UvA-DARE

There are many assumptions about the contents of media coverage of the far-right and its possible effects. Yet, these assumptions lack empirical arguments. In this contribution, the authors provide an answer to the important question how media cover the far-right. They do so from a longitudinal perspective, using systematic quantitative content analysis of three Dutch newspapers. The assumption that Dutch newspapers follow a ‘minimal attention policy’ does not hold. Throughout the research period newspaper coverage of the far-right becomes increasingly versatile and although concentrating on negative aspects of the far-right, increasingly open to positive support for the far-right. In the last three years of the research period, the majority of the coverage concerns Fortuyn’s populist right party LPF. Whereas at first the coverage of this party fits in the general picture of coverage on the far-right, towards the end of the research period the coverage of LPF is increasingly diverse and less concentrated on negative aspects.


Introduction and research question
"U bent de schuldige. De pers heeft ons vier jaar lang doodgezwegen. Hoe kon de kiezer dan van ons bestaan weten?" 1 (It's your fault. The press ignored us for four years. How could voters even know that we exist?) This is how Hans Janmaat, leader of the far-right party Centrum Democraten, explained the 1998 elections results, when the party lost all three of their seats in parliament. He then aggressively pushed the reporter out of the room, stating "zo gaan we voortaan met de pers om!"(this is how we deal with the press from now on!). Besides the emotion and bitterness, this analysis is not exceptional. It actually is a widely accepted hypothesis among right and left wing politicians, journalists, social scientists and other stakeholders that there is a relationship between media attention and (the lack of) far-right electoral success 2 . Strangely enough, media attention to the far-right has hardly been investigated empirically.
Despite the volume of research on success factors of the far-right (i.e. Lubbers and Scheepers 2001, Ignazi growth of populist/far-right is followed by a period of electoral decline from the last quarter of the year until 2004, when a period of stabilisation seems to embark.

Media coverage of the far-right
In order to provide an answer to our research question we must first investigate which aspects of the newspaper coverage are relevant. For that, clues can be found in sociological and political science literature.
In this section we critically investigate the value of these clues, which will lead to a more specific elaboration of our research question: how do Dutch newspapers cover the far-right?

The level of media attention
In his study on far-right voting, Lubbers expected to find a relationship between (a) the amount of articles on the far-right and on their main target in the nineties: asylum seekers and (b) far-right voting (Lubbers 2001: 166, Lubbers & Scheepers 2001. An increase in the amount of coverage (on both indicators) was supposed to positively relate to the relative amount of far-right voters. These hypotheses have been tested for two Dutch and one German newspaper(s). The outcomes of Lubbers' study show opposite effects on both variables in the two countries. In the German case, an increase in media attention to the far-right was related to an increase of far-right voting, as predicted. But an increase of the amount of articles on the far-right in the Netherlands was related to a decrease of far-right votes. While the frequency of articles on asylum seekers in the Netherlands was related to an increase in far-right votes, in Germany it had a negative effect (Lubbers: 2001: 193, Lubbers & Scheepers 2001. The study of Lubbers and Scheepers only focused on the number of articles rather than the contents of these articles. Hence it fails to consider the contents that may help to solve the puzzle of the contradictory findings. The increasing amount of articles in Dutch newspapers on the moment of decreasing support for the parties CD and CP'86 in April and May 1994 may well be explained by a negative content of news coverage 3 . That would be a reversal of Lubbers' hypothesis: "the higher the media attention to extreme right wing parties, the larger the likelihood of voting for an extreme right-wing party " (2001: 166). In this light, the contrasting results of Lubbers' two case studies form a puzzle that can only be solved by a study of the contents of the coverage on the far-right. The question of the amount of media attention to the farright must be extended by the question of how newspapers cover the far-right.

The contents of media attention: far-right actors and far-right ideology
In one of the scarce studies on the relationship between the media and the far-right, Van Donselaar characterises this relationship as "hostile" ( Van Donselaar 1998: 39). In the documentation on the issue, the very aggressive quotes of far-right leaders about journalists are easy to find 4 , but Van Donselaar also states that the media were very harsh on the far-right during the election campaigns of 1994 (ibid: 41).
Multiple under cover journalists published shocking articles and books on Centrum Democraten enforcing an "image of bandits and Nazism" (ibid: 41, see also Rensen 1994). Based on an interview study and media output, Van Donselaar proposed an interaction model of the media and the far-right, that includes three media strategies: 'systematically ignoring' the far-right, 'stigmatising' the far-right and 'judging consequences of far-right ideology' 5 ( Van Donselaar 1998: 46). The first strategy is what is also called the 'minimal attention policy'. It suggests that for our purpose we must look at formal aspects of attention, meaning size and position of coverage. The assumed existence of the minimal attention policy suggests that if media report on the far-right they will keep it to a minimal size and inconspicuous place in the newspaper. Stigmatising the far-right suggests it is important to consider possible different roles the farright is featured in news articles. Do these enable stigmatising or not? The third strategy suggests analysis of the presence of far-right ideology in the coverage, which is confirmed by Walgrave and De Swert who found a relation between the level of attention to typical far-right issues in the media and far-right voting percentages (Walgrave & De Swert 2004: 495).
Journalists are mostly reserved about covering the far-right, while the far-right seeks ways to make it into the press ( Van Donselaar 1998: 77-78). Although Van Donselaar did not research the coverage systematically, his interaction model does point to the tension between the news value of the far-right and the fact that many journalists are cautious to overexposing the far-right because of possible electoral effects.
From these considerations we derive that a study of the contents of media coverage of the far-right must focus on four aspects of media attention:  For substantial attention we distinguish the passive or active representation of far-right actors, the situational role in which they are represented and the presence of far-right ideological standpoints.

• Substantial attention:
o To what level is the far-right represented passively (mentioned) and actively (paraphrased and quoted)?
o In which situations does the far-right play a role?
o Which far-right ideological issues are represented in the newspaper coverage?

The content of media attention: support of stakeholders
For both voters' opinions on political issues and their evaluation of political parties, the confirmation, denial or (possible) legitimation of these opinions by third parties is proven important. In research on far-right violence, it has become clear that the direction and amount of this violence, depends on the visibility, resonance and legitimation, as "mechanisms of positive and negative feedback processes (Koopmans & Olzak 2004: 223)" in the media. Also, the reactions of politicians to right wing violence depend on who the violence is directed to and how this is reported in the media (Koopmans 2001). Important stakeholders, of whom this feedback descends, are politicians, journalists and experts whose message is almost exclusively accessible through the media. Research on far-right voting shows that far-right parties must be evaluated as 'normal democratic parties' by voters, in order to increase the chance of being considered to be voted for (Van der Brug, Fennema & Tillie, 2005: 546, 561).
These outcomes point out that it matters how non-far-right stakeholders (such as other politicians, journalists, experts and other voters) relate to the far-right or far-right standpoints. The primal source of information on these attitudes for voters is the media coverage of the far-right. Therefore we analyse a third aspect of media coverage of the far-right: support attention. This means we will analyse the proportion of coverage containing positive or negative support from stakeholders for the far-right, or for far-right ideological standpoints. Attention to support is measured in the news coverage here, although support attention can also be found in editorials, opinions and letters to the editors. First the relative amount of coverage on the far-right that contains support of any kind at all provides an answer to the question, if the far-right is portrayed in the coverage as important enough to deal with at all or not. Then the balance between positive and negative support gives insight in how much the far-right is presented as a 'normal democratic party'. A high level of negative support implies an emphasis on the abnormal position of the far-right, while a greater share of positive support points to a more normal representation in the coverage. This leads to two research questions: • Support attention: o What part of the newspaper coverage contains non-far-right actors' attitudes towards the far-right?
o How do non-far-right actors relate to the far-right actors and to far-right ideological standpoints in the newspaper coverage?

Longitudinal variation
As the electoral history of the Dutch far-right is full of ups-and-downs and contains multiple key events, some variation in media attention can be expected. Therefore the second question we answer in this chapter is whether there were any changes in the media attention to the Dutch far-right between 1986 and 2004? Our main interest in the longitudinal variation therefore, is whether this variation developed in a particular direction.

Quality and popular newspapers
Drawing from the work of Steward, Mazzoleni & Horsfield (2003) we postulate three hypotheses about quality and popular newspapers and test them, looking at differences between the popular De Telegraaf and quality papers de Volkskrant and NRC Handelsblad. A general difference between quality and popular newspapers revealed in various studies is that quality newspapers pay more attention to politics and political parties (Scholten & Ruigrok, 2006: 10, 24;Kleinnijenhuis et al., 2007: 98). It would be logical to expect that this also applies to specific political parties, in this case far-right political parties. Therefore we expect the following: Telegraaf.

Left-leaning and right-leaning newspapers
Although the period of polarisation in Dutch society is over, newspapers still carry traces of political  (Hallin & Mancini, 2004: 180). We therefore expect to find differences between the left-leaning Volkskrant and right-leaning NRC Handelsblad and Telegraaf: H 4 de Volkskrant paid more attention to volume and formal attention than NRC Handelsblad and de Telegraaf.
H 5 de Volkskrant paid more substantial attention to the far-right than NRC Handelsblad and de Telegraaf.
H 6 de Volkskrant paid more support attention to the far-right than NRC Handelsblad and de Telegraaf.

Data
Three Dutch national daily newspapers were used for this study. De Volkskrant (progressive/liberal), NRC Handelsblad (conservative/liberal) and De Telegraaf (popular). (Ward 2005: 126). These newspapers were selected because they all belong to the most widespread newspaper titles in the country, each more or less cater to a distinct readership (Kaiser 1994: 128). The research period reaches from 1986 to 2004. The year 1986 was chosen as a starting point because that year roughly marks the point in several countries where the far-right was in a similar position of low voting percentages and started to grow (again) (Mudde 1998: 49, 132, 189). The period reaches until 2004 because that is the nearest election year before the start of this study and it includes interesting new developments on the far-right side of the political spectrum with the follow-up of the 'old' far-right parties with new 'populist right wing' parties (see Van der Brug & Fennema 2003, Van Praag 2003. We choose to delineate our research by concentration on election periods, defined as two months preceding national or European parliamentary elections and the month directly after the elections. In this way, the corpus is considered to cover the entire campaign period as well as the reactions to and effects of the electoral results. This total period was expected to contain the most dense coverage of the far-right, and is the most likely to include changes in the newspaper coverage of the far-right because of changes in the political situation or relevant events that are most likely to occur around elections 6 . Drawing the samples from complete dense periods also avoids the bias of fragmentation, that is may result from aselect drawn samples (Wester & van Selm 2006: 128).
Within these set periods we collected all news and background articles containing at least one far-right actor or the far-right as a general term. The electronic data source Lexis Nexis Academic (LNA) as well as Micro Fiche archives of Radboud University and the National Library were used for data collection. All relevant articles were drawn from theses sources using a key word list and data collection protocol developed previously for this purpose, taking historical, sociological and political science reviews into account (Schafraad, Wester & Scheepers 2006). The use of this instrument allows for a structured, relevant, reproducible collection of all relevant material.
In principle, we took the complete sub-universes of each paper in each election year as our corpus, except in those cases where the size of the universe allowed drawing a random sample. In those cases, we used Morris' formula for small population sampling to draw a random sample 7 . A total of 1932 articles featuring at least one far-right actor where coded. After data collection, it turned out that the EP elections of 1989, 1994, 1999 were too small to be considered individually. The 1999 case (n = 21) was excluded from the analysis and the 1989 and 1994 EP data were integrated in the national parliament election case of the same year. This yields seven election years between 1986 -2004 on which we collected and analysed newspaper articles.

Measurement
In section 3 we have argued the importance of including certain aspects of the contents of newspaper coverage of the far-right in the research. Here we explain how these aspects were operationalised in our instrument.
The two aspects of formal attention were measured by coding (1) the size of articles in three categories and (2) whether the article featured on the front page or not.
The first aspect of substantial attention, the question how far-right actors are represented in the coverage, has been operationalised as follows: (1)  Six well trained coders and the first author coded 1281 newspaper articles. Reliability of the coding was calculated twice, shortly after the coders received training and after the coding work was all done. After the first calculations the coders received extra training, to deal with complicating issues found in the data. A three-fold reliability calculation was used, including percentage of agreement and two different calculations of Scott's π (Scott 1954, Wester & van Selm 2006. This procedure was chosen because many of the crosstabs had a strong deviation of a normal distribution, while the formulas assume a more or less normal distribution, which is an issue, although often occurring, content analysists have not yet developed a proper solution to, yet (Neuendorf 2002: 151). By applying our three-fold procedure, however, we argue to have dealt with each of the weaknesses of the individual techniques 8 .
Ten percent of the data was double coded. And reliability was calculated for each variable. Most variables had a sufficient (%agreement >.9, π between .67 and .80), or good (π > .80) reliability score on all three measures. The variables 'far-right actor named', 'far-right actor paraphrased', 'role in elections', 'role in 8 Formulas reliability: Percent agreement: Total number of agreements/n = pa Scott's pi: (total number of agreements) -(expected number of agreements) Scott 1955, Neuendorf 2002: 154, Wester & van Selm 2006 For a full report on the evaluation of the reliability figures readers can contact the first author. political actuality' and 'non-far-right actors with a negative attitude towards the far-right' had an insufficient score on one or more of the measures. This means part of the variation found for these variables in the data may be caused by coding errors and must be considered with greater care. 9

Volume attention
Volume attention was measured as the average number of articles per day and was calculated as follows.
For each election year we calculated the number of news days based on 13 week periods, except for 1989 and 1994 (18 and 19 weeks). In cases where we used random samples for further analyses, the actual number of news articles in the population had to be estimated, based on the proportions in the sample.
The actual number of news articles was then divided by the number of days. A T-test was used to test the significance of the mean differences between election years and between newspapers at .05 level.
Overall, the Dutch newspapers printed one article per day that at least mentions the far-right in one way or another. Throughout the research period the volume attention fluctuated quite considerably. In some of the electorally more successful years it was above average, but in other 'success' years it decreased greatly, 9 All calculated reliability levels: *) these levls have a negative bias due to the abnormal distribution in the cross-tabs and were therefore not considered in the evaluation of the reliability of the variable (>.25 difference between two π calculations). **) the absolute disagreement for 'role internal negative' was 1. π values ought to be between 0 and 1 (Scott 1955). For other variables marked with ** similar distributions were found. These levels were not considered in the evaluation of the reliability of the variable.
compared to the previous year, as in 1989. In 1999 there was so little media attention to the far-right that we had to exclude that year from further analysis. *) the differences between years and between newspapers have been tested, p < 0,05. The differences discussed in the text are all significant. **) TK + EP is 18 (105 news days in 1989) or 19 (111 news days in 1994) weeks instead of 13. ***) In further analysis the election year 1999 is excluded because of its sample being too small.
There are significant differences between newspapers, especially between de Volkskrant and NRC Handelsblad on one side, and the popular paper De Telegraaf on the other, which is not surprising since the latter has less news content in general (Scholten & Ruigrok 2006: 8). Its average number of articles on the far-right is only half of the average number in de Volkskrant and NRC Handelsblad.

Formal attention
Visibility is the central idea behind the formal aspect of attention to the far-right. Formal attention was measured as the size of articles, front page appearance, headline appearance and prominence of the farright in articles.

Size of articles
The second variable measuring formal attention is the size of the article. Well over 50% of the newspaper coverage on the far-right are longer articles (400 or more words), while 14% are short reports (less than 150 words) ( Table 2). This means the far-right is subject not only to 'short news' type of sections in newspapers. In the eighties the percentage of long articles was less than 40%, but from 1998 on it rose rapidly to sometimes more than 60% (61% in 1998 and 63% in 2004). The trend of the share of short reports has a reversed tapering, going down from 23% in 1986 to a mere 6% in 2004. As Appendix 1 shows, de Volkskrant and NRC Handelsblad both printed significantly more long articles on the far-right than De Telegraaf (56% and 58% vs 37%). De Telegraaf printed mostly medium length articles on the far-right, at 45% that was often 15% more than both other newspapers. It also printed slightly more short articles on the far-right (Tel 18%, vs VK 13% and NRC 12%).

Position of the article in the newspaper
The next aspect of formal attention is the position of the article in the newspaper, measured as the percentage of the coverage of the far-right published on the front page (see Fig. 1). Only 11% of the coverage of the far-right can be found on the front page of newspapers, a figure that was fairly stable over the years. In the nineties it was slightly less (8% in 1994, 9% in 1998, while in 1989 it was at its highest with 17%). This did not differ between newspapers 10 .

Prominence of the far-right in the coverage
The fourth aspect of formal attention that has been measured is the prominence of the far-right in the coverage. Figure 1 below shows that until 1994, in over two-thirds of the coverage that includes far-right actors, the far-right had a prominent role. When the far-right was in the papers, it was usually a central

Far-right in the headlines
The last aspect of formal attention is the percentage of the coverage mentioning the far-right in the headlines, which means it immediately draws the attention of the reader to the political colour of a central actor in the coverage. This variable works chiefly as an additional variable to the previous one. Figure 1 shows, as expected, similar trends for prominence and far-right as headline news, although the percentage of coverage with the far-right in the headlines was higher than that of far-right prominence in 1986 and 1989. Until the mid-nineties over two-third of the coverage featured the far-right in the headlines; after this the share dropped to 17% in 2004.
In de Volkskrant articles on the far-right had the far-right in their headlines more often (54%) than both other newspapers (Tel 44%, NRC 40%).

Substantial attention
Next the substantial attention in the coverage was considered. With substantial attention the focus lies on far-right actors and far-right ideologies. Substantial attention has been operationalised via four dimensions.
First the representation of the far-right actor(s), second the role in which the far-right actor(s) feature, third explicit far-right stigmatising ( Van Donselaar, 1998: 52) and fourth the presence of far-right ideological standpoints in the coverage.

Representation of far-right actors
Far-right representation is the first aspect of substantial attention that has been measured.  The data were collected in an extended period around Election Day. Therefore an overview is given of the percentage of coverage wherein the far-right had a role in the elections (possibly besides other roles) in Table 3, before we describe the attention to situated roles in which the far-right featured. As we expected, a relatively large share of the coverage deals with the elections, or is in fact election news. 57% of all coverage is about the far-right in elections. The percentage varies over the years between 30% in 2004 and 82% in 1989. Generally, roles in elections gained relatively most attention in de Volkskrant (61%) and least in NRC Handelsblad (53%). Attention to these roles in De Telegraaf is 57% (see Appendix 4).
Other roles of the far-right, as featured in the newspaper coverage, have been analysed as well. When we look at Table 4 it is interesting to note that the coverage dealt as much with far-right-actors in legal conflict (15%) and extra-parliamentary roles (13%) as in 'normal' (parliamentary) roles in political actuality (18%).
Only 4% of the coverage contained the far-right in normal internal roles (meaning party conferences, leadership elections, or other party events), while negative internal roles (internal dispute, quarrel, party splits) are present in 10% of the coverage.
In years when there was far-right representation in parliament (especially 1994, 2003-2004), the share of articles on the far-right in political actuality, as expected, was greater. In 1994 when Centrumdemocraten obtained three seats in parliament (and 77 local council seats) there was an increase to 13%, and  Table 4 shows that until the late nineties the far-right featured relatively frequently in legal conflicts (18% in 1998, 35% in 1986). In the 2000s that figure dropped below 10%. A similar pattern can be found for the far-right in extra-parliamentary roles (rallies, petitions, but also racist violence). It was at its peak in 1986, with 41%, concerning a great deal of confrontation in the form of anti-fascist rallies against the newly chosen Centrum Partij council members in various cities.
Attention to far-right actors in normal internal situations was continuously about 5% of the coverage, with a dip to almost nothing in 1998 (1%) and 2002 (3%). Newspapers paid much more attention to negative internal situations, peaking at 21% of the coverage in 1994. Here too, there was a divide between the twentieth and twenty-first centuries, with an exception for 2003 (those early elections were held after the coalition break-up due to internal conflicts in LPF). The decrease of the attention to internal conflicts started in 1998, the first election year after the peak year of 1994 (dropped from 21% to 7%). In normal internal situations 6 5 4 1 3 5 2 4 In negative internal situations 14 12 21 7 6 9 2 10 Some of the roles of the far-right variables are handled differently by each of the three newspapers in one or more years (see appendix 4). In the three newspapers roles in extra-parliamentary activity and in internal affairs did not feature differently in any significant way. Concerning roles in political actuality there is a difference between de Volkskrant on one side and NRC Handelsblad and De Telegraaf on the other side.
VK paid much less attention to the far-right in roles in political actuality than the other two papers (VK 9%, NRC 23% and Tel 20%). The other intriguing difference is between VK and NRC on one side and Tel on the other side and concerns the far-right in legal conflict. Both quality papers (VK 17%, NRC 16%) paid more attention to these roles than the popular newspaper (11%) The trends described above become clearer on a somewhat aggregated level. Therefore we recoded the specific roles into two groups, one that we assume portrays the far-right positively as legitimate 'normal democratic' parties (role in political actuality and positive internal) and one that emphasises their negative, illegitimate side (in legal conflict and negative internal) 11 . This dichotomy makes visible the level to which  The policy might be similar, but the image clearly changed. All three dailies contained similar percentages of such associations, there were no between-newspaper differences found (see Appendix 5).

Far-right ideological standpoints
An important final aspect of substantial attention is the presence of genuine far-right ideological standpoints (Mudde, 1998;Ignazi, 2006). The presence of each of these seven was checked in the coverage, but a first indicator of the attention to far-right ideology is the total share of the coverage which contains any of these standpoints, and how much of the coverage does not contain any far-right standpoint at all. These figures are presented in appendix 6. About 33% of all coverage included one or more of the far-right ideological standpoints, in most years roughly varying between 20 and 30%. The exceptions were 1989 (36%) and especially 1994 (54%). In these years, the attention to far-right ideologies was above average. For the rest of the research period there was not much difference between the years, or it could be that the relatively low percentage in 1986 might have suggested less attention to far-right standpoints in the preceding period.
Looking at the attention to specific ideological standpoints reveals a more subtle variation in the news coverage. Not all seven central aspects of far-right ideology received the same amount of attention. In Table 5 the variation between ideological standpoints over the years is presented. Anti-progressive, antidemocratic and traditionalist standpoints hardly featured in the coverage at all. The four other ideological standpoints can be ranked as follows: The most prominent featured ideological standpoints were those containing xenophobic or exclusionist sentiments; 27% of the coverage of the far-right contained such standpoints: 9% of the coverage contained harsh law and order standpoints: almost equally prevalent were anti-establishment (or populist) standpoints (5%) and nationalistic standpoints (4%). The most controversial standpoints also received most attention in the newspaper coverage.
Anti-democratic standpoints featured in a small portion of the coverage in the 80s and 90s, as well as in 2004. The anti-establishment and populist far-right standpoints only gained attention from 1998 onwards.
Nationalist standpoints were found in the coverage throughout the research period, but mainly in 1989-1998 (7-10%). As one would expect, by far the most prominent standpoints were the xenophobic or exclusionist standpoints. The peak in the attention to these was also in the period 1989-1998 (>30%), but attention to these standpoints remained on a high level throughout the entire 18 year research period.
Harsh law and order standpoints gained more and more attention from 1994 on (4%), and became almost as prominent as xenophobic or exclusionist standpoints in 2003 (16%) and 2004 (13%).
As we can see in Table 5,   Law & Order* 2 0 4 9 9 16 13 9 Traditionalist 0 1 0 1 1 1 2 1 *) differences significant at p < 0,05 There are very few differences between the three newspapers concerning the attention paid to these seven far-right ideological standpoints. There are only two exceptions. The only exception to the neglect of antiprogressive standpoints was de Volkskrant. De Telegraaf paid twice as much attention to law and order standpoints (16%) than de Volkskrant (8%) and NRC Handelsblad (6%). The difference between extra attention to xenophobic and exclusionist standpoints in de Volkskrant (30%), as shown in Appendix 7, is only significant compared to De Telegraaf (23%)

Support attention to the far-right
The fourth aspect of media attention to the far-right is that of support attention. It refers to the degree that the far-right and their standpoints are (de-)legitimated by non-far-right actors. This dimension of newspaper contents is measured in two ways. First we look at the relative attention to favourable and unfavourable attitudes towards the far-right or their ideological standpoints. A second indicator is the attention to public debate and among non-far-right actors about the way one should deal with the far-right.
Non-far-right actors' attitudes towards the far-right The first measure, as said, is the extent to which far-right actors are confronted with favourable or unfavourable attitudes of non-far-right actors (NFRA). In half of the coverage these attitudes were not present at all (51%, see Table 6). The remaining half contained mainly exclusively negative attitudes towards the far-right, or far-right standpoints (34%).
When we look at the developments throughout the research period, three periods can be distinguished, as one can see in the table. The first period from 1986 to 1994 was characterised by a dominant and (almost) exclusive presence of NFRA with unfavourable attitudes towards the far-right (35-65%). The coverage contained almost no favourable NFRA attitudes in that period. Although this situation was at its most extreme in 1986, and there was some opening for alternative voices in 1994, the dominance of unfavourable attitudes and the marginality of favourable attitudes towards the far-right remained throughout the whole period.
In the second period (1998)(1999)(2000)(2001)(2002)(2003) the share of the coverage that contained favourable NFRA attitudes started to break out of its marginal position. Exclusive unfavourable attitudes towards the far-right were still the most frequent and remained a majority (still 28% as compared to 12% for exclusively favourable attitudes in 2003). However, the share of articles that include favourable attitudes towards the far-right increased in this period from 10% to 17% ('exclusively favourable' and 'both favourable and unfavourable' combined). They were no longer an exception. The share of coverage with exclusively unfavourable attitudes in 2002 and 2003 was greater than in 1998, with a peak of 36% in 2002, meaning Fortuyn had a great deal of negative response in the press.
The third period contains only one election year in this study. In 2004 the share of exclusively unfavourable attitudes towards the far-right almost halved compared to 2003 (from 28% to 15%), while the share of the coverage including exclusively favourable attitudes towards the far-right almost doubled from 12 to 21%.
The share of articles with both favourable and unfavourable attitudes increased even more sharply and peaked with 19% in that year. This shift means a turn-around of the situation in the previous 17 years. Between newspapers there are modes, but also significant differences. De Volkskrant paid least attention to favourable attitudes (7%) and most to unfavourable attitudes (38%). De Telegraaf paid most attention to favourable attitudes (15%). NRC contained most articles with both favourable and unfavourable attitudes in them (8%), twice as much as the other two newspapers (VK 4%, Tel 3%, see Appendix 8).
Dealing with the far-right as a topic in the coverage In general, the far-right is not considered as a 'normal democratic party'. Moreover among politicians and political debaters, civic organisations as well as activists there is a debate going on about how to fight the far-right in the most effective way. Such debates, whatever the strategy under discussion (from criminalisation, to a so called 'cordon sanitaire' or even draconian measures, or the opposite -inclusion in the 'normal political community'), indicate that the far-right receives separate treatment, even if the outcome of such a debate is to include them in the normal political community, because neither socialist, liberal nor confessional parties are subject to such debates ( Van Donselaar and Rodrigues, 2006;Jacobs & Swyngedouw, 2002). Therefore coverage that pays attention to such debates is also an indicator that the far-right receives a 'different treatment' or that there is controversy about them among non-far-right actors.
Overall a mere 7% of the coverage paid attention to debates on how to deal with the far-right. In the Dutch newspapers this issue received decreasing attention. Whereas it was a relatively relevant issue in 1986 (27%), after the mid-nineties it ceased to be so. It was not totally absent, but with an attention figure of about 5% after 1994, it cannot be said to be a 'hot issue' in the news coverage. There were no significant between-newspaper differences (Appendix 9).

Conclusions and discussion
In this section we will provide answers to the three research questions. While doing that, we will also discuss the value of the preliminary hypotheses that we engaged and used to sharpen our focal point.
Following this, we will first give some general conclusions about media attention to the far-right in Dutch newspapers (RQ1), then look at longitudinal variation (RQ2) and close with the expected differences between types of newspapers (H1-6).

General conclusions
The fact that we found between a few dozens to hundreds of news articles on the far-right in each election year means that the far-right was not neglected in the coverage. Even more significantly, in most of this coverage the far-right is often a prominent actor in headline or body text. In practically all of these articles there were one or more far-right actors, who are actively represented in fifty percent of the total coverage.
The following aspects of substantial attention deepen this analysis more thoroughly. One of the first intriguing findings is that far-right actors featured as much in roles in legal conflict as in political actuality, which is the normal every-day role of a political party. Furthermore, stigmatising associations have been In order to make such statements, we would first need an account of far-right activity and compare that to what was reported in the news. Still, it is clear that we have found that media focused their attention on certain recognisable aspects of the far-right. This focus results in a media image that sets the far-right apart from 'normal parties' by emphasising its controversial characteristics, hence this media image can be described as that of 'the controversial outsider'. In the following section we take a closer look at the development of that media image.

Longitudinal variation
First of all, the volume attention fluctuated greatly. There were two periods of increased attention, 1994 and the "long election year 2002(Brants & Van Praag, 2005", both peaks were simultaneous with a rise in electoral popularity. In 1994, most attention went to Hans Janmaat and his CD, with the numerous incidents in and around the party. In 2002, Fortuyn continuously drew media attention. Media attention was at an all-time high, and continued to be so, also due to the disastrous participation of the LPF in the coalition government, which fell before the end of the year and led to new elections in January, 2003 (see Kleinnijenhuis et al., 2003). With an exception of the election year 2002, formal attention decreased steadily throughout the research period.
In terms of substantial and support attention, we see a double edged pattern. On one side there are tendencies that suggest increasingly more open and diverse media attention to the far-right. There was increasing attention to the far-right in roles in political actuality (and decreasing attention to roles in legal conflict). A greater diversity of far-right standpoints received attention in more than 5% of the coverage after 1998. Not only xenophobic or exclusionist standpoints, but also law and order standpoints and antiestablishment or populist standpoints came more to the fore. Last but not least, next to the serious attention to unfavourable attitudes towards the far-right, more and more attention was given to favourable attitudes to the far-right from 1998 on.
On the other side of the coin, attention to controversial roles and stigmatising associations may have decreased in the second half of the research period, but did not become exceptional. Together with the unfavourable attitudes of NFRA, they remained important aspects of the media attention to the far-right.
All together, this means that media attention to the far-right did change significantly since 1986. In 1986 it was relatively limited and almost exclusively focused on negative aspects of the far-right. During the nineties there was increasing diversity in media attention, something that continued in the new century.
However, the media image of the controversial outsider continued to be used, at least until 2003. The arrival of Fortuyn did not cause an immediate change of media attention to the far-right, other than in volume. Some aspects that were in a process of change already, changed faster. However, Fortuyn also was the focal point for strong counter-reactions, which is reflected in the coverage, as the aspects of media attention that contribute to the controversial outsider image received increased attention as well. It was not until 2004, that these aspects suddenly received much less attention, while the diversity tendency continued. Therefore, it seems that the media image of the far-right in 2004 can no longer be characterised just as that of the controversial outsider.

Quality and popular newspapers
In the introduction of this chapter, we distinguished de Volkskrant and NRC Handelsblad as quality newspapers and De Telegraaf as a popular newspaper. Based on previous research, we formulated three hypotheses about the differences in media attention to the far-right in the quality papers and the popular paper in our Dutch case study. Here we consider our findings on newspaper differences to see if these hypotheses hold.
Hypothesis 1 stated that VK and NRC were expected to pay both more volume and formal attention to the far-right, than Tel. Indeed, Dutch quality newspapers paid significantly more volume attention to the farright, than the popular Tel did. Concerning formal attention, we did not find a clear distinction between quality and popular papers. As we explained in section 3.2, due to missing data we cannot interpret the findings on front page appearance here. Of the other two variables, headline appearance and prominence, one does support the hypothesis (headline appearance) and one does not (prominence). In conclusion, hypothesis 1 was supported on 2 out of three variables. In terms of volume attention it is confirmed, in terms of formal attention, it found mixed support.
Hypothesis 2 stated that VK and NRC were expected to pay more attention to (a wider range) of substantial attention aspects than Tel. In the analysis we focused this hypothesis on the range of roles (H2a), the range of far-right standpoints (2b). In terms of attention to roles in which the far-right featured, we found one significant difference relevant in terms of hypothesis 2. VK and NRC did indeed pay more attention to the far-right in legal conflict, than Tel did. All other differences were insignificant. Together with this one significant difference they do not refute hypothesis 2a, but they do not form convincing support for it either.
In terms of attention to far-right standpoints, we did find some differences, but these seem to express the individual characteristics of the newspapers first. Here too, the differences found are no direct refutation of hypothesis 2b, but form no support for it either.
The third hypothesis concerned support attention. As, contrary to the German and Flemish cases, the indicator of attention to debate on how to deal with the far-right is not directly related to one particular issue, we did not expect to find particular differences on that indicator. Therefore, we looked only at the attention to favourable and unfavourable attitudes of NFRA towards the far-right. We counted slightly more reported attitudes in the quality newspapers. Unfavourable attitudes featured slightly more in VK and favourable attitudes in Tel. This means that the hypothesis formulated was too general. Specific attitudes turned out to feature in particular newspapers. Hypothesis 3 has to be refuted for the Dutch case.
Left-leaning and right-leaning newspapers The second dimension of comparing newspapers was along the lines of political leaning. As we expected, more attention was given to critical aspects of the far-right in the ideologically most distant newspaper, de Volkskrant, based on the ideas of Hallin & Mancini (2003). Again we formulated three general hypotheses to test this general expectation. In order to test these hypotheses in a valid way, differences found between the two quality newspapers are taken as more crucial, to be able to make statements about political differences, because we need to distinguish quality -popular differences from left-leaning -rightleaning differences.
Hypothesis 4 expected more volume and formal attention in the left-leaning VK than in the right-leaning two newspapers. Differences turned out to be relatively small. Concerning volume attention, hypothesis 4 found some support, but the difference between VK and NRC is too small for convincing support. In terms of formal attention there is support, again on one indicator, which again is not enough. Hypothesis 4 therefore, found only weak support in the Dutch case.
The fifth hypothesis that expected more substantial attention in VK can be split into three parts. The first concerns roles of the far-right. Here there were few differences, but one very particular one, namely that VK paid least attention to the far-right in its 'natural habitat' as a political party, in political actuality. This finding refutes H5a, but is a relevant difference between left-leaning and right-leaning newspapers. H5b concerns stigmatising associations. These were expected more in the left-leaning VK, which also occurred in the corpus, but the differences were very small. Last, H5c expected more attention to far-right standpoints in the left-leaning VK. Here we did find a clear difference on some standpoints that indeed distinguish VK from NRC and Tel. It was the only newspaper that reported anti-progressive standpoints. It paid slightly more attention to xenophobic or exclusionist standpoints than NRC, and significantly more than Tel. However, attention to law & order standpoints featured most in Tel. Apparently there is a significant difference between the left-leaning VK and the right-leaning two, but not in the simple opposition of more or less attention to standpoints, but for specific standpoints, where the left-leaning newspaper seems to emphasis the traditional or 'typical' far-right standpoints, and the right-leaning newspapers emphasis the more conservative or populist standpoints.
The final hypothesis, H6, expected more attention to NFRA attitudes in the left-leaning newspaper than in the right-leaning ones. Here we found a similar difference as in H5c. The left-leaning VK paid most attention to unfavourable attitudes, while the right-leaning NRC paid least attention to unfavourable attitudes and the right-leaning Tel paid twice as much attention to favourable attitudes than the other two.
This again, does not exactly support H6, but does suggest a significant difference in emphasis between leftand right-leaning newspapers.

Discussion
In this paper we have shown that with the possible exception of 1986, newspapers did not practice a minimal attention policy. Whether it is a conscious policy or not, throughout most of the research period the coverage is narrowly focussed on negative aspects. Janmaat's accusation that he and his party have been ignored by the media is unjust, but the media attention was relatively negative for most of the research period (in general, journalists did not have to 'dig deep' to find newsworthy negative facts around especially CD and CP'86). In general the media attention can best be described as producing a image of controversial outsiders; this image was one dimensional in 1986 and became more nuanced during the nineties and first years of the new century.
Differences between newspapers were relatively small. Individually they did not produce very different media images. In emphasises there were some differences, both between quality and popular newspapers as well as between left-leaning and right-leaning newspapers. As this case study was carried out on no more than three titles, we need to see if these differences, and their slight degree are representative for the various types of newspapers, or rather are individual differences, in the other countries as well.