How campaigns polarize the electorate

Abstract

The minimal effect theory of campaign studies stipulates that intense political competition during campaigns assures and reinforces the initial party choice of the electorate. We find that this reinforcement is two-fold. During the campaign, the party preference of the voters’ in-group party increases while the party preference of the voters’ out-group parties decreases. Voters’ preference for their most preferred party (MPP) increases during the election campaign, while their preference for their least liked party decreases during the campaign (LPP). Across parties voters experience an increase in their preference for their most preferred party and a decrease for their least liked party as the campaign progresses. These trends show that the political campaign polarizes the electorate by increasing the affective distance between in-group party and out-group party preferences, thereby resulting in stronger political polarization after the campaign than before the campaign. The data utilized in this study is a large six-wave panel-study of Danish voters’ party preferences during the Danish parliamentary election of 2011. Thus, the analysis provides evidence of the minimal effect theory and of political polarization within a multi-party context.

Keywords

campaigns Denmark minimal effect theory multi-party polarization

Introduction

The minimal effect theory from the pioneer studies on campaign effects posited that the campaign had a significantly limited effect on voter choice as campaigns only reinforce an initial choice by mobilizing predisposition and party identification (Berelson et al., 1954; Lazarsfeld et al., 1944; Schmitt-Beck, 2007). Recent research, however, suggests that elite party cues and their party campaigns play an important role in understanding the development of political polarization among the electorate (Druckman et al., 2013; Iyengar et al., 2012).

This article combines the theory of minimal effect with the idea of political polarization between the in-group party and the out-group parties within a multi-party context. We analyze how party preferences develop among the in-group and out-group parties during the campaign of the 2011 Danish general election. We show increasing in-party preference as well as decreasing out-party preference during the campaign, thus suggesting that the campaign both strengthens voters’ party choice and fuels political polarization between parties.

The minimal effect theory within a multi-party system

If there is any universal effect from campaigns, the minimal effect theory seems to be an appropriate candidate for explaining this effect. As the campaign progresses, voters are intensely reminded of their initial vote preference through a mobilization of their political predisposition and their link to a party preference. As predisposition and party identification were considered as relatively stable factors, so, too, was the vote choice regardless of the campaign. In this sense, the campaign reassured voters of their choice by activating a stable set of predispositions in the vote choice (Berelson et al., 1954; Campbell et al., 1960). Due to the high level of information regarding politics, policies and candidates, campaigns provide voters with information that helps them vote according to their predispositions (Brady and Johnston, 2006; Gelman and King, 1993; Holbrook, 1996; Johnston et al., 1992; Popkin, 1991; Schmitt-Beck and Farrell, 2002). During the high intensity campaign period when all candidates and parties are campaigning, it could be argued that all of the pieces of information provided will balance each other out and hence equalize the impact of the campaign and result in, more or less, a status quo.

Furthermore, studies highlight that the selective nature of voters’ perceptions and recollections of the campaign tend to reinforce the intermediated choice. For example, liberals tend to focus on liberal news channels (exposure) and tend to process and remember primarily liberal frames (perception) of the issues, which ultimately serve only to reinforce rather than to challenge their vote choice (Festinger, 1957; Hansen, 2004; Iyengar, 1990; Kinder and Sears, 1985; Leeper, 2014). For example, voters inclined to vote ‘no’ on a referendum tend to learn the factual premises supporting their position prior to the referendum and vice versa (Hansen, 2004). Similarly, Hansen and Pedersen (2008) find that voters tend to explore the website of the parties with which they are inclined to agree with. Therefore, what basically happens during a campaign is that voters seek confirmation of their initial hunch regarding party choice.

In sum, the minimal effect theory argues that political campaigns reinforce initial party choice by mobilizing party attachment and party identification.

The minimal effect and the degree of political polarization are well documented within the US two-party system. However, these two concepts may not apply to multi-party systems in a similar vein. Elections in multi-party systems differ markedly from elections in two-party systems. Although many issues may be at play, the competition in two-party systems is one-dimensional as there is one battleground, namely that between the two parties. In multi-party systems, there are multiple battlegrounds within multidimensional systems. While proximate parties compete for the same votes, they are also cooperating, to a greater extent, either to form a government coalition or to form an alternative government in opposition. Multi-party systems provide voters not only with more choices but also with more parties to attract voters within the group of cooperating parties. Voters are essentially tied into a complex web of information on multiple parties and issues that on Election Day must result in a single vote being cast for one among the many competing parties and candidates. Whereas voters’ choices in two-party systems are among party A, party B and abstention, voters in multi-party systems must choose among several parties that, almost to the same degree, closely represent their views on the relevant policy issues. Accordingly, party choice is simply not that simple within multi-party systems.

Following from this, with respect to measuring voter choice within multi-party systems, the intended vote choice provides only a single statement about the complex process of choosing between multiple parties. Intended vote choice says nothing about the strength of the party choice, whether the choice was a close call between many parties or whether the choice was strong and without doubt. To unfold the vote choice in multi-party settings, we ask voters to assess all parties on a scale from 0 (dislike) to 10 (like). This measure of party preference allows us to investigate vote choice in much more detail (van der Brug et al., 2007). However, in our case, as we have eight running parties, thus we have eight dependent variables. To overcome this problem, we introduce Most Preferred Party (MPP). MPP is a party preference (on a 0 to 10 dislike-to-like-scale) for the party for which the voter has the strongest preference immediately after the election. For almost all voters this corresponds to the party they actually voted for. In this way we can collapse the eight party variables into one dependent variable (MPP). Furthermore, it also gives us the advantage that we can follow the development on MPP throughout the campaign as we have up to six waves for each individual in our panel-study.

Hypotheses

On the basis of the minimal effect theory and the additional considerations as stated herein, we propose three hypotheses on the development of party preferences and general party preference formation within a multi-party setting.

The first hypothesis (H₁) is the enhancement hypothesis, which argues that MPP will increase during the campaign. This hypothesis is fueled by our expectation from the minimal effect theory, stipulating that voters seek confirmation of an initial vote hunch along with increased selective perception of party cues as information over the campaign intensify (Druckman et al., 2013; Nicholson, 2012). Furthermore, according to social identity theory, this reasoning is strengthened with respect to in-group affiliation, which directly corresponds to stronger biases when assessing opponents (Iyengar et al., 2012).

The enhancement hypothesis is also supported by the fact that as the campaign progresses fewer voters tend to be in doubt about what to vote, even though as much a 20% of the voters were still in doubt the day before the election under investigation (Hansen 2014b). From the enhancement hypothesis we would expect that polarization increases with the decision on which party to support. Furthermore, undecided voters still have preferences for each of the running parties, which suggest that most of the undecided voters have a good hunch on what they will end up voting.

The second and third hypotheses are concerned with political polarization (Dilliplane, 2014; Fiorina and Abrams, 2009; Hetherington, 2009; Zaller, 1992). Following Iyengar et al. (2012), we operationalize political polarization as the affective distance between in-group party and out-group parties. Voters not only favor more strongly the party or party bloc which they prefer, but they also express stronger opposition to the opposing parties. That is, as the campaign progresses, we find political polarization along party lines as the preference for the in-group party increases and the preference for the out-group parties decreases more significantly. The reasons for political polarization during the campaign include selective exposure and the perception of party cues as determined by initial party attachment (Nicholson, 2012). Coined by the concept of the media’s “echo chambers” in Capella and Jamieson (2009) where stereotypes of opponents are being repeated and echoed in chambers, the nature of especially negative campaigns tends to increase polarization as it reinforces a partisan bias in regard to assessing in-group and out-group political viewpoints (Arceneaux and Johnson, 2013; Iyengar et al., 2012; Levendusky, 2013). In a multi-party system organized into two governing blocs, hypothesis 2 (H₂), the bi-polarization hypothesis, posits that voters’ increased party preference for the main party within the bloc for which they intend to vote corresponds to a decreasing party preference for the main party in the opposing bloc. The main party in either bloc is the largest party and the one promoting a candidate for Prime Minister.

Hypothesis 2 resembles what happens in two-party systems. However, hypothesis 2 must be supplemented to account for the character of multi-party systems. Even if there are two main parties within a multi-party system that have Prime Minister candidates, it is relevant to consider what happens to party preferences across the entire political spectrum. Hence, we propose hypothesis 3 based on the differences in party preference across the campaign, that is, the multi-polarization hypothesis (H₃). This hypothesis proposes that party preferences polarize as the campaign progresses for all voters across all eight running parties irrespectively of their final vote choice. Voters will increase party preference for their own party and decrease the preference for the party they dislike the most.

Method and context

The data are collected within the context of the Danish multi-party system. In Denmark, election dates are not fixed as it is the prerogative of the Prime Minister to call for an election, and she may call for one at any time within the four-year-term period. Elections are usually called three weeks in advance, which makes election campaigns short and intensive. The election system is highly proportional, divided into 10 districts and 92 multi-member constituencies. The electoral threshold is 2%, 135 seats are distributed in the 10 districts and 40 seats adjustment seats. The vote can either be for a candidate running for a party or the party. About 50% cast a vote for a candidate (Bengtsson et al., 2014; Elklit, 2008). The gross electoral volatility was 32.7% between the 2007 and 2011 election and has been at about that level since 1994 (Hansen, 2014a).

As parties dominate candidate recruitment and monopolize representation in parliament, there are no public primaries, all candidates have an indispensable party label, and all relevant parties and candidates are known well in advance of the campaign. The eight parties represented in parliament prior to the election were also represented after the election, hence no change in the party system. The election that enabled this data collection was called on 26 August 2011 and held on 15 September 2011. All parties competing in the election supported either the incumbent Liberal Prime Minister or the Social Democratic challenger; it was the latter who won the election (for more information on the election and Danish parties, see Kosiara-Pedersen, 2012; Stubager et al., 2013).

The data applied are derived from the research project titled Online Panel of Electoral Campaigning (OPEC), which was composed of a six-waves panel-design and several experiments (Hansen, 2008; Hansen et al., 2012; Hansen and Pedersen, 2012, 2014). In this study, we utilize the full OPEC data as a six-wave panel-study whereby we are able to follow the campaign very closely at the individual level.

The questions used to assess Most Preferred Party (MPP) consider the preference in each wave for the party the voter eventually ended up having the strongest preference for. For almost all voters this corresponds to the party they actually voted for. The party preference question is the classic question on whether the voter likes/dislikes the parties, measured on a 0 to 10 scale, where 0 stands for ‘dislikes’ and 10 reflects ‘likes very much’.¹ These questions show the extent to which voters like each of the eight parties represented in parliament. We also introduce the Least Preferred Party (LPP), i.e. the party preference in each wave for the party the voter like the least by the end of the campaign. The few respondents who replied “don’t know” are excluded (this amounts to less than 2.4%). In combination these two variables, MPP and LPP, provide an assessment of how the voters’ party preferences develop during the campaign; i.e. party preferences are measured by the 0 to 10 dislike/like scale throughout the paper.

Analyses

In the analyses, we first show the development of MPP among the electorate at large as a pure descriptive analysis. We then turn to the analyses of the enhancement hypothesis (H₁), bi-polarization hypothesis (H₂) and multi-polarization hypothesis (H₃) using fix effects time-series analysis to make the most use of the six-waves panel-study.

Development of sympathy for most and least preferred party

First, Figure 1 shows the development of MPP on the left axis across each of the six waves as the campaign progresses. MPP increases significantly from 7.05 to 8.43 during the campaign; i.e. voters like their preferred party more and more as the campaign progresses. Second, Figure 1 shows the development for the Least Preferred Party (LPP) with the scale on the right axis. LPP significantly decreases from 2.47 prior to the intensive campaign period to 0.64 immediately after the election; hence, a marked increase in voters’ dislike of their least preferred party as the campaign progresses.

The largest increases happen for both MPP and LPP in the weeks before the election is called and in the final week of the election campaign. One interpretation of this finding is that when the election is announced, voters prepare for the election by stepping up for their preferred party by liking them more and disliking the strongest opponent more as well. In the final week of the campaign, voters in doubt about what to vote eventually have to come to a decision which goes hand in hand with a final increase in like of MPP and dislike of LPP. (See Appendix Table 1 for descriptive statistics on the development presented in Figure 1).

Figure 1.

Party preference across the campaign, N = 2506–8289 in each wave.

The findings of increasing MPP and decreasing LPP in combination imply that the campaigns have a polarization effect on party preferences. There is a larger affective distance between the most and least preferred parties. If we measure political polarization as the distance between the MPP and LPP (Figure 1), we find that polarization before the campaign was 4.58 (7.05–2.47) and that this is increased to 7.79 (8.43–0.64) after the campaign, thus confirming the significant affective polarization effect of the campaign. These effects will be analyzed at the individual level in the following sections with a time-series fix effect model.

Enhancement of most preferred party (MPP) and least preferred party (LPP)

To provide a better understanding of the development at the level of the individual voter we apply a time-series fixed effect model to our six-wave panel-study. This allows us to account for individual heterogeneity regardless of whether these individual variables are observed or unobserved in the data. In this case it could be basic socio-demographic variables such as education, gender and age which can be assumed to be stable over the short period of a three week campaign, but it could also be variables such as party attachment or various psychological traces that would be expected to influence how the voters react to the campaign. The fixed effect model works by adding a dummy variable for each individual and thus provides an estimation of the within-subject effect of the campaign controlled for individual characteristics whether it is observed or not. This implies that the fixed effect models control for time-invariant characteristics thus providing the net effect of the campaign for each voter over time. Another advantage is that the model also copes with missing data e.g. if a respondent misses a wave the fixed effect allows to control for this and keep the individual in the model. All in all, this model is a conservative, or pure, test of the campaign impact on the voter because we control for any effect from individual characteristics (Wooldridge, 2009).

Figure 2 presents predicted values of MPP (left axis) and LPP (right axis) from two models. The dependent variables are MPP and LPP, respectively, and the waves, included as dummies, are the only independent variables. The models apply individual fixed effects and use individual and wave robust cluster standard errors (see Appendix Table 2 for entire models). Hence, whereas Figure 1 presents the descriptive aggregate development in MPP and LPP, Figure 2 provides the aggregate development in MPP and LPP with the control for any changes in composition of the individual voter in each wave.

Figure 2.

Predicted party preference across the campaign using fixed effect time-series model, N = 28,130/28,020 voters = 8434.

Figure 2 provides a similar development of MPP and LPP as the campaign progresses as the aggregate picture of Figure 1. We see increasing MPP and decreasing LPP, i.e. a polarization of party preferences during the campaign. But whereas Figure 1 was a description of the development in party preferences, the predicted values in Figure 2 are controlled for individual time invariant-factors. Therefore we can be quite certain that it is the campaign which causes the change in party preferences. The models also allow for an interpretation of each coefficient from the model. For example, for a given voter, as the campaign moves from the time point before the campaign to the campaign start, MPP increases by 0.38 on the 0 to 10 scale, i.e. from 7.36 to 7.74 on the top curve between the first two waves in Figure 2.

Hypothesis 1 claims that MPP would increase as the campaign progresses. We found evidence both in the description in Figure 1 and through the time-series fixed effect model in Figure 2. Furthermore, the figures also presented evidence of the polarization of party preferences as the distance between MPP and LPP increases as the campaign progresses even with the strong time-invariant control that is applied in Figure 2. This implies that the campaign has a polarization effect on voters across their party preferences.

Bi-polarization of party preferences

The analyses have, thus far, analyzed how much voters like or dislike their most and least preferred parties without taking into consideration the specific preference for the various parties within the multi-party system nor each individual voter’s specific party choice. In the following analysis, we focus on the two largest parties within the Danish multi-party system, the Social Democrats and the Liberals. They are ideologically placed on each side of the political left-right spectrum and each head one of the two blocs in parliament. The Social Democrats are the largest party left of center, while the Liberals are the main party in the right of center on the traditional, economic left-right dimension. Furthermore, these two parties present the only two Prime Minister candidates in this election. This allows us to test the bi-polarization hypothesis (H₂) that voters during the campaign have increased party preference for the main party within the bloc they eventually end up voting for, and vice versa, that voters decrease their party preference for the main party in the opposing bloc.

The dependent variables in Figure 3 are party preferences for the Social Democrats and for the Liberals and the independent variables are dummies for each of six waves. The models are divided into four separate models: right-leaning voters’ preference for the Liberals, right-leaning voters’ preference for the Social Democrats, left-leaning voters’ preference for the Liberals, and left-leaning voters’ preference for the Social Democrats.² Right-leaning voters are those voting for the four center-right and right parties, whereas the left-leaning voters are those voting for the four center-left and left parties. Again, the strong control for the effect of the campaign is applied through the times-series fixed-effect model. (See Appendix Table 3 for entire models).

Figure 3.

Predicted preference for Social Democats and Liberals using fixed effect time-series model, N = 28,130/28,020 voters = 8434.

The preferences of the right-leaning voters (solid lines) are as expected by the bi-polarization hypothesis. During the campaign the right-leaning voters increasingly like the main party in the bloc they voted for and dislike the main party in the opposing bloc. Right-leaning voters are clearly bi-polarized by the campaign.

The left-leaning voters, on the other hand, are quite stable when it comes to their preference for the main party in the bloc they vote for, i.e. the Social Democrats. Furthermore, they experience a slight increase in their preference for the main opposition party, i.e. the Liberals. The Liberals did increase their general support during the campaign which might be one explanation for this slight increase in the preference for their main opponent.

In sum, the bi-polarization hypothesis (H₂), stating that voters during the campaign have increased preference for the main party within the bloc they eventually end up voting for and decreasing preference for the main party in the opposing bloc, is supported for the right of center voters but not for the left of center voters within the Danish multiparty system.

Multi-polarization of party preference

However, as previously argued, multi-party systems work quite differently from two-party systems. To account for this, we propose and test the multi-polarization hypothesis (H₃) which argues that all voters experience a polarization of their party preferences through an increased support for their most preferred party and a decreased support for their least preferred party.

Figure 4 shows the difference between the voters’ most and least preferred party in each wave during the campaign depending on what party they voted for. Again, the estimates are predicted probabilities from times-series fixed models divided on each party with the absolute difference between most and least preferred party as dependent variable and waves as the only independent variables. (See Appendix Table 4 for entire models).

Figure 4.

Predicted difference preference for parties using fixed effect time-series model, N = 28,138/28,020 voters = 8434.

Figure 4 clearly supports the multi-polarization hypothesis. All voters experience an increase in the difference between the party they liked the most and the party they disliked the most as the campaign progresses. The increase is apparent for all parties at all stages from before the intensive campaign period to immediately after Election Day. Figure 4 also shows that left-leaning voters, all colored red, have a stronger level of individual polarization than right-leaning voters, all colored blue, before the campaign as well as during the campaign. This implies that even though they show little development in their preference for the main party within their bloc, the Social Democrats (Figure 3), during the course of the campaign, show clear polarization when it comes to their own party choice. The explanation for this may be found in the specific context of this campaign where left-leaning voters were not all thrilled about the Social Democrats and their performance during the campaign.

Another interesting finding from the eight models presented in Figure 4 is that the two most extreme parties on the traditional, economic, redistribution dimension, the Red-Green Alliance placed on the far left and Liberal Alliance placed on the far right, both experience the largest increases in the polarization as the campaign progresses. An obvious explanation is that the political distance between the most and least preferred party is larger for these voters than for voters placed further towards the center of the political space. In addition, this could indicate that when voters have fewer parties to choose from their party preferences become clearer. Voters towards the ends of the political spectrum experience a smaller level of cross pressures from various parties when compared to the center of the political space. Their party choice may thus be firmer (Hansen et al., 2012). Transferred to a two-party system this would suggest that moderates experience a smaller degree of polarization than the hardliners within both parties (Levendusky, 2013).

Following on from the argument that political distance matters, it could also be mentioned that among the left-leaning voters there is a clear tendency towards a larger degree of polarization, the further towards the extreme the party is placed. The four red lines are placed in the order of the parties’ place on the traditional, redistributive left-right dimension, where the voters of the party placed most to the left, the Red-Green Alliance, experience the largest degree of polarization, as mentioned above. They are followed by the voters of the Socialist People’s Party, the Social Democrats and the Social Liberals. The same pattern is not found among the right-leaning voters which might be due to less clear place of these parties on the left-right dimension, where they are placed closer to each other than the left-of-center parties.

Conclusion: Campaigns increase political polarization

Voters must navigate in a crowded political space within multi-party systems. However, on Election Day, they must compress their political preferences into a single mark on the ballot. We have developed and tested two variables, Most Preferred Party (MPP), which is the voter’s preference for the party the voter prefers the most immediately after the election, and Least Preferred Party (LPP), which is the voter’s preference for the party the voter dislikes the most immediately after the election. Both of these variables are measured six times across the campaign and thus provide a simple but compact operationalization of party preference in a multi-party system. We suggest that this operationalization would be fruitful for other analysis of party preferences in the multi-party context (see also de Sio and Weber, 2014).

During campaigns, the number of competing political messages is at its highest. The way that voters comprehend these mutable sources of often conflicting information seems to be by increasing their preference for their most preferred party and by decreasing their preference for parties they dislike the most. Hence, the political campaign results in a polarization of voters’ party preferences across the political spectrum. By building on theories such as the minimal theory of campaign effects, campaigns reinforce voters’ initial party choice, and social identity theory of in-group/out-group polarization suggests that these findings are entirely plausible. That is, campaigns do indeed matter, but for many voters the primary effect is reinforcement of initial party preference rather than a cause of party switching.

The enhancement hypothesis, H₁, (Figures 1 and 2) is confirmed. This hypothesis stipulates that voters’ preference for their most preferred party (MPP) increases during the election campaign, while their preference for their least liked party decreases during the campaign. This is tested descriptively in Figure 1 and through the times-series fixed-effect model where all individual invariant variables are controlled for in Figure 2.

The bi-polarization hypothesis, H₂, (Figure 3) is supported by the right-leaning voters as their preference for the main party in their own bloc (Liberals) increases and their preference for the main party in the opposing bloc (Social Democrats) decreases as the campaign progress. However, it is not supported by the left-leaning voters.

Finally, we find support for the multi-polarization hypothesis, H₃, (Figure 4) and argue that all voters experience an increase in their preference for their most preferred party and a decrease for their least liked party as the campaign progresses.

Political polarization is not just a two-party phenomenon. The analyses presented here clearly show that multi-party campaigns also fuel political polarization. The polarization is marked within the two blocs but also among the individual parties indicating that it is not the two-bloc nature of the case analyzed here that causes the polarization. Polarization also takes place within multiparty systems characterized by multiparty (minority) governments, a consensual nature of parliament with many (almost) unanimous legislative decisions and a high degree of cooperation.

Furthermore, political polarization takes place even within a campaign context almost entirely dominated by positive campaigning. Negative campaigning is very limited and the media climate is dominated by strong public service ideals of balanced and impartial news as well as the inclusion of all parties (Hansen and Pedersen, 2014). Even within such a campaign context that impedes strong political polarization, we find these clear effects. While this suggests that the selective perception (and less selective exposure) of party cues and voters’ search for confirmation of initial vote choice may be the factors that drive polarization in our case, it also suggests that we could expect to see even greater polarization in multiparty systems with a higher level of negative campaigning.

Should we be concerned about the polarization effect of campaigns? We have seen that polarization drives the losing side to experience increased dissatisfaction with democracy (Iyengar et al., 2012); however, if this is due to intensive campaigning, which fades away between elections, we may be less concerned. Even with increasing polarization, we still see voters that are increasingly willing to change parties. Hence, it can be concluded that political polarization does not necessary lead to political entrenchment in the long run.

Footnotes

Appendix

Table 4.

Modeling the difference between MPP and LPP across the campaign for each of the parties (models for Figure 5).

	Difference between MPP and LPP for Social Democrats	Difference between MPP and LPP for Social Liberals	Difference between MPP and LPP for Socialist People’s Party	Difference between MPP and LPP for Red-Green Alliance	Difference between MPP and LPP for Conservative People’s Party	Difference between MPP and LPP for Danish People’s Party	Difference between MPP and LPP for Liberals	Difference between MPP and LPP for Liberal Alliance
Campaigns starts	0.48^***	0.74^***	0.54^***	0.38^***	0.24^*	0.49^***	0.66^***	0.63^***
	(0.06)	(0.08)	(0.12)	(0.07)	(0.12)	(0.13)	(0.07)	(0.08)
Week 1	0.49^***	0.78^***	0.52^***	0.32^**	0.61^***	0.40^*	0.75^***	0.67^***
	(0.07)	(0.10)	(0.13)	(0.10)	(0.13)	(0.17)	(0.08)	(0.10)
Week 2	0.51^***	1.02^***	0.66^***	0.52^***	0.77^***	0.65^***	0.79^***	0.68^***
	(0.07)	(0.09)	(0.13)	(0.08)	(0.14)	(0.15)	(0.07)	(0.10)
Week 3	0.72^***	1.19^***	0.94^***	0.62^***	1.03^***	0.77^***	0.99^***	0.96^***
	(0.07)	(0.10)	(0.13)	(0.08)	(0.14)	(0.17)	(0.08)	(0.09)
Just after election	0.93^***	1.56^***	1.37^***	1.02^***	1.71^***	1.50^***	1.54^***	1.26^***
	(0.06)	(0.08)	(0.11)	(0.07)	(0.12)	(0.12)	(0.06)	(0.08)
Constant	6.90^***	6.25^***	6.13^***	7.25^***	6.06^***	6.06^***	6.09^***	7.31^***
	(0.04)	(0.05)	(0.07)	(0.05)	(0.08)	(0.09)	(0.04)	(0.05)
R2	0.08	0.21	0.15	0.11	0.23	0.14	0.17	0.17
R2_within	0.08	0.21	0.15	0.11	0.23	0.14	0.17	0.17
R2_between	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
R2_overall	0.02	0.06	0.04	0.03	0.07	0.04	0.05	0.04
Corr_i_u	–0.00	–0.00	0.00	–0.01	–0.01	–0.00	–0.01	–0.00
Rho	0.72	0.69	0.72	0.72	0.67	0.71	0.71	0.73
No. individuals	1358	774	372	750	373	371	1237	610
Observations	5271	3000	1480	3000	1499	1484	4856	2399

Note: Fixed effects model, Unstandardized beta coefficients, Robust standard error in parentheses *p < 0.05; **p < 0.01; ***p < 0.001. Ref. before the campaign.

Acknowledgements

A previous version was presented at MPSA 2012, Chicago, USA and at workshops at the Center for Voting and Parties 2012, University of Copenhagen. Thanks to the editors and reviewers at Party Politics, and to Martin Vinæs Larsen and Bolette Danckert, University of Copenhagen.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The project is financed by the Danish Council for Independent Research (grant no. 09-063837).

Notes

References

Arceneaux

Johnson

(2013) Changing Minds or Changing Channels? Chicago, IL: The University of Chicago Press.

Bengtsson

Hansen

Harðarson

. (2014) The Nordic Voter: Myths of Exceptionalism. Colchester: ECPR-Press.

Berelson

Lazarsfeld

McPhee

(1954) Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago, IL: The University of Chicago Press.

Brady

Johnston

(2006) Capturing Campaign Effects. Ann Arbor, MI: The University of Michigan Press.

Campbell

Converse

Miller

. (1960) The American Voter. Chicago, IL: The University of Chicago Press.

Cappella

Jamieson

(2009) Echo Chamber. New York: Oxford University Press.

de Sio

Weber

(2014) Issue yield: A model of party strategy in multidimensional space. American Political Science Review 108(4): 870–885

Dilliplane

(2014) Activation, conversion, or reinforcement? The impact of partisan news exposure on vote choice. American Journal of Political Science 58(1): 79–94.

Druckman

Peterson

Slothuus

(2013) How elite partisan polarization affects public opinion formation. American Political Science Review 107(1): 57–79.

10.

Elklit

(2008) Denmark: Simplicity embedded in complexity (or is it the other way around?) In: Gallagher

Mitchells

(eds) The Politics of Electoral Systems. Oxford: Oxford University Press, pp. 453–471.

11.

Festinger

(1957) Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.

12.

Fiorina

Abrams

(2009) Disconnect: The Breakdown of Representation in American Politics. Norman, OK: University of Oklahoma Press.

13.

Gelman

King

(1993) Why are American presidential election campaign polls so variable when votes are so predictable? British Journal of Political Science 23(1): 409–451.

14.

Hansen

(2004) Deliberative Democracy and Opinion Formation. Odense: University Press of Southern Denmark.

15.

Hansen

(2008) The Effect of Political Campaigns – Overview of the research Online Panel of Electoral Campaigning (OPEC). Paper presented at the XV NOPSA Conference, University of Tromsø, 6–9 August 2008.

16.

Hansen

(2014a) Indledning. In: Hansen

Kosiara-Pedersen

(eds) Folketingsvalgkampen 2011 i perspektiv. Copenhagen: DJØF-forlag, pp. 26–36.

17.

Hansen

(2014b) Mediernes meningsmålinger. In: Hansen

Kosiara-Pedersen

(eds) Folketingsvalgkampen 2011 i perspektiv. Copenhagen: DJØF-forlag, pp. 119–132.

18.

Hansen

Pedersen

(2008) Negative campaigning in a multiparty system. Scandinavian Political Studies 31(4): 408–427.

19.

Hansen

Pedersen

(2012) Efficiency of different recruitment strategies for web panels. International Journal of Public Opinion Research 24(2): 238–249.

20.

Hansen

Pedersen

(2014) Valgkampens negative kampagnebudskaber. In: Hansen

Kosiara-Pedersen

(eds) Folketingsvalgkampen 2011 i perspektiv. Copenhagen: DJØF-forlag, pp. 133–152.

21.

Hansen

Kosiara-Pedersen

Pedersen

(2012) Online Panel of Electoral Campaigning (OPEC) – Facts about the Surveys. Copenhagen: University of Copenhagen, Department of Political Science, Centre for Voting and Parties (CVAP) Working Paper Series no. 4.

22.

Hetherington

(2009) Putting polarization in perspective. British Journal of Political Science 39(2): 413–448.

23.

Holbrook

(1996) Do Campaigns Matter? Thousand Oaks, CA: Sage Publications.

24.

Iyengar

(1990) Shortcuts to political knowledge: The role of selective attention and accessibility. In: Ferejohn

Kuklinski

(eds) Information and Democratic Processes. Chicago, IL: University of Illinois Press, pp. 160–185.

25.

Iyengar

Sood

Lelkes

(2012) Affect, not ideology: A social identity perspective on polarization. Public Opinion Quarterly 76(3): 405–431.

26.

Johnston

Blais

Brady

. (1992) Letting the People Decide: Dynamics of a Canadian Election. Stanford, CA: Stanford University Press.

27.

Kinder

Sears

(1985) Public opinion and political action. In: Lindzey

Aronson

(eds) The Handbook of Social Psychology. New York: Random House, pp. 659–741.

28.

Kosiara-Pedersen

(2012) The 2011 Danish parliamentary election: A very new government. West European Politics 35(2): 415–424.

29.

Lazarsfeld

Berelson

Gaudet

(1944) The People’s Choice: How the Voter Makes up his Mind in a Presidential Campaign. New York: Columbia University Press.

30.

Leeper

(2014) The informational basis for mass polarization. Public Opinion Quarterly 78(1): 27–46.

31.

Levendusky

(2013) How Partisan Media Polarize America. Chicago, IL: The University of Chicago Press.

32.

Nicholson

(2012) Polarizing cues. American Journal of Political Science 56(1): 52–66.

33.

Popkin

(1991) The Reasoning Voter: Communication and Persuasion in Presidential Elections. Chicago, IL: The University of Chicago Press.

34.

Schmitt-Beck

(2007) New modes of campaigning. In: Dalton

Klingemann

H-D

(eds) The Oxford Handbook of Political Behavior. Oxford: Oxford University Press, pp. 744–764.

35.

Schmitt-Beck

Farrell

(2002) Studying political campaigns and their effects. In: Farell

Schmitt-Beck

(eds) Do Political Campaigns Matters? Campaign Effects in Elections and Referendums. London: Routledge, pp. 1–21.

36.

Stubager

Hansen

Goul Andersen

(eds) (2013) Krisevalg: Økonomien og folketingsvalget 2011. Copenhagen: DJØF-Forlag.

37.

van der Brug

van der Eijk

Franklin

(2007) The Economy and the Vote: Economic Conditions and Elections in Fifteen Countries. Cambridge: Cambridge University Press.

38.

Wooldridge

(2009) Introductory Econometrics: A Modern Approach. 4th edition. Mason, OH: South-Western Cengage Learning.

39.

Zaller

(1992) The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.