Scholarly article on topic 'Campaign Effects and Second-Order Cycles: A Top-Down Approach to European Parliament Elections'

Campaign Effects and Second-Order Cycles: A Top-Down Approach to European Parliament Elections Academic research paper on "Political Science"

Share paper
Academic journal
European Union Politics
OECD Field of science

Academic research paper on topic "Campaign Effects and Second-Order Cycles: A Top-Down Approach to European Parliament Elections"


European Union Politics

DOI: 10.1177/1465116507082812 Volume 8 (4): 509-536 Copyright© 2007 SAGE Publications Los Angeles, London, New Delhi and Singapore


■ campaign mobilization

■ electoral cycle

■ European Parliament elections

■ second-order elections theory

■ spatial representation

Campaign Effects and Second-Order Cycles

A Top-Down Approach to European Parliament Elections

♦ Till Weber

European University Institute, Italy ABSTRACT

Second-order elections theory explains cyclical losses by national government parties in elections to the European Parliament (EP) through strategic protest voting owing to performance deficits in policy-making. This paper confronts the conventional bottom-up view with a top-down approach to second-order elections. Ultimately, the electoral cycle is driven not by instrumental voting behaviour but by party strategies oriented towards governmental power in the member states of the European Union. Based on survey data from the European Election Studies of 1999 and 2004, firstorder campaign mobilization is shown to determine the prospects of government parties in second-order elections. Mobilization itself depends on the quality of spatial representation in terms of distinct programmatic alternatives, which governments are unable to provide during the midterm. Although this process can be traced on the left—right dimension, parties prevent it with regard to integration issues by systematic demobilization. After all, EP elections are still second order, but first-order politics exert their influence through cyclical campaign mobilization and not through strategic protest voting.


Voting behaviour in elections to the European Parliament (EP) is motivated by national rather than European concerns. This so-called second-order thesis has enjoyed high popularity in electoral research since Reif and Schmitt came up with the initial idea in 1980. According to their conceptual framework, voters use EP elections to voice dissatisfaction with their government's performance by strategic protest voting. As this behaviour is said to reflect the curves of government popularity, it explains the typical second-order effect. The results of EP elections follow the national electoral cycles, with government parties losing support in a systematic manner.

The second-order elections model comprises two important elements: the general 'multi-level' argument relating EP elections to national politics, and the specific 'protest' argument explaining exactly how this dependency affects election results. Whereas the 'multi-level' argument remains undisputed, this paper questions the 'protest' argument upon which the widely accepted micro-foundations of second-order elections theory rest. The conventional bottom-up view focusing on strategic voting behaviour is confronted with a top-down model of the electoral cycle in which the effects of domestic party competition are of crucial importance. The central argument is threefold.

First of all, voters do not produce second-order cycles at their discretion. Rather, their behaviour is animated by first-order campaign effects. Whereas in the run-up to national elections parties mobilize their voters by means of information and persuasion, the mobilization deficit at the midterm provokes higher defection rates. Secondly, mobilization itself depends on the quality of spatial representation in terms of distinct programmatic alternatives. Only if parties fulfil this leadership function are voters willing to follow them. Finally, with their programmatic flexibility restricted by retrospective evaluations at the midterm, national government parties do not have the mobilization capacity at their disposal to avoid defeat by the opposition.

In the following these arguments will be further elaborated. First I describe the conventional second-order theory and develop the alternative approach in detail. Then I set up a model accounting for mobilization effects with regard to left-right and integration issues. This model consists of three parts: the first dealing with cyclical information flows; the second with patterns of spatial representation; and the third with underlying processes of persuasion and opinion polarization. Finally I perform a stepwise test of the model with survey data from the European Election Studies (EES) of 1999 and 20041 and interpret the statistical results.

The conventional second-order elections theory

Right after the first direct EP elections in 1979, Reif and Schmitt (1980) published their seminal article 'Nine Second-Order National Elections'. This meaningful title captures a good part of our present knowledge about the specific logic of EP elections.

First of all, although EP elections are held concurrently across the European Union (EU) every five years and give rise to a common legislative body, these elections are primarily of a national character. National parties stand for election, the elections are held under national electoral laws, news coverage is done by national media systems, and national issues dominate the campaigns. However, although they resemble national parliamentary elections in these respects, EP elections cannot be assumed to function in the same way. Reif and Schmitt provided a set of differences, but the main one lies in the restricted powers of the EP. Unlike national parliaments, it cannot generate a government, and its legislative competencies are not (yet) comparable with those of its national counterparts.

Given the generally low stakes and the importance of domestic politics in EP elections, Reif and Schmitt classify them as 'second-order elections'. Hence, EP elections are comparable to subnational elections of various kinds. As Reif (1984: 245) puts it, for electoral research the EU is a common subsystem of its nation-states. Besides lower turnout and better prospects for small parties owing to low public interest and weaker incentives for tactical voting, this peculiarity entails the almost inevitable defeat of national government parties. According to Reif and Schmitt, some former voters for the government express dissatisfaction with their government's performance by abstaining or by supporting an opposition party in EP elections. These voters follow Hirschman's (1970) 'voice' strategy, i.e. they try to exert pressure on the government by temporarily withdrawing their support.

The present article deals with this 'less-at-stake dimension' of the second-order model and, among this set of hypotheses, with the argument about government losses especially. Certainly, the original piece by Reif and Schmitt (1980) is comprehensive, resembling a 'framework' rather than a model, as its subtitle indicates. It covers a wide range of issues, including electoral procedures, EP election campaigns, and aspects of political and social change. However, not only has the 'less-at-stake dimension' attracted greater scholarly attention (e.g. Van der Eijk and Franklin, 1996), it also touches upon the most obvious features of EP elections in need of explanation. Furthermore, this is where the original account comes closest to a fully developed model, which owes its present-day popularity to the well-known 'cycle argument'.

As Reif and Schmitt proposed, the routine government losses can be modelled as a function of the timing of the EP elections in the national legislative period. Losses should be highest around the midterm and fade towards the beginning and the end of the period, therefore leading to a classic electoral cycle (Goodhart and Bhansali, 1970). Since Reif (1984) first demonstrated the explanatory power of the cyclical model with EP election results, numerous studies have successfully replicated his findings over the years. To name but a few, the results of work by Ferrara and Weishaupt (2004), Marsh (2005) and Schmitt (2005) generally confirm the model for the EP elections of 1999 and 2004, which are of interest to us.

All these contributions support the conventional second-order theory on macro-empirical grounds. But what can be said about the theoretical foundation of the cycle approach? Why should strategic protest voting against the government happen mostly at the midterm? Moreover, why do incentives for protest voting develop at all? Reif and Schmitt (1980: 9f.) argue that EP election results follow the popularity curves of the national governing parties. However, this answer merely avoids the problem of explaining the course of government popularity.

The search for a more profound argument explaining these phenomena identified performance problems in policy-making as the central variable. For example, governments are unable to maximize economic performance continuously (Nordhaus, 1975) or they simply fail owing to problems of 'overload' (King, 1975). Reif and Schmitt have related such possible systemic trends to actual voting behaviour in second-order elections. At least this fits nicely with their own allusions to increasing disappointment among the electorate after a government takes over (Reif and Schmitt, 1980: 10).

To sum up, the conventional second-order theory generally proposes:

(a) the second-order elections cycle is driven by strategic protest voting;

(b) strategic protest is a reaction to performance deficits of the government.

A top-down approach to second-order elections

Based on the above account, the conventional second-order theory offers a plausible explanation of government losses in EP elections. However, this explanation cannot be considered logically compulsive. The predictive power of the model completely depends on the empirical fulfilment of the driving forces behind the cyclical effect: do governments really fail, and do voters really get disappointed? There may be reasons to expect similar phenomena in modern democracies, but they cannot be deduced from the theory itself. The

level of abstraction Reif and Schmitt chose as a starting point for their model is so demanding that, brought down to earth, the theory merely states that 'unpopular governments lose elections'.

Certainly this is a strong criticism, but it cannot be doubted that all further hypotheses are based on an indistinct systemic level that is beyond the range of the theory. Borrowing terminology developed in the tradition of Lorenz (1937) and Tinbergen (1963), protest voting is a 'proximate' but not an 'ultimate' cause of government losses. The model tells us about the stimulus for fewer voters ticking government parties in the polling booth. However, it explains neither how this stimulus 'evolves' and where its origins lie, nor why it emerges mainly at the midterm and follows a cyclical course over time.

Although I raise a fundamental objection to the second-order model, this does not refute its general logic. Being 'second-order' simply means that an election is dominated by forces emanating from another, 'higher-order' level. In this context, second-order EP elections are influenced by first-order politics in the EU nation-states. Reif and Schmitt define these first-order forces as protest voting, but this view is not theoretically exclusive. Because the multilevel argument does not depend on the protest argument, other factors may be introduced into the model so long as they stem from the national arenas. This provides an opportunity to reformulate the model in a way that gives not only proximate but also ultimate answers to the problem of government losses.

My theoretical argument consists of three elements that - if combined - lead to an alternative explanation of the second-order effect: assume that (1) the electoral cycle is driven by campaign mobilization in the first-order arenas, (2) effective mobilization requires party competition via distinct policy positions, and (3) the government's ability to provide such positions is restricted at the midterm when policy outcomes are subject to retrospective evaluation. It logically follows that government parties share a disadvantage in EP elections, with vote losses determined by the national mobilization cycles. But before we take this step, the individual arguments should be elaborated in more detail.

To begin with the first element, the conventional second-order theory explains EP election results according to a bottom-up model of the electoral process. Voters intentionally produce a second-order cycle, and it is down to them alone to 'switch off' government losses towards the end of the period. However, by focusing on the sophisticated strategies of voters, the theory is rendered blind to any form of independent action by parties. As some recent analyses show, European electoral politics are best understood as a reciprocal dependency between parties and voters, or even as a unidirectional exertion of influence from the 'elitist' top to the 'popular' bottom (Arceneaux, 2006; Gabel and Scheve, 2007; Ray, 2003; Steenbergen et al., 2007). According

to this top-down approach, important processes of opinion and preference formation are completed long before bottom-up influence may set in.

With the notable exception of Toka (2007), who focuses on different information environments, this perspective is widely missing in comparative EP elections research. This is astonishing because an elitist approach allows for an almost self-evident interpretation of the cyclical government losses. Obviously, the cycle closely follows the campaign efforts parties undertake with regard to parliamentary elections in their home countries. These efforts are highest during the months leading up to election day and the period after, and decline during the midterm. Of course, this timing serves a strategic purpose. Parties try to bring their own voters to the polls and to win over undecided and formerly opposed voters when governmental power is at stake in the EU member states.2

According to this reformulation, second-order effects result from a systematic mobilization deficit at the midterm. With regard to differences in turnout, this logic is intuitively appealing: the higher the first-order mobilization in a nation-state, the higher the second-order turnout in EP elections occurring 'incidentally' during a first-order campaign. However, explaining not only variation in turnout but also the drop in government support requires a further step. Why should a general mobilization deficit affect government support in particular? This question leads us to the second element of the alternative explanation.

As argued above, the actual independent variable of the second-order model is policy effectiveness. What matters for vote choice are the (at best) obscure shortcomings of government performance, while other major determinants of voting behaviour are neglected. Thus, the model may be called 'apolitical' because a vote on governmental performance does not imply a choice between policy alternatives. Of course, the parsimony of the model is intended. It is exactly the credo of large parts of electoral research that EP elections simply mirror national elections (e.g. Schmitt and Mannheimer, 1991) - except for the reactions to the government's failure. But do government parties fail only in one dimension - the technical implementation of their promises?

At this point it is important to recall the two-step logic of the second-order model. Considering an election as 'second order' to a first-order level does not explain the nature of the relationship between the two. Just as second-order effects are not necessarily provoked by strategic protest, voting behaviour in EP elections is not necessarily about performance. Rather, the factors that suggest themselves are those that in the main characterize party competition in the EU member states. More precisely, if we assume that

political mobilization accounts for the second-order effect, the relevant impact should emanate from national election campaigns. In my model, these campaigns affect voting behaviour via the top-down definition of competing policy alternatives.

The importance of this proposition is underlined by the Responsible Party Model developed by Schattschneider (1942). In this classic work, party competition via distinct policy positions forms the central prerequisite of an effective democratic process. In Schattschneider's own (1960: 141) words: 'Democracy is a competitive political system in which competing leaders and organizations define the alternatives of public policy in such a way that the public can participate in the decision-making process'. Schattschneider does not allege that parties are interested in democratic efficacy itself. Rather he assumes that a working representational system emerges as a by-product of the partisan struggle for power, which fits the American experience of an at times polarizing race between the two major parties.

Empirical applications of Schattschneider's ideas in the European context confirm that it is indeed the distinction between political platforms that characterizes significant elections (Miller et al., 1999; Schmitt and Thomassen, 1999). It is only when voters are offered a real choice at the polls that they can attach meaning to the parties' programmes and cast their votes on the basis of their policy preferences. From this point of view, a working democratic process relies on the leadership aspect of political representation already elaborated by Dalton (1985) and tellingly referred to as 'representation from above' by Esaiasson and Holmberg (1996). Both pieces emphasize the crucial role parties play in developing well-defined political positions and communicating them to the voters.3

Following this line of research, the concept of representation is used here as a means of top-down campaign mobilization; i.e. neither as a self-contained mechanism nor to characterize some form of reciprocal feedback between voters and parties. Thus, if parties are able to mobilize voters by the quality of representation they offer, the electoral connection in EP elections should work best close to or concurrent with national elections. On the other hand, if first-order campaign effects are mainly absent (the situation at the midterm), the European arena is denied its share in the profits.

Turning to the third element of the model, the interaction of campaign mobilization with democratic representation gives an answer to the theoretical problem of cyclical government losses. Up to this point, electoral prospects can be defined as the ability to mobilize via distinct policy positions. This ability varies systematically across parties and through the electoral cycle: once a government's initial 'honeymoon' is over, its policies will be evaluated

retrospectively on the basis of outcomes, as Kousser (2004) shows for EP elections. In the spatial terminology, these outcomes can be interpreted as the extent to which the political status quo has been moved in the direction announced during the preceding election campaign. Exactly because this campaign has been fought with distinct positions, the government is highly unlikely to realize its initial position during the midterm. Meanwhile, opposition parties are less restricted in their programmatic flexibility.

Although Kousser (2004) and most of the literature on retrospective voting focus on the general (policy-neutral) state of the economy, my model of voting is not the only one to recognize the inherent interaction of economic performance and policy choice. For example, Kiewiet (1981) shows how individual preferences over particular policies determine the extent and direction of economic voting. Johnston and Pattie (2001) also demonstrate that retrospective evaluations go beyond mere economic issues and depend on the perceived connection between policy choice and outcomes. Therefore, the question of government performance is secondary for my argument. Because retrospective voting affects spatial flexibility, the important variable is choice, not quality alone.

As this argument suggests, government parties suffer a strategic disadvantage at the midterm. However, once the next first-order campaign has started, both government and opposition parties advertise future policies that are then subject to prospective evaluation. For the government in particular, the campaign serves to overcome unfavourable retrospective evaluations, to update support for its long-term objectives, and to 'prime' advantageous criteria for prospective voting (see Bartels, 2006; Druckman, 2004; Miller and Krosnick, 1996). Consequently, electoral prospects should return to normal during a first-order campaign.

To conclude: I have argued that the government's programmatic flexibility is severely restricted during the midterm owing to prevailing retrospective evaluations of policy outcomes. I have also proposed that campaigning effectively means providing the public with distinct policy alternatives. As a result, the second-order effect in EP elections can be explained by the government's limited ability to mobilize voters via appealing political positions during the midterm.

From this discussion, counter-versions to the propositions derived from the conventional second-order elections theory above can be formulated.

(a) the second-order elections cycle is driven by changes in the level of political mobilization generated by first-order campaigning;

(b) vote losses are a reaction to the government's inability to provide adequate representation in terms of distinct and persuasive programmatic positions.

Information flows and persuasion effects

Testing the specific effect of first-order campaigning on voting behaviour in EP elections requires a more general model of political communication. Of particular use is Zaller's (1992) theory of mass opinion, which describes opinion formation as a top-down process in which political actors communicate their messages to the public. Basically, Zaller distinguishes persuasive messages, which contain the statements actors want the public to believe, and cueing messages, which provide contextual information allowing people to link a persuasive message to their political predispositions. Both types of message are handled by the public in a two-step process of reception and acceptance, with actual persuasion taking place at the second stage.

Opinion formation in election campaigns can be understood in these terms, although things are a bit more complicated here. First of all, parties send messages on political issues to the voters that then constitute the party platforms. Hence, the initial task for parties is to make sure voters receive information on their platforms. But this is not the full story. In the case of party platforms especially, persuasion is involved even at this early stage. Positions must be communicated in a credible way, so that voters actually believe the party is serious about its promises. Only then does the 'real' business of persuasion begin, as parties try to make voters accept their messages, i.e. to adopt the party positions.

Reception depends on the level of cognitive engagement (awareness) with the issue in question. The higher the voters' awareness, the higher is the probability they will be exposed to and comprehend political messages. With regard to acceptance, the relationship is reversed: voters with higher awareness tend to resist attempts of persuasion more often. This is because voters need information to link the content of a message to their political predispositions. If they lack this information, they are more likely to accept messages even if they are not congruent with their predispositions.

The consequences of this basic mechanism vary according to different elite constellations. If there is a mainstream position among the major parties, agreement will be high in the electorate as well. Additionally, higher awareness supports agreement because the probability of reception rises and alternative cueing messages - the basis of potential resistance - are largely absent. In contrast, if parties are divided over an issue and communicate opposing messages, mass opinion will be polarized along the lines of political predispositions. Higher awareness now leads to a more pronounced polarization because contextual information allows voters to evaluate messages in light of their predispositions.

Concerning the messages themselves, intensity and familiarity both enhance the probability of reception. At the next stage, however, intensity

implies a higher probability of acceptance and familiarity a lower probability. Familiarity means people are well prepared to resist a message, whereas intensity may overcome this resistance. Furthermore, intensity should be more effective among the less aware who have little means to offer resistance, and - almost exceptionally - the same applies for familiarity: although familiarity gives even less aware voters the chance to resist, this is overcompensated for by their relatively higher probability of receiving the message in the first instance.

Two dimensions of party competition

Although the above summary of the Zaller model generated some precise expectations of general campaign effects, it is not possible draft specific hypotheses unless the relevant issue sets are known, including their central properties (intensity and familiarity). These questions are now discussed for the two dimensions of party competition that are dominant in European electoral research: the traditional left-right dimension and the EU-specific pro/contra-integration dimension.

Today the left-right dimension is considered to be the main survivor of the historical cleavage system (Lipset and Rokkan, 1967). A systematic elaboration of its structural properties was provided by Fuchs and Klingemann (1989). Following Luhmann (1981), they regard the left-right dimension as a generalized medium of communication. For individuals, it serves an orientation function, allowing them to relate a wide range of issue positions to a single dimension. For the political system, it serves a communication function, facilitating communication - although the involved persons might possess different notions of the specific meaning.

Apart from its prevalence among Western mass publics, the left-right dimension offers advantageous conditions for political mobilization. Of course parties do not proclaim a left-right position per se. Rather, they communicate a specific selection from the elements of the collective schema, thus preparing the menu for their voters. This is possible because 'individual left-right schemata can be seen as incomplete reflections of the collective schema of left and right' (Fuchs and Klingemann, 1989: 207). Again, my argument suggests understanding mobilization as a top-down definition of distinct policy alternatives.

Whereas I expect strong campaign effects on the left-right dimension, the picture is different with regard to the integration dimension. First of all, this dimension lacks the structural properties described for left-right above. Unlike left-right positions, European integration fails to structure ideological

competition between parties (Thomassen and Schmitt, 1999). Integration seems to be a non-structural issue compared with the issues constrained by left-right (Thomassen, 1999). However, this is not the main reason for cautious expectations of campaign effects on the integration dimension.

One might expect that EP elections especially should deal with European integration, but this dimension has always been the problem child of electoral research. Although the integration process began as a declared elite project, it was assumed to be legitimized by a 'permissive consensus' of the people (Lindberg and Scheingold, 1970). Mass opinion has diversified since, but the major parties have adhered to their pro-European positions anyway (Van der Eijk and Franklin, 2004). The political supply side in questions of integration resembles a cartel.

Hix and Lord (1997) explain this peculiarity with a pattern of strategic assimilation of the major parties. Whereas the traditional support of these parties is defined in terms of left and right, the integration dimension runs orthogonal to this basic alignment. Therefore national parties (as well as the European party federations) contain a potential source of internal conflict. To avoid negative electoral consequences, parties converge to a common pro-European position, so preventing conflict and competition. Mobilization capacities are restrained by the risk of internal dissent.

This view is not uncontested. Hooghe and Marks (1999) suggest that the two dimensions are strongly related in certain policy fields, and Gabel and Anderson (2004) even see them merged in one single dimension. However, as Gabel and Hix (2002: 954) argue, mapping the dimensions of conflict between parties does not tell us anything about the initial problem of conflict within parties or their federations. On the contrary, it is exactly the process of strategic assimilation that may conceal this troublesome feature of European party systems. At the very least there seem to be increasing signs of negative consequences for the competitive strength of parties affected by internal dissent (Ferrara and Weishaupt, 2004; Steenbergen and Scott, 2004), a phenomenon already expected by Reif and Schmitt (1980: 11).


Knowing the structure of the European political space, we can now identify the Zallerian properties of the two dimensions of party competition. On the left-right dimension, with its broad range of topics and its favourable structural properties, I assume high intensity and high familiarity. On the integration dimension, with its lack of structural identity and its potential for intra-party dissent, I assume low intensity and low familiarity. Accordingly,

the hypotheses concerning first-order campaign effects in EP elections read as follows:

H1 Left-right: Campaign effects mainly occur close to national elections. H2 Integration: Campaign effects are generally weak. H3 Left-right: Higher awareness leads to weaker campaign effects. H4 Integration: Higher awareness leads to stronger campaign effects. H5 Left-right: Campaign effects polarize mass opinion. H6 Integration: Campaign effects streamline mass opinion.

Additionally, the model considers the size of the competing parties. Bigger parties should get more news coverage, which raises intensity. On the other hand, intensity might suffer because big parties must design their messages for a broad electorate, following Kirchheimer's (1966) 'catch-all' logic. Given the above discussion, I expect this logic on the left-right dimension only. However, the same net effect might result on the integration dimension because small parties are less affected by the risk of internal divisions (Gabel and Scheve, 2007). For this reason, the following analysis deals with an open empirical question on both dimensions. The third logical possibility, a quadratic relation, with medium-sized parties doing best or worst, gained no support.

H7a Both dimensions: Party size enhances campaign effects. H7b Both dimensions: Party size reduces campaign effects.

Apart from electoral size, parties are also characterized by the size of their issue sets. Because campaign communication relies on the medium of generalized dimensions, single-issue parties should do poorly:

H8 Both dimensions: Single-issue parties generate weaker campaign effects than ideological platform parties.

I employ a three-step procedure to test these hypotheses in a Zallerian model and relate them to our problem of government losses in EP elections. The first part deals with cyclical information flows as measured by the certainty of voters' perceptions of party positions. The second part evaluates the mobilization capacity of national parties in EP elections by focusing on spatial representation in terms of Euclidean distances between voter and party positions. The third part analyses processes of persuasion and opinion polarization as a link between information flows and representation, thus clarifying the logic behind poor government performance.

Modelling communication of party positions

As the discussion of the Zaller model showed, even the mere communication of party positions involves a process of persuasion, so the propositions concerning both reception and acceptance must be considered at this stage. Voters will assign a certain position to a party only if they are convinced the party is going for these policies. Hence, the initial task is to measure voters' certainty about party positions. Here I follow Van der Eijk (2001), who designed a measure to record the perceptual agreement among voters on the exact position of a party. Thus, the certainty of individual voters is reflected in the agreement of the whole electorate.

The relation to my hypotheses is clear. If parties advertise their platforms intensively, perceptual agreement among voters is supposed to rise. If campaign efforts slow down, public perception should diffuse. Comparable learning processes are best known from analyses of US election campaigns, but have also been detected in the European context (Andersen et al., 2005; Arceneaux, 2006). However, this should not be confused with the hypothesis on opinion polarization. Perceptions of party positions are expected to converge during a campaign, whereas mass opinion - i.e. voters' own policy positions - should diverge across but not within parties.

The measure of agreement is applicable to ordered rating scales such as the 10-point left-right and integration dimensions in the European Election Studies (EES) data sets. Because it is insensitive to the location of the mean, it is superior to standard deviations or related measures.4 However, in spite of this and some other favourable properties, the measure has one flaw: it does not take missing values into account. Especially in the case of questions referring to knowledge this is an important point. Response could be subject to self-selection, so uncertain respondents simply produce missing values and agreement is finally overestimated. To neutralize the undesirable data variation resulting from self-selection, the share of missing values is added as a control variable.

Van der Eijk's measure was calculated for all the parties included in the batteries on left-right and integration positions from the EES of 1999 and 2004. Descriptive statistics support the general assumption of differences between the two dimensions: mean agreement on the left-right dimension (0.45) is higher than on the integration dimension (0.31). However, the dynamics involved in my hypotheses can be tested only in a multivariate regression design with several independent variables.

Following Reif (1984), the effect of cyclical campaign mobilization is expressed by the time-span between an EP election and the preceding national election, relative to the length of the legislative period in a nation-state. This

gives a measure of closeness to any of the two surrounding national elections, ranging from 0 to 1. The closer EP elections are held to a national election, the higher should be the effect of partisan contention. To test this hypothesis, the cycle position is considered in its simple and squared form. The simple term represents the initial (supposedly downward) course of the cycle, while the squared term should raise the curve towards the end of the legislative period.

Political awareness is represented by a four-point interest scale, averaged per country and election year, where 1 stands for low and 4 for high interest in politics. I am aware that, for Zaller, measures of knowledge are superior to measures of interest for representing political awareness. Unfortunately, the EES data set for 1999 contains only a very narrow, binary knowledge item, and the data set for 2004 contains none at all. Therefore we have to be satisfied with the 'limited' but still related effect of political interest (Zaller, 1992: 43).

As to the control variables, single-issue parties are denoted by a dummy.5 Party size is operationalized as the share of valid votes at the preceding national election. The related hypotheses (news coverage and mass appeal) imply that party size is revealed before the EP elections. The choice of the preceding national election as a reference point reflects this assumption. The size variable and the share of missing values for perceptual agreement are both scaled 0 to 1. The share of missing values for Belgium (1999) and Great Britain (2004) is calculated on the basis of regional subsamples.

Initial inspection of the data identified several outliers not explained even by the set of control variables. Robust ordinary least squares (OLS) regression (Huber-bisquare transformation) is used to deal with this problem. Although keeping outliers in implies estimation bias, simply dropping them is hard to justify on theoretical grounds. Robust regression is an attempt to minimize the consequences of both problems simultaneously.6 Through an iterative, dual weighting procedure, the influence of outliers is gradually reduced. Additionally, cases are inversely weighted by the number of parties per country and election year while overall n is kept constant. This gives equal weight to each political system irrespective of the number of parties included in the data set.

The regression results in Table 1 reveal many expected effects, but also a big surprise. First of all, the campaign cycle runs as expected on the left-right dimension. Parties succeed in forming a relatively uniform mass opinion about their platforms at the time of national elections, whereas the midterm is characterized by an information deficit (H1). During a campaign, voters learn about left-right party positions, so the foundation stone for 'enlightened preferences' (Gelman and King, 1993) is laid. The electoral connection

Table 1 Perceptual agreement through the electoral cycle

Left-right Integration

Cycle position (simple) -0.334** 0.262**

(0.120) (0.084)

Cycle position (squared) 0.317** -0.384**

(0.121) (0.086)

Political awareness -0.172** 0.073**

(0.035) (0.028)

Party size -0.228** 0.046

(0.074) (0.051)

Single-issue party -0.193** 0.041

(0.027) (0.021)

Share of missing values (dimension-specific) -0.528** 0.051

(0.084) (0.053)

Constant 1.097** 0.092

(0.104) (0.085)

Observations 264 258

Adjusted R2 .311 .139

F statistic 21** 8**

Notes: Robust OLS coefficients (Huber-bisquare transformation) with standard errors in parentheses. Cross-sectional, pooled analysis of the European Election Studies 1999 and 2004. The dependent variable is Van der Eijk's (2001) measure of perceptual agreement (party-level aggregates of individual-level data). Cases are inversely weighted by the number of parties per country and election while n is kept constant. ** significant at 0.01

in national parliamentary elections is supposed to benefit from this process, but most EP elections are held under unfavourable conditions for a meaningful vote. Just like government losses in EP elections, the clarity of party positions is a function of proximity to a national election.

Turning to the integration dimension, one finds the surprise promised above: not only does the cycle not follow the left-right pattern (H2); it actually runs in the opposite direction! Agreement is highest at the midterm and lowest at election time in the member states. Parties do not mobilize voters during domestic election campaigns; they demobilize them by taking vague and ambiguous positions on integration issues. This takes the Hix-Lord argument to its extremes: parties seem to be so afraid of internal conflict over European integration that they try to neutralize this issue whenever important elections are approaching. Only at the midterm are voters granted a rest from this sorry spectacle.

Furthermore, the higher absolute value of the coefficient for the squared term suggests that the logic of demobilization is most effective in the run-up to a first-order election and that voters seem to 'recover' very soon once that election has passed. This fits nicely with the assumptions of ultimately office-motivated party leaders and a deliberate attempt at 'artificial' disinformation, going beyond the mere 'de-politicization of the issue of European integration' (Van der Brug and Van der Eijk, 2007: 277).

The effects of political awareness conform to my hypotheses. The intense campaign information on the left-right dimension gets through especially to unaware voters with little means to resist, so awareness is negatively related to agreement (H3). The reverse is true on the integration dimension (H4), confirming earlier findings on the role of political interest in mediating elite influence on voter positions towards European integration (Ray, 2003). It probably needs quite some awareness to cope with strategic disinformation.

Party size negatively affects agreement on left-right (H7). Obviously big parties have to obey the rules of 'catch-all' politics whereas small parties can mobilize their voters with precise policy announcements. As expected, single-issue parties do poorly on the highly generalized left-right dimension (H8). However, neither party size nor the single-issue dummy shows noteworthy effects on the integration dimension. Rather ironically, the big platform parties do not do any better than marginal single-issue parties in communicating positions on European integration (nor, arguably, do they want to).

On the integration dimension, the influence of the share of missing values tends to support the suspicion of self-selection, although it is not significant. On the left-right dimension, the expected effect is even reversed. Probably the high familiarity of left-right issues reduces the incentives for self-selection.

Spatial representation

Recapitulating the preceding section, there is considerable evidence of campaign effects on voter perceptions of party positions. However, it should be noted that cyclical learning is only a necessary but insufficient condition to explain government losses in EP elections. Counter-proposition (a) of the alternative second-order model predicts a cyclical course of first-order campaign effects, which is supported by the evidence presented so far. Beyond this evidence, however, counter-proposition (b) requires campaign mobilization to affect not only perceptions but also preferences themselves. When parties communicate distinct political positions to their voters, mass opinion should polarize on the left-right scale and effective representation should come about.

To deal with these questions, actual voting behaviour is taken into account. Concerning second-order elections, we can distinguish core voters (who stick with their party) and defectors (who turn away from their party) and compare the policy positions of the two groups. The basic assumption is that core voters are in a state of partisan mobilization at any time during the election period whereas the behaviour of defectors reflects a mobilization deficit. The core voters' positions therefore provide a baseline that serves to identify perceptual and preferential changes underlying defection. But what kind of defection should be evaluated against this baseline?

According to the conventional second-order theory, strategic protest voting against the government is the form of defection responsible for the second-order effect. However, as my theoretical discussion demonstrated, protest voting is not the only possible motive - an argument also supported by empirical data. As Heath et al. (1996) show, government losses can equally be induced by true conversion towards opposition parties. To address this empirical question, I propose a comprehensive voter typology incorporating EES measures of behaviour, utility and attitudes (see Weber, 2007). These measures concern party choice in the EP election (EE), the preceding national election (NE1) and a hypothetical national election the day after the interviews (NE2). With these instruments, we can distinguish the voter types presented in Table 2.

Empirical application of this typology to the EES data sets of 1999 and 2004 shows that - as predicted by the second-order model - protest voting is important in quantitative terms. However, cyclical government losses in particular are induced by true conversion towards opposition parties and by

Table 2 A voter typology for European Parliament elections

Type Characteristics

Core voters Stick with one party

Strategic protest voters Exert pressure on their party in EE by switching or


Sincere Euro-switchers Switch from strategic voting in NE1 to sincere voting in EE

Converts Start a lasting preference change at EE

Alienated Turn away from the whole party system from EE on

Indifferent Mix strategies by repeated switching and/or abstaining

Uninterested Do not care for nor vote in EE

Dissatisfied Do not regard EU democracy as appropriate nor vote in EE

Serial non-voters Do not vote in any of the three elections

Note: Details in Weber (2007).

alienation from the whole party system (see Weber, 2007). This result supports the idea of a top-down approach to second-order elections: electoral mobilization should be more likely to affect sincere voter orientations than strategic motives, given that the latter require at least some crude but still conscious consideration. Therefore, the following analysis of the representation process in EP elections considers converted and alienated voters in addition to core and protest voters.

Based on Schattschneider's Responsible Party Model, I have argued that mobilization depends on the quality of democratic representation. If this is correct, core voters should be better represented than any of the defecting groups. Following Achen (1978), I measure representation by the Euclidean distances between parties (voter perceptions) and voters (self-assessments) on the two 10-point scales, averaged per voter type.7 Additionally, representation is evaluated for the three elections mentioned above (EE, NE1 and NE2) so that signs of strategic voting or changing preferences can be observed over time. The figures in Table 3 always refer to the party a respondent voted for (or intends to vote for) in the respective election.

The first result from Table 3 is clear: in all cases, representation on the left-right dimension is better than on the integration dimension. Even in EP elections this pattern does not change. Furthermore, core voters are better represented than are the other groups. For core voters, the quality of representation is sufficient to mobilize them for at least three elections in a row.8 Protest voters, by contrast, are further away from their NE1 party. They react to this deficit by strategically supporting a party they actually dislike in the EE, which is reflected by the distance measure. For converts, the initial representation deficit is much more pronounced. It leads them to abandon

Table 3 Patterns of programmatic (in)congruence

Left-right Integration


Core voters 1.10 1.11 1.09 1.76 1.75 1.74

Strategic protest

voters 1.31 1.91 1.37 1.93 2.18 1.87

Converts 2.22 1.57 1.72 2.38 2.07 2.09

Alienated 1.60 abstained abstained 2.22 abstained abstained

Notes: Euclidean distances between parties (voter perceptions) and voters (self-assessments) on 10-point scales. Distances are averaged per voter type and election (first and second national election, EP election).

Data source: European Election Studies 1999 and 2004.

their NE1 party and support a closer party in the long run. The NE1 distance for alienated voters is higher compared with protest voters but lower compared with converts. Obviously protest voting is no longer a reasonable option, but long-term change is a step too far.

Two conclusions can be drawn from this analysis. First, the distinction between the two dimensions is confirmed. Ironically, a distance measure producing alienation on the left-right dimension is still better than the maximum that core voters achieve on the integration dimension! Secondly, there is reason to assume that second-order voting behaviour is determined by the quality of representation by national parties. Because conversion and alienation strike government parties harder than opposition parties and these motives come with higher distance measures, government losses can be ascribed to a representation deficit. According to the alternative second-order model, this deficit results from the poor performance level that parties achieve in mobilizing their voters at the midterm.

Persuasion and polarization

In spite of everything, counter-proposition (b) is not exhausted yet. According to Schattschneider (1942), congruence between parties and voters depends on distinct programmatic alternatives. A similar argument can be made with Zaller (1992): as the careful reader will have noticed, hypothesis 5 has been passed over without comment. However, it is exactly the polarization of mass opinion (at least on the left-right dimension9) in reaction to diverging party positions that would complete my picture. So far I have shown that parties can change both voter perceptions of their positions and the quality of spatial representation. But does representation depend only on perceptions, or does it also involve processes of persuasion? This is a decisive step given that the second-order effect ultimately reflects the inability of the government to polarize mass opinion.

Needless to say, tracking opinion change with cross-sectional data is no easy task. Although time-series data have long been shown to be useful in this context (Abramowitz, 1978), cross-sectional data require other methods to pinpoint change. Fortunately we can go back to the baseline measure discussed in the previous section: with core voters showing no sign of declining mobilization, their positions can be contrasted with the dynamics in the defector groups. Because the size of these groups varies in a cyclical manner, opinion change can be derived logically if we find systematic differences indicating opinion polarization. In this case, it should be possible to predict deviations of defector groups by core voter positions, with defectors drifting towards the centre of the political space.

Figure 1 shows the predictions obtained by regressing the mean left-right perceptions and self-assessments of the defector groups on those of the core voter groups. Each data point consists of the value for the core voters sticking with a specific party and the value for the defectors turning away from the same party.

To begin with party positions, we find the expected phenomenon: the more extreme the core voters' view of their parties, the more moderate is the perception of the defector groups in relative terms. Hence parties mobilize their voters by promoting a public image of distinct, diverging platforms. Whereas core voters receive these messages, defectors are at least temporarily unaffected by the communication process and fail to update their perceptions of party positions.

Although differential perceptions span only about one point on the 10-point left-right scale, they have profound consequences for voters' own positions. The more extreme the assessment by core voters of their own preferences, the more moderate are the self-assessments of the defector groups in relative terms. This difference is up to four points on the scale. Voters accept persuasive messages and follow their parties towards the extremes of the political space when confronted with effective campaign mobilization (H5). If the latter is missing, voters swing back to their original positions at the centre and defect from their former parties.

One might object to this interpretation because the distance between voters and parties could be affected by ex post rationalization. That would mean that voters project their own positions onto the party they have voted for and seek a contrast with the other parties (Granberg and Holmberg, 1988; Van der Brug, 1996). However, if voters take relatively moderate positions at the midterm and parties go for more extreme positions during a campaign, then why do voters not continue to project their positions onto parties? Instead, voters follow their parties if and only if the latter embark on extreme strategies. This must be a process of persuasion; projection can be ruled out logically.

Concerning the differences between voter types, the pattern found above is confirmed. The largest swing-back is associated with conversion, followed by alienation and protest voting. Not only does the representation deficit highlighted in the previous section correlate with a general state of depolarization, but both phenomena feature the same specific subgroup line-up. Figure 2 verifies this finding by employing a direct test. Instead of comparing perceptions or self-assessments between voter groups, voters' perceptions of party positions are now plotted against their own self-assessments. This procedure is used to reveal the spatial pattern of the general congruence deficit set out in Table 3.

Whereas core voters are appropriately represented by their NE1 parties, protest voters, alienated voters and converts depart from this ideal bit by bit.

Perceptions of party positions

Core voters

---Protest voters-----Alienated

........ Converts - y = x


Core voters

---Protest voters-----Alienated

........ Converts - y = x

Figure 1 The polarization of mass opinion.

Notes: Regression lines obtained as described for Table 1. All parameters significant at .01.

Perceptions and self-assessments compared

Left 23 45 6789 Right


— Core voters —

- Alienated

— Protest voters --■ Converts

Figure 2 Polarization and representation.

Notes: Regression lines obtained as described for Table 1. All parameters significant at .01.

This goes so far that converts see their former party always at the centre of the space, almost independently of their own position! At the extremes of the dimension, this implies a significant congruence gap of about four points on the left-right scale. As the four regression lines make clear, the congruence deficit cannot be explained simply by random variation in party and voter positions in defector groups. Rather, there is a systematic relation between voter positions and congruence: the more extreme a defector, the higher is the distance from the former party.

The conclusion from this analysis is clear: if voters are not provided with distinct policy platforms, they are not willing to stick with their parties. In a strict sense there are two processes at work. The self-assessments of some extreme voters are fairly stable, but they are confronted with a relatively moderate perception of party positions at the midterm. The resulting in-congruence provokes defection. For other voters, both perceptions and self-assessments are subject to swing-back. Interestingly, defection occurs even though a new congruence position emerges at the centre of the space. This confirms my assumption that effective representation means congruence induced by mobilization via distinct policy alternatives, not just any form of congruence. Because opposition parties share the freedom to polarize,

government losses in EP elections can be traced back to this top-down logic of the democratic process.


Irrespective of all the theoretical and methodological divergences, this paper joins a long line of electoral research. The results of elections to the European Parliament can be interpreted according to the logic of second-order elections insofar as national rather than European factors account for the cyclical government losses in the second-order arena. However, although the debate is only about the set of relevant factors within this framework, the exact choice has profound implications. The conventional second-order theory proposes a bottom-up model in which strategic protest voters voice dissatisfaction with their government's performance. In contrast, it can be shown that a top-down model of first-order campaign mobilization accounts for the second-order effect.

Once parties are assigned an active role, their strategic interest in maximizing support towards election day in their domestic arenas leads to a campaign cycle. As a result, high defection rates in EP elections held during the midterm can be explained by a mobilization deficit. More specifically, parties fail both in communicating their policy positions and in persuading voters to adopt them. This has far-reaching consequences because electoral mobilization can be shown to work only if parties present distinct programmatic alternatives to the public. Because prevailing retrospective evaluations of policy outcomes restrict the capacity of national government parties to follow this logic during the midterm, their cyclical losses ultimately reflect the intensity and effectiveness of domestic campaigning.

For democracy at the national level this is good news. As the pressure of political competition prompts parties to promote a clear choice, programmatic voter-party congruence is brought about and representation can be said to function effectively. However, the European arena may benefit only indirectly from national campaigns, and even these mere by-products are missing in most cases. Without the orientation guide provided by political competition, voters are generally lost with regard to EU politics.

Furthermore, the virtues of campaign effects can be found only in connection with traditional socioeconomic issues. In contrast, questions of European integration are systematically kept out of political competition. Because the major parties are plagued by internal dissent over integration, they not only avoid the issue but actually try to demobilize voters at election time in the EU member states by taking vague and ambiguous positions with regard to EU-specific affairs.

The implications for democracy in the EU are rather depressing. Political representation by the European Parliament is dominated by a systematic bias whose roots lie in the logic of party competition in the national arenas. Therefore, minor institutional reform may not serve as a remedy for the infamous 'democratic deficit'. However, major reform strategies aiming at the full parliamentarization of the EU would turn the basic logic upside down: if Europe became the first-order arena, mobilization deficits would necessarily emerge in the nation-states; second-order effects are adaptable.

Notes -

I am indebted to Bernhard Wessels for inspiring discussions throughout the development of this research. Additional thanks go to Thilo Bodenstein, Michael Bolle, Mark Furness, Berthold Rittberger and four anonymous reviewers for their helpful comments on earlier versions of this paper. The survey data utilized for the quantitative analysis were originally collected by the European Election Study research group (see The usual disclaimer applies.

1 These surveys were conducted across the EU following the EP elections of 1999 and 2004. Only Malta was not covered in 2004. Furthermore, important variables are missing for Northern Ireland (both elections) and for Belgium, Lithuania, Luxembourg and Sweden (2004). These countries had to be dropped. The data sets were pooled to obtain a sufficient number of cases to test for a typical quadratic/cyclical effect. Individual-level cases were double-weighted so each country has the same weight and party support reflects the official results of the EP elections. The EES weighting had to be corrected on several occasions.

2 Please note that this argument differs from the 'campaign dimension' of the conventional second-order model. Reif and Schmitt (1980: 13f.) refer exclusively to campaign efforts on the occasion of EP elections, and not to the more general logic explicated here.

3 One might suspect a trace of the 'directional' theory of party competition (Rabinowitz and Macdonald, 1989) in this argument on top-down representation. However, my model of voting adheres to the concept of programmatic positions (rather than direction and intensity) and thus remains in the Downsian spatial paradigm. On the other hand, it may be considered compatible with directional theory when the latter is blended with mobilization effects endogenous to the electoral process (see Iversen, 1994).

4 The measure is constructed by decomposing the empirical distribution into weighted layers and assessing the degree of unimodality on the basis of triads of categories. Complete agreement is represented by 1, perfect polarization by -1 and an even distribution by 0. For details see Van der Eijk (2001).

5 The 'single issues' identified are: corruption, pensions, immigration, hunting, ethnic or regional minority concerns, ethnocentric nationalism, religious matters, and anti-EU protest. Of course the last issue is not included for the integration dimension.

6 Note, however, that the results remained robust both without any outlier treatment and with outlier elimination according to the influence of individual cases measured by Cook's D.

7 Achen shows that the mean is preferable to correlational measures. Unlike Achen, I do not use squared values, which would add to the sensitivity of the mean to extreme values, and I consider party-voter dyads instead of representative-constituent dyads, which suits the reality of European party systems better than the US-centred representative model (Esaiasson, 1999).

8 There are slight differences between the elections because intra-government and intra-opposition switching is allowed. Obviously the consequences are irrelevant.

9 Because 'competition' on the integration dimension is characterized by demobilization, whatever the result it could not be attributed to mobilization effects. Therefore the following procedure is applied to the left-right dimension only (H6 is dropped).


Abramowitz, Alan I. (1978) 'The Impact of a Presidential Debate on Voter Rationality', American Journal of Political Science 22(3): 680-90.

Achen, Christopher H. (1978) 'Measuring Representation', American Journal of Political Science 22(3): 475-510.

Andersen, Robert, James Tilley and Anthony F. Heath (2005) 'Political Knowledge and Enlightened Preferences: Party Choice through the Electoral Cycle', British Journal of Political Science 35(2): 285-302.

Arceneaux, Kevin (2006) 'Do Campaigns Help Voters Learn? A Cross-National Analysis', British Journal of Political Science 36(1): 159-73.

Bartels, Larry M. (2006) 'Priming and Persuasion in Presidential Campaigns', in Henry E. Brady and Richard Johnston (eds) Capturing Campaign Effects, pp. 78-112. Ann Arbor: University of Michigan Press.

Dalton, Russell J. (1985) 'Political Parties and Political Representation: Party Supporters and Party Elites in Nine Nations', Comparative Political Studies 18(3): 267-99.

Druckman, James N. (2004) 'Priming the Vote: Campaign Effects in a U.S. Senate Election', Political Psychology 25(4): 577-94.

Esaiasson, Peter (1999) 'Not All Politics Is Local: The Geographical Dimension of Policy Representation', in Warren E. Miller, Roy Pierce, Jacques Thomassen, Richard Herrera, Sören Holmberg, Peter Esaiasson and Bernhard Wessels, Policy Representation in Western Democracies, pp. 110-36. Oxford and New York: Oxford University Press.

Esaiasson, Peter and Sören Holmberg (1996) Representation from Above: Members of Parliament and Representative Democracy in Sweden. Aldershot: Dartmouth.

Ferrara, Federico and J. Timo Weishaupt (2004) 'Get Your Act Together: Party Performance in European Parliament Elections', European Union Politics 5(3): 283-306.

Fuchs, Dieter and Hans-Dieter Klingemann (1989) 'The Left-Right Schema', in M. Kent Jennings, Jan W. van Deth, Samuel H. Barnes, Dieter Fuchs, Felix

J. Heunks, Ronald Inglehart, Max Kaase, Hans-Dieter Klingemann and Jacques Thomassen, Continuities in Political Action: A Longitudinal Study of Political Orientations in Three Western Democracies, pp. 203-34. Berlin and New York: de Gruyter.

Gabel, Matthew J. and Christopher J. Anderson (2004) 'The Structure of Citizen Attitudes and the European Political Space', in Gary Marks and Marco R. Steen-bergen (eds) European Integration and Political Conflict, pp. 13-31. Cambridge: Cambridge University Press.

Gabel, Matthew J. and Simon Hix (2002) 'Defining the EU Political Space: An Empirical Study of the European Election Manifestos, 1979-1999', Comparative Political Studies 35(8): 934-64.

Gabel, Matthew J. and Kenneth Scheve (2007) 'Mixed Messages: Party Dissent and Mass Opinion on European Integration', European Union Politics 8(1): 37-59.

Gelman, Andrew and Gary King (1993) 'Why are American Presidential Election Campaign Polls So Variable When Votes Are So Predictable?', British Journal of Political Science 23(4): 409-51.

Goodhart, Charles A.E. and Rajendra J. Bhansali (1970) 'Political Economy', Political Studies 18(1): 43-106.

Granberg, Donald and Sören Holmberg (1988) The Political System Matters: Social Psychology and Voting Behavior in Sweden and the United States. Cambridge and Paris: Cambridge University Press and Maison des Sciences de l'Homme.

Heath, Anthony F., Roger Jowell, John Curtice and Bridget Taylor (1996) 'The 1994 European and Local Elections: Abstention, Protest and Conversion', CREST Working Paper No 46. London and Oxford: SCPR and Nuffield College.

Hirschman, Albert O. (1970) Exit, Voice, and Loyalty. Responses to Decline in Firms, Organizations, and States. Cambridge, MA: Harvard University Press.

Hix, Simon and Christopher Lord (1997) Political Parties in the European Union. New York: Macmillan.

Hooghe, Liesbet and Gary Marks (1999) 'The Making of a Polity: The Struggle over European Integration', in Herbert Kitschelt, Peter Lange, Gary Marks and John D. Stephens (eds) Continuity and Change in Contemporary Capitalism, pp. 70-97. Cambridge: Cambridge University Press.

Iversen, Torben (1994) 'The Logics of Electoral Politics: Spatial, Directional, and Mobilization Effects', Comparative Political Studies 27(2): 155-89.

Johnston, Ron and Charles Pattie (2001) 'Dimensions of Retrospective Voting: Economic Performance, Public Service Standards and Conservative Party Support at the 1997 British General Election', Party Politics 7(4): 469-90.

Kiewiet, D. Roderick (1981) 'Policy-Oriented Voting in Response to Economic Issues', American Political Science Review 75(2): 448-59.

King, Anthony (1975) 'Overload: Problems of Governing in the 1970s', Political Studies 23(2-3): 284-96.

Kirchheimer, Otto (1966) 'The Transformation of Western European Party Systems', in Joseph LaPalombara and Myron Weiner (eds) Political Parties and Political Development, pp. 177-200. Princeton, NJ: Princeton University Press.

Kousser, Thad (2004) 'Retrospective Voting and Strategic Behavior in European Parliament Elections', Electoral Studies 23(1): 1-21.

Lindberg, Leon and Stuart Scheingold (1970) Europe's Would-Be Polity: Patterns of Change in the European Community. Englewood Cliffs, NJ: Prentice Hall.

Lipset, Seymour Martin and Stein Rokkan (1967) 'Cleavage Structures, Party Systems and Voter Alignments: An Introduction', in Seymour Martin Lipset

and Stein Rokkan (eds) Party Systems and Voter Alignments: Cross-National Perspectives, pp. 1-64. New York: The Free Press.

Lorenz, Konrad (1937) 'Biologische Fragestellungen in der Tierpsychologie', Zeitschrift für Tierpsychologie [Ethology] 1(1): 24-32.

Luhmann, Niklas (1981) 'Der Politische Code. "Konservativ" und "Progressiv" in Systemtheoretischer Sicht', in Niklas Luhmann, Soziologische Aufklärung 3. Soziales System, Gesellschaft, Organisation, pp. 267-86. Opladen: Westdeutscher Verlag.

Marsh, Michael (2005) 'The Results of the 2004 European Parliament Elections and the Second-Order Model', in Oskar Niedermayer and Hermann Schmitt (eds) Europawahl 2004, pp. 142-58. Wiesbaden: VS-Verlag.

Miller, Joanne M. and Jon A. Krosnick (1996) 'News Media Impact on the Ingredients of Presidential Evaluations: A Program of Research on the Priming Hypothesis', in Diana C. Mutz, Paul M. Sniderman and Richard A. Brody (eds) Political Persuasion and Attitude Change, pp. 79-99. Ann Arbor: University of Michigan Press.

Miller, Warren E., Roy Pierce, Jacques Thomassen, Richard Herrera, Sören Holmberg, Peter Esaiasson and Bernhard Wessels (1999) Policy Representation in Western Democracies. Oxford and New York: Oxford University Press.

Nordhaus, William D. (1975) 'The Political Business Cycle', Review of Economic Studies 42(2): 169-90.

Rabinowitz, George and Stuart Elaine Macdonald (1989) 'A Directional Theory of Issue Voting', American Political Science Review 83(1): 93-121.

Ray, Leonard (2003) 'When Parties Matter: The Conditional Influence of Party Positions on Voter Opinions about European Integration', Journal of Politics 65(4): 978-94.

Reif, Karlheinz (1984) 'National Electoral Cycles and European Elections 1979 and 1984', Electoral Studies 3(3): 244-55.

Reif, Karlheinz and Hermann Schmitt (1980) 'Nine Second-Order National Elections - A Conceptual Framework for the Analysis of European Election Results', European Journal of Political Research 8(1): 3-44.

Schattschneider, Elmer E. (1942) Party Government. New York: Holt, Rinehart & Winston.

Schattschneider, Elmer E. (1960) The Semisovereign People: A Realist's View of Democracy in America. New York: Holt, Rinehart & Winston.

Schmitt, Hermann (2005) 'The European Parliament Elections of June 2004: Still Second-Order?', West European Politics 28(3): 650-79.

Schmitt, Hermann and Renato Mannheimer (1991) 'About Voting and Non-Voting in the European Elections of June 1989', European Journal of Political Research 19(1): 31-54.

Schmitt, Hermann and Jacques Thomassen (eds) (1999) Political Representation and Legitimacy in the European Union. Oxford and New York: Oxford University Press.

Steenbergen, Marco R. and David J. Scott (2004) 'Contesting Europe? The Salience of European Integration as a Party Issue', in Gary Marks and Marco R. Steen-bergen (eds) European Integration and Political Conflict, pp. 165-92. Cambridge: Cambridge University Press.

Steenbergen, Marco R., Erica E. Edwards and Catherine E. de Vries (2007) 'Who's Cueing Whom? Mass-Elite Linkages and the Future of European Integration', European Union Politics 8(1): 13-35.

Thomassen, Jacques (1999) 'Political Communication between Political Elites and Mass Publics: The Role of Belief Systems', in Warren E. Miller, Roy Pierce, Jacques Thomassen, Richard Herrera, Sören Holmberg, Peter Esaiasson and Bernhard Wessels, Policy Representation in Western Democracies, pp. 33-58. Oxford and New York: Oxford University Press.

Thomassen, Jacques and Hermann Schmitt (1999) 'Partisan Structures in the European Parliament', in Richard S. Katz and Bernhard Wessels (eds) The European Parliament, the National Parliaments, and European Integration, pp. 129-48. Oxford and New York: Oxford University Press.

Tinbergen, Nikolaas (1963) 'On Aims and Methods of Ethology', Zeitschrift für Tierpsychologie [Ethology] 20(4): 410-33.

Toka, Gabor (2007) 'Information Effects on Vote Choices in European Elections', in Michael Marsh, Slava Mikhaylov and Hermann Schmitt (eds) European Elections after Eastern Enlargement: Preliminary Results from the European Election Study 2004, pp. 141-79. Mannheim: CONNEX.

Van der Brug, Wouter (1996) Where's the Party? Voters' Perceptions of Party Positions. Amsterdam: Amsterdam University Press.

Van der Brug, Wouter and Cees van der Eijk (2007) 'European Elections, Domestic Politics and European Integration', in Wouter Van der Brug and Cees van der Eijk (eds) European Elections and Domestic Politics. Lessons from the Past and Scenarios for the Future. Southbend, IN: University of Notre Dame Press.

Van der Eijk, Cees (2001) 'Measuring Agreement in Ordered Rating Scales', Quality & Quantity 35(3): 325-41.

Van der Eijk, Cees and Mark N. Franklin (eds) (1996) Choosing Europe? The European Electorate and National Politics in the Face of Union. Ann Arbor: University of Michigan Press.

Van der Eijk, Cees and Mark N. Franklin (2004) 'Potential for Contestation on European Matters at National Elections in Europe', in Gary Marks and Marco R. Steenbergen (eds) European Integration and Political Conflict, pp. 32-50. Cambridge: Cambridge University Press.

Weber, Till (2007) 'Exit, Voice, and Cyclicality: Testing the Micro-Logic of Second-Order Elections Theory', mimeo, Freie Universität Berlin.

Zaller, John R. (1992) The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.

About the author

Till Weber is a researcher in the Department of Political and Social Sciences, European University Institute, Via dei Roccettini 9, 50014 San Domenico di Fiesole, Italy. Fax: +39 055 4685298 E-mail: