Keywords

5.1 What We Expect from Democratic Elections

In a democracy, the laws are supposed to correspond to the “preferences and will” of the citizens of the democracy. (The phrase is from Rehfeld 2009: 229, who refers to the specification of this relationship as “the central normative problem” of representative democracy.) The election of representatives is a substitute for small-scale direct democracy in which such correspondence would emerge naturally from citizens’ personal participation in deliberation and policymaking. In large democracies, today most policymaking is made by representatives, who have the time and incentive to learn the nuances of complex issues, not through direct participation. Elections are supposed to be “responsive” in the sense of choosing representatives to reflect the desires of their constituents.

Of course, reflecting the desires of the constituents is not the only virtue in policymaking. On one hand, even the constituents themselves may wish their representatives to make policies on the basis of reflection and deliberation about their interests, not as a snapshot of momentary opinions. On the other hand, considerations of justice and equality, closely woven into the fabric of democracy itself, may sometimes conflict with constituent preferences. Still, when representatives act contrary to constituents’ wishes, such deviation must be explicitly justified and understand as exceptional (Pitkin 1967: 163–164).

For elections to serve the purposes of responsiveness, they must first meet the standards of democratic authorization. Democratic elections are acts of authorization, by which citizens choose representatives to be policymakers to act for them. The context of this act of authorization needs to meet certain standards of freedom of deliberation, adequate information, fairness of access and aggregation, acceptable choices, and so forth for the citizens, such that the authorization is meaningful and authentic. Moreover, the chosen representatives must have real policymaking power. Without meeting these standards, the elections cannot serve democratic purposes. (See the analysis of the “menu of manipulation” and the “chain of democratic choice” in Schedler 2002.) Elections that are in various ways manipulated to eliminate uncertainty and remove the autonomous effectiveness of citizens’ choices exemplify democratic failure, not democratic responsiveness.

Given the freedom from manipulation, we also expect elections to be in some fashion “responsive” to the constituents as they connect citizens and policymakers. But exactly how do we know whether elections have been responsive? The simplest answer is that the election outcome should reflect the distribution of the votes, which are authorizing representatives to make policies on the citizens’ behalf. (See the standard vote-seat representation studies in the tradition of Rae (1967 ) and Lijphart (1994) and also see the somewhat different approach in Hajnal 2009.) As more citizens vote for a given party or candidate, this party or candidate should be more likely to have influence in policymaking. It should have more seats in the legislature or more cabinet posts in the government, or whatever institutional resources help shape policy. When fewer citizens support a given party or candidate, its share of the policymaking resources should diminish.

As we shall see in a moment, there are alternative shapes that this responsiveness relationship can take, each supported by a different strand of democratic theory. But it should always be the case for democratic electoral responsiveness that elections connect more citizens with greater influence in policymaking. If gains for a party or candidate trigger intervention by the military, this is obviously a failure of electoral responsiveness. Less obviously, but still significantly, if a party or candidate that comes in first in the votes ends up with fewer policymaking resources than the second-place candidate, we may consider electoral responsiveness diminished (e.g., Strøm 1990: 73, on party votes and cabinet portfolios; Powell 2000, Chap. 6 on plurality losers).

One complexity in the vote-influence relationship concerns announced pre-election coalitions between parties or candidates. Two or more parties may announce that if they jointly win office they will govern together or jointly implement an explicit policy program (see Powell 2000 and Golder 2006). As the voters are informed that this connection between the parties has been formed, it seems appropriate then to treat these parties as a unit in assessing electoral responsiveness. (This will be especially important in a majoritarian vision of responsiveness, as explained below.)

The votes have great significance in democratic elections, as authentic expressions of the will of the electorate. Yet, correspondence between votes and representation in government has limitations as an indicator of responsiveness to constituents’ preferences. Ideally, parties promise to carry out bundles of policies if elected; citizens vote for their preferred bundle; the vote distribution then reflects the preference distribution; parties are committed to make policies that correspond to these promises. This is the familiar “mandate” model of policy representation.

One problem lies in the less well-informed voter, who may have trouble in determining which party is closest to his or her preferences, casting capricious votes or being unduly influenced by irrelevant factors. Another difficulty lies in the complexity of citizen opinions across multiple issues. Most troubling, perhaps, is the fact that voters can only choose between alternatives that are offered in a given election, which they may convert into voting choices in various ways (additive, lexicographic) as permitted by the configuration of party choices available. If some popular alternatives, or important combinations of issue positions, are not being offered in the election, or if parties with similar positions split the votes, or if parties are ambiguous in promises, or if a party’s candidates do not homogeneously stand for the same policies, or if none of the candidates seem trustworthy, then vote distributions may be misleading as conveyers of preference. Interactions between voter preferences, configurations of party offerings, and rules for aggregating votes into representation create potential for substantively non-responsive outcomes.

If the goal of democracy is to induce policymakers to do what citizens would like them to do, then simple connections between votes and officeholding can be misleading. Rather, we should measure citizens’ preferences directly and compare them to the commitments (and, perhaps, the later actions) of policymakers. Responsiveness of elections to preferences could be more directly captured in this way. This approach, too, has its difficulties, particularly in depending on survey methodology to ascertain and aggregate citizens’ preferences and in finding a reliable way to compare their preferences to the positions and actions of policymakers.

In the current essay, I approach electoral responsiveness as a hypothesis, or first cut approach, to preference responsiveness. I suggest that we first consider, as much of the literature does, electoral responsiveness as the connection between votes and officeholding. After investigating several normative ideals of this electoral connection, we shall then consider the theoretical and empirical conditions that lead from electoral responsiveness to preference responsiveness. Note that we are always interested in the role of elections in creating or inducing policymakers to be responsive, not simply in correspondences that emerge by accident or through the benevolence of dictators (Pitkin 1967: 232–234).

5.2 What Kind of Democratic Electoral Responsiveness? The Alternative Visions of Majoritarianism and Proportionality

There are at least two major alternative visions of democratic electoral responsiveness. They seem to be connected with the desirability of concentrating or dispersing policymaking power in a democracy. They are both democratic in conceiving of responsiveness as the rewarding of electoral gains with increased officeholding or influence. But one of these envisions the ideal connection as majoritarian. In a two-party context as long as one party remains below the 50 % support threshold, it has no claim on office or influence. As soon as it crosses the threshold that marks support of 50 % of the electorate, it is entitled to assume complete control of policymaking. This ideal connection is shown by the solid line in Fig. 5.1. American research on electoral responsiveness has generally conceptualized responsiveness in this way, with a large boost in office victories concentrated around the 50 % threshold (Gelman and King 1994 ; Katz 1997). The power to make policies is concentrated in one party, which has received the support of a majority of voters.

Fig. 5.1
figure 1

Two ideals of democratic responsiveness. Source Powell 2000

Full size image

On the other hand, European conceptions of electoral responsiveness have tended to focus on proportionality. In this ideal, policymaking authority is more dispersed. (More generally, see Lijphart 1984.) Crossing the 50 % threshold leads to only a marginal increase in officeholding probabilities or influence. As we see depicted in the dashed line in Fig. 5.1, 45 % of the votes should be associated with 45 % of the government and 55 % of the votes should be associated with 55 % of the government. This concern with proportionality as the fair consequence of democratic elections was strongly articulated by John Stuart Mill in 1861 and the invention of the proportional representation family of election rules was designed to deal with the perceived unfairness of majoritarianism [Mill (1861) 1958: 102]. Although analyses of consequences of these rules have been primarily applied to vote-seat connections (as in Rae 1967 and Lijphart 1994), the increasing importance of the executive in modern policymaking directs attention to connections to government as well (Taylor and Lijphart 1985; Pinto-Duschinsky 1998; Powell 2000; Vowles 2004; Forestiere 2009).

5.3 Electoral and Governmental Institutions and Electoral Responsiveness

The depiction of alternative conceptions of democratic electoral responsiveness brings us to the institutions that shape these connections. In parliamentary systems, the legislature elects the government, so we shall begin with the electoral and policymaking rules in parliamentary systems.

Many of the parliamentary systems can be fairly easily characterized as primarily majoritarian or proportional in their electoral and legislative institutions. The majoritarian systems use single-member district election rules (of the plurality or majority kind) and typically have concentrated policymaking power in the primary legislative body. The proportional systems use some kind of multimember, PR election rules and have legislatures that provide substantial influence to opposition political parties. (There are, of course, many variations on PR election rules, such as the various thresholds for representation, and these interact with the number of parties and other contextual features, which are also in part shaped by them.) As Kaare Strøm has suggested, the legislative institutions, such as committee powers, can help us identify stronger opposition influence and relatively dispersed policymaking power (1990: 70–73) (see Powell 2000, Chap. 2).

The single-member district election rules commonly produce single-party (or, possibly, pre-election coalition) majority governments following directly from the election itself. The elections are, in Strøm’s language, decisive for government formation (1990: 72–74). Because of concentrated executive-legislative power, these governments enjoy largely unchecked policymaking power. Most of the time, the plurality party or pre-election coalition forms the unchecked majority government, in a way that is “responsive” in the American or majoritarian sense. Although occasionally a splitting of the vote will bring victory to the runner-up party, a distinctively non-responsive result by all standards, this occurs relatively rarely (about 10 % of the time) although across many countries (Powell 2000). The flaw from a theoretical point of view is that very often these majority governments are created by the election rules for plurality parties winning 40 % or even less of the vote. Actual vote majorities are very unusual. But the structure of the relationship corresponds to the solid, majoritarian line in Fig. 5.1, with the key threshold moved left.

On the other hand, the PR election rules typically lead to minority or multiparty governments that must also share at least some policymaking power in the legislature with the opposition parties. (See Strøm 1990, and Powell 2000 on opposition influence.) The plurality party in the legislature, which is usually the plurality vote winner, does tend to enjoy a greater share in cabinet governments than its proportion of the vote. This is because in most parliamentary systems, the government selection and final policymaking rules within the legislature are usually based on simple majorities. (In some countries, some issues, including changes in a constitution, may require super-majorities.) Minority governments, based on less than a majority of legislative seats, are particularly dependent on outside parties to endure and to pass legislation and budgets, dispersing policymaking power, more like the dashed line in Fig. 5.1. Taking account of opposition influence in policymaking through strong committees and other institutions further disperses power. In this sense, we can say that the institution-linked hypotheses about the type of electoral responsiveness relationship seem to work rather well. Each type of institution seems to be fairly successful in generating the kind of responsiveness posited for it normatively in Fig. 5.1.

A slightly different perspective on the responsiveness relationship is dynamic and retrospective. We can evaluate the responsiveness of elections in terms of the effect of the votes on the retention or eviction of incumbents. This feature may be said especially to distinguish democratic elections from authoritarian ones. In 153 legislative elections in parliamentary systems, the incumbents lost votes more often than not, and when the incumbents lost 5 % of the vote, they were completely replaced 49 % of the time (Powell 2000: 48). Authoritarian regimes are unlikely to allow such vote losses or to permit loss of power if they occur.

Here, again, we can take a majoritarian or proportional perspective. In the majoritarian perspective, we want voters to be able to retain the incumbents completely if a majority of voters approve of them, to replace the incumbents completely if a majority of voters disapprove. From the proportional perspective, the incumbents’ influence should be enhanced proportionally if more voters support them, but reduced proportionally to the degree the voters turn against them.

The empirical evidence from parliamentary systems is, again, that the two types of systems work pretty much as normatively expected. In the majority systems, the fate of incumbent parties is pretty much an “all or nothing” situation, with the probability of retention closely tied to voter support. Incumbents who gain votes are almost always retained in office; those that lose votes tend to be replaced, although there is some “lumpiness” in the gross pattern, because incumbents may have varying margins from the last election. In the PR systems, lots of governments (about a third of them) change some, but not all, of the parties in the coalition, but the likelihood that the entire government will change decreases as the government gains votes increases as it loses votes. There are, of course, individual elections and, especially, individual party experiences that do not fit the general patterns. If the vote is split, the right way in a majoritarian country, a party can lose votes and even its plurality status and still hold its legislative majority (New Zealand in 1981). Individual parties, or even whole governments, can lose support and still build a winning legislative coalition. But overall, the pattern holds quite strongly.

Presidential systems, of course, deliberately divorce executive and legislative institutions and elections. The presidency itself can be conceived as a majoritarian institution, as it is usually held by a single party that is the most influential actor in policymaking, especially in policy initiation. (However, legislative and administrative presidential powers vary greatly, as shown by Shugart and Carey 1992, Chaps. 7–8.) However, the separate elections for the legislature very frequently result in “divided government,” where one party or coalition controls the presidency and another controls the legislature, a circumstance that both requires and hinders accommodative bargaining. Divided government would be opposed by majoritarians, but favored by proportionalists.

Interestingly enough, most electoral responsiveness analyses in presidential systems seem to focus on the presidential election, rather than trying to take account of presidential and legislative outcomes simultaneously. However, there is a substantial literature on institutional arrangements, such as timing of elections—concurrence/non-concurrence—as well as aggregation rules, that affect the probability and consequences of divided government (Shugart and Carey 1992; Samuels and Hellwig 2008).

Even considering presidential electoral responsiveness in isolation, the presidential election rules can make a substantial difference. The most common distinction is between plurality and majority runoff presidential elections. Colomer reports that in Latin American presidential races 1945–2000, the elections under plurality rules resulted in the winner gaining a majority of votes in only 30 % of the elections (2001: 107). With complex intermediate institutions such as the American Electoral College the aggregation of votes may even result in the vote winner losing the election, as in the US election of 2000.

5.4 Electoral Responsiveness Itself as an Empirical Connection: Do Electorally Responsive Elections Result in Governments Producing Policies that Voters Want?

The concept of electoral responsiveness that involves simply the connection (static or dynamic) between votes and officeholding has both advantages and limitations in the context of democratic theory. On one hand, the vote has unique significance as the expression of citizen choice; it authorizes representatives who have the collective power to make public policies. Insofar as we think of democracy as any peaceful and equal involvement of citizens in policymaking, we may not be concerned with the quality of that involvement. Whether citizens exercise their votes irresponsibly, influenced by whims or candidate personality or misinformed stereotypes, matters less than the fact of peaceful and equal participation, which replaces the various coercive, or unequal, or oligarchic alternatives. As long as the electoral outcomes are responsive in the sense(s) discussed above, connecting more votes with some greater representation in policymaking, that may suffice.

On the other hand, there is a powerful line of justification for democracy that resides in the degree to which democratic processes systematically induce substantive correspondence between the issue preferences of citizens and the policies made by their governments. In this line of thought, more than any other form of government we know, democracy should systematically induce the policymakers to do what citizens want them to do and avoid what the citizens dislike (Dahl 1989: 95). Democratic representation theory is thus a multistage theory that connects citizens’ issue preferences to their voting choices and those voting outcomes to policymakers’ commitments and actions. The stage of election responsiveness is a necessary part of this theory, but it is insufficient by itself, especially as it disregards the substantive content of responsiveness.

If citizens’ votes and the aggregation rules of the election laws bring a plurality party to power in one country or one election and deprive it of office in another, it can matter a lot what substantive policies that party embodies. If all the parties offer unattractive policies, or if the plurality winner is created because more attractive parties split the vote, electoral responsiveness might not generate substantive responsiveness. If all the parties offer similar, attractive policies, it is not so important substantively which one comes to office. In the latter case, an apparent failure of electoral responsiveness, bringing a second-place vote winner to power may still result in preference responsiveness.

Rather than thinking of responsiveness to votes and responsiveness to preferences as two different approaches to the role of elections in democracy, it seems more useful to consider responsiveness to votes as one of the links in a theory of responsiveness to preferences. I leave it to the reader to decide whether one has a greater claim to an identity with democracy itself. Personally, I tend to feel that a theory of elections and democracy has to incorporate responsiveness to both votes and preferences. In order to articulate this, however, we must consider more fully the stages in a substantive theory of democratic representation.

A substantive theory begins with citizens’ preferences. For elections to connect these preferences to the policymakers, the party system must offer attractive choices, with at least one party committed to policies embodying the preferences of most of the citizens. For majoritarians, this implies a party (or candidate) at the position of the median citizen, as in a one-dimensional space in which voters vote for the closest party that position can defeat any other and minimizes the number of citizens distant from the policy. (That the empirical world of policy preferences may often be multidimensional and not easily reduced to a single pair of choices places a complex burden on majoritarians in multiparty contexts.) For proportionalists, this implies multiple parties with most citizens having a fairly close party, reflecting the full diversity of citizen preferences whatever this might be. Full-blown theories of substantive representation thus imply theories of party competition specifying factors that predict parties’ campaign commitments. And the desirable distribution varies with the normative perspective, as it did in the vote authorization perspective.

Similarly, given the distribution of party commitments and the voter preferences, the voters should chose to vote for substantively proximate political parties (which we know empirically that they do not always do), and the rules of aggregation of votes should be responsive in the sense of accurately reflecting the votes into the legislative seats (or executive winner in presidential systems). This stage is similar to the electoral authorization responsiveness process, including the proportional and majoritarian differences, but the focus is on representing the desired distribution of citizens’ preferences, captured by the party commitments, not the votes as such. Moreover, for parliamentary systems in which the legislature selects the executive, we need to add a theory of government formation with also a substantive dimension. Governments should be centered at the citizen median and either represent that median (majoritarian version) or reach out from it to include other major citizen positions (proportional version).

In considering the responsiveness of presidential elections, a key issue is whether the election winner could have defeated each of the other candidates individually. This is the outcome known as Condorcet winning, which is itself a majoritarian criterion. Insofar as the election is run only once, our assessment has to be based on surveys of voter preferences. (If we assume preferences are based on a single dimension, the median candidate is the only Condorcet winner, and some analyses focus on the median candidate, as determined by various measures of candidate position, such as expert surveys, which may be more generally available, as in Colomer 2001.)

The key distinction in the presidential election rules is, again, between plurality elections and majority runoffs. Theoretically, with multiple credible candidates, the plurality elections are more vulnerable to winning by candidates who are non-Condorcet winners (or even Condorcet losers). We can see this intuitively in the example of a three-party race in which two leftist candidates split the vote and elect a rightist candidate who would have been defeated if paired against either of them. Colomer’s analysis of the experience of Latin American presidents seems to bear this out, with the median candidate much more likely to prevail in runoff elections (107). However, even with a majority runoff, a potential Condorcet winner can be eliminated on the first round, leading to an outcome that is less than ideally responsive on the second. (See the French 1995 election discussed by Colomer 2001: 95.) Thus, the majority (or high plurality threshold) runoff elections are more likely to produce outcomes that are relatively responsive to preferences.

It is interesting that both majoritarian and proportional visions lead to the normative standard that the government should be at the position of the citizen median, although they differ in whether the government should be concentrated at that position or building around it. There are fairly well-developed theories of party competition and government formation that lead to the empirical expectation that governments will be close to the citizen median under both majoritarian and proportional election rules. Majoritarian election rules (especially single-member district plurality) are expected to produce two-party systems (Duverger’s Law), and two-party competition is predicted by Downs to produce convergence of parties to the position of the median voter (Duverger 1954 ; Downs 1957; Cox 1997). (But see Ezrow (2008, 2010), who finds no systematic centrist tendencies in disproportional systems.) Two-party elections will produce legislative majorities that select governments of the larger party. Even with more than strictly two parties, these rules will often produce legislative majorities and governments based on them, as we saw in the electoral responsiveness discussion analysis above, focusing on the plurality party.

Proportional representation election rules will generally produce multiparty systems, depending on the social configuration and the thresholds (Duverger 1954; Cox 1997). Party competition will lead parties, in the most likely outcome, to spread across the spectrum as the voters are spread (Cox 1990, 1997, Chap. 12). (But see Calvo and Hellwig 2011.) Unless there are too many parties relative to the threshold, as happened in the low-information conditions of some of the early Eastern European elections, PR usually leads to accurate representation in the legislature of the parties above the threshold. If those parties represent the distribution of citizen preferences in appropriate weights, as will happen if the parties are disbursed and citizens vote for the closest party, then the parties in the legislature will reflect the citizen preference distribution. The position of the median legislative party should be close to the citizen median.

Elections in PR systems seldom produce single-party legislative majorities (Powell 2000; Mitchell and Nyblade 2008: 206), so theory of substantive representation in PR systems needs to include a theory of government formation in such situations. The process of government formation theoretically advantages the median legislative party (which is included in about 80 % of governments), thus tying the government to the citizen median (Laver and Schofield 1990: 113; Mitchell and Nyblade 2008: 210). (However, asymmetrical additions of other parties, especially the plurality party, when that is different from the median party, may systematically pull the government somewhat away from the median.)

In an analysis of the alternative paths to policy representation under majoritarian (strong) institutions and PR (permissive) institutions, Cox (1997: 237) argues that the advantage depends on possible coordination failures at electoral and government formation levels. Such failures are possible under either type of system. If “non-Duvergerian” (multiparty) results appear frequently in the majoritarian systems, they can perform “erratically” if the center candidates/parties fail to coordinate and their divisions allow extremist parties to win elections. These extremist parties can then form governments far from the median. Cox argues that election-level coordination failures in PR (“permissive”) systems are less likely to distort the legislative balance. Yet, a second stage of government formation will still be needed and coordination could fail at that stage, leading to less centrist governments.

Because there are theoretically plausible paths to preference representation, as well as possibilities for coordination failure, under both majoritarian and proportional institutions, a lively empirical literature has emerged as to which approach is in fact more successful in generating ideological congruence between the median citizen, the legislative median, and, especially, parliamentary governments. Studies in the 1980s and 1990s by Huber and Powell (1994), Powell and Vanberg (2000), Powell (2000), McDonald et al. (2004), and McDonald and Budge (2005) found that PR systems produced, on average, greater ideological congruence. (Also see Kim et al. 2010.) But these findings were challenged, using new measures of congruence and a later time period, by Blais and Bodet (2006) and Golder and Stramski (2010). Reanalyzing various methods and data, Powell (2009) replicated both sets of findings and argued that the difference depended on the time period of the study and the greater congruence in the majoritarian systems after the mid-1990s. This change seems to be caused, in turn, by declining party system polarization in the SMD systems.

Party system polarization has an interactive effect on ideological congruence. In conditions of low party system polarization, any electoral system seems to generate good congruence. Greater party system polarization makes good congruence less likely, and because of the key role of the plurality party, the effect is especially large in majoritarian systems (Kim et al. 2010; Powell 2011). From a theoretical point of view, the connection that is most problematic in the majoritarian systems is the Downsian theory of party competition that predicts party ideological convergence in the majoritarian systems with small numbers of parties. Sometimes this is true, but often it is not. When it is not and the parties are more polarized, congruence breaks down. In the PR systems, the theoretical paths hold very much as predicted, with very good vote-seat responsiveness creating quite good congruence between median voter and the median legislator; the process of government formation does lead governments to be more distant, on average, than the median legislator. Party system polarization is a problem, making it harder to form coalitions across the median, but overrepresentation of the largest party in the government coalitions in proportional systems does not undercut congruence as badly as the pure plurality governments in the majoritarian systems.

There remain several problems with ideological congruence analysis that have not been adequately explored. One concerns minority governments. As Strøm (1990) emphasized and explored, parliamentary governments in Western democracies have very frequently been minority governments—that is, the parties in the cabinet do not command a majority of seats in the legislature, relying on “outside” parties to pass legislation and budgets and to sustain them against votes of no confidence. These arrangements may either involve forming new “ad hoc” coalitions on different issues or consistent reliance on a particular partner. Analyses of government ideological congruence have typically treated these governments as like any other, but their dependence on others may mean that their real effective positions from a policy point of view may be rather different. Powell (2000) estimates “policymaker” congruence, as well as government congruence, giving more weight to opposition parties under minority governments; Carey and Hix (2011) use the position of the median legislative party rather than the government when estimating the distance of minority “governments” from the median voter. Obviously, neither solution is ideal nor do these take account of the circumstances and implications of legislative deadlock for median voters favoring change.

All these analyses of ideological congruence focus on the distance between the (estimated) ideological position of the median voter or median citizen and the ideological position of the government, usually taken immediately after the election. McDonald and Budge, however, argue that one should also take into account a longer time frame (2005: 130–135). (Also see, e.g., Forestiere 2009.) It may be the case that the PR systems generate more congruent governments after the average election than the majoritarian systems. But across a long series of elections, the greater majoritarian distances, sometimes to the left of the median voter and sometimes to the right, average out to a level of congruence more comparable to the PR systems. Moreover, they argue, as it takes substantial time for government policy commitments to result in altered policy outcomes, the majoritarian distortion in policy outcome is not as severe as the distortion in distance between median voter and government commitment (2005).

There is no doubt that the results from the longer time perspective are relevant to the congruence problem. Moreover, the improvement in average preference responsiveness in the long run in the majoritarian systems is reassuring for the justification of democracy itself. However, it seems clear that the election-by-election level of congruence (sometimes called distortion) is also relevant. Citizens may not take a longer time perspective; they may feel that democracy means more immediate citizen–government congruence. Moreover, not all citizens will still be around until the changing cycle of dominance evens the balance. Electoral and preference responsiveness analyses usually focus on the election as a unit for good reason.

McDonald and Budge’s concern with the actions of policymakers, rather than just their electoral promises or perceived positions, is well-taken in its own right. While this is yet one stage further from electoral “responsiveness,” it is the ultimate democratic promise. While substantive congruence between the positions of median voters and their perception of the position of the government may be valued in itself, voters expect politicians to keep their promises and are likely unhappy when they do not. (See Stokes 2001 on voters’ reactions to presidents’ policy reversals of campaign promises after Latin American elections in the 1980s and 1990s.) Because of the multitude of factors that shape real policy outcomes, it is difficult to estimate the effect of voter preferences, as transmitted through elections and parties, on those outcomes, let alone to determine the advantages of alternative institutional arrangements. One recent attempt is Kang and Powell (2010), who do find significant effects of the estimated median voter position on redistributive welfare spending in a multivariate error correction analysis. Interestingly, they find no significant advantage to the PR systems, despite the more accurate congruence of the ideological positions of the governments.Footnote 1 These growing efforts to explore congruence of actions as well as policy positions draw attention to the role of voters and elections in sanctioning officeholders who fail to behave as expected, which can put a slightly different perspective on the concept of responsiveness itself.

5.5 Alternative Concepts of Responsiveness: Sensitivity to Sanctions

In his subtle and interesting analysis of representation and democracy, Andrew Rehfeld uses “responsiveness” in a somewhat different way as a property of “representative decision makers” (2009: 222–223). He distinguishes the extent to which the decision makers are more or less responsive to electoral sanctions. Those less responsive to sanctions, making policies regardless of the threat of being thrown out of office, he calls “gyroscopic” (following Mansbridge 2003), responding to their own internal mechanisms. Those who are more responsive to sanctions, he calls “induced” policymakers.

As is widely understood, the threat of eviction from office in future elections can be an important inducement to officeholders to keep their promises and to anticipate what citizens may do in the next election. In a broad and general sense, democratic elections should always contain this possibility of holding officeholders accountable for their actions in office. However, as Mansbridge and Rehfeld point out, citizens may or may not prefer policymakers who are more sensitive to the next election than to their own internal gyroscope. Sometimes, they may want policymakers who are committed to a general theory of the public good or to the deeply felt policy directions for which they were elected (which may also be a mechanism of correspondence to voter opinion). Voters may be suspicious of those politicians who trim their promises to every shift in public opinion polls. At other times, the citizens may well want to exercise retribution against parties they perceive as having betrayed the public trust or having exhibited massive incompetence.

This concern with sensitivity to sanctions raises important issues. Both the Duverger-Downs theory of majoritarian convergence and concentrated policymaking power and its more permissive, reflective, proportionally oriented counterpart assume that the substantively congruent governments that they should generate will keep their election promises as best they can. But the temptations of political power are many and the pressures from the well-off and well-organized in the society are difficult to resist. The concentrated power of majoritarian governments facilitates open, unchecked abuse of the weak and the minorities. But the dispersed, shared power of coalition governments and inclusive institutions is often opaque, rather than transparent, facilitating hidden abuses of all kinds.

Without wishing further to complicate the concept of responsiveness itself, we need at least to take account of the role of sensitivity to electoral sanctions in our account of the role of electoral responsiveness in democracy. There is evidence that voters are more likely to hold governments accountable for poor policy outcomes, such as poor economic performance, where it is easier for them to assign responsibility or read signals of incompetence. Voters may also hold governments more accountable for past performance when lower party system polarization makes voters less concerned about the future implications of substantive policy proposals (Hellwig 2010). The research on “economic voting” in parliamentary systems has grown rapidly in recent years and becomes increasingly sophisticated. (See the review and analysis in Duch and Stevenson 2008.) It seems possible that the conditions of clear responsibility that facilitate economic voting also make officeholders more sensitive to possible election sanctions. Incumbent officeholders may also be less sensitive to electoral sanctions when they can control the timing of elections, as they can in some parliamentary systems.

Recent comparative work on parliamentary and presidential systems suggests that for voters, the opportunity to use the electoral weapon directly against the chief executive, especially in concurrent executive and legislative elections, further facilitates economic voting (Hellwig and Samuels 2007). (Also see Hellwig 2010; Samuels 2004, and Stokes 2001.) It is possible, although only speculative at the moment, that this facilitation of electoral retribution makes policymakers more sensitive to the possibility of future sanctions.

The current common wisdom of electoral system design suggests a trade-off between responsiveness, to either votes or preferences, and accountability (sensitiveness to sanctions). However, recent work by Carey and Hix (2011) explores this representation-accountability frontier and argues for a “sweet spot,” in low-magnitude PR systems, which would seem to facilitate both good (proportional) electoral responsiveness and relatively transparent coalition governments.

5.6 Why Democratic Elections Perform Imperfectly

Elections are essential in representative democracies. Yet, even if elections are free from constraint and manipulation, they are seldom free from criticism. The problems of responsiveness facilitate this unhappiness.

Even if we limit our concept of responsiveness to the correspondence between distributions or changes in votes and distributions or changes in government, there are difficulties. There are empirical difficulties. (See the literature reviewed in Powell 2004.) A “party” may mean different things to voters in different regions. Different geographic distributions of votes have different implications under alternative election rules. “Too many” political parties relative to the election threshold can distort vote representation under any set of rules. Single-member district election rules are especially sensitive to the distribution of votes and number of parties. PR systems seldom elect majorities, creating dependence on legislative rules (usually favoring majorities) and legislative bargaining to complete the link to policymakers.

Beyond these empirical difficulties, which creative and context-sensitive institutional design can help alleviate, there is the tension between the alternative normative visions of majoritarianism and proportionalism. Should the winners of a majority or plurality of votes be given unchecked policymaking power or only the largest share? Should the second-place finishers get nothing or a proportional (second largest) share? No magic of institutional design can resolve this fundamental normative difference, which also shapes the relative evaluations of democratic performance by voters for the winners and losers (Anderson and Gillroy 1997). The institutions associated with each vision perform relatively well by one standard, but not by the other.

Moreover, thoughtful analysts of the role of elections in democratic responsiveness seldom want to stop with the vote-policymaker connection alone. For, if the value of electoral responsiveness to votes lies in part in votes as indicators of citizens’ preferences, then there are many ways in which that indication can go astray. The theories that connect preferences and policies through electoral vote responsiveness depend on the behavior of voters and on the commitments of political parties. From this point of view, the majoritarian and proportional visions are in part hypotheses about the roles of voter choice, party competition, vote aggregation, and legislative behavior in government formation. Successfully connecting citizen preferences to government commitments can go astray when any of the theorized linking connections break down. The empirical research suggests that the majoritarian connections are especially vulnerable to party system polarization, although both approaches have their vulnerabilities and can on occasion produce governments distant from the median voter. As observers analyze the congruence failures in their own system, the imperfections of elections in systematically inducing congruence will emerge repeatedly.

Producing governments connected to the median voter through electoral responsiveness and preference congruence is a significant achievement for democratic representation. Yet, this framing itself seems incomplete without integrating the potential sanctioning role of elections that encourages policymakers to fulfill their commitments. It is not only logically incomplete, but is tangled with the majoritarian leanings of accountability theory. Carey and Hix’s exploration of the “accountability-representation frontier” and their focus on a “sweet spot” in institutional design that optimizes government closeness and small, transparent coalitions is an effort to take this into account. However, connecting citizen preferences and implemented policies, or changes in each, is empirically challenging and political science has yet far to go here. It seems clear that there are circumstances when other factors, such as changing needs and resources as the economy fluctuates, dominate public policy and obscure the role of citizens’ long-term preferences. Again, there are plenty of occasions when elections will seem to perform imperfectly in every system.