A Two-Stage Theory of Discussant Influence on Vote Choice in Multiparty Systems

Rüdiger Schmitt-Beck; Julia Partheymüller

doi:10.1017/S0007123414000301

A Two-Stage Theory of Discussant Influence on Vote Choice in Multiparty Systems

Published online by Cambridge University Press: 21 October 2014

Rüdiger Schmitt-Beck and

Julia Partheymüller

Article contents

Abstract
DISCUSSANT INFLUENCE AT ELECTIONS
ELECTORAL CHOICES IN MULTIPARTY SYSTEMS
A TWO-STAGE THEORY OF DISCUSSANT INFLUENCE AT ELECTIONS IN MULTIPARTY SYSTEMS
HYPOTHESES
DATA AND METHODS
RESULTS
CONCLUSION
Footnotes
References

Rights & Permissions

Abstract

This article addresses two aspects of social network influence on voters’ electoral choices that are not well understood: the role of party systems as institutional contexts and the relationship between social pressure and information sharing as mechanisms of influence. It argues that in the cleavage-based multiparty systems of Western Europe, discussant influence at elections occurs in two stages. First, discussants place social pressure on voters to opt for parties from the same ideological camp. Secondly, by providing information, discussants influence which parties voters eventually choose out of these restricted ‘consideration sets’. The study tests these assumptions using a panel survey conducted at the 2009 German federal election. The first proposition is clearly confirmed, and the evidence supports the second proposition, although less unequivocally.

Information

Type: Articles
Information: British Journal of Political Science , Volume 46 , Issue 2 , April 2016, pp. 321 - 348
DOI: https://doi.org/10.1017/S0007123414000301 [Opens in a new window]
Copyright: © Cambridge University Press 2014

Voters’ electoral choices do not develop in a social vacuum. When making up their minds about which candidate or party to support at the ballot box, citizens refer to the members of their social networks. Disagreeing over electoral preferences with other people stimulates them to reconsider their decisions, and often leads voters to adapt their preferences to those of their discussants. As a consequence, voting behaviour appears contagious. This ‘social logic’ of votingFootnote ¹ has been demonstrated by numerous studies.Footnote ² While this thriving literature has made unequivocally clear that the phenomenon of discussant influence at elections is a substantive one, it is in various ways not yet well understood. This article addresses two questions that extant research has not yet solved: (1) how discussant influence works in the context of West European multiparty systems and (2) how social pressure and information sharing relate to one another as mechanisms of discussant influence.

Studies on discussant influence have rarely addressed issues of institutional context. Most analyses of voters’ electoral interdependence have focused on candidate voting in the American two-party system. Few studies have been conducted in countries with multiparty systems, and they have usually construed electoral influence in the same way as American studies. With few exceptions, this research thus failed to take into account the more complex conditions of party competition and choice in such systems.Footnote ³

In the two-party US system, political competition is dualistic and confrontational. Electoral disagreement between a voter and a discussant inevitably involves being on different sides of the fence, and therefore entails antagonism and discord.Footnote ⁴ In West European multiparty systems, party competition is more complex. Reflecting socio-political cleavages, these systems are structured in polar ideological campsFootnote ⁵ or directionsFootnote ⁶ that usually contain several different parties. Accordingly, voters can experience at least two different types of electoral disagreement with their discussants, depending on whether it concerns parties from the same or the opposing ideological camp. The latter type is dualistic and therefore resembles antagonistic disagreement in the American sense. But the former concerns ideologically adjacent parties that differ in details, but share fundamental political values and sometimes even form electoral alliances. In such cases, disagreement between voters and discussants can be assumed to be considerably less intense, and the resulting conflict (if any) to be less divisive. That disagreement in multiparty systems is thus a matter of degrees may have important implications for discussant influence that have not yet been addressed by extant research.

The second shortcoming addressed by our article concerns the mechanisms of discussant influence. While it is clear that voters often adapt their electoral preferences to those of their discussants, it is a matter of dispute why they do so. To explain the influence of social networks on voting behaviour, two competing lines of theories have been proposed. One emphasizes cognitive processes of information sharing between voters, while the other focuses on the affect-laden processes of social pressure and conformity.Footnote ⁷ The available evidence suggests that there is some truth to both the social pressure and the information-sharing perspectives. However, the relationship between them is unclear. While many studies have adopted one or the other perspective as an untested premise of their research,Footnote ⁸ attempts to identify which of them better describes the influence taking place in voters’ social networks have been rare,Footnote ⁹ and none has thus far investigated the conditions under which one or the other of the two mechanisms prevails.

With regard to both shortcomings of current research, we seek to break new ground by proposing a two-stage theory of discussant influence. For the specific context of West European multiparty systems, it states under which conditions the two approaches better capture the influence process taking place. It maintains that both mechanisms are important in these systems, although to varying degrees at different stages of layered decision-making processes that begin long before elections. We contend that in the long run, and mainly by means of continuous normative pressure, discussants restrict the party alternatives taken into account by voters to ideologically coherent ‘consideration sets’, whereas the ultimate choices of particular parties out of these pre-filtered sets at specific elections are more strongly guided by information obtained during campaigns through social communication. We test this theory using the case of the 2009 German federal election.

The following sections develop the two-stage theory of discussant influence in more detail and derive testable hypotheses. Then, we outline our research strategy, including technical information on the data used and the variables included in our models. The next section presents our results and interprets them in light of our hypotheses. In the final section we summarize our findings and discuss their implications.

DISCUSSANT INFLUENCE AT ELECTIONS

While scholars of electoral behaviour have long neglected voters’ interdependence at elections,Footnote ¹⁰ during the past two decades numerous studies have demonstrated its relevance beyond reasonable doubt. However, while it is clear that voters adjust their preferences to those of their discussants, it is less clear why they do so. Two competing explanations have been proposed to explain this phenomenon: information sharing and social pressure.Footnote ¹¹

The first perspective is inspired by Downs’ view of voters as instrumental seekers of cost-efficient information on which they can base their electoral choices.Footnote ¹² Here, discussant influence at elections is thought to come about as a consequence of voters’ search for electoral orientation, the provision of information by other persons and ensuing processes of learning on the part of voters. Social communication is believed to affect electoral decisions by providing voters with useful electoral guidance.Footnote ¹³ While this theory implies that discussant influence has essentially cognitive roots, the competing view emphasizes rather unpolitical affective processes relating to individuals’ interest in the stability of their social ties.Footnote ¹⁴ It dates back to the classic Columbia studiesFootnote ¹⁵ and conceives of voters as human beings who are inherently social and place a high value on maintaining harmonious relationships with their peers.Footnote ¹⁶ According to this perspective, normative group pressure leads to behavioural conformity on the part of individuals who adopt the preferences of others in order to avoid damaging their mutual relationship.Footnote ¹⁷

Many studies in the field have not bothered with theoretical reasoning on mechanisms by which discussant influence may come about, but have contented themselves with demonstrating its existence. Others have adopted either social pressureFootnote ¹⁸ or information sharingFootnote ¹⁹ as explicit, though untested, theoretical premises. Only recently have researchers, most notably Sinclair in her study of American elections,Footnote ²⁰ started to investigate which of the competing views better describes the influence actually taking place in social networks.

How can one decide which of the two mechanisms better explains discussant influence? According to Sinclair, this requires understanding how the attributes of both the discussants and their relationships with voters moderate their influence.Footnote ²¹ Each of the two theories implies specific expectations with regard to factors that should facilitate or impede discussant influence. For instance, if it comes about as a result of social pressure, it should vary depending on the intimacy of the relationship between the voter and discussant.Footnote ²² More powerful influence in primary relationships that are characterized by a climate of confidence, emotional appreciation and mutual regard (as well as a high density of interaction) indicates the operation of social pressure and conformity. Some studies have indeed reported discussant influence on electoral preferences to be particularly intense between spouses, followed by other family bonds and friendships. Compared to these ‘strong’ ties, secondary relationships like neighbours and co-workers appear less conducive to discussant influence.Footnote ²³

In contrast, from the point of view of the information-sharing model, ‘it is the message that has the potential to persuade, not the messenger’.Footnote ²⁴ Whereas simple cues during everyday interactions should be sufficient to trigger the social pressure mechanism, information-sharing effects can only come about as a ‘consequence of explicit social communication’.Footnote ²⁵ In line with this ‘conversion through conversation’ hypothesis,Footnote ²⁶ a discussant’s influence on a voter should vary in accordance with certain characteristics of the communication that takes place between them. The available evidence indicates that there is some truth to this perspective. For instance, studies found discussants to be more influential when engaging in frequent political talks with votersFootnote ²⁷ and when communicating their views effectively (that is, clearly enough that they are easily discernible by their addressees).Footnote ²⁸ Other research suggests that network partners are more influential when voters perceive them to be knowledgeable about political matters and trustworthy in the sense of having basically similar political views.Footnote ²⁹ These are exactly the attributes emphasized by Downs (and more recently by Lupia and McCubbins) as qualities that should make discussants particularly attractive to voters as sources of electoral advice.Footnote ³⁰

Hence, there is evidence to support both the social pressure and information-sharing models of discussant influence. How these findings can be theoretically reconciled is unclear. In the following we propose an answer to this question in the context of West European multiparty systems. In contrast to previous studies, it construes the two perspectives as complementary, rather than competitive. Conceding that influence processes can be characterized by both mechanisms, it specifies the conditions under which each can be expected to dominate.

ELECTORAL CHOICES IN MULTIPARTY SYSTEMS

Compared to candidate elections in the American two-party system, parliamentary elections in West European multiparty systems are more complex, layered contests. Whereas in the United States each vote gained by one of the candidates is one lost for its (usually) sole competitor, elections in Western Europe are characterized by competition between two cleavage-based ideological camps at one level, and between parties of the same ideological direction within camps at another level.Footnote ³¹ In the former case, the contest is ‘between enemies on different sides of the cleavage boundary’, in the latter it is ‘between friends in the same cleavage block’.Footnote ³² According to the directional theory of voting, this contrast is reflected in the way voters look at elections. For them, it makes a fundamental difference whether the parties that compete for their votes belong to the same or opposing ideological camps.Footnote ³³ Competition between parties of opposing ideological directions is antagonistic and resembles the dualism of American electoral contests. In contrast, competition between parties of the same ideological camp concerns alternatives that share important basic values and is thus rather a matter of degrees. In fact, such parties often campaign as proto-coalitions, coordinate their electioneering, and form governments together when they obtain a joint majority.

Some recent studies have sought to take these conditions into account by modelling electoral decision making in multiparty systems as a two-stage process of funnelling down alternatives to the eventual choice.Footnote ³⁴ They start from the psychological premise that in choice situations with multiple alternatives, decision processes are more complex than under conditions of simple binary choices. Their inspiration comes from psychological social judgement theoryFootnote ³⁵ and marketing research.Footnote ³⁶ The studies assume that voters come to grips with this complexity by first forming a consideration set consisting of a limited number of parties that are deemed, in principle, electable. They further assume that the party that is eventually supported at the ballot box is chosen in a second step from this pre-filtered reservoir of alternatives.

Research on stepwise decision making generally suggests that in the first stage, non-compensatory, heuristics-based decision strategies are used to narrow down the range of alternatives. This implies that only alternatives that have certain characteristics are kept under consideration, whereas those lacking them are immediately excluded.Footnote ³⁷ Regarding vote choices, this means that only parties sharing one, or at most a small number of, core characteristics have a chance to remain in the consideration set. Several studies of West European elections have shown that left-right ideology is crucially important for circumscribing individuals’ consideration sets.Footnote ³⁸

When it comes to choosing a party out of this constrained set of alternatives, a different mechanism kicks in that appears to be more strongly guided by cognitive processes. Picking a party from a restricted range of alternatives that differ little in their basic directional orientation requires more careful and detailed scrutiny of the available options. The final choice is therefore assumed to be made more carefully by means of compensatory strategies for evaluating and ordering the remaining alternatives.Footnote ³⁹ This requires more thorough and systematic thinking, which in turn presupposes a sufficient base of rather specific and detailed political information. This may concern short-term factors related to the current election like leaders’ personalities, recent political events or the issues of the day,Footnote ⁴⁰ but also parties’ electoral prospects, programmatic platforms and coalition signals to which electors need to refer when aiming to cast their votes strategically.Footnote ⁴¹

The first step of decision making that leads to the exclusion of parties from consideration sets thus appears rather simple and refers mainly to the basic heuristic of ideological direction. In contrast, choosing a single party out of a limited set of alternatives that resemble each other in their fundamental political values and vary only in details implies more differentiated cognitive processes. It also requires rather specific electoral information. In the following, we propose that discussant influence plays a role in each of these stages of electoral decision making, although in different ways – primarily via social pressure at the first stage, and more strongly by means of information sharing at the second.

A TWO-STAGE THEORY OF DISCUSSANT INFLUENCE AT ELECTIONS IN MULTIPARTY SYSTEMS

It is commonly assumed that processes of discussant influence at elections are set in motion when voters experience disagreement over electoral preferences with members of their social networks. Interacting with a discussant supporting another candidate or party stimulates voters to reconsider and ultimately adapt their preferences.Footnote ⁴² In American research, disagreement over candidate preferences is typically equated with political discord and conflict.Footnote ⁴³ Conversations with non-like-minded individuals have accordingly been characterized as ‘dangerous’, in contrast to interactions with like-minded persons, who are ‘safe’.Footnote ⁴⁴ Obviously this understanding of electoral disagreement makes good sense in the context of an inherently dualistic two-party system like that of the United States, where most existing research on discussant influence at elections has been conducted. In American elections there are usually only two contenders, and the winner takes all. The essence of electoral disagreement can thus be efficiently expressed by the simple formula that whoever is not with me is against me.

In a multiparty system, however, being exposed to a discussant who supports another party can mean different things, depending on whether this party belongs to the same or the opposing ideological camp. For such contexts we therefore propose to distinguish between two types of electoral disagreement. The first concerns the ideological direction of parties: Does a discussant support a party from the same side or from across the ideological divide? The second type of disagreement is nested within the first one: If a discussant prefers a party from the same spectrum as a voter, is it the same party or a different one?

Most studies of discussant influence in European multiparty systems have ignored these complexities and simply adopted the American approach. They determined electoral disagreement by distinguishing whether voters supported the same or any other party than their discussants.Footnote ⁴⁵ Some studies also chose to focus on party blocks instead of single parties for determining electoral agreement and disagreement. They referred to criteria such as incumbency and opposition,Footnote ⁴⁶ electoral alliancesFootnote ⁴⁷ or ideological camps.Footnote ⁴⁸ These studies recognized that in West European party systems, party blocs are a functional equivalent to single parties in the American two-party system. However, they neglected the possibility that discussant influence might also play a role in choices between parties from the same bloc. Moving beyond this research, our theory of discussant influence in multiparty systems offers an approach to deal with the fact that under such conditions electoral disagreement is a matter of degrees.

At its core is the claim that both types of electoral disagreement – between party blocs and parties from the same bloc – trigger processes of discussant influence, although in different ways and at different stages of the funnel-shaped decision process outlined above. Whether the parties supported by voters’ discussants are ideologically congruent or incongruent to their own preferences can be assumed to be crucial for the formation and maintenance of ideologically coherent consideration sets. In contrast, whether a discussant prefers the same party as a voter or another party from the same ideological camp should make a difference for the ultimate selection of a particular party out of this choice set. Importantly, we further assume that different mechanisms of discussant influence characterize the two stages. We expect social pressure to be more important for filtering down electoral alternatives to ideologically consistent consideration sets. Information sharing should be especially relevant for making up one’s mind about which party ultimately to choose out of a consideration set. We see the first stage as a long-term process based on continuous social interaction and potentially spanning entire electoral cycles. The latter we assume to be rather short term and primarily driven by voters’ political conversations during campaigns.

Basic ideological leanings to the left or right are important elements of many West Europeans’ political identityFootnote ⁴⁹ and thus carry a significant affective flavour.Footnote ⁵⁰ We therefore assume that social pressure is a major reason for the ideologically uniform structure of consideration sets. Especially within their intimate relationships, discussants will tend not to take it lightly if a voter’s electoral behaviour deviates from their basic political value orientation. As a consequence, many individuals will find themselves confronted with external social norms signalling that persons they hold in high esteem are not inclined to tolerate them straying from the ‘fold to which they belong’.Footnote ⁵¹ By threatening them with withdrawal from a relationship that voters do not want to see damaged, spouses, relatives and friends can exert substantial influence on these persons’ decision making at elections.

Research into discussant influence on partisanship supports this argument. In one of the few studies that compared discussant influence on partisanship and candidate evaluations at an election, MacKuen and BrownFootnote ⁵² found American voters’ partisanship to be strongly constrained by their friends’ party identifications. They concluded from their research that a ‘political self-definition […] represents a […] serious matter. Conversational content is not sufficient to generate change; instead, individuals respond to their friends’ fundamental loyalties. This pattern suggests that self-identification depends on imitative rather than cognitive learning, with the citizen adopting friends’ characters rather than reacting to what friends say’.Footnote ⁵³ In West European multiparty systems, cleavage-based ideological identities fulfil partly equivalent functions to partisanship in the United States.Footnote ⁵⁴ It can thus be assumed that citizens similarly strongly care about their close associates’ fidelity to parties on either the ideological left or right. By means of affect-laden normative pressure, they consequentially constrain voters to parties from the same ideological spectrum.

Yet discussants can be assumed to take a more relaxed stance with regard to which party is eventually chosen from this limited set. This is no longer about a polar confrontation between contradictory worldviews, but about details and degrees on the basis of a fundamental correspondence with regard to basic political values. Therefore, at this stage the value of discussants as providers of cost-efficient information should come to the fore. When analysing discussant influence on voters’ candidate evaluations, MacKuen and Brown found that ‘[c]urrent, transitory, information plays the greatest part in influencing attitude change’.Footnote ⁵⁵ According to their research, voters’ candidate attitudes were not responsive to their friends’ basic political loyalties, but rather to ‘hot communication – that is, the content of currently ongoing conversations, whether they reflect stable political biases or not’.Footnote ⁵⁶ This suggests that in Western Europe, discussant influence with regard to choosing a party out of a consideration set occurs by way of information sharing rather than social pressure.

In sum, we assume that discussant influence at elections takes two different forms in West European multiparty systems, corresponding to successive stages of decision making. We expect social pressure to be more important with regard to the first stage, which limits voters to parties from the same ideological spectrum, as they will conform to the behavioural norms to which they are exposed by their discussants. We further expect information sharing to be relevant mainly at the ensuing stage, when voters make up their minds about which parties to choose out of their consideration sets.

HYPOTHESES

The remainder of this article aims to provide evidence for our contention that in ideologically structured multiparty systems like those of Western Europe, both social pressure and information sharing are important mechanisms of discussant influence, although to varying degrees at different stages of electoral decision processes. To gain leverage for substantiating this claim empirically, we refer to patterns of discussant influence that are characteristic of each of the two mechanisms.Footnote ⁵⁷ If discussant influence comes about as a result of social pressure exerted by a discussant on a voter, it should vary by the intimacy of the relationship between these individuals.Footnote ⁵⁸ Since we expect social pressure to be especially important as a mechanism of discussant influence at the first stage of decision making, when discussants constrain the range of alternatives considered for possible electoral support by voters, we propose the following testable hypothesis.

Hypothesis 1: Discussant influence is stronger for primary relationships (most notably for spouses, followed by relatives and friends) than for secondary relationships. These differences between relationships are more pronounced at the first than at the second stage of decision making.

If characteristics of the social communication between a discussant and a voter moderate the influence exerted by the discussant, information sharing is its likely explanation.Footnote ⁵⁹ This may concern the frequency and clarity of such communication as well as how voters perceive their discussants with regard to their expertise and trustworthiness as political information sources. Since we assume that information sharing is especially important at the second stage of decision making, when discussants influence which parties voters choose from their ideologically coherent consideration sets, we hypothesize:

Hypothesis 2a: A discussant’s influence is higher the more often he or she discusses political matters with a voter. This effect is more pronounced at the second than at the first stage of decision making.

Hypothesis 2b: A discussant’s influence is higher the more effectively he or she communicates his or her electoral preference to a voter. This effect is more pronounced at the second than at the first stage of decision making.

Hypothesis 2c: A discussant’s influence is higher the more knowledgeable in political matters he or she is perceived to be by a voter. This effect is more pronounced at the second than at the first stage of decision making.

Hypothesis 2d: A discussant’s influence is higher the more trustworthy in political matters he or she is perceived to be by a voter. This effect is more pronounced at the second than at the first stage of decision making.

DATA AND METHODS

Data

An excellent case for testing our hypotheses is Germany. Its party system is rooted in socio-political cleavages and structured ideologically and by party size.Footnote ⁶⁰ Parties have long traditions of representing the interests of particular social groups that are reflected in distinct positions on the left-right axis of ideological identities.Footnote ⁶¹ All five parties represented in the national legislature (Bundestag) can be located on the right or left of the ideological spectrum. Each direction is represented by one larger and one or two smaller parties. The system’s gravitation centres are the Christian Democrats (CDU/CSU) and the Social Democrats (SPD), two large parties of the centre-right and centre-left, respectively. They represent opposing sides of the two major historical cleavages of German society: religion and class. All federal governments since 1949 have been led by one of these two parties. Like the CDU/CSU, the market-liberal FDP is located on the right. The Greens and the Left are both located to the left of the SPD.

The parties’ left-right placements are reflected in their election manifestos,Footnote ⁶² and voters are well aware of them.Footnote ⁶³ Ideological directions are also an important driving force behind the parties’ politics of coalition building,Footnote ⁶⁴ both with regard to proto-coalitions during campaigns and government formation. Since the 1980s, the main confrontation has been between a ‘bourgeois’ or ‘black-yellow’ coalition of the FDP and a ‘red-green’ alliance of Social Democrats and Greens.Footnote ⁶⁵ In recent decades, electoral competition in Germany has thus been strongly structured by ideological confrontations, pitting two camps – one left, one right – against each other.Footnote ⁶⁶

Our analysis is based on a pre- and post-election panel survey conducted as part of the 2009 German Longitudinal Election Study (GLES). A random sample of 6,008 German citizens aged eighteen and above was interviewed by telephone during the 2009 German federal election campaign, of which about two-thirds were re-interviewed after the election (N=4,027).Footnote ⁶⁷ The pre-election wave contained a sequence of questions developed to efficiently map respondents’ political core discussion networksFootnote ⁶⁸ in the context of a national election study. It prompted respondents’ most important political discussion partners and elicited perceptional data for up to two discussants.Footnote ⁶⁹

Our analysis applies a dyadic perspective. Its units of analysis are pairs of voters and discussants.Footnote ⁷⁰ Our active dataset is restricted to cases for which our research question is substantially meaningful – that is, dyads in which the respondent in the pre-election wave (1) stated a preference for one of the five parties in the federal parliament on the part of his or her discussant and (2) stated an intention to vote for one of these parties or was undecided. Respondents who did not engage in any political conversations (and who thus had no one to influence them) could not be included in the analysis. For the same reason, dyads in which respondents did not receive partisan communications from discussants had to be excluded.Footnote ⁷¹ Dyads in which respondents were likely or certain non-voters also had to be excluded from the models. The resulting number of dyads available for the cross-sectional analyses was 3,661 and 2,757 for the panel analyses (corresponding number of respondents: 2,581 and 1,930).Footnote ⁷²

Dependent Variables

Our analysis includes four dependent variables: two cross-sectional, based on pre-election data on vote intentions, and two with a pre- and post-election panel structure, which take into account respondents’ actual party choices. They are based on comparisons between discussants’ perceived pre-electoral party preferencesFootnote ⁷³ and respondents’ own vote intentions (cross-section) or recalled vote choices (panel). One cross-sectional and one panel variable register whether or not a voter favoured a party of the same ideological direction as the one supported by his or her discussant.Footnote ⁷⁴ The two ideological camps consist of parties situated to the right of centre (CDU/CSU and FDP) and left of centre (SPD, Greens and Left). The other two dependent variables refer only to dyads that are homogenous in terms of parties’ ideological direction and differentiate by whether or not voters supported the same party as their respective discussants.Footnote ⁷⁵

Strategy of Analysis

A survey-based study of discussant influence like ours has to deal with a number of methodological challenges. Following Fowler et al., establishing causality requires safeguards against several potential fallacies.Footnote ⁷⁶ The first is unrepresentative egocentric network data. Smaller networks like those analysed in our study run the risk of unwarranted generalization to the entire universe of voters’ politically relevant associates. Information on discussants beyond the second one was not elicited in our study, and it is unclear whether the findings are generalizable beyond the first and second discussants. However, comparative studies suggest that German voters’ discussant networks are, on average, rather small.Footnote ⁷⁷ Moreover, our data display substantial variation in all discussant attributes relevant for our study (Table 1), and the distributions largely conform to those registered on the basis of network generators allowing for more discussants.Footnote ⁷⁸ We are therefore confident that the information loss caused by truncated networks does not significantly bias our findings.

Table 1

Descriptives for Main Independent Variables

The second problem is misperception of discussant attributes. It concerns only those attributes for which, for theoretical reasons, true (rather than perceived) states are expected to be relevant for political behaviour,Footnote ⁷⁹ in particular discussants’ party preferences.Footnote ⁸⁰ Regarding discussants’ electoral preferences, American research based on snowball samples found some evidence of projection, leading to perceptions that are biased toward the respondents’ own preferences. But overall accuracy has nonetheless been found to be quite high.Footnote ⁸¹ Similar findings have been recorded by German snowball studies, with one important difference: German voters display a marked tendency to evade with ‘don’t know’ answers when in doubt about discussants’ true electoral preferences.Footnote ⁸² Dyads in which this is the case cannot be included in our analysis, which diminishes the risk that our findings are confounded by projection.

The third problem concerns the possibility of selection biasFootnote ⁸³ as a consequence of the homophily principle, which predicts that similar individuals will seek each other out as interaction partners.Footnote ⁸⁴ Either through politically motivated choices, or by accident as a by-product of selection on other attributes that are correlated with political affinities, this may lead to electoral agreement between voters and discussants. However, several recent long-term studies suggest that at least the amount of direct political selectivity reflected in dyadic similarities is rather small in comparison to influence. For most people, politics does not seem to be a major concern in discussant choice, but once connected members of social networks begin to converge in their party or candidate preferences.Footnote ⁸⁵ In addition, it must be kept in mind that the availability of like-minded associates depends on supply. Many voters will not be able to construct congenial networks, despite trying, since selectivity is constrained by the composition of the contexts out of which discussants can be chosen.Footnote ⁸⁶ Although not unequivocal, the state of research is thus encouraging with regard to our interpretation of electoral agreement as a result of influence rather than selection.

Nonetheless, our analysis needs to take precautions against the possibility of reciprocal causation in order to build a strong case for testing our theory of discussant influence. As McClurg recently pointed out, there is no single ideal solution to this problem; researchers can usually only resort to an approach ‘akin to building a circumstantial case’:Footnote ⁸⁷ to use the available data in a way that rules out alternative interpretations as far as possible, even if some residual ambiguity remains. In particular, this requires controlling for demographic attributes and basic political orientations that may lead respondents to select people with similar attributes as interaction partners.Footnote ⁸⁸ However, this strategy of course cannot be considered fully equivalent to randomization and experimental designs, since there is always the possibility of systematic confounders being excluded from the model because they are unmeasured or even unknown. As an additional safeguard to reduce the potential ambiguities of our findings, we therefore rely not only on cross-sectional but also on panel analysis with lagged dependent variables when modelling social influence. The panel approach is much less vulnerable to both reciprocal causation and projection problems, since it models change, rather than static relationships.Footnote ⁸⁹ The panel models essentially tell us which factors strengthen or weaken the likelihood that voters will converge toward the preferences previously demonstrated by their discussants. Since there was rather little overall turnover during the campaign (11 per cent for the same camp condition, 8 per cent of which was convergent; 16 per cent for the same party condition, 10 per cent of which was convergent), these models pose a particularly tough test for our hypotheses (see Table 2 for marginals). Arguably, they underestimate the true amount of electoral influence, since they register only short-term movements during the last weeks of the campaign. This concerns in particular the hypothesized first stage of the decision process, which we see as a continuous long-term process of which our panel data can provide only a quick snapshot. In contrast, the cross-sectional analyses indicate upper bounds of effect sizes, albeit with less certainty, as they are possibly more strongly affected by selection and projection bias.

Table 2

Party Preferences in Dyads

Our analysis proceeds in three steps. Using probit analysis, we first model whether voters supported parties from the same or the opposite ideological camp as their respective discussants. In the second step we model the probability of supporting the same or different parties, given the same ideological direction. Dyads in which voters and discussants favour parties from opposite ideological camps are excluded at this stage.Footnote ⁹⁰ To assess the validity of our hypotheses we need to compare these two models with regard to the strength of the estimated effects – this is the third and most important step of the analysis. All effects should be statistically relevant, but some of them in the first model, and others in the second. In limited dependent variable models this is not as straightforward as in the linear model, because the regression slopes may be affected by unobserved heterogeneity.Footnote ⁹¹ Therefore we calculate population averaged marginal effects (AME) and discrete differences following the observed-value approach as recommended by Hanmer and Kalkan.Footnote ⁹² These quantities are not affected by unobserved heterogeneityFootnote ⁹³ and offer an intuitive interpretation, indicating the average increase in the probability of Y=1 given a one-unit change in X. In addition, from generating the marginal effects via the simulation method we automatically get bounds of uncertainty not merely for these quantities but also for derived differences in effects between the two models (differences and corresponding 95 per cent confidence intervals are displayed in the last two columns of Tables 3 and 4). This allows us to test whether the impact of the various independent variables varies significantly between the models that represent the two stages of our process model.

Table 3

Moderators of Discussant Influence: Cross-Sectional Models (Probit Estimates for Dyads)

Note: The AME for dummy variables indicates the discrete change from the base level. ∆AME is the difference in AME between the camp and the party model; the 95 per cent confidence interval indicates whether this difference significantly differs from zero. *p<0.05; **p<0.01; ***p<0.001 (cluster-robust standard errors).

Table 4

Moderators of Discussant Influence: Panel Models (Probit Estimates for Dyads)

Note: The AME for dummy variables indicates the discrete change from the base level. ∆AME is the difference in AME between the camp and the party model; the 95 per cent confidence interval indicates whether this difference significantly differs from zero. * p<0.05; ** p<0.01; *** p<0.001 (cluster-robust standard errors).

Independent Variables

To test our hypothesis about the relevance of different types of relationships, we include dummy variables in the models for the following relationships: spouses (including life partners), relatives and friends as primary relationships, and neighbours as well as the residual category of ‘other’ relationships as secondary relationships. Another secondary relationship, co-workers, serves as the implied reference category.Footnote ⁹⁴ Our measure of the frequency of political conversations is a dummy variable indicating whether the respective discussant was the person to whom respondents talked most often about politics and thus was named first in the network (coded 1), or not (0).

To test the extent to which effective communication on the part of a discussant increases his or her electoral influence, we refer to response latencies with regard to respondents’ reports on discussants’ vote intentions – that is, the time it took respondents to reproduce these preferences during interviews. A swift response to this question indicates easy accessibility of the discussant’s vote intention in the respondent’s memory. This measurement is based on the assumption that the unambiguousness of the discussant’s electoral preference is reflected by the strength with which the mental concepts of a discussant and a party are associated in the long-term memory of a respondent. If a discussant expresses a party preference with great clarity, a strong mental association between person and party can be established that can be easily – and thus quickly – retrieved during an interview.Footnote ⁹⁵

Discussants’ perceived political expertise is measured by a question asking how well the respective discussant was versed with regard to political matters (0=not at all, 1=not so well, 2=well, 3=very well). In line with Downs’ and Lupia and McCubbins’ claim that voters assess discussants as politically trustworthy if they share their general political worldviews, we refer to respondents’ overall assessments of the frequency of opinion differences during political conversations (0=often, 1=sometimes, 2=rarely, 3=never) to measure discussants’ perceived political trustworthiness. Dissent during political talks signals to a voter that this discussant is not on the same political wavelength and therefore should not be listened to, whereas its absence indicates that he or she is a political soulmate whose electoral advice can be confidently followed.Footnote ⁹⁶ Table 1 provides descriptives for all independent variables, based on the dyads available for the cross-sectional and panel analyses, respectively. It shows that panel attrition has affected the distributions remarkably little.Footnote ⁹⁷

Control Variables

In addition to discussant influence, electoral similarity between voters and their discussants can also be a consequence of actively seeking out like-minded discussants. Political predispositions in the sense of stable identities deeply rooted in voters’ personalities are the most important potential motivators of political selectivity. To the degree that associates are chosen on political terms, two types of identities can be expected to be of particular importance in Germany: partisanship and ideology.Footnote ⁹⁸ To control for the potential impact of party attachment, we include in our models a dummy variable that indicates whether respondents supported a party at the ballots with which they identified (coded 1), or not (0=out-party voting, independent or undecided).Footnote ⁹⁹ To assess the role of ideological identification with regard to selectivity, a standard left-right scale, folded at the midpoint, would be ideal. Unfortunately, this instrument is not available in our dataset. As a proxy we use a thermometer scale for respondents’ feelings regarding a grand coalition as a possible outcome of the upcoming election (scaled −5 to +5); using other data it can be shown that such an alliance between the parties dominating the two ideological camps is most popular among centrist voters.Footnote ¹⁰⁰

Political interest (five-point scale from 0=‘outcome of election considered as personally not important at all’ to 4=‘very important’) is included to account for the reception and processing of campaign information from other sources than social communication. Moreover, the models contain a bloc of dummy variables indicating which party the respective discussant preferred (SPD, FDP, Greens, Left; implicit reference category: CDU/CSU). This accounts for differential opportunities to get in touch with supporters of parties in accordance with the parties’ varying sizes.Footnote ¹⁰¹ Furthermore, respondents’ demographic characteristics – age (in years), sex (1=male, 0=female) and education (1=secondary education completed, 0=lower education level) – are taken into account as control variables. Finally, the dependent variables of the cross-sectional models are added to the panel models as additional controls.Footnote ¹⁰²

RESULTS

Discussant influence is highly conditional. Under some circumstances it becomes particularly pronounced, while under others it is attenuated.Footnote ¹⁰³ From the specific patterns such conditionalities assume, we can infer the relevance of the two mechanisms of discussant influence at the various stages of the decision process.Footnote ¹⁰⁴ Our expectation is that social pressure should be more evident at the first stage, when ideologically coherent consideration sets are formed, while patterns indicating the effects of shared information should more strongly characterize the second stage, when a particular party is chosen out of this limited set of alternatives. Table 3 shows the results of the cross-sectional models, and Table 4 the corresponding estimates from the panel models. Both tables contain the same two blocs of independent variables. Types of relationships are included to indicate the role of social pressure. The frequency of political conversations and the clarity of communicated messages as well as the perceived political expertise and trustworthiness of discussants indicate the importance of information sharing.

Before looking at our findings in detail, it is important to note that our analysis confirms all premises from which we developed our hypotheses. Discussant influence at elections does indeed vary by the type of relationship, but also by the frequency of political conversations and the clarity of communicated messages, as well as the perceived political expertise and trustworthiness of discussants. Hence, like previous studies we find evidence for both mechanisms of discussant influence. However, testing our hypotheses requires moving beyond these basic observations by inspecting the differences between the first and second stages of the decision process.

Larger AME estimates for primary than for secondary relationships suggest that social pressure is stimulating voters to bring their electoral preferences in line with those of their discussants. In contrast, if information sharing is the operating mechanism of discussant influence, ‘there should be no heterogeneous effects based on the intensity of personal relationships’.Footnote ¹⁰⁵ However, both the cross-sectional and panel analyses find significant differences of that kind. In most analyses, voters’ preferences are much more likely to correspond to those of their spouses than to secondary relationships (which do not differ between themselves). Three out of four models also indicate the particularly influential role of relatives, although to a lesser degree than spouses. Friends differ from secondary relationships only in the cross-sectional model that predicts preferences for a party of the same ideological camp, and the effect is weaker again than for relatives.

Importantly, according to Table 4, the likelihood that voters will converge in their choices to the parties preferred by their discussants instead of sticking with another, ideologically adjacent party is not affected by the nature of relationships, whereas movements toward parties from the same ideological camp appear to be responsive to relationships. Coming from panel models, these observations are especially robust, and they are well in line with Hypothesis 1. But to test this expectation properly we need to compare the patterns of relationships across the two stages of the decision process. If Hypothesis 1 is correct, we should see significantly larger differences between primary and secondary relationships for the first stage than for the second stage. In other words, the difference between intimate and functional relationships with regard to voters’ likelihood to conform to their discussants should be more sizable for the distinction between voting for a party of the same and not the opposing ideological direction than for the distinction between voting for the same party and favouring another party from the same camp.

For both spouses and relatives such a pattern is indeed registered by the cross-sectional and in particular also the panel analysis. For friendships our finding is similarly clear-cut only in the cross-sectional model. While friends’ influence appears considerably stronger than that of co-workers at the first stage of the decision process, no such difference is found for the second stage. Correspondingly, the difference between the two models is even larger for friends than for spouses and relatives. However, friendship does not appear to spur short-term change toward parties of the same ideological leaning more strongly than workplace ties. Overall, Hypothesis 1 is thus clearly confirmed for two of the three types of primary relationships distinguished in our analysis, but less unambiguously so for friendship. That our short-term panel analysis supports our expectation of stronger effects in the first step of decision-making for spouses and relatives is particularly remarkable since our theory assumes this step to be rather long term in nature. Table 4 suggests that the mechanism of social pressure exerted within these relationships also works on the short run, although probably less massively than on the long run.

The other independent variables in the models concern aspects of information sharing between discussants and voters. These effects should be stronger for the second stage of the decision process, when voters make up their minds about which party to choose from those contained in their consideration sets. As Tables 3 and 4 show, the likelihood of voters adopting the preferences of their discussants indeed appears higher when politics is discussed more often, but only for voters’ decisions between the same party that their discussant supports and other parties from the same ideological camp. This suggests that information sharing and concomitant learning on the part of voters play not only a less important role, but are in fact irrelevant at the first stage of decision making, which is well in line with Hypothesis 2a. This result is clearly confirmed by the panel analysis and is thus particularly robust.

The expectation that effective communication on the part of discussants increases the likelihood of adopting their electoral preferences at the second stage of decision making is confirmed by the cross-sectional analysis. That the clarity of communication appears to be an important facilitator of electoral influence at that stage – but not at the previous one, when consideration sets are demarcated – is in accordance with Hypothesis 2b. It is another indication that in multiparty systems, discussant influence via information sharing is more characteristic of the second than the first stage of electoral decision making. However, this evidence is less robust, as it does not show up in the panel analysis.

Hypotheses 2c and 2d expect similar patterns with regard to discussants’ perceived political expertise and trustworthiness. In line with extant research, Tables 3 and 4 show that discussants who are endowed with substantial political knowledge exert considerably more influence than those with lower expertise. The degree to which discussants are deemed politically trustworthy appears to be an even more important condition for electoral influence. However, comparing models, we see that the pattern is opposite to the one expected by our hypotheses, which are thus disconfirmed. In both the cross-sectional and panel models, the effects of discussants’ political expertise and trustworthiness are stronger for the first than for the second stage of decision making. One possible reason for this unexpected outcome is that these two indicators might be less clearly related to information sharing as a mechanism of discussant influence than we assumed. To test for this possibility, we ran additional models (tables not shown) without these two indicators, which considerably strengthened the effects connected to types of relationships. While this cannot fully explain our unexpected findings, it suggests that as moderators of discussant influence, these variables are not sufficiently clearly indicative of processes of information transfer. They seem to confound the two mechanisms of discussant influence. Their operational utility for differentiating these mechanisms thus appears questionable.

In sum, when modelling German voters’ electoral decision making in two stages, one delimiting a range of electable parties that is constrained to the same ideological direction and the other determining which party from this restricted set is ultimately chosen, discussant influence appears relevant at both stages. However, different mechanisms of influence seem to characterize the two steps. Our evidence on these mechanisms is mostly, though not unequivocally, in line with our hypotheses. The expectation that social pressure is more important at the first stage is clearly confirmed, especially for spouses and relatives. Support for the proposition that information sharing is more important at the second stage is not quite as unambiguous. With regard to two of the relationships that indicate information sharing as a mechanism of influence – concerning the intensity and clarity of the discussants’ political communication – the expected patterns emerged. However, possibly at least in part due to an ambiguous quality of these two predictors as indicators for distinguishing the two mechanisms of discussant influence, our findings concerning discussants’ political expertise and trustworthiness are not in line with our expectations.

CONCLUSION

Extant research into discussant influence at elections has been dominated by studies from the United States in two ways. First, a large majority of existing analyses has focused on candidate elections in the American two-party system. More importantly, studies of elections in other countries have, with few exceptions, adopted this research as a model and conceptualized discussant influence in the same way. The specific institutional settings of these contexts have rarely been taken into account. With regard to West European democracies, it is therefore unclear how the specific conditions of cleavage-based and ideologically structured multiparty systems, which are characterized by typical patterns of opposition and affinity between parties, moderate discussant influence. Another shortcoming of the existing literature concerns the relationship between different mechanisms of discussant influence. Both cognitive and affective mechanisms appear important, but when (and under which conditions) is not clear.

We have sought to address both deficiencies by developing a two-stage theory of discussant influence in West European multiparty systems. In essence it is derived from two theoretical building blocs. The first is the idea that in multiparty systems voters decide in a stepwise fashion, first narrowing down the range of alternatives to a set of parties from one side of the ideological spectrum and then choosing a party out of this restricted set.Footnote ¹⁰⁶ This construal of electoral decision making is linked to the two dominant views on the mechanisms of discussant influence at elections: that voters comply with social pressure exerted by significant others, and that they learn from information provided by persons with whom they interact.Footnote ¹⁰⁷ Our theory of stepwise discussant influence maintains that social pressure and ensuing conformity are especially important at the first stage of the electoral decision process, when they lead voters to eliminate all parties from their consideration set that do not belong to the same ideological camp as the party supported by their discussants. It further contends that at the second stage, when a choice has to be made out of this restricted set of ideologically coherent parties, alternatives need to be weighed up more carefully and systematically so that useful information provided by discussants becomes more important as a basis for their influence at that stage. We construe the first stage as a process that is rather long term, potentially spanning entire electoral cycles and driven by the continuous interaction of voters with members of their social networks, whereas the latter is rather short term and is primarily fuelled by voters’ political conversations during campaigns.

Using the German federal election of 2009 as an example, we subjected this theory to an empirical test. To develop a test criterion for distinguishing between the two mechanisms of discussant influence, we adopted a logic proposed by SinclairFootnote ¹⁰⁸ and referred to implied patterns of moderating factors that can be considered typical for each of them. Our findings were mostly in line with the expectations derived from our two-stage theory.

Primary relationships appeared as important facilitators of discussant influence, especially at the stage when consideration sets are formed by excluding parties from the ‘wrong’ ideological camp. This suggests that social pressure is more important at the first stage of electoral decision making than at the second. In multiparty systems, social pressure is of particular importance when voters narrow the range of electable alternatives down to ones that are ideologically acceptable to discussants. Importantly, these findings emerged even in our short-term panel analysis, although our theory assumes these processes to be predominantly long term in nature. Since we applied a fine-grained lens at differences between relationships, we were additionally able to show that the various types of primary relationships are not equivalent. Spouses appeared particularly important, followed by relatives, whereas findings were somewhat ambiguous for friends. At least in the short run, social pressure among friends does not seem to spur convergence to parties from the same ideological camp. Perhaps friendship, as an archetypical chosen relationship, is more strongly affected by selection bias. In any case, simply subsuming these three types of relationships to the global category of ‘strong ties’ would neglect important differences with regard to discussant influence.Footnote ¹⁰⁹

Regarding influence via social information flows, on the other hand, strong moderating effects were expected for factors directly related to social communication. Frequent political talksFootnote ¹¹⁰ and the clarity of discussants’ political communicationsFootnote ¹¹¹ indeed appeared to facilitate discussant influence at the stage of party choice, but not with regard to the formation of consideration sets. Voters also appeared particularly open to influence from discussants deemed particularly knowledgeable and trustworthy.Footnote ¹¹² But here the observed patterns were contrary to our expectations, since they were more marked for the first than for the final stage of decision making. However, this may be due in part to ambiguities about the discriminative quality of these indicators with regard to the two mechanisms of discussant influence. Overall, our results provide strong (though not fully unequivocal) support for the assumption that information sharing is a particularly important mechanism of discussant influence when it comes to deciding which party to support at the ballots out of those contained in voters’ pre-filtered consideration sets.

Clearly, our study is not without limitations. Most notably, its evidence could not be derived from direct observation of the two processes of social influence, but from their implications. Our measures indicate qualities of relationships, interactions, and communications between respondents and discussants, but it would be desirable to also be able to register the specific content conveyed during political conversations. However, such data are difficult to come by, and large-scale surveys may be too crude for that purpose. In order to get more precise readings of what happens between citizens when they influence one another, it seems more promising to combine surveys with qualitative methods, such as focus groupsFootnote ¹¹³ or participant observation,Footnote ¹¹⁴ and with experimental evidence.Footnote ¹¹⁵ More direct evidence would also be welcome with regard to our contention that voters decide in stages with different temporal logics. Our results are suggestive, but in order to observe how voters’ decision processes evolve over time, data that combine short- and long-term panel waves would be necessary. In addition, more extensive measures for voters’ egocentric networks would be welcome, encompassing a larger number of discussants with more personal and relational attributes, ideally enriched with snowball surveys of discussants.Footnote ¹¹⁶

Arguably, the most serious limitation of our study is its restriction to one country. The scope of our theory extends to parliamentary elections in all multiparty systems in which parties are arrayed on the left-right continuum as the single-most important dimension of organizing political conflict. As soon as several electorally important parties can, with some constancy, be located on either side of the ideological spectrum (and provided left-right identities are a meaningful element of citizens’ political personalities), the sequential social influence processes outlined by our theory should be observable at elections. This clearly pertains to most West European democracies, where such systems grew out of socio-political cleavages dating back several centuries.Footnote ¹¹⁷ Whether other attributes of party systems also moderate these processes,Footnote ¹¹⁸ and whether similar processes can also be expected in multiparty systems in other parts of the world, is open to further theorizing and empirical inquiry.

Footnotes

University of Mannheim (emails: schmitt-beck@uni-mannheim.de, julia.partheymueller@mzes.uni-mannheim.de). For important inspiration and support in preparing this article we are indebted to Hans Hassell, Josephine Hörl, Oana Lup, Patricia Moy, Thomas Plischke, Thomas Poguntke, Anne Schäfer, Richard Traunmüller, Bernhard Weßels, Line Winterhoff as well as three anonymous reviewers. The data used in this article are available for free download at http://www.gesis.org/gles. Online appendices and data replication sets are available at http://dx.doi.org/doi: 10.1017/S0007123414000301

¹ Zuckerman Reference Zuckerman2005.

² E.g., Campus, Pasquino, and Vaccari Reference Campus, Pasquino and Vaccari2008; Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004; Huckfeldt and Sprague Reference Huckfeldt and Sprague1991, Reference Huckfeldt and Sprague1995; Huckfeldt et al. Reference Huckfeldt, Beck, Dalton and Levine1995; Ikeda Reference Ikeda2010; Johnston and Pattie Reference Johnston and Pattie2006, 127–43; Knoke Reference Knoke1990; Levine Reference Levine2005; Magalhães Reference Magalhães2007; Pattie and Johnston Reference Pattie and Johnston2000, Reference Pattie and Johnston2001, Reference Pattie and Johnston2002; Richardson and Beck Reference Richardson and Beck2007; Schmitt-Beck Reference Schmitt-Beck2000, Reference Schmitt-Beck2004; Sinclair Reference Sinclair2012; Zuckerman, Dasovic, and Fitzgerald Reference Zuckerman, Dasovic and Fitzgerald2007.

³ For a rare exception, which refers to the Japanese multiparty system, see Ikeda and Richey (Reference Ikeda and Richey2009).

⁴ Klofstad, Sokhey, and McClurg Reference Klofstad, Sokhey and McClurg2013.

⁵ Bartolini and Mair Reference Bartolini and Mair1990.

⁶ Rabinowitz and Macdonald Reference Rabinowitz and Macdonald1989.

⁷ Ikeda Reference Ikeda2010; Leighley Reference Leighley1990; Sinclair Reference Sinclair2012, 7–12.

⁸ E.g., Berelson, Lazarsfeld, and McPhee Reference Berelson, Lazarsfeld and McPhee1954; Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004; Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; Lazarsfeld, Berelson, and Gaudet Reference Lazarsfeld, Berelson and Gaudet1968; Zuckerman, Dasovic, and Fitzgerald Reference Zuckerman, Dasovic and Fitzgerald2007.

⁹ Sinclair Reference Sinclair2012.

¹⁰ For an intellectual history of the ‘social logic’ of voting cf. Zuckerman, Dasovic, and Fitzgerald (Reference Zuckerman, Dasovic and Fitzgerald2007, 1–31).

¹¹ Ikeda Reference Ikeda2010; Leighley Reference Leighley1990; Sinclair Reference Sinclair2012, 7–12.

¹² Downs Reference Downs1965, 207–37.

¹³ Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; Zuckerman, Dasovic, and Fitzgerald Reference Zuckerman, Dasovic and Fitzgerald2007.

¹⁴ Cf. Abrams, Iversen, and Soskice Reference Abrams, Iversen and Soskice2011.

¹⁵ Berelson, Lazarsfeld, and McPhee Reference Berelson, Lazarsfeld and McPhee1954; Lazarsfeld, Berelson, and Gaudet Reference Lazarsfeld, Berelson and Gaudet1968.

¹⁶ For a general discussion of humans’ ‘need to belong’, cf. Baumeister and Leary (Reference Baumeister and Leary1995).

¹⁷ Sinclair Reference Sinclair2012.

¹⁸ E.g., Berelson, Lazarsfeld, and McPhee Reference Berelson, Lazarsfeld and McPhee1954; Lazarsfeld, Berelson, and Gaudet Reference Lazarsfeld, Berelson and Gaudet1968.

¹⁹ E.g., Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004; Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; Zuckerman, Dasovic, and Fitzgerald Reference Zuckerman, Dasovic and Fitzgerald2007.

²⁰ Sinclair Reference Sinclair2012.

²¹ Sinclair Reference Sinclair2012, 12–13, 105–7, 151–3.

²² Berelson, Lazarsfeld, and McPhee Reference Berelson, Lazarsfeld and McPhee1954; Lazarsfeld, Berelson, and Gaudet Reference Lazarsfeld, Berelson and Gaudet1968; Lazer et al. Reference Lazer, Rubineau, Chetkovich, Katz and Neblo2010; McClosky and Dahlgren Reference McClosky and Dahlgren1959; Sinclair Reference Sinclair2012.

²³ Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; Lazer et al. Reference Lazer, Rubineau, Chetkovich, Katz and Neblo2010; Levine Reference Levine2005; Schmitt-Beck Reference Schmitt-Beck2000; Sinclair Reference Sinclair2012; Zuckerman, Dasovic, and Fitzgerald Reference Zuckerman, Dasovic and Fitzgerald2007.

²⁴ Sinclair Reference Sinclair2012, 10.

²⁵ Sinclair Reference Sinclair2012, 2.

²⁶ Johnston and Pattie Reference Johnston and Pattie2006, 127.

²⁷ Kenny Reference Kenny1994, Reference Kenny1998; Knoke Reference Knoke1990; Schmitt-Beck Reference Schmitt-Beck2000, 257–320.

²⁸ Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004, 68–97.

²⁹ Huckfeldt Reference Huckfeldt2001; Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; Johnston and Pattie Reference Johnston and Pattie2006, 136–7; Kenny Reference Kenny1994, Reference Kenny1998; Levine Reference Levine2005; Richey Reference Richey2008; Schmitt-Beck Reference Schmitt-Beck2000, 280–5.

³⁰ Downs Reference Downs1965, 230–3; Lupia and McCubbins Reference Lupia and McCubbins1998.

³¹ Van der Eijk and Franklin Reference van der Eijk and Franklin2009, 87–117.

³² Bartolini and Mair Reference Bartolini and Mair1990, 3.

³³ Rabinowitz and Macdonald Reference Rabinowitz and Macdonald1989. See also Baldassari Reference Baldassarri2013, 70; Zakharova and Warwick Reference Zakharova and Warwick2014.

³⁴ De Vries and Rosema Reference de Vries and Rosema2009; Karlsen and Aardal Reference Karlsen and Aardal2012; Oscarsson, Gilljam, and Granberg Reference Oscarsson, Gilljam and Granberg1997; Paap et al. Reference Paap, van Nierop, van Heerde, Wedel, Franses and Alsem2005; Plischke Reference Plischke2014; Plischke and Bergmann Reference Plischke and Bergmann2012; Shikano Reference Shikano2003; Steenbergen and Hangartner Reference Steenbergen and Hangartner2008; Wilson Reference Wilson2008.

³⁵ Sherif and Hovland Reference Sherif and Hovland1961.

³⁶ Roberts and Lattin Reference Roberts and Lattin1991; Shocker et al. Reference Shocker, Ben-Akiva, Boccara and Nedungadi1991.

³⁷ Gertzen Reference Gertzen1992.

³⁸ Karlsen and Aardal Reference Karlsen and Aardal2012; Oscarsson, Gilljam, and Granberg Reference Oscarsson, Gilljam and Granberg1997.

³⁹ Gertzen Reference Gertzen1992.

⁴⁰ Karlsen and Aardal Reference Karlsen and Aardal2012; Oscarsson, Gilljam, and Granberg Reference Oscarsson, Gilljam and Granberg1997; Plischke and Bergmann Reference Plischke and Bergmann2012.

⁴¹ Bytzek and Roßteutscher Reference Bytzek and Roßteutscher2011; Meffert et al. Reference Meffert, Huber, Gschwend and Pappi2011.

⁴² Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; McPhee, Ferguson, and Smith Reference McPhee, Ferguson and Smith1963.

⁴³ Klofstad, Sokhey, and McClurg Reference Klofstad, Sokhey and McClurg2013.

⁴⁴ Eveland and Hively Reference Eveland and Hively2009.

⁴⁵ E.g., Hopmann Reference Hopmann2012; Huckfeldt, Ikeda, and Pappi Reference Huckfeldt, Ikeda and Pappi2005; Schmitt-Beck Reference Schmitt-Beck2000.

⁴⁶ Magalhães Reference Magalhães2007.

⁴⁷ Campus, Pasquino, and Vaccari Reference Campus, Pasquino and Vaccari2008.

⁴⁸ Coffé and Need Reference Coffé and Need2010; Lup Reference Lup2010.

⁴⁹ Dalton, Farrell, and McAllister Reference Dalton, Farrell and McAllister2011.

⁵⁰ Baldassari Reference Baldassarri2013, 70; Sniderman, Brody, and Tetlock Reference Sniderman, Brody and Tetlock1991, 140–63.

⁵¹ Lazarsfeld, Berelson, and Gaudet Reference Lazarsfeld, Berelson and Gaudet1968, 73.

⁵² MacKuen and Brown Reference MacKuen and Brown1987.

⁵³ MacKuen and Brown Reference MacKuen and Brown1987, 484.

⁵⁴ Fleury and Lewis-Beck Reference Fleury and Lewis-Beck1993; Shively Reference Shively1972; van der Eijk and Niemöller Reference van der Eijk and Niemöller1983.

⁵⁵ MacKuen and Brown Reference MacKuen and Brown1987, 484.

⁵⁶ MacKuen and Brown Reference MacKuen and Brown1987, 472.

⁵⁷ Cf. Sinclair Reference Sinclair2012, 12–3.

⁵⁸ Sinclair Reference Sinclair2012, 12.

⁵⁹ Cf. Sinclair Reference Sinclair2012, 8–10.

⁶⁰ Niedermayer Reference Niedermayer2006.

⁶¹ Knutsen Reference Knutsen1995.

⁶² Bytzek and Roßteutscher Reference Bytzek and Roßteutscher2011.

⁶³ Dalton, Farrell, and McAllister Reference Dalton, Farrell and McAllister2011, 134; Neundorf Reference Neundorf2012; Roßteutscher and Scherer Reference Roßteutscher and Scherer2011.

⁶⁴ E.g., Martin and Stevenson Reference Martin and Stevenson2001.

⁶⁵ The CDU/CSU and FDP governed the country from 1983–98 and have again since 2009. The SPD and Greens were in power from 1998–2005. The 2005–09 electoral period is partly an exception. After a polarized campaign that once again sharply pitted the traditional alliances against each other at the 2005 federal election, none of them reached a majority of seats. Therefore, a grand coalition between the two large parties was the only feasible option. However, it was always clear that neither of the involved parties sought a continuation of this collaboration.

⁶⁶ The Left has thus far only participated in SPD-led coalitions at the state level, but has not been deemed acceptable as a coalition partner at the federal level. This is a peculiarity of German politics in the aftermath of German unification, since the Left is a successor organization of the former East German socialist state party (SED) and pursues rather fundamentalist policy stances.

⁶⁷ Rattinger, Hans, Sigrid Roßteutscher, Rüdiger Schmitt-Beck and Bernhard Weßels, GLES, Component 2: Rolling Cross-Section Campaign Study with Post-election Panel Wave (ZA5303, version Pre1.6). Field period: 29 July to 26 September 2009. The data are available for free download at http://www.gesis.org/gles.

⁶⁸ Marsden Reference Marsden1987.

⁶⁹ The network generator was preceded by a question on the general frequency of discussions ‘with other persons, such as family members, friends or acquaintances, about the parties and the federal election’ that filtered off all non-discussants. Those discussing politics were asked: ‘If you think of the person to whom you talked most often during the past week about the parties or the federal election, what is your relationship to that person – is it your spouse or partner, a relative, a friend, a co-worker or a neighbor?’ Respondents were then asked to assess this discussant’s political expertise, degree of opinion congruence in political conversations and perceived vote intention, for which response latencies were registered in addition to verbal responses. Subsequently, respondents were asked if there were other persons with whom they also discussed ‘the parties or the federal election’. If so, for that person of the remaining group with whom they discussed politics most often, the same question sequence was asked.

⁷⁰ Cf. Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004; Huckfeldt and Sprague Reference Huckfeldt and Sprague1995.

⁷¹ This concerns discussants perceived as non-voters as well as discussants of whose party preference respondents were unaware (‘don’t know’ answers) or who were perceived as being undecided. Research on the turnout effects of discussants found respondents with such discussants to be statistically indistinguishable from voters without any discussants (Partheymüller and Schmitt-Beck Reference Partheymüller and Schmitt-Beck2012). Still, the possibility cannot be ruled out that discussants who are perceived as non-voters might send potentially influential partisan messages to voters. However, discussants’ perceived abstention was elicited by the party preference question (cf. footnote 73), and our data do not contain any other measures of possible partisan leanings. We therefore cannot investigate the possibility that non-voting discussants might be influential with regard to party choice.

⁷² The various conditions resulting in the exclusion of cases from the analyses concerned varying shares of respondents and were, in effect, partly overlapping. The most important condition was abstention from political conversations, which was the case for 34 per cent of all respondents. Six per cent indicated that they were non-voters; 2 per cent aimed to vote for one of the small non-parliamentary parties; 15 per cent were not aware of a party preference on the part of any of their discussants or perceived them to be non-voters; and 1 per cent had only discussants who supported one of the small non-parliamentary parties. Additional cases had to be dropped due to item non-response, mostly resulting from refusals to answer questions on respondents’ own or discussants’ vote intentions or choices.

⁷³ The question read: ‘For which party do you think this person will presumably vote at the federal election on 27 September, or do you think that he or she will stay home from the polls?’

⁷⁴ Coded as: 1 = respondent prefers the same party or a party from the same ideological direction as the party favoured by the discussant; 0 = respondent prefers a party from the opposite ideological direction as the party favoured by the discussant (pre-election cross-sectional analysis also: discussant favours a party, respondent is undecided).

⁷⁵ Coded as: 1 = respondent prefers same party as discussant, 0 = respondent favours other party than discussant, but from same ideological direction (pre-election cross-sectional analysis also: discussant favours a party, respondent is undecided).

⁷⁶ Fowler et al. Reference Fowler, Heaney, Nickerson, Padgett and Sinclair2011, 440–5.

⁷⁷ Cf. Schmitt-Beck Reference Schmitt-Beck2000, 168.

⁷⁸ Cf. Schmitt-Beck Reference Schmitt-Beck2000, 163–71.

⁷⁹ Fowler et al. Reference Fowler, Heaney, Nickerson, Padgett and Sinclair2011, 444.

⁸⁰ Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004, 68–97.

⁸¹ Huckfeldt and Sprague Reference Huckfeldt and Sprague1995, 124–45; Huckfeldt, Sprague, and Levine Reference Huckfeldt, Sprague and Levine2000; Huckfeldt et al. Reference Huckfeldt, Beck, Dalton, Levine and Morgan1998.

⁸² Koßmann Reference Koßmann1996; Pappi and Wolf Reference Pappi and Wolf1984; cf. also Schmitt-Beck Reference Schmitt-Beck2000, 217.

⁸³ Shalizi and Thomas Reference Shalizi and Thomas2011.

⁸⁴ McPherson, Smith-Lovin, and Cook Reference McPherson, Smith-Lovin and Cook2001. In small networks dominated by more intimate ties like those in our analysis, this problem may be of particular concern.

⁸⁵ A year-long panel study that started its fieldwork before networks were even created demonstrated such a pattern for a sample of American college students (Lazer et al. Reference Lazer, Rubineau, Chetkovich, Katz and Neblo2010). Klofstad (Reference Klofstad2011) reported similar findings with regard to political participation. Bello and Rolfe (Reference Bello and Rolfe2014) analysed agreement in party preferences using panel data collected over nine months prior to the 2010 British general elections and found discussant influence to surpass selection considerably. Research on assortative mating has come up with mixed findings on selection and influence among spouses (Alford et al. Reference Alford, Hatemi, Hibbing, Martin and Eaves2011; Coffé and Need Reference Coffé and Need2010; Klofstad, McDermott, and Hatemi Reference Klofstad, McDermott and Hatemi2012, Reference Klofstad, McDermott and Hatemi2013). But strong long-term evidence in favour of influence instead of selection explanations of spousal similarity was provided by analyses of biennial panel data from the British and German household panel studies, which trace how husbands’ and wives’ party attachments converged over periods of ten and sixteen years, respectively (Zuckerman, Dasovic, and Fitzgerald Reference Zuckerman, Dasovic and Fitzgerald2007, 71–90).

⁸⁶ Huckfeldt and Sprague Reference Huckfeldt and Sprague1995.

⁸⁷ McClurg Reference McClurg2011, 358–9.

⁸⁸ Cf. Fowler et al. Reference Fowler, Heaney, Nickerson, Padgett and Sinclair2011, 443; Sinclair Reference Sinclair2012, 80.

⁸⁹ Cf. Fowler et al. Reference Fowler, Heaney, Nickerson, Padgett and Sinclair2011, 444–5; Sinclair Reference Sinclair2012, 80.

⁹⁰ At first sight, a Heckman selection model may seem to suit such an analysis. However, our data do not allow for the inclusion of an appropriate instrument variable to meet the exclusion restriction for a Heckman selection model. The lagged dependent variables of the two panel models are not identical, which also precludes such a modelling strategy. Moreover, it must be borne in mind that the bias introduced by an incorrectly specified selection model may be severe, and in fact larger than the bias of a non-selection model, even when selection is present (Brandt and Schneider Reference Brandt and Schneider2007; Little Reference Little1985; Little and Rubin Reference Little and Rubin1987).

⁹¹ In limited dependent variable models, assumptions are imposed on the distribution of the error in order to identify the mean and variance of the latent variable. As a consequence, the size of the β-coefficients depends on the actual size of the error variance, i.e., the amount of unobserved heterogeneity, and cannot be compared across different models or groups (Mood Reference Mood2010).

⁹² To calculate the AME, each of the other independent variables is held at its observed value (Hanmer and Kalkan Reference Hanmer and Kalkan2013). The probability that Y = 1 is calculated for each respondent at the observed values reflecting his or her individual characteristics, and then averaged across all respondents.

⁹³ For a comparison and assessment of various measures using Monte Carlo simulation methods, see Mood (Reference Mood2010).

⁹⁴ Co-workers are the most frequent type of secondary relationships in our sample (cf. Table 1). More importantly, both an American study by Mutz and Mondak (Reference Mutz and Mondak2006) and comparative findings by Schmitt-Beck (Reference Schmitt-Beck2000, 231) suggest that the workplace is a context with a particularly high likelihood of encountering political disagreement. Hence, co-workers are optimal as a contrast to the other types of relationships.

⁹⁵ Cf. Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004, 76–85. Response times were registered in milliseconds by the interviewers. To control for the possibility of erroneous time measures, interviewers were asked to assess whether their measurements were valid. We also only consider such indications to be valid if they are within a range of two standard deviations above or below the mean response time for the analysed question. Furthermore, the (validated) indications were standardized with reference to the individual baseline speed of each respondent, defined as a mental baseline speed of an individual that is independent of the specific content of a question. This was achieved by using a bivariate regression of the target reaction time on the baseline reaction time, which was determined as the overall time needed for the entire interview. The residuals of this estimation are used as response latency measures in the analyses (cf. Mayerl and Urban Reference Mayerl and Urban2008).

⁹⁶ Downs Reference Downs1965, 230–3; Lupia and McCubbins Reference Lupia and McCubbins1998. See also Sinclair Reference Sinclair2012, 7.

⁹⁷ Unsurprisingly, panel respondents were somewhat more educated and more interested in the election outcome than panel dropouts (panel respondents: 39 per cent completed secondary education, 37 per cent very strong interest; panel dropouts: 35 per cent completed secondary education, 26 per cent very strong interest). They also included a higher share of persons identifying with a party and voting in line with this attachment (50 per cent v. 39 per cent). We checked whether the findings reported below were robust to panel attrition by re-running the cross-sectional models only for respondents who took part in the post-election wave of the survey. The results of this sub-group analysis were almost identical to those based on all first-wave respondents (see Section A1 of the online appendix).

⁹⁸ Like in other West European democracies, partisanship and ideology in Germany developed as mental derivatives of the traditional anchoring of party competition in socio-political cleavages. Although initially rooted in the social categories from which these cleavages emerged (and which they still mediate), they have over time become much more powerful than sociodemographic attributes as predictors of electoral behaviour (cf., e.g., Thomassen Reference Thomassen2005; Weßels Reference Weßels2000). When these two (interrelated) forms of political identity are taken into account, it is difficult to conceive of additional factors that could plausibly be expected to give rise to political selectivity. As both predispositions are strongly related to electoral behaviour, including them as controls imposes a particularly tough test on any hypothesis stating other backgrounds of vote choice.

⁹⁹ The standard German instrument used for measuring partisanship is documented and discussed in Weisberg (Reference Weisberg1993, 724–6). Comparative research has shown that partisanship in Germany is similarly stable to that in the United States and other countries (Schickler and Green Reference Schickler and Green1997).

¹⁰⁰ Using the GLES face-to-face pre-election study (Hans Rattinger, Sigrid Roßteutscher, Rüdiger Schmitt-Beck and Bernhard Weßels: German Longitudinal Election Study, Component 1: Pre-election cross-section (ZA5300, version 5.0.0). Field period: 10 August to 26 September, 2009), we validated this assumption by comparing the left-right positions of respondents who ranked the red-green coalition, black-yellow coalition or grand coalition best. According to this analysis, those favouring the two within-camp coalition options are either more to the right (CDU/CSU-FDP) or left (SPD-Greens) of those favouring a block-straddling coalition of CDU/CSU and SPD (means 6.7, 4.5 and 5.6 on an 11-point left-right scale).

¹⁰¹ Huckfeldt, Ikeda, and Pappi Reference Huckfeldt, Ikeda and Pappi2005.

¹⁰² Formally, the models estimated in the first and second steps of our analysis can be specified as: Cross-sectional models:

The dependent variables in the cross-sectional models, $Y_{{ij}}^{{Camp,t=0}} $ and $Y_{{ij}}^{{Party,t=0}} $ , depict the two forms of agreement within a dyad of respondent i and discussant j at time t=0 (pre-election interview). Accordingly, $Y_{{ij}}^{{Camp,t=1}} $ and $Y_{{ij}}^{{Party,t=1}} $ are the dependent variables in the panel models and represent agreement within a dyad of respondent i and discussant j at time t=1 (post-election interview). Φ is the cumulative density function of the standard normal distribution, the link function of the probit model. The matrices D _Pijh and D _Iijk contain the independent variables that indicate social pressure (h=5) and information sharing (k=4) within the dyad of respondent i and discussant j. The matrices C _Ril and C _Dijm contain the control variables. The former includes l=6 individual characteristics of respondent i, and the latter captures m=1 features of the jth discussant of respondent i (the party supported by the discussant). The panel models include the lagged dependent variables, $Y_{{ij}}^{{Camp,t=0}} $ or $Y_{{ij}}^{{Party,t=0}} $ , as additional control variables. The vectors β _h,γ _k, δ _l, ζ _m and η comprise the corresponding coefficients and α is the constant. AMEs derived from these models are compared in the third and essential step of the analysis. For both cross-sectional and panel analyses, the AMEs pertaining to D _Pijh are expected to be larger in Model 1 than in Model 2, whereas the reverse is expected for D _Iijk.

¹⁰³ Pattie and Johnston Reference Pattie and Johnston2002.

¹⁰⁴ Sinclair Reference Sinclair2012, 12–13, 105–7, 151–3.

¹⁰⁵ Sinclair Reference Sinclair2012, 10.

¹⁰⁶ Oscarsson, Gilljam, and Granberg 1997; Pieters and Verplanken Reference Pieters and Verplanken1995; Plischke and Bergmann Reference Plischke and Bergmann2012; Steenbergen and Hangartner Reference Steenbergen and Hangartner2008; Wilson Reference Wilson2008.

¹⁰⁷ Ikeda 2010; Leighley Reference Leighley1990; Sinclair Reference Sinclair2012.

¹⁰⁸ Sinclair Reference Sinclair2012, 12–13.

¹⁰⁹ Cf. Eveland, Morey, and Hutchens Reference Eveland, Morey and Hutchens2011.

¹¹⁰ Knoke 1990; Kenny Reference Kenny1994, Reference Kenny1998.

¹¹¹ Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004, 68–97.

¹¹² Downs 1965, 230–3; Huckfeldt Reference Huckfeldt2001; Huckfeldt and Sprague Reference Huckfeldt and Sprague1995; Levine Reference Levine2005; Lupia and McCubbins Reference Lupia and McCubbins1998.

¹¹³ E.g., Klofstad Reference Klofstad2011.

¹¹⁴ E.g., Walsh Reference Walsh2004.

¹¹⁵ E.g., Parker, Parker, and McCann Reference Parker, Parker and McCann2008.

¹¹⁶ Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004; Huckfeldt and Sprague Reference Huckfeldt and Sprague1995.

¹¹⁷ Cf., e.g., Bartolini and Mair Reference Bartolini and Mair1990; Dalton Reference Dalton2008; Klingemann Reference Klingemann2005; Knutsen Reference Knutsen1998.

¹¹⁸ For instance, party families within the same ideological camps, varying party sizes, experiences of electoral proto-coalitions and coalition governments, or the ideological polarization of party systems might additionally structure the ‘social logic’ of voters’ choices within multiparty systems.

References

Abrams, Samuel, Iversen, Torben, and Soskice, David. 2011. Informal Social Networks and Rational Voting. British Journal of Political Science 41:229–257.Google Scholar

Alford, John R., Hatemi, Peter K., Hibbing, John R., Martin, Nicholas G., and Eaves, Lindon J.. 2011. The Politics of Mate Choice. Journal of Politics 73:362–379.Google Scholar

Baldassarri, Delia. 2013. The Simple Art of Voting. The Cognitive Shortcuts of Italian Voters. Oxford: Oxford University Press.Google Scholar

Bartolini, Stefano, and Mair, Peter. 1990. Identity, Competition and Electoral Availability: The Stabilisation of European Electorates, 1885–1985. Cambridge: Cambridge University Press.Google Scholar

Baumeister, Roy F., and Leary, Mark R.. 1995. The Need to Belong: Desire for Interpersonal Attachments as a Fundamental Human Motivation. Psychological Bulletin 117 (3):497–529.CrossRef Google Scholar PubMed

Bello, Jason, and Rolfe, Meredith. 2014. Is Influence Mightier Than Selection? Forging Agreement in Political Discussion Networks During a Campaign. Social Networks 36:134–146.Google Scholar

Berelson, Bernard R., Lazarsfeld, Paul F., and McPhee, William N.. 1954. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago, IL: University of Chicago Press.Google Scholar

Brandt, Patrick T., and Schneider, Christiana J.. 2007. So the Reviewer Told You to Use a Selection Model? Selection Models and the Study of International Relations. Unpublished manuscript. Available from http://polisci2.ucsd.edu/cjschneider/working_papers/pdf/Selection-W041.pdf, accessed 25 March 2013.Google Scholar

Bytzek, Evelin, and Roßteutscher, Sigrid. 2011. Holpriger Start einer Wunschehe? Die Regierungsbildung der schwarz-gelben Koalition (Bumpy Start of a Desired Marriage? Government Formation of the Black-yellow Coalition). In Zwischen Langeweile und Extremen: Die Bundestagswahl 2009 (Between Boredom and Extremes: The 2009 German Federal Election), edited by Hans Rattinger, Sigrid Roßteutscher, Rüdiger Schmitt-Beck and Bernhard Weßels, 265–80. Baden-Baden: Nomos.Google Scholar

Campus, Donatella, Pasquino, Gianfranco, and Vaccari, Cristian. 2008. Social Networks, Political Discussion, and Voting in Italy: A Study of the 2006 Election. Political Communication 25:423–444.CrossRef Google Scholar

Coffé, Hilde, and Need, Ariana. 2010. Similarity in Husbands’ and Wives’ Party Family Preference in The Netherlands. Electoral Studies 29:259–268.Google Scholar

Dalton, Russell J. 2008. The Quantity and Quality of Party Systems. Comparative Political Studies 41 (7):899–920.CrossRef Google Scholar

Dalton, Russell J., Farrell, David M., and McAllister, Ian. 2011. Political Parties and Democratic Linkage: How Parties Organize Democracy. Oxford: Oxford University Press.Google Scholar

de Vries, Catherine E., and Rosema, Martin. 2009. Taking Voters’ Consideration Set into Consideration: Modelling Electoral Choice in Two Stages. Paper presented at the 5th ECPR General Conference, 10–12 September, Potsdam, Germany.Google Scholar

Downs, Anthony. 1965. An Economic Theory of Democracy. Boston, MA: Addison Wesley.Google Scholar

Eveland, William P., Morey, Alyssa C., and Hutchens, Myiah J.. 2011. Beyond Deliberation: New Directions for the Study of Informal Political Conversation from a Communication Perspective. Journal of Communication 61 (6):1082–1103.Google Scholar

Eveland, William P. Jr., and Hively, Myiah H.. 2009. Political Discussion Frequency, Network Size, and ‘Heterogeneity’ of Discussion as Predictors of Political Knowledge and Participation. Journal of Communication 59 (2):205–224.Google Scholar

Fleury, Christopher J., and Lewis-Beck, Michael S.. 1993. Anchoring the French Voter: Ideology Versus Party. Journal of Politics 55 (4):1100–1109.Google Scholar

Fowler, James H., Heaney, Michael T., Nickerson, David W., Padgett, John F., and Sinclair, Betsy. 2011. Causality in Political Networks. American Politics Research 39:437–480.Google Scholar

Gertzen, Heiner. 1992. Component Processes of Phased Decision Strategies. Acta Psychologica 80 (1–3):229–246.Google Scholar

Hanmer, Michael J., and Kalkan, Kerem O.. 2013. Behind the Curve: Clarifying the Best Approach to Calculating Predicted Probabilities and Marginal Effects from Limited Dependent Variable Models. American Journal of Political Science 57 (1):263–277.Google Scholar

Hopmann, David N. 2012. The Consequences of Political Disagreement in Interpersonal Communication: New Insights from a Comparative Perspective. European Journal of Political Research 51 (2):265–287.Google Scholar

Huckfeldt, Robert. 2001. The Social Communication of Political Expertise. American Journal of Political Science 45 (2):425–438.Google Scholar

Huckfeldt, Robert, and Sprague, John. 1991. Discussant Effects on Vote Choice: Intimacy, Structure, and Interdependence. Journal of Politics 53 (1):122–158.CrossRef Google Scholar

Huckfeldt, Robert, and Sprague, John. 1995. Citizens, Politics, and Social Communication. Information and Influence in an Election Campaign. Cambridge and New York: Cambridge University Press.Google Scholar

Huckfeldt, Robert, Beck, Paul A., Dalton, Russell J., and Levine, Jeffrey. 1995. Political Environments, Cohesive Social Groups, and the Communication of Public Opinion. American Journal of Political Science 39 (4):1025–1054.CrossRef Google Scholar

Huckfeldt, Robert, Beck, Paul A., Dalton, Russell J., Levine, Jeffrey, and Morgan, William. 1998. Ambiguity, Distorted Messages, and Nested Environmental Effects on Political Communication. Journal of Politics 60:996–1030.Google Scholar

Huckfeldt, Robert, Ikeda, Ken’ichi, and Pappi, Franz U.. 2005. Patterns of Disagreement in Democratic Politics: Comparing Germany, Japan, and the United States. American Journal of Political Science 49 (3):497–514.Google Scholar

Huckfeldt, Robert, Johnson, Paul, and Sprague, John. 2004. Political Disagreement. The Survival of Diverse Opinions within Communication Networks. Cambridge: Cambridge University Press.Google Scholar

Huckfeldt, Robert, Sprague, John, and Levine, Jeffrey. 2000. The Dynamics of Collective Deliberation in the 1996 Election: Campaign Effects on Accessibility, Certainty, and Accuracy. American Political Science Review 94:641–651.Google Scholar

Ikeda, Ken’ichi. 2010. Social Networks, Voting and Campaign Participation in Japan. The Interpersonal Political Environment and the Autonomous Dimension of Social Networks. In Political Discussion in Modern Democracies. A Comparative Perspective, edited by Michael R. Wolf, Laura, Morales and Ken’ichi Ikeda, 162–182. London and New York: Routledge.Google Scholar

Ikeda, Ken’ichi, and Richey, Sean. 2009. The Impact of Diversity in Informal Social Networks on Tolerance in Japan. British Journal of Political Science 39 (3):655–668.Google Scholar

Johnston, Ron, and Pattie, Charles. 2006. Putting Voters in Their Place. Geography and Elections in Great Britain. Oxford: Oxford University Press.Google Scholar

Karlsen, Rune, and Aardal, Bernt. 2012. Political Values, Party Set, and Issue Ownership: How Stable and Dynamic Factors Affect Vote Choice in Multiparty Systems. Paper presented at the 2nd EPSA Annual Conference, Berlin, Germany, 21–23 June.Google Scholar

Kenny, Christopher B. 1994. The Microenvironment of Attitude Change. Journal of Politics 56:715–728.Google Scholar

Kenny, Christopher B.. 1998. The Behavioral Consequences of Political Discussion: Another Look at Discussant Effects on Vote Choice. Journal of Politics 60 (1):231–244.Google Scholar

Klingemann, Hans-Dieter. 2005. Political Parties and Party Systems. In The European Voter. A Comparative Study of Modern Democracies , edited by Jacques Thomassen, 22–63. Oxford: Oxford University Press.Google Scholar

Klofstad, Casey A. 2011. Civic Talk. Peers Politics, and the Future of Democracy. Philadelphia, PA: Temple University Press.Google Scholar

Klofstad, Casey A., Sokhey, Anand, and McClurg, Scott D.. 2013. Disagreeing About Disagreement: How Conflict in Social Networks Affects Political Behavior. American Journal of Political Science 57 (1):120–134.Google Scholar

Klofstad, Casey A., McDermott, Rose, and Hatemi, Peter K.. 2012. Do Bedroom Eyes Wear Political Glasses? The Role Politics in Human Mate Attraction. Evolution and Human Behavior 33:100–108.Google Scholar

Klofstad, Casey A., McDermott, Rose, and Hatemi, Peter K.. 2013. The Dating Preferences of Liberals and Conservatives. Political Behavior 35:519–538.Google Scholar

Knoke, David. 1990. Networks of Political Action: Toward Theory Construction. Social Forces 68 (4):1041–1063.Google Scholar

Knutsen, Oddbjørn. 1995. Value Orientations, Political Conflicts and Left-Right Identification: A Comparative Study. European Journal of Political Research 28 (1):63–93.Google Scholar

Knutsen, Oddbjørn. 1998. Expert Judgements of the Left-Right Location of Political Parties: A Comparative Longitudinal Study. West European Politics 21 (2):63–94.Google Scholar

Koßmann, Ingo. 1996. Meinungsbildungsprozesse in egozentrierten Netzwerken (Opinion-formation Processes in Egocentric Networks). Frankfurt: Lang.Google Scholar

Lazarsfeld, Paul F., Berelson, Bernard, and Gaudet, Hazel. 1968. The People’s Choice. How the Voter Makes up His Mind in a Presidential Campaign. 3rd edition. New York and London: Columbia University Press.Google Scholar

Lazer, David, Rubineau, Brian, Chetkovich, Carol, Katz, Nancy, and Neblo, Michael. 2010. The Coevolution of Networks and Political Attitudes. Political Communication 27:248–274.Google Scholar

Leighley, Jan E. 1990. Social Interaction and Contextual Influence on Political Participation. American Politics Quarterly 18:459–475.Google Scholar

Levine, Jeffrey. 2005. Choosing Alone? The Social Network Basis of Modern Political Choice. In The Social Logic of Politics. Personal Networks as Contexts for Political Behavior, edited by Alan Zuckerman, 132–151. Philadelphia, PA: Temple University Press.Google Scholar

Little, Roderick J. A. 1985. A Note About Models for Selectivity Bias. Econometrica 53 (6):1469–1474.Google Scholar

Little, Roderick J. A., and Rubin, Donald B.. 1987. Statistical Analysis with Missing Data. New York: Wiley.Google Scholar

Lup, Oana. 2010. The Role of Political Discussion in Developing Democracies. Evidence from Hungary. In Political Discussion in Modern Democracies. A Comparative Perspective, edited by Michael R. Wolf, Laura Morales and Ken’ichi Ikeda, 183–200. London and New York: Routledge.Google Scholar

Lupia, Arthur, and McCubbins, Mathew D.. 1998. The Democratic Dilemma. Can Citizens Learn What They Need to Know? Cambridge and New York: Cambridge University Press.Google Scholar

MacKuen, Michael, and Brown, Courtney. 1987. Political Context and Attitude Change. American Political Science Review 81 (2):471–490.Google Scholar

Magalhães, Pedro C. 2007. Voting and Intermediation: Informational Biases and Electoral Choices in Comparative Perspective. In Democracy, Intermediation, and Voting on Four Continents, edited by Richard Gunther, Jose R. Montero and Hans-Jürgen Puhle, 208–254. Oxford: Oxford University Press.Google Scholar

Marsden, Peter V. 1987. Core Discussion Networks of Americans. American Sociological Review 52:122–131.CrossRef Google Scholar

Martin, Lanny W., and Stevenson, Randolph T.. 2001. Government Formation in Parliamentary Democracies. American Journal of Political Science 45 (1):33–50.Google Scholar

Mayerl, Jochen, and Urban, Dieter. 2008. Antwortreaktionszeiten in Survey-Analysen: Messung, Auswertung und Anwendungen (Response Latencies in Survey Research: Measurement, Analysis, Applications). Wiesbaden: VS Verlag für Sozialwissenschaften.Google Scholar

McClosky, Herbert, and Dahlgren, Harold E.. 1959. Primary Group Influence on Party Loyalty. American Political Science Review 53:757–776.CrossRef Google Scholar

McClurg, Scott. 2011. Porous Networks and Overlapping Contexts. Methodological Challenges in the Study of Social Communication and Political Behavior. In The Sourcebook for Political Communication Reserach. Methods, Measures, and Analytical Techniques, edited by Erik P. Bucy and Lance R. Holbert, 346–364. New York and London: Routledge.Google Scholar

McPhee, William N., Ferguson, Mark, and Smith, Robert B.. 1963. A Theory of Informal Social Influence. Glencoe and London: Free Press/Collier Macmillan.Google Scholar

McPherson, Miller, Smith-Lovin, Lynn, and Cook, James M.. 2001. Birds of a Feather: Homophily in Social Networks. Annual Review of Sociology 27:415–444.Google Scholar

Meffert, Michael F., Huber, Sascha, Gschwend, Thomas, and Pappi, Franz U.. 2011. More Than Wishful Thinking: Causes and Consequences of Voters’ Electoral Expectations About Parties and Coalitions. Electoral Studies 30 (4):804–815.Google Scholar

Mood, Carina. 2010. Logistic Regression: Why We Cannot Do What We Think We Can Do, and What We Can Do About It. European Sociological Review 26 (1):67–82.Google Scholar

Mutz, Diana, and Mondak, Jeffery. 2006. The Workplace as a Context for Cross-Cutting Political Discourse. Journal of Politics 68:140–155.Google Scholar

Neundorf, Anja. 2012. Die Links-Rechts-Dimension auf dem Prüfstand: Ideologische Einstellungen und Wahlverhalten im vereinten Deutschland 1990 bis 2008 (The Left-right Dimension on the Test Bench: Ideological Attitudes and Voting Behaviour in Unified Germany 1990 to 2008). In Wählen in Deutschland (Voting in Germany). PVS special issue 45, edited by Rüdiger Schmitt-Beck, 227–251. Baden-Baden: Nomos.Google Scholar

Niedermayer, Oskar. 2006. Das Parteiensystem Deutschlands (The Party System of Germany). In Die Parteiensysteme Westeuropas (The Party Systems of Western Europe), edited by Oskar Niedermayer, Richard Stöss and Melanie Haas, 109–133. Wiesbaden: VS Verlag für Sozialwissenschaften.Google Scholar

Oscarsson, Henrik, Gilljam, Mikael, and Granberg, Donald. 1997. The Concept of Party Set. Paper presented at the Annual Meeting of the Swedish Political Science Association, Uppsala, Sweden, 5–7 October.Google Scholar

Paap, Richard, van Nierop, Erjen, van Heerde, Harald J., Wedel, Michel, Franses, Philip H., and Alsem, Karel J.. 2005. Consideration Sets, Intentions and the Inclusion of Don’t Know in a Two-Stage Model for Voter Choice. International Journal of Forecasting 21 (1):53–71.Google Scholar

Pappi, Franz U., and Wolf, Gunter. 1984. Wahrnehmung und Realität sozialer Netzwerke. Zuverlässigkeit und Gültigkeit der Angaben über beste Freunde im Interview (Perception and Reality of Social Networks. Reliability and Validity of Statements about Best Friends in Survey Interviews). In Soziale Realität im Interview (Social Reality in Survey Interviews), edited by Heiner Meulemann and Karl-Heinz Reuband, 281–300. Frankfurt and New York: Campus.Google Scholar

Parker, Suzanne L., Parker, Glenn R., and McCann, James A.. 2008. Opinion Taking within Friendship Networks. American Journal of Political Science 52:412–420.Google Scholar

Partheymüller, Julia, and Schmitt-Beck, Rüdiger. 2012. A ‘Social Logic’ of Demobilization: The Influence of Political Discussants on Electoral Participation at the 2009 German Federal Election. Journal of Elections, Public Opinion & Parties 22 (4):457–478.Google Scholar

Pattie, Charles, and Johnston, Ron. 2000. People Who Talk Together Vote Together: An Exploration of Contextual Effects in Great Britain. Annals of the Association of American Geographers 90:41–66.Google Scholar

Pattie, Charles, and Johnston, Ron. 2001. Talk as a Political Context: Conversation and Electoral Change in British Elections: 1992–1997. Electoral Studies 20:17–40.Google Scholar

Pattie, Charles, and Johnston, Ron. 2002. Political Talk and Voting: Does it Matter to Whom One Talks? Environment and Planning 34:1113–1135.Google Scholar

Pieters, Rik G. M., and Verplanken, Bas. 1995. Intention-Behaviour Consistency: Effects of Consideration Set Size, Involvement and Need for Cognition. European Journal of Social Psychology 25 (5):531–543.Google Scholar

Plischke, Thomas. 2014. Wann Wähler entscheiden. Abläufe von Entscheidungsprozessen und der Zeitpunkt der Wahlentscheidung (When Voters Decide. Processes of Decision-making and the Time of Voting Decisions). Baden-Baden: Nomos.Google Scholar

Plischke, Thomas, and Bergmann, Michael. 2012. Entscheidungsprozesse von Spätentscheidern bei der Bundestagswahl 2009 (Late-deciders’ Processes of Decision-making at the 2009 German Federal Election). In Wählen in Deutschland (Voting in Germany). PVS special issue 45, edited by Rüdiger Schmitt-Beck, 489–513. Baden-Baden: Nomos.Google Scholar

Rabinowitz, George, and Macdonald, Stuart E.. 1989. A Directional Theory of Issue Voting. American Political Science Review 83 (1):93–121.Google Scholar

Richardson, Bradley, and Beck, Paul A.. 2007. The Flow of Political Information: Personal Discussants, the Media, and Partisans. In Democracy, Intermediation, and Voting on Four Continents, edited by Richard Gunther, José R. Montero and Hans-Jürgen Puhle, 183–207. Oxford: Oxford University Press.Google Scholar

Richey, Sean. 2008. The Autoregressive Influence of Social Network Political Knowledge on Voting Behavior. British Journal of Political Science 38:527–542.Google Scholar

Roberts, John H., and Lattin, James M.. 1991. Development and Testing of a Model of Consideration Set Composition. Journal of Marketing Research 28 (4):429–440.Google Scholar

Roßteutscher, Sigrid, and Scherer, Philipp. 2011. Ideologie und Wertorientierungen (Ideology and Value Orientations). In Zwischen Langeweile und Extremen. Die Bundestagswahl 2009 (Between Boredom and Extremes: The 2009 German Federal Election), edited by Hans Rattinger, Sigrid Roßteutscher, Rüdiger Schmitt-Beck and Bernhard Weßels, 265–280. Baden-Baden: Nomos.Google Scholar

Schickler, Eric, and Green, Donald. 1997. The Stability of Party Identification in Western Democracies: Results from Eight Panel Surveys. Comparative Political Studies 30 (4):450–483.Google Scholar

Schmitt-Beck, Rüdiger. 2000. Politische Kommunikation und Wählerverhalten. Ein Internationaler Vergleich (Political Communication and Electoral Behavior. An International Comparison). Wiesbaden: VS Verlag.Google Scholar

Schmitt-Beck, Rüdiger. 2004. Political Communication Effects: The Impact of Mass Media and Personal Conversations on Voting. In Comparing Political Communication. Theories, Cases, and Challenges, edited by Frank Esser, 293–322. New York and London: Routledge.Google Scholar

Shalizi, Cosma R., and Thomas, Andrew C.. 2011. Homophily and Contagion Are Generically Confounded in Observational Social Network Studies. Sociological Methods & Research 40 (2):211–239.Google Scholar

Sherif, Muzafer, and Hovland, Carl I.. 1961. Social Judgment. Assimilation and Contrast Effects in Communication and Attitude Change. New Haven, CT: Yale University Press.Google Scholar

Shikano, Susumu. 2003. Construction of Choice Sets and Its Influence on Voting Decision: Application of the Probabilistic Choice Set Model for Voter Choice under Two-Ballot System in Germany and Japan. Paper presented at the ECPR joint sessions of workshops, Edinburgh, Scotland, 28 March–2 April.Google Scholar

Shively, Phillips W. 1972. Party Identification, Party Choice, and Voting Stability: The Weimar Case. American Political Science Review 66:1203–1225.Google Scholar

Shocker, Allan D., Ben-Akiva, Moshe, Boccara, Bruno, and Nedungadi, Prakash. 1991. Consideration Set Influences on Consumer Decision-Making and Choice: Issues, Models, and Suggestions. Marketing letters 2 (3):181–197.Google Scholar

Sinclair, Betsy. 2012. The Social Citizen. Peer Networks and Political Behavior. Chicago and London: University of Chicago Press.CrossRef Google Scholar

Sniderman, Paul M., Brody, Richard A., and Tetlock, Philip E.. 1991. Reasoning and Choice. Cambridge, MA: Cambridge University Press.Google Scholar

Steenbergen, Marco R., and Hangartner, Dominik. 2008. Political Choice Sets in Multi-Party Elections. Paper presented at the workshop ‘The Politics of Change’, Amsterdam, The Netherlands, 13–14 June.Google Scholar

Thomassen, Jacques, ed. 2005. The European Voter: A Comparative Study of Modern Democracies. Oxford: Oxford University Press.Google Scholar

van der Eijk, Cees, and Niemöller, Broer. 1983. Electoral Change in The Netherlands. Amsterdam: CT Press.Google Scholar

van der Eijk, Cees, and Franklin, Mark N.. 2009. Elections and Voters. Basingstoke and New York: Palgrave Macmillan.Google Scholar

Walsh, Katherine Cramer. 2004. Talking about Politics: Informal Groups and Social Identity in American Life. Chicago, IL: University of Chicago Press.Google Scholar

Weisberg, Herbert F. 1993. Political Partisanship. In Measures of Political Attitudes, edited by John P. Robinson, Phillip R. Shaver and Lawrence S. Wrightman, 681–736. San Diego, CA: Academic Press.Google Scholar

Weßels, Bernhard. 2000. Gruppenbindung und Wahlverhalten: 50 Jahre Wahlen in der Bundesrepublik (Group Attachment and Voting Behaviour: 50 Years of Elections in the Federal Republic of Germany). In 50 Jahre Empirische Wahlforschung in Deutschland. Entwicklung, Befunde, Perspektiven, Daten (50 Years of Empirical Electoral Research in Germany. Development, Findings, Perspectives, Data), edited by Markus Klein, Wolfgang Jagodzinski, Ekkehard Mochmann and Dieter Ohr, 129–158. Wiesbaden: Westdeutscher Verlag.Google Scholar

Wilson, Carole J. 2008. Consideration Sets and Political Choices: A Heterogeneous Model of Vote Choice and Sub-National Party Strength. Political Behavior 30 (2):161–183.Google Scholar

Zakharova, Maria, and Warwick, Paul V.. 2014. The Sources of Valence Judgments: The Role of Policy Distance and the Structure of the Left-Right Continuum. Comparative Political Studies 47:1–26.Google Scholar

Zuckerman, Alan S., ed. 2005. The Social Logic of Politics: Personal Networks as Contexts for Political Behavior. Philadelphia, PA: Temple University Press.Google Scholar

Zuckerman, Alan S., Dasovic, Josip, and Fitzgerald, Jennifer. 2007. Partisan Families. The Social Logic of Bounded Partisanship in Germany and Britain. Cambridge, MA: Cambridge University Press.Google Scholar

Table 1Descriptives for Main Independent Variables

Table 2Party Preferences in Dyads

Table 3Moderators of Discussant Influence: Cross-Sectional Models (Probit Estimates for Dyads)

Table 4Moderators of Discussant Influence: Panel Models (Probit Estimates for Dyads)

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 1

PDF 69.5 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 2

File 6.1 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 3

File 120 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 4

File 167 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 5

File 167 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 6

File 120 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 7

File 120 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 8

File 120 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 9

File 171 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 10

File 171 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 11

File 120 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 12

File 171 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 13

File 171 Bytes

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 14

File 20.3 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 15

File 5.6 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 16

File 2.7 MB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 17

File 5.2 MB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 18

File 2.1 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 19

File 16.5 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 20

File 17.2 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 21

File 16.1 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 22

File 16.3 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 23

File 16.7 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 24

File 16.9 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 25

File 16.9 KB

Schmitt-Beck and Partheymüller Supplementary Material

Supplementary Material 26

File 7.8 MB

Article contents

A Two-Stage Theory of Discussant Influence on Vote Choice in Multiparty Systems

Abstract

Information

DISCUSSANT INFLUENCE AT ELECTIONS

ELECTORAL CHOICES IN MULTIPARTY SYSTEMS

A TWO-STAGE THEORY OF DISCUSSANT INFLUENCE AT ELECTIONS IN MULTIPARTY SYSTEMS

HYPOTHESES

DATA AND METHODS

Data

Dependent Variables

Strategy of Analysis

Independent Variables

Control Variables

RESULTS

CONCLUSION

Footnotes

References

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Schmitt-Beck and Partheymüller Supplementary Material

Save article to Kindle

Save article to Dropbox

Save article to Google Drive

Reply to: Submit a response

Your details

You have entered the maximum number of contributors

Conflicting interests