WHEN THE GLOVES COME OFF: INTER-PARTYVARIATION IN NEGATIVE CAMPAIGNING

ANNEMARIE S. WALTERDepartment of Political Science

University of AmsterdamOudezijds Achterburgwal 237

1012 DL AmsterdamThe Netherlandsa.s.walter@uva.nl

CATHERINE E. DE VRIESDépartment de Science Politique et Relations Internationales

Université de Genève40 Boulevard du Pont d’Arve

CH-1211 Genève 4, Switzerlandcatherine.devries@unige.ch

August 10, 2011

Abstract

This study examines inter-party variation in negative campaigning within multipartycompetition. It advances existing research in two ways. First, it contributes to thetheoretical understanding of the determinants of negative campaigning in a multi-party context by differentiating between parties’ vote-seeking and office-seeking in-centives and stating the implications of office-seeking for the use of negative cam-paigning. Second, it empirically tests these new hypotheses by employing contentanalysis-data from ten Dutch electoral campaigns between 1981 and 2010. The anal-ysis highlights that negative campaigning within a multiparty system is a riskier strat-egy compared to a two-party context as parties face both electoral backlash effects aswell as post-election coalition bargaining costs.

Key Words: Negative Campaigning, Multiparty Competition, Election Campaigns,Netherlands and Political Parties.

1 Introduction

When the gloves come off has long been a sports idiom used in the boxing ring. In recent years, it

is used more and more frequently by journalists and political commentators to describe election

campaigns in which politicians go head to head in television debates and attack their main rivals

in ads. Scientific inquiry into the extent, nature and impact of negative campaigning originated

in the US context (see e.g. Benoit et al. 2003; Geer 2006), but later inspired research on British

election campaigns (van Heerde-Hudson 2011; Sanders and Norris 2005). The use of negative

campaigning within continental European multiparty systems has received much less scholarly

attention (the exception being recent studies on Danish campaigns, see Elmelund-Præstekær

2008, 2010; Hansen and Pedersen 2008). Most scientific work on electoral campaigns in con-

tinental Europe focuses more generally on campaign styles or on political advertising (see e.g.

Kaid and Holtz-Bacha 1995, 2006). As such, our understanding of the use of negative campaign-

ing in multiparty competition remains incomplete. This lack of understanding is worrisome

as most European party systems are characterized by multiple parties and by coalition govern-

ments (see Hobolt and Karp 2010), both factors we will argue affects the extent and scope of

negative campaigning.

This study tackles the issue of negative campaigning in multiparty competition. Specifically,

it explores inter-party variation. Gaining an insight into which parties are more likely to go

negative and why this is the case is central to grasping the use of negative attacks in election

campaigns. By examining variation among different parties within multiparty competition, the

article brings to bear a new theoretical framework for comprehending inter-party variation in

negative campaigning that combines both vote- and office-seeking incentives of parties. This

framework provides a useful analytical tool for understanding differences among parties in the

use of negative attacks within different institutional settings. The analysis also presents new data

on negative campaigning from a multiparty system par excellence, namely the Netherlands, from

1981 to 2010. The Dutch case provides an interesting testing ground for exploring inter-party

variation in negative campaigning as it is one of the most proportional electoral systems, char-

acterized by a highly fragmented party system (Andeweg and Irwin 2005). Examining negative

campaigning in this context thus provides a laboratory for the development and application of

1

new theoretical approaches on the topic.

Overall, this study enhances present research on negative campaigning in two important

ways. First, it advances existing explanatory models of inter-party variation in negative cam-

paigning originating from the US context that focus solely on parties’ vote-seeking objectives.

The few existing studies on negative campaigning in a multiparty system so far are predomi-

nately guided by this vote-seeking perspective (Elmelund-Præstekær 2008, 2010; Hansen and

Pedersen 2008). We claim that within multiparty competition parties’ office-seeking goals are

equally important. Second, we not only argue but also empirically substantiate that negative

campaigning constitutes a riskier strategy in a multiparty system compared to a two-party con-

text. Not only do parties face potential electoral costs of negative campaigning through back-

lash effects, they also face possible coalition bargaining costs. Negative campaigning, especially

when directed towards previous or future coalition partners, may jeopardize a party’s position

within post-election coalition negotiations. Consequently, within multiparty competition neg-

ative campaigning is mainly the prerogative of parties willing to take considerable electoral and

coalition bargaining risks.

The remainder of the article is structured as follows. In the next section, we briefly intro-

duce the concept of negative campaigning employed in this study. Subsequently, we present

our theoretical expectations regarding the relationship between party characteristics and neg-

ative campaigning based on parties’ vote- and office-seeking objectives. This is followed by a

discussion of the Dutch case. After outlining our data, coding procedures, operationalizations

and method, we present the empirical results. We conclude by drawing several general lessons

from our analysis.

2 Negative Campaigning and Multiparty Competition

The Concept of Negative Campaigning

Before we outline the factors that account for inter-party variation in negative campaigning, it is

important to develop a precise definition of the concept. The literature on negative campaigning

defies easy summary (Lau et al. 1999; Lau, Sigelman and Rovner 2007). It is voluminous yet fails

2

to yield consensus regarding a clear-cut definition of the term. Despite this lack of definitional

consensus, two types of definitions can be distinguished – evaluative and directional. The first

type employs a largely evaluative approach to the concept (e.g. Jamieson 1992; Mayer 1996).

From this perspective, not all criticism directed at an opponent can be classified as negative

campaigning; rather, only those critiques that are unfair, illegitimate, dishonest and deal with

trivial issues constitute negative campaigning. In this view, negative campaigning essentially

equates to lying about the undesirable characteristics of rivals (Davis and Ferrantino 1996).

Recently, scholars have moved away from this evaluative understanding of negative cam-

paigning to a definition that stresses the directional meaning of the concept, i.e. whether a party

or candidate says something negative about an opponent (e.g. Djupe and Peterson 2002; Sigel-

man and Shiraev 2002; Geer 2006). Here negative campaigning is defined as a strategy for polit-

ical parties to distinguish themselves from other political parties by attacking or criticizing their

opponents (e.g. Lau, Sigelman and Rovner 2007; Geer 2006). Political parties that use negative

campaigning try to become voters’ preferred party by diminishing positive feelings and emo-

tions directed towards opposing candidates or parties (Budesheim, Houston and DePaola 1996;

Lau, Sigelman and Rovner 2007; Westen 2007). The opposite strategy would be one of positive

campaigning where parties engage in acclamation or self-appraisal to appear more desirable

than their opponents (Budge and Farlie 1983; Benoit et al. 2003). In this view, negative cam-

paigning is one of several means available to political parties to win elections.

We employ the directional definition in this study for three reasons. First, a discussion re-

garding the ethics of negative attacks is not the objective of this article. The debate regarding

dirty politics focuses on the normative implications of negative campaigning for democracy.

This in itself is an interesting topic, but links up to a very different scientific discussion than the

one to which we intend to contribute. Our study provides an empirical examination of inter-

party variation in negative campaigning and sets out to explain which parties are more likely

to employ negative attacks. In doing so, we wish to refrain from a normative judgment regard-

ing the use or content of negative campaigning. Second, the directional perspective enables

researchers to more easily distinguish between positive and negative campaigning. Simply put,

from the directional perspective negative campaigning equates to critique of opponents in order

to convince voters to vote for one’s own party, while positive campaigning indicates that a party

3

highlights its own strengths and accomplishments in order to persuade voters. This distinction

becomes blurred within the evaluative perspective. For example, dishonesty and triviality are

seen as exclusive to negative campaigning, but one can easily imagine positive campaigning

strategies that stretch the ”truth” (e.g. Geer 2006). Finally, the evaluative definition of negative

campaigning introduces measurement error. For instance, what constitutes an unfair or dis-

honest critique? And which issues are trivial? These judgments ultimately lie in the eyes of the

beholder and thus cannot provide a solid basis for reliable and valid measurement (Mark 2006).

By defining negative campaigning on the basis of its directional nature rather than its evalu-

ative meaning, we avoid problems relating to theoretical clarity and empirical measurement.

Consequently, we understand negative campaigning as a critique directed at an opposing party,

regardless of the kind or accuracy of the criticism.

Inter-Party Variation in Negative Campaigning: Some Expectations

Two of the central questions for scholars in the field of negative campaigning are which parties

decide to go negative and why (e.g. Theilmann and Wilhite 1998; Hale, Fox and Farmer 1996;

Sigelman and Buell 2003; Skaperdas and Grofman 1995). The literature dealing with these ques-

tions originates from the US context and starts from the assumption that parties or candidates

are primarily vote-seeking. In this view, negative campaigning is a campaign strategy to increase

one’s vote share but also entails risks. These risks are potential boomerang or backlash effects.

One can think of these in the following way: parties engage in negative campaigning in order

to reduce positive judgments or emotions of voters towards rivals while at the same time try-

ing to increase positive judgments or emotions towards one’s own party. Negative campaigning

here resembles a careful balancing act: campaigns that are too negative or too harsh risk yield-

ing negative returns as they may instigate damaging feelings towards the attacker rather than

towards the targeted party or candidate (Johnson-Cartee, Copeland and Johnson 1991). Such

boomerang or backlash effects allow for the possibility that negative campaigning leads to a loss

rather than a boost in votes.

This raises the following question: Which parties or candidates are willing to bear the po-

tential risks of negative campaigning and why? Two party characteristics stand out in the US

4

literature: incumbency status and a party’s competitive standing in the polls.1 First of all, the

incumbency status of a party or a candidate matters. Challengers, i.e. parties or candidates cur-

rently not in office, are expected to be more willing than incumbents to take the risk of negative

campaigning. This is mainly because challengers have to overcome the natural advantages of

incumbency that their rivals possess (Swint 1998; Lau and Pomper 2004; Druckman, Kifer and

Parkin 2009). Incumbents can promote themselves, their work and their accomplishments on

the basis of their official position and duties. In addition, incumbents are likely more visible in

the media and therefore can benefit from name recognition and established reputations (Kaid

and Holtz-Bacha 2006). Finally, incumbents have a policy record that can serve as a strong basis

for positive campaigning (Benoit 1999). Challengers, by contrast, have to provide reasons why

voters should vote the incumbents out of office this time around (Kahn and Kenney 2004; Hale,

Fox and Farmer 1996).

Second, US studies highlight the importance of the competitive standing of parties or candi-

dates for the decision to go negative (Kahn and Kenney 2004; Hale, Fox and Farmer 1996; Skaper-

das and Grofman 1995). Parties or candidates use negative campaigning at the moment in the

campaign when its potential benefits are likely to be the highest. This is when the competition

between the parties or candidates increases. When there is no clear winner, parties or candidates

are more willing to endure the potential electoral costs associated with negative campaigning

(Kahn and Kenney 2004; Hale, Fox and Farmer 1996). Similarly, authors argue that parties trail-

ing behind in the polls are more inclined to go negative than front-runners (Benoit 1999; Damore

2002; Sigelman and Buell 2003).2 From these studies we can also deduce that challenger parties

1Within the US literature, authors also highlight electoral context effects, such as electoral rewards and the time-point in a race. Electoral rewards are important as certain elections are considered more critical for parties or can-didates than others. This is the case for delegate-rich US primaries, for example. The more delegates that are up forgrabs, the more likely candidates are going to take the risk of negative campaigning. Furthermore, parties or can-didates are more likely to attack other candidates prior to high stakes events, such as the New Hampshire primaryand other “make or break contests” that yield extensive media coverage (Haynes and Rhine 1998, 700). The argumentconcerning the time-point in the race is that negative campaigning will be more extensive as election day approaches(Damore 2002; Kahn and Kenney 2004) since the harmful side-effects of negative campaigning are likely to be smalleras there is little time left to lose voters due to backlash effects. In primaries, for example, negative campaigning is atits peak the closer one gets to candidate nominations (Peterson and Djupe 2005). The focus of this study is on varia-tion in negative campaigning across parties; thus, these factors are less relevant as they should predict the same levelof negativity for all parties.

2Peterson and Djupe (2005) use other measures of competitiveness, arguing that it is hard to determine whethernegativity drives closeness or whether closeness shapes the campaign decision to go negative. Their alternative mea-sures include the number of quality candidates, the number of challengers in general and the amount of spending.A quality candidate is a candidate that has previously held office; the more quality candidates take part in a race, themore competitive the election is expected to be. The same holds for the number of challengers; when this pool is

5

that are losing in the polls should be even more likely to resort to negative campaigning than

those that are ahead. Although negative campaigning is electorally risky, challenger parties that

are trailing behind in the polls may have no other viable option (e.g. Druckman, Kifer and Parkin

2009).

Drawing on the current US literature which assumes negative campaigning to be partly risky

due to possible electoral costs dude to backlash effects, we deduce three hypotheses:

Challenger Party Hypothesis (H1): Challenger parties are more likely to engage in

negative campaigning than incumbent parties.

Losing Party Hypothesis (H2): Parties that are losing in the polls are more likely to

engage in negative campaigning than parties that are gaining in the polls.

Losing Challenger Party Hypothesis (H3): Challenger parties that are losing in the

polls are more likely to engage in negative campaigning than challenger parties that

are gaining in the polls.

These hypotheses all relate to parties’ vote-seeking behavior which is important but most

likely not fully sufficient when exploring inter-party variation in negative campaigning, espe-

cially within multiparty competition. There is no question that obtaining votes is the chief ob-

jective for all parties, regardless of the institutional setting, but other goals also influence par-

ties’ campaigning strategies, especially in multiparty contexts. Building on the seminal work by

Müller and Strøm (1999), we assume parties to be rational actors whose behavior is guided by

three overarching political objectives: office, policy and votes. Within a two-party system, es-

pecially one with majority electoral districts, parties can concentrate on vote-seeking goals as a

majority of votes enables them to achieve office and to implement policy as a result. In contrast,

parties operating within a multiparty system have to carefully balance their vote-, office- and

policy-seeking objectives. In these systems, obtaining the most parliamentary seats does not au-

tomatically translate into government office or policy influence. Winning an election does not

large, the competitiveness of the election increases as it becomes more important to present voters with a clear anddistinctive profile (Peterson and Djupe 2005). Competitiveness may also relate to the amount of campaign spending.Here the expectation is that higher spending campaigns are more hotly contested campaigns. As a result, candidateswill be more willing to take risks and to engage in negative campaigning (Peterson and Djupe 2005). These factors areclearly relevant, but since inter-party variation is the object of this study the number of candidates and their “quality”are context effects that do not differ across parties. Of course, the amount of spending could be important, but incase of the Netherlands, object of our study, parties and races are also state subsidized which makes this factor lessrelevant.

6

necessarily translate into office in multiparty competition; rather, it makes a party a key player

in the post-election coalition bargaining process. Indeed, recent research demonstrates that

since the end of the Second World War two-thirds of all governments in Western Europe were

coalition governments (Hobolt and Karp 2010).3 Due to the necessity of post-election coalition

bargaining, negative campaigning constitutes a potentially riskier strategy for parties operat-

ing in a multiparty context. While parties in a two-party system need to worry about possible

backlash effects, parties in a multiparty system also need to worry about potential post-election

coalition bargaining costs. Given that elections within multiparty systems are fought in the wake

of coalition negotiations, negative campaigning may endanger a party’s goal of obtaining office.

Coalition bargaining impinges “on campaigns and advertising strategies because campaigners

may for example refrain from negative advertising against future or former coalition partners”

(Elmelund-Præstekær 2008, 6). A campaign that is too aggressive, too rough and too negative

may damage the opportunity to govern together, thus hurting parties’ office-seeking objectives

(e.g. van der Brug and van der Eijk 2005; Kaid and Holtz-Bacha 2006; Hansen and Pedersen 2008).

So, which parties are expected to engage in negative campaigning in a multiparty system?

We contend that next to vote-seeking incentives this depends on the coalition potential of a

party, i.e. a party’s office-seeking behavior. In line with Sartori (2005) and Lijphart (1999), we

understand a party’s coalition potential not only as the extent to which a party has participated

in previous government coalitions, but perhaps more importantly by the degree to which a party

is viewed as a viable coalition partner. Parties with low coalition potential have less to lose from

negative campaigning as their chances of being part of the government are slim to none from

the beginning. Parties with high coalition potential, by contrast, may be less inclined to attack

possible coalition partners.

The literature on coalition formation points to four key factors that contribute to the coali-

tion potential of parties: party size, distance to median party, government experience and new

party status. First, the size of the party, defined as the proportion of parliamentary seats it con-

3Hobolt and Karp (2010) show that between 1944-2009, two-thirds of governments in Western Europe were gov-erned by coalitions. Analyzing data collected across 479 governments in 17 West European countries, they also showa growing trend in the last twenty years for executive power to be shared between two or more parties. In only abouta third of cases, did a single party hold executive power, and in most of these cases, the executive governed with aminority of seats in the legislature, implying that the party was still reliant on the support of other parties to stay inoffice and promote its legislative agenda.

7

trols, influences negative campaigning negatively as larger parties are more likely to take part

in government coalitions. This is because size reflects electoral popularity and government par-

ticipation signals responsiveness to voters (Warwick 1996). Moreover, parties want to maximize

their power and thus want to form coalitions in which they have to share ministerial posts with

as few other political parties as possible. Extending this logic to negative campaigning, the coali-

tion bargaining costs of negative campaigning are expected to increase with party size.

The second factor affecting a party’s coalition potential is the distance to the median party.

Parties that control the median position are most likely to be part of a government coalition

(Warwick 1996). In a one-dimensional space, the median party occupies the dominant bargain-

ing position and is thus most likely to be part of a government coalition. Although political space

may be multi-dimensional, most scholars agree that the left/right ideological dimension is the

key conflict dimension in most advanced industrial democracies (Pierce 1999; Gabel and Huber

2000). Consequently, the median position on the left/right dimension constitutes the dominant

bargaining position and a party’s coalition potential decreases as it moves further away from

this dominant position (Bartolini 1998; Laver and Schofield 1998). Parties on the outskirts of the

political spectrum, i.e. extremist parties, lack this kind of coalition potential. Such parties, for

example the Flemish right extremists of Vlaams Belang, are often considered ideologically unac-

ceptable to govern and are therefore left out of government coalitions (van Spanje and van der

Brug 2007). In addition, parties closer to the median party in the system have more possible

coalition options. These parties can form government coalitions with both left- as well as right-

leaning parties (this may be especially the case in the Netherlands which is characterized by

broad-based oversize coalitions). Extremist parties, on the other hand, have only one coalition

opportunity, regardless of the direction of the party competition. We thus expect the likelihood

of negative campaigning to increase as the distance between a party and the median party in-

creases.

The third factor is government experience, i.e. the extent to which parties have already par-

ticipated in previous government coalitions (Lijphart 1999; Sartori 2005; Warwick 1996). Previ-

ous government experience matters for coalition potential because it demonstrates that parties

are able to govern. Parties that have built a reputation for being good coalition partners are

known for abiding by coalition accords and for working effectively with other parties (Warwick

8

1996). The costs of attempting to form a coalition with these parties are perceived to be lower

compared to parties with no government experience. Consequently, parties with frequent gov-

ernment experience have a high degree of coalition potential which in turn increases the possi-

ble coalition bargaining costs of negative campaigning.

A fourth and final factor is new party status. New parties are those that have not yet partic-

ipated in elections. Their coalition potential is expected to be low, and negative campaigning

may be less risky as a result. Additionally, these parties have no reputation to lose and are free

to challenge the establishment in every way. Compared to parties with existing linkages to vot-

ers and other parties in the system, new parties may face less harmful side-effects of negative

campaigning (for similar argumentation, see Elmelund-Præstekær 2010).

In addition to the three hypotheses based on the vote-seeking perspective, we introduce four

hypotheses based on parties’ office-seeking behavior:

Party Size Hypothesis (H4): Parties with a larger number of parliamentary seats are

less likely to engage in negative campaigning than those with fewer parliamentary

seats.

Median Party Hypothesis (H5): Parties that are closely positioned to the median

party position within the system are less likely to go negative than parties that are

positioned far away from the median party.

Government Experience Hypothesis (H6): Parties that have more government ex-

perience are less likely to go negative than parties that have less or no government

experience.

New Party Hypothesis (H7): New parties are more likely to engage in negative cam-

paigning than established parties.

3 The Dutch Case

We evaluate our expectations regarding inter-party variation in negative campaigning using new

data on negative campaigning from ten recent Dutch electoral campaigns. With one of the most

proportional electoral systems in the world (Andeweg and Irwin 2005), the Dutch case provides

an interesting testing ground. Dutch party competition is characterized by a high level of po-

9

litical fragmentation. As a result, the Dutch context allows us to study the use of negative cam-

paigning among a large number of parties. Finally, studying negative campaigning in the con-

text of a European multiparty system provides an exciting laboratory for the development of new

theoretical approaches regarding inter-party variation and for the testing of existing ones.

Before turning to the data, operationalization and method, a brief description of Dutch elec-

tion campaigns is useful. Dutch electoral campaigns are relatively short, especially compared

to US presidential races. For most parties the official campaign starts about four weeks prior to

Election Day (Brants, Kok and van Praag 1982). Moreover, measured by international standards

Dutch elections are quite inexpensive (Andeweg and Irwin 2005; van Praag 2005). The main

source of financing for parties is membership dues. Parties that are represented in parliament

also receive state subsidies which can be used for campaigns. The Netherlands is reluctant to

allow the solicitation of private funds, and overall Dutch election campaigns are still considered

to be in the process of professionalizing (Anker 1998; van Praag 2005; Andeweg and Irwin 2005).

These campaign characteristics affect the role television plays in Dutch election campaigns.

For example, the purchasing of television air time is a new development and still relatively un-

common. Political parties that are represented in Dutch parliament are granted free air time

on the national public television stations by the Dutch government. Since the introduction of a

new media law in 2008, also new parties taking part in the elections are granted airtime, but only

when they compete in all electoral districts (Mediawet 2008). These party election broadcasts are

called Zendtijd Politieke Partijen (ZPPs). Parties are allowed to air about twenty broadcasts an-

nually plus an additional six in the last two weeks of the election campaign. The broadcasts last

approximately three minutes each and thus resemble political advertisements (Mediawet 2008).

Until 1998, the Dutch public and commercial broadcasters refused to sell air time. Although to-

day air time is available on some channels, this is a still a fairly rare phenomenon, partly due to

limited finances of political parties. Consequently, the ZPPs remain the most important way of

airing party election broadcasts (Andeweg and Irwin 2005).

10

4 Data and Coding Procedure

In order to determine the impact of party characteristics on the extent of negative campaign-

ing, we need to develop a measure of negative campaigning per party per election. We derive

this measure from systematic content analysis of ten parliamentary election campaigns in the

Netherlands between 1981 and 2010. 4 The sources used for the content analysis are party elec-

tion broadcasts (ZPPs). These largely party controlled campaign sources were chosen to ensure

that we measure the behavior intended by a party. The research period covers four weeks prior

to election day for each campaign. In addition, we restrict our analysis to political parties that

have gained parliamentary representation, i.e. who at least received one parliamentary seat.

The data were gathered from television records of the public broadcasting companies and the

electronic media archive of the Netherlands Institute of Sound and Vision. In total, our sample

includes 146 party broadcasts aired during our research period5.

The transcripts of the ads serve as the basis for the analysis.6 The unit of analysis is not the

source (e.g. the ad), but the appeals made. Appeals are any mention of self praise or criticism

directed at a rival. In order to construct a measure of negative campaigning, every appeal is

classified as either positive or negative. The coding method is based on the procedure devel-

oped by Geer (2006). Positive appeals consist of any self praise of a party and negative appeals

are any criticism against a rival.7 The tone of the appeal, i.e. positive or negative, is the depen-

dent variable in the analysis. An appeal provides a fine-grained measure of negative or positive

campaigning and thus avoids problems often associated with the coding of complete ads. Cod-

ing complete ads is problematic as ads often contain both positive and negative campaigning

(Benoit et al. 2003; Geer 2006). Furthermore, we only coded spoken and written text. Some-

times negative appeals are made implicitly through the tone with which a subject is mentioned

or through a combination of text and visuals. Since visual aspects often resonate differently with

4We include all elections campaigns for which the Dutch media archive has collected and stored the politicalcampaign ads, these are the parliamentary elections of 1981, 1982, 1986, 1989, 1994, 1998, 2002, 2003, 2006 and themost recent one from 2010.

5See the additional reviewer appendix for detailed information about the specific party broadcasts that were in-cluded.

6As there was no archive from which the transcripts of these sources could be retrieved, they were produced bythe researchers before the content analysis was conducted.

7Benoit (2003) makes a further distinction; he classifies the negative appeals in attacks and counterattacks (de-fenses). This approach is ignored here as the difference between attacks and counterattacks is irrelevant for themeasurement of campaign negativity and prevents an endless search for the first attacker.

11

different people, the visuals are not coded in this analysis (for a similar argumentation, see Geer

2006).

In total, 3702 appeals were coded of which over one quarter constituted negative appeals.8

The content analysis is executed by a group of native Dutch speaking graduate political science

students who were trained intensely in several training sessions. Geer’s coding method proved

very reliable in this study. The intercoderreliability was determined on the basis of the coding of

a random sample of ten percent of all appeals. Overall, the intercoderreliability proved very high

and the most difficult coding category was to determine the unit of analysis, i.e. what actually

constitutes an appeal or not, and does a text segment consist of one or multiple appeals. The

Krippendorfs Alpha was 0.77 for the unit of analysis, i.e. is a text segment a appeal or not, and

was 0.92 for the tone of the appeal, i.e. is an appeal positive or negative.

5 Operationalization of Independent Variables and Method

After reviewing the coding of our dependent variable, i.e. a negative versus a positive appeal,

we provide an overview of the operationalization of the main independent variables and the

method employed in the empirical analysis. The variable Challenger Status is a dummy indi-

cating if a party is in opposition at the time of the election campaign or is part of the governing

coalition (1=challenger; 0=incumbent). We operationalize a party’s strategic standing by their

standing in the election polls. That is to say, we create a variable Loss in Polls which measures the

difference (i.e. loss or gain) between the vote share a party would receive according polls prior to

the election and the vote share a party obtained during the previous elections in percent.9 As the

changes in vote percentages, i.e. the losses or gains between the concurrent polls and the pre-

vious elections, will be larger in magnitude for larger parties, we divided the changes by the size

of the party in the previous election. Unfortunately, we lack information about the exact airing

8The overall level of negativity in the Netherlands is not exceptional, but is somewhat lower in comparison withfigures from two party systems like the US. Geer (2006) finds on the basis of US party election broadcasts in theperiod 1960-2000 an overall level of 32 per cent, van Heerde-Hudson (2011) finds on the basis of British party electionbroadcasts in the period 1964-2005 an overall level of approximately 40 per cent. Elmelund-Praestekaer (2010) reportsan overall level of 39 per cent for Danish election in the period 1994-2007, however this level of negativity is based onother sources, coding and calculation method.

9We use the results of the national poll of the main Dutch polling organization TNS/NIPO polling institute at thestart of the election campaign.

12

of the different party broadcasts, so we could not examine the effect of parties’ standing in the

polls in a dynamic fashion. In order to explore the extent to which negative campaigning among

challenger parties is mediated by a party’s competitive standing in the election campaign, we

also include an interaction term. This variable Challenger*Loss allows us to discern if challenger

parties that are lagging behind in the polls are more likely to go negative than opposition parties

that are stable or are actually gaining in the polls.

Party Size is operationalized as the percentage of seats a party would receive according to

the polls four weeks prior to the election. The variable New Party is a dummy variable indicating

that a party did not previously compete in an election campaign (1=new; 0=old). Distance to

Median Party is operationalized as the absolute distance between a party’s left/right placement

subtracted from the left/right placement of the median party within the system. This variable

ranges from a maximum distance of 3.16 to a minimum distance of 0.21 from the median party

position. These party placements are based on voters’ mean placements of parties on a 11-

point left/right scale using Dutch parliamentary election surveys from 1981 through 2010.10 The

variable Government Experience captures the number of government coalitions that a party took

part in during the post-war II period. This variable ranges from 0 government coalitions to 26

government coalitions for the Dutch Christian Democrats in the 2010 election.

We employ a logistic regression model to examine if these different factors account for inter-

party variation in negative campaigning. Our dependent variable, that is to say the tone of an

appeal made by a party, is dichotomous in nature (i.e. negative or positive) so ordinary least

squares regression is ruled out. In addition, we pool data from different election campaigns and

thus have repeated observations for parties in different election campaigns. In order to account

for this clustering in the data, we run our models using White’s heteroskedastic standard errors.

Finally, we added election campaign dummies to account for the possible differences in nega-

tive campaigning across different campaigns. Dutch scholars argue that negative campaigning

entered Dutch elections only very recently (van Praag 2007). We thus expect the sign for election

campaign dummies for the most recent campaigns to be positive.

10To ensure the validity of these measurements, we cross-validated the voters’ left/right placements from recentelections data from expert surveys, namely the Chapel Hill Expert (CHES) data from 2002 and 2006 (Hooghe et al.2010). The mean voter scores from 2002 and 2006 correlated highly with the CHES data: Pearson’s R coefficientsbetween both measures amount to 0.93 and to 0.90 respectively (both significant at the p ≤ 0.01 level).

13

6 Empirical Analysis

Can the different party characteristics based on party’s vote- and office-seeking objectives ac-

count for inter-party variation in negative campaigning? Put differently, do factors like chal-

lenger status or government experience affect parties’ decisions to go negative? Table 1 pro-

vides an overview of the effects of party characteristics specified in hypotheses H1 though H7

on the degree of negative campaigning. We present results for five different models. Model 1,

coined the Vote-Seeking Model, is based on the current US literature and estimates the impact of

vote-seeking incentives on the extent of negative campaigning, namely challenger status, loss in

the polls and interaction term between these two predictors (see hypotheses H1 through H3).11

The second model, labeled the Office-Seeking Model, introduces the predictors derived from the

coalition formation literature which are expected to be especially important within multiparty

competition (see hypotheses H4 through H7). Model 3 shows the results of the Combined Model

accounting for the variation in negative campaigning across parties using both the vote-seeking

and the office-seeking explanations. Finally, models 4 and 5 include party size as a predictor in

the office-seeking model rather than government experience. Due to to high correlations be-

tween party size and government experience, we cannot enter these variables simultaneously.

The high correlations between both predictors are not surprising in the Dutch context which

is characterized by broad-based over-sized coalitions in which large parties are very frequently

part of the government coalition (Andeweg and Irwin 2005). A comparison of the results of these

different specifications in models 2 and 3 versus 4 and 5 reveals that both predictors yield similar

results.

TABLE 1 ABOUT HERE

With the exception of the Losing Challenger Party Hypothesis, we find strong corroboration

for most of our theoretical conjectures. It is important to note that the coefficients provided in

table 1 present the change in the log odds of a party engaging in negative campaigning versus

positive campaigning as a result of a one-unit increase in the independent variable. In other

words, a positive coefficient indicates an increased likelihood of negative campaigning. Re-

11The models with interactions all include the elements that constitute the interaction terms, i.e. the constitutiveterms, in order to guarantee correct estimation of the interaction effects (Brambor, Clark and Golder 2006).

14

viewing the Vote-Seeking Model presented in model 1, we find a significant positive effect for

challenger parties. This is in line with the Challenger Party Hypothesis stating that parties in

opposition are more likely to engage in negative campaigning. In the case of parties lagging be-

hind in the polls, the result is in the expected direction and statistically significant. This suggests

that the relative gain or loss of a party at the beginning of the election campaign compared to

its previous share of parliamentary seats clearly accounts for the inter-party variation in nega-

tive campaigning. This finding is especially important when one considers that post-election

coalition bargaining and broad-based coalition governments make polling results less relevant

for individual parties. Being a winner or a loser in the polls is much less clear-cut in multiparty

competition compared to two-party competition. Even so, we find that losing parties are more

willing to wage a negative campaign to bolster their electoral standing. This being said, the in-

teraction between a challenger party and loss in the polls does have a statistically significant

effect, but is not in the expected direction. The result indicates that the positive influence of

challenger status on the likelihood of negative campaigning does not increase when these par-

ties trail behind in the polls. Thus, while we find limited evidence for the Losing Challenger Party

Hypothesis, there is evidence for the Challenger Party Hypothesis.

Models 2 and 4 present the results for the Office-Seeking Model; model 2 includes a variable

for government experience of a party while model 4 includes a variable for size of a party. These

models provide evidence for the Median Party, Government Experience and Party Size Hypothe-

ses. The results indicate that the likelihood of negative campaigning decreases for larger parties,

for parties close to the median party in terms of left/right positioning, and for parties that have

been part of government coalitions. The New Party variable is not statistically significant, indi-

cating that the newness of a party does not seem to affect a party’s decision to go negative. One

possible explanation for this result is that new parties may lack the financial resources and the

experience to wage a professional campaign (for a similar argumentation, see Abbe et al. 2001).

The results presented in the combined models presented in models 3 and 5 provide evidence

for five out of our seven hypotheses: the Challenger, Losing Party, Party Size, Median Party and

Government Experience Hypotheses. These results indicate that both the vote-seeking and office-

seeking explanations matter for understanding inter-party variation in negative campaigning.

In order to gain a sense of the contribution of the different party characteristics to the overall

15

fit of the combined model (3), we calculated log-likelihood ratio tests which compare the Com-

bined Model including all predictors to a nested Vote-Seeking Model and Office-Seeking Model.

The tests indicate that the overall fit of the complete model is significantly better than the vote-

seeking model (1) and the office-seeking model (2) individually – the improvements in terms

of log-likelihood is 155.47 (p ≤ 0.01) and 49.64 (p ≤ 0.01) respectively. These results provide

support for the utility of including both vote-seeking and office-seeking related variables when

examining variation in negative campaigning across parties.

Note that we corroborated these findings by conducting several robustness checks. First,

we inspected if our results were driven by certain parties or certain campaigns by performing

our analyses while dropping one party or one electoral campaign at a time. This provided sub-

stantively identical results to those reported in table 1. Hence, we are quite confident about the

robustness of the results we present in table 1.

As suggested above, the interpretation of logistic regression coefficients is less than straight-

forward given that they convey changes in log odds. To express the results in a more intuitive

manner, we conduct post-estimation simulations to gain a sense of the effect of each variable in

terms of changes in predicted probabilities. These simulations are performed using CLARIFY,

which employs Monte Carlo simulations to convert raw statistical results into more intuitive

quantities without changing the statistical assumptions (King, Tomz and Wittenberg 2000). Ta-

ble 2 shows the changes in predicted probabilities, i.e. the likelihood of a party to go negative,

under a given set of values for the independent variables. By manipulating the values of each

independent variable while keeping the others constant at their respective means, we are able

to observe changes in the predicted probabilities of a party employing negative attacks. The ta-

ble provides moderate changes – changes from one standard deviation below to one standard

deviation above the mean – as well as the 95 percent confidence intervals.

TABLE 2 ABOUT HERE

The results of the simulations indicate that a challenger party is 21.4 percent more likely

than an incumbent party to use a negative appeal. This finding supports the idea that chal-

lenger parties naturally have to oppose incumbents and their policies in order to reap electoral

benefits. These parties must make it clear to voters why they should support them instead of the

16

incumbent party this time around. The loss in the polls has an even stronger effect compared to

challenger status, namely 36.2 percent. These results provide evidence for the idea that losing

parties have an incentive to engage in negative campaigning in order to discredit competitors

and sway public opinion their way. The changes for distance to the median party government

experience and party size are slightly smaller in size, the likelihood of negative campaigning

changes by about five percent. These findings are in line with our theoretical predictions sug-

gesting that parties with high levels of coalition potential, i.e. large parties and parties that are

frequently part of government coalitions, are less likely to wage a negative campaign as its use

may jeopardize their future involvement in government coalition bargaining. When parties are

further removed from the median party in the party system, however, the likelihood of negative

campaigning increases. The change in the predicted probabilities of negative campaigning for

new party status is very small and statistically insignificant.

7 Discussion and Conclusion

This study expands our knowledge regarding inter-party variation in negative campaigning and

broadens existing theoretical models to include features of multiparty competition. It provides

an important stepping stone for the study of negative campaigning outside the Anglo-Saxon

context of two-party systems by examining variation in negative campaigning across parties in

a multiparty system par excellence, the Netherlands, over almost three decades. The study ad-

vances existing research in two ways. First, it extends the current state-of-the-art that assumes

parties to be primarily vote-seeking to include factors that relate to parties’ office-seeking behav-

ior in an explanation of inter-party variation in negative campaigning. We argue that the chief

difference between negative campaigning in two-party compared to multiparty competition is

that campaign strategies in a multiparty context are always crafted in the wake of post-election

coalition bargaining. Consequently, negative campaigning constitutes a more perilous strategy

within a multiparty system as it potentially jeopardizes a party’s credibility in government coali-

tion bargaining. Parties that have low coalition potential are able to take considerable electoral

and coalition bargaining risks will make more use of negative campaigning. Second, it is ex-

pands existing empirical work on negative campaigning in multiparty contexts while exploring

17

cross-party variation in negative campaigning (next to recent valuable studies exploring Dan-

ish campaigns, see Elmelund-Præstekær 2008, 2010; Hansen and Pedersen 2008). By examining

Dutch election campaigns with a almost 30 year time frame, this study validates the Danish

results as it also finds that negative campaigning is used more by challenger parties (Elmelund-

Præstekær 2010). In addition, the new hypotheses based on office-seeking objectives are tested.

The empirical results indicate that a combined model including vote-seeking as well as office-

seeking related factors provides the most leverage for an explanation of inter-party variation in

negative campaigning. Negative campaigning is most extensive among challengers, small par-

ties, parties trailing behind in the polls or on the outskirts of the political system, and parties

that have not frequently been part of government coalitions. We do not find evidence for the

idea that new parties are more likely to employ negative campaigning. This result may be due

to the fact that new parties lack sufficient financial resources and experience to wage a profes-

sionally run campaign. Our evidence regarding a party’s strategic standing in the polls is slightly

more mixed. Although we do find a direct effect of loss in the polls in a multiparty setting which

is in line with current findings from the US, we do not find evidence for an indirect effect. The

level of negative campaigning for challenger parties that are lagging behind in the polls is not

different from challenger parties that are gaining in the polls. This finding could be explained

through the fact that challenger status in itself already makes parties wage negative campaigns

as these parties lack the natural advantage of holding office and thus find themselves in a posi-

tion where they need to convince voters not to cast their ballot for incumbent parties.

Building on the results of this study, we can outline three avenues for further research. First,

future work on inter-party variation in negative campaigning within multiparty systems should

provide a more in-depth look into the importance of office-seeking goals in constraining the

use of negative campaigning in multiparty systems. For example, one could think of conducting

in-depth interviews with campaign managers and party officials in order to explore further if

indeed post-election coalition bargaining serves as a natural ceiling for the use of negative cam-

paigning in multiparty competition. Second, additional avenues for further research include the

analysis of other campaign materials, such as internet clips or websites (see for example Druck-

man, Kifer and Parkin 2009). This would allow a for an in-depth comparison of negative cam-

paigning across a variety of media outlets. Finally, it would be interesting to study the impact

18

of party system characteristics, such as the level of fragmentation or the degree of proportional-

ity, on the extent of negative campaigning. This would require data collection on the degree of

negative campaigning across a large set of countries and elections which of course is both rather

costly and time-consuming.

Overall, this study provides valuable insights into negative campaigning within multiparty

competition. Although the findings show that negative campaigning in a multiparty system

shares many features similar to those found in the US context, it highlights one vital difference,

namely the impact of post-election coalition bargaining. Therefore, this study breaks theoretical

as well as empirical ground in the study of negative campaigning. It demonstrates that negative

campaigning within a multiparty system constitutes a riskier strategy compared to a two-party

context, as parties operating in a system with multiple rivals face both electoral backlash effects

and post-election coalition bargaining costs. As a result, negative campaigning in multiparty

competition is the prerogative of parties willing to take considerable electoral and coalition bar-

gaining risks.

19

References

Abbe, Owen G., Paul S. Herrnson, David B. Magelby and Kelly D. Patterson. 2001. Are Profes-

sional Campaigns More Negative? In Playing Hardball: Campaigning For The U.S. Congress,

ed. Paul S. Herrnson. Upper Saddle River: Prentice Hall, 70-91.

Andeweg, Rudy B. and Galen A. Irwin. 2005. Governance and Politics of the Netherlands. Hound-

mills: Palgrave Macmillan.

Anker, Hans. 1998. Een Bijzondere Verkiezingscampagne? In Stemmen in Stromenland: De

Verkiezingen van 1998 Nader Bekeken, ed. Peter Kramer, Tom van der Maas and Leonard Orn-

stein. Den Haag: SDU Uitgevers, 114-127.

Bartolini, Stefano. 1998. Coalition Potential and Governmental Power. In Comparing Party Sys-

tem Change, ed. Paul Pennings and Jan Erik Lane. London: Routledge, 40-61.

Benoit, William. L. 1999. Seeing Spots: A Functional Analysis of Presidential Television Advertise-

ments, 1952-1996. London: Praeger.

Benoit, William. L., John. P. McHale, Glenn. J. Hansen, P. M. Pier and John. P. McGuire. 2003.

Campaign 2002: A Functional Analysis of Presidential Campaign Discourse. Lanham: Rowman

and Littlefield.

Brambor, Thomas, William Roberts Clark and Matt Golder. 2006. “Understanding Interaction

Models: Improving Empirical Analyses.” Political Analysis 14:63–82.

Brants, Kees, Walter Kok and Philip van Praag. 1982. De Strijd om de Kiezersgunst, Verkiez-

ingscampagnes in Nederland. Amsterdam: Kobra.

Budesheim, Thomas L., David A. Houston and Stephen J. DePaola. 1996. “Persuasiveness of

In-Group and Out-Group Political Messages: The Case of Negative Political Campaigning.”

Journal of Personality and Social Psychology 70:523–534.

Budge, Ian and Dennis J. Farlie. 1983. Explaining and Predicting Elections: Issue Effects and Party

Strategies in Twenty-Three Democracies. London: Allen and Urwin.

20

Damore, David F. 2002. “Candidate Strategy and the Decision to Go Negative.” Political Research

Quarterly 55:669–685.

Davis, Michael L. and Michael Ferrantino. 1996. “Towards a Positive Theory of Political Rhetoric:

Why Do Politicians Lie?” Public Choice 33:1–33.

Djupe, Paul A. and David. A. M. Peterson. 2002. “The Impact of Negative Campaigning: Evidence

from the 1998 Senatorial Primaries.” Political Research Quarterly 55:845–860.

Druckman, James N., Martin J. Kifer and Michael Parkin. 2009. “Campaign Communications in

U.S. Congressional Elections.” American Political Science Review 103:343–366.

Elmelund-Præstekær, Christian. 2008. “Negative Campaigning in a Multiparty System.” Repre-

sentation 44:27–39.

Elmelund-Præstekær, Christian. 2010. “Beyond American Negativity: Toward a General Under-

standing of the Determinants of Negative Campaigning.” European Political Science Review

2:137–156.

Gabel, Matthew. J. and John Huber. 2000. “Putting Parties in their Place.” American Journal of

Political Science 44:94–103.

Geer, John G. 2006. In Defense of Negativity: Attack Ads in Presidential Campaigns. Chicago:

Chicago University Press.

Hale, John F., Jeffrey C. Fox and Rick Farmer. 1996. “Negative Advertisements in U. S. Senate

Campaigns: The Influence of Campaign Context.” Social Science Quarterly 77:229–343.

Hansen, Kasper. M. and Rasmus T. Pedersen. 2008. “Negative Campaigning in a Multiparty Sys-

tem.” Scandinavian Political Studies 31:229–343.

Haynes, Audrey and Staci L. Rhine. 1998. “Attack Politics in Presidential Nomination Campaigns:

An Examination of the Frequency and Determinants of Intermediated Negative Messages

against Opponents.” Political Research Quarterly 51:691–721.

Hobolt, Sara B. and Jeffrey Karp. 2010. “Voters and Coalition Governments.” Electoral Studies

29:299–307.

21

Hooghe, Liesbet, Ryan Bakker, Anna Brigevich, Catherine E. de Vries, Erica E. Edwards, Gary

Marks, Jan Rovny, Marco R. Steenbergen and Milada A. Vachudova. 2010. “Reliability and

Validity of Measuring Party Positions: The Chapel Hill Expert Surveys of 2002 and 2006.” Euro-

pean Journal of Political Research 49:687–703.

Jamieson, Kathleen Hall. 1992. Dirty Politics: Deception, Distraction and Democracy. Oxford:

Oxford University Press.

Johnson-Cartee, Karen S., Gary Copeland and Eric E. Johnson. 1991. Negative Political Advertis-

ing: Coming of Age. London: Lawrence Erlbaum.

Kahn, Kim F. and Patrick J. Kenney. 2004. When do Candidates Go Negative? In No Holds Barred,

Negativity in U. S. Senate Campaigns, ed. Lynda L. Kaid and Christina Holtz-Bacha. Upper

Saddle River: Pearson, 19-37.

Kaid, Lynda L. and Christina Holtz-Bacha. 1995. Political Advertising in Western Democracies.

London: Sage.

Kaid, Lynda L. and Christina Holtz-Bacha. 2006. The Sage Handbook of Political Advertising.

London: Sage.

King, Gary, Michael Tomz and Jason Wittenberg. 2000. “Making Most of Statistical Analyses:

Improving Interpretation and Presentation.” American Journal of Political Science 44:341–355.

Lau, Richard R. and Gerald M. Pomper. 2004. Negative Campaigning: An Analysis of U.S. Senate

Elections. Lanham: Rowman and Littlefield.

Lau, Richard R., Lee Sigelman, Caroline Heldman and Paul Babbitt. 1999. “The Effects of Neg-

ative Political Advertisements: A Meta-Analytical Review.” American Political Science Review

93:851–875.

Lau, Richard R., Lee Sigelman and Ivy B. Rovner. 2007. “The Effects of Negative Political Cam-

paigns: A Meta-Analytic Reassessment.” Journal of Politics 69:1167–1209.

Laver, Michael and Norman Schofield. 1998. Multiparty Government: The Politics of Coalition in

Europe. Ann Arbor: Michigan University Press.

22

Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Performance in Thirty-Six

Countries. New Haven: Yale University Press.

Mark, David. 2006. Going Dirty: The Art of Negative Campaigning. Lanham: Rowman and Little-

field.

Mayer, William G. 1996. “In Defense of Negative Campaigning.” Political Science Quarterly

111:437–455.

Mediawet. 2008. “Wet van 29 December 2008 tot Vaststelling van een Nieuwe Mediawet.” Staats-

blad van het Koninkrijk der Nederlanden 548:1–79.

Müller, Wolfgang and Kaare Strøm. 1999. Policy, Office or Votes?, How Political Parties in Western

Europe Make Hard Decisions. Cambridge: Cambridge University Press.

Peterson, David A. M. and Paul A. Djupe. 2005. “When Primary Campaigns Go Negative: The

Determinants of Campaign Negativity.” Political Research Quarterly 58:45–54.

Pierce, Roy. 1999. Mass-Elite Issue Linkages and the Responsible Party Model of Representa-

tion. In Policy Representation in Western Democracies, ed. Warren Miller, Roy Pierce, Jacques

Thomassen, Richard Herrera, Soren Holmberg, Peter Esaiasson and Bernhard Wessels. Ox-

ford: Oxford University Press, 9-32.

Sanders, David and Pippa Norris. 2005. “The Impact of Political Advertising in the UK General

Election.” Political Research Quarterly 58:525–536.

Sartori, Giovanni. 2005. Parties and Party Systems: A Framework for Analysis. Cambridge: Cam-

bridge University Press.

Sigelman, Lee and Emmett H. Buell. 2003. “You Take the High Road and I’ll Take the Low Road?

The Interplay of Attack Strategies and Tactics in Presidential Campaigns.” Journal of Politics

65:518–531.

Sigelman, Lee and Eric Shiraev. 2002. “The Rational Attacker in Russia? Negative Campaigning

in Russian Presidential Elections.” Journal of Politics 64:45–62.

23

Skaperdas, Stergios and Bernhard Grofman. 1995. “Modeling Negative Campaigning.” American

Political Science Review 89:49–61.

Swint, Kerwin C. 1998. Political Consultants and Negative Campaigning: The Secrets of the Pros.

Lanham: University Press of America.

Theilmann, John and Allen Wilhite. 1998. “Campaign Tactics and the Decision to Attack.” Journal

of Politics 60:1050–1062.

van der Brug, Wouter and Cees van der Eijk. 2005. Welke Effecten Hadden de Campagnes Nu

Eigenlijk. In Politiek en Media in Verwarring: De Verkiezingscampagne in het Lange Jaar 2002,

ed. Kees Brants and Philip van Praag. Amsterdam: Het Spinhuis, 244-267.

van Heerde-Hudson, Jennifer. 2011. “The Americanization of British Party Advertising? Negativ-

ity in Party Election Broadcasts, 1964-2005.” British Politics 6:52–77.

van Praag, Philip. 2005. De Veranderende Nederlandse Campagnecultuur. In Politiek en Media

in Verwarring: De Verkiezingscampagne in het Lange Jaar 2002, ed. Kees Brants and Philip van

Praag. Amsterdam: Het Spinhuis, 21-43.

van Praag, Philip. 2007. De Verkiezingscampagne: Professioneler en Sneller. In Een Verdeeld

Electoraat: De Tweede Kamerverkiezingen van 2006, ed. Kees Aarts, Henk van der Kolk and

Martin Rosema. Utrecht: Het Spectrum, 97-119.

van Spanje, Joost and Wouter van der Brug. 2007. “The Party as Pariah: Ostracism of Anti-

Immigration Parties and its Effect on their Ideological Positions.” West European Politics

30:1022–1040.

Warwick, Paul V. 1996. “Coalition Government Membership in West European Parliamentary

Democracies.” British Journal of Political Science 26:471–499.

Westen, Drew. 2007. The Political Brain: The Role of Emotion in Deciding the Fate of the Nation.

New York: Public Affairs.

24

8 Tables and Figures

Table 1: The Effects of Party Characteristics on Negative CampaigningModel 1 Model 2 Model 3 Model 4 Model 5

Vote-Seeking Office-Seeking Combined Office-Seeking Combined[Experience] [Experience] [Party Size] [Party Size]

Challenger Party 2.08** 1.82** 2.26**(0.14) (0.16) (0.16)

Loss in Polls 1.22** 1.13** 1.45**(0.26) (0.26) (0.28)

Challenger*Loss -1.14** -1.07** -1.39**(0.26) (0.26) (0.28)

New Party 0.17 0.10 0.18 0.20(0.14) (0.15) (0.13) (0.15)

Distance Median Party 0.01** 0.01** 0.01** 0.01**(0.00) (0.00) (0.00) (0.00)

Government Experience -0.09** -0.02*(0.01) (0.01)

Party Size -0.03** -0.02**(0.00) (0.00)

Constant -2.83** -0.86** -2.73** -0.98** -3.33**(0.19) (0.15) (0.22) (0.16) (0.25)

LogLikelihood -1908.04 -1960.96 -1883.22 -2032.01 -1878.41Pseudo R2 0.11 0.09 0.12 0.05 0.12CorrectlyClassified 75.45 74.28 74.28 74.20 73.99N 3702 3702 3702 3702 3702

Notes: Table entries are logistic regression coefficients with robust standard errors in parenthe-ses. All models also include election year dummies which were suppressed in the table due tospace limitations (these results are available upon request). ** significant at p ≤ 0.01; * signifi-cant at p ≤ 0.05 (one-tailed).

25

Table 2: Changes in Predicted ProbabilitiesChange 95 % CI

Challenger Party 21.4** [ 17.1 ; 25.4]Loss in Polls 36.2* [22.2 ; 49.5]New Party 0.5 [-3.1 ; 4.5]Distance Median Party 5.7** [2.0 ; 9.6]Government Experience -3.9* [-1.1 ; -8.8]Party Size -6.6** [-0.6 ; -11.4]

Notes: Table entries in the second column are changes in predicted probabilities when the re-spective independent variable is moved from one standard deviation below to one standard de-viation above the mean (ceteris paribus) and 95 percent confidence intervals in the third col-umn. These changes are based on model 3 and in the case of party size on model 5 reported intable 1. ** significant at p ≤ 0.01; * significant at p ≤ 0.05 (one-tailed)

26

IntroductionNegative Campaigning and Multiparty CompetitionThe Dutch CaseData and Coding ProcedureOperationalization of Independent Variables and MethodEmpirical AnalysisDiscussion and ConclusionTables and Figures