approximate parliamentarism, and we would expect to see a similar form of government accountability in both types of systems. But the collective, “responsible parties,” policy-oriented sense of accountability that is rooted in normative political theory can occur under presidentialism only under these limited conditions. When such conditions are not met, and when the electoral and party systems separate the bases for selecting members of each branch, we have a more clearly distinct “presidentialist” dynamic. For example, even when elections are concurrent, voters are often likely to sanction incumbents in both branches of government to different degrees, depending on institutional and party-system factors. And when elections are nonconcurrent, national factors matter relatively little in both executive and legislative elections.
For legislative elections at least, Shugart and Carey (1992) suggested that nonconcurrence and/or localizing electoral rules can be a normatively good thing because there is nothing prima facie wrong with the idea of giving legislators an opportunity to campaign on their constituency service and voters an opportunity to base their vote on local rather than national concerns. In this way the institutional “separation of purpose” under presidentialism (Samuels and Shugart 2003) can enhance representation and accountability in ways that parliamentary elections cannot. Thus although the state of the national economy matters relatively little in nonconcurrent legislative elections, and national clarity of responsibility variables do not affect this relationship, local variables may systematically affect the vote and thus promote accountability (for example, whether voters can recall the incumbent's name or the extent to which they credit the incumbent with providing constituency service). The degree to which presidentialism allows both national and local accountability simultaneously merits exploration.
Yet if one disputes Shugart and Carey's suggestion that separating the bases for electing the executive and legislative branches can be a normatively good thing for democracy because it permits different forms of accountability, and instead holds to the Linzian view that both branches should generally be held accountable for the same things, then the results here reinforce the conclusion that the core problem of for government accountability under presidentialism is the potential lack of electoral linkage between the president and his or her legislative contingent. Party-system and other institutional factors shape the strength of this linkage, so the findings in this paper thus reinforce the critical role that institutions and party-system configurations can play in shaping democratic performance. Future research ought to explore these differences in more detail.
APPENDIX 2: ELECTORAL DATA SOURCES
For Argentina, Bolivia, Brazil, Chile, Costa Rica, the Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela, the sources were as follows.
- Inter-American Development Bank. 2002. Democracies in Development. Washington, DC: IDB.
- Political Database of the Americas. 2003. “Statistical Data: Electoral Results.” Various dates. http://www.georgetown.edu/pdba/. Retrieved November 2003.
- Magar, Eric, and Kevin J. Middlebrook. 2000. “Statistical Appendix: National Election Results, 1980s and 1990s, for Argentina, Brazil, Chile, Colombia, El Salvador, Peru, and Venezuela.” In Conservative Parties, the Right, and Democracy in Latin America, ed. Kevin Middlebrook. Baltimore: Johns Hopkins University Press.
For particular countries, I also used the following.
- Benin: personal correspondence, Professor Shaheen Mozaffar, Bridgewater State University.
- Bolivia: Gamarra, Eduardo, and James Malloy. 1995. “The Patrimonial Dynamics of Party Politics in Bolivia.” In Building Democratic Institutions: Party Systems in Latin America, ed. Scott Mainwaring and Timothy R. Scully, Stanford, CA: Stanford University Press.
- Brazil: Jairo Nicolau. 2003. “Dados Eleitorais do Brasil.” www.iuperj.br/deb/port/. Retrieved November 2003.
- Chile: (a) Chilean government Web site, www.elecciones. gov.cl. Retrieved November 2003. (b) Cruz-Coke, Bernardo. 1984. Historia Electoral de Chile, 1925–1973. Santiago: Editorial Judicial de Chile.
- Costa Rica: personal correspondence, Michelle Taylor-Robinson, Texas A&M University.
- Cyprus: Nordsieck, Wolfram. 2003. “Parties and Elections in Europe.” www.parties-and-elections.de/indexe.html. Retrieved November 2003.
- Ecuador: Mejía-Acosta, Andrés. 2000. “Weak Coalitions and Policy Making in the Ecuadorian Congress (1979–1996).” Presented at the 2000 meeting of the Latin American Studies Association, Chicago.
- Korea: (a) Keesings Contemporary Archives (various). (b) Foreign Broadcast Information Service. 1992 “Daily Report, East Asia.” March 26, 1992, pp. 22–23. (c) Moriss, Peter. 1996. “Electoral Politics in South Korea.” Electoral Studies 15 (December): 550–62. (d) Kang, W. T., and H. Jaung. 1999. “The 1997 Election in Korea.” Electoral Studies 18 (December): 599–608.
- Malawi: (a) Elections Around the World. 2003. www.electionworld.org. Retrieved November 2003. (b) Nohlen, Dieter, Michael Krennerich, and Bernhard Thibaut, ed. 1999. Elections in Africa: A Data Handbook. Oxford: Oxford University Press.
- Panama: personal correspondence, Carlos Guevara-Mann, University of Notre Dame.
- Peru: Tuesta Soldevilla, Fernando. 2001. Perú Político en Cifras: 1821–2001. Lima: Fundación Friedrich Ebert.
- Uruguay: No author. 2000. Elecciones 1999/2000. Montevideo: Ediciones de la Banda Oriental.
- United States: Ornstein, Norman, et al. Vital Statistics on Congress, 2001–2002. Washington, DC: American Enterprise Institute.
- Venezuela: Kornblith, Miriam, and Daniel H. Levine. 1995. “Venezuela: The Life and Times of the Party System.” In Building Democratic Institutions: Party Systems in Latin America, ed. Scott Mainwaring and Timothy R. Scully. Stanford, CA: Stanford University Press.
For assistance and comments on previous versions I thank Octavio Amorim Neto, Teri Caraway, John Carey, Bill Clark, John Freeman, Timothy Hellwig, Gabriel Negretto, Scott Morgenstern, Shaheen Mozaffar, Alejandro Poiré, Wendy Schiller, Phil Shively, and the participants at presentations at the University of Michigan, Fundação Getúlio Vargas—Rio de Janeiro, New York University, University of California at Irvine, Brown University, Instituto Tecnológico Autónomo de México, Centro de Investigaciones y Docencias Económicas in Mexico City, University of Minnesota, and the 2003 APSA meetings in Philadelphia.
%20-%20https://sketch.com%20--%3e%3ctitle%3e3ECBBF50-FD08-412A-943D-2CFC5E025A16%3c/title%3e%3cdesc%3eCreated%20with%20sketchtool.%3c/desc%3e%3cg%20id='Stylesheet'%20stroke='none'%20stroke-width='1'%20fill='none'%20fill-rule='evenodd'%3e%3cg%20transform='translate(-926.000000,%20-668.000000)'%20id='BUTTONS'%3e%3cg%20transform='translate(100.000000,%20597.000000)'%3e%3cg%20id='BTN-/-RIGHTS'%20transform='translate(816.000000,%2065.000000)'%3e%3cg%20id='ICON-/-Rights'%20transform='translate(10.000000,%206.000000)'%3e%3cpath%20d='M13.5752841,0%20L1.34943182,0%20C0.59375,0%200,0.586666667%200,1.33333333%20L0,22.6666667%20C0,23.4133333%200.620738636,24%201.34943182,24%20L17.6505682,24%20C18.40625,24%2019,23.3866667%2019,22.6666667%20L19,5.36%20L13.5752841,0%20Z%20M13.5752841,1.89333333%20L17.0838068,5.36%20L13.5752841,5.36%20L13.5752841,1.89333333%20Z%20M1.34943182,22.6666667%20L1.34943182,1.33333333%20L12.1988636,1.33333333%20L12.1988636,5.36%20C12.1988636,6.10666667%2012.7926136,6.69333333%2013.5482955,6.69333333%20L17.6235795,6.69333333%20L17.6505682,22.6666667%20L1.34943182,22.6666667%20Z'%20id='Shape'%20fill='%23595959'%20fill-rule='nonzero'%3e%3c/path%3e%3ccircle%20id='Oval'%20stroke='%23595959'%20stroke-width='1.5'%20fill='%23595959'%20cx='9.5'%20cy='13.5355339'%20r='5.5'%3e%3c/circle%3e%3cpolyline%20id='Path-2'%20stroke='%23FFFFFF'%20stroke-width='1.5'%20transform='translate(9.500000,%2013.035534)%20rotate(45.000000)%20translate(-9.500000,%20-13.035534)%20'%20points='8%2016.0355339%2011%2016.0355339%2011%2010.0355339'%3e%3c/polyline%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/g%3e%3c/svg%3e)
Comments
No Comments have been published for this article.