Contagious Parties: Anti-Immigration Parties and Their Impact on Other Parties’ Immigration Stances in Contemporary Western Europe

1 This should not be confused with the classic controversy over ‘contagion from the left’ versus ‘contagion from the right’ (Duverger, 1954; Epstein, 1967).

2 These two types are often confused. A more restrictive immigration policy is widely considered as ‘rightist’, whereas the ideal of the multicultural society is usually regarded as an idea of the left. As shown empirically in this article, this relation is not as straightforward as it prima facie seems. Recently, many parties of the left have shifted to more restrictive immigration policies, while several parties with a right-wing profile have become less strong advocates of the ideal of cultural unity.

3 ‘Left’ and ‘right’ not only refer to a traditional economic axis here, but also to a broader dimension that encompasses clusters of issue positions, as, for example, Kitschelt and McGann (1995) note. The ‘issues that divide the Left and the Right are linked in ways contingent upon time and space’ (Kitschelt and McGann, 1995: 44).

4 However, the National Front (FN) was electorally unsuccessful in the early 1980s. This calls into question whether the first-mentioned effect can be considered as ‘contagion’ in the sense of this article. In the context of French politics, the actions of Social Democrats like Fabius may be interpreted as attempts to hurt the centre—right by legitimizing the National Front rather than as contagion effects. Such actions dovetail with those of the Social Democratic president Mitterrand in the 1980s. Not only did Mitterrand urge the leaders of the national broadcasting corporations to devote more attention to FN party leader Le Pen in 1982, he also changed the electoral rules to a system of proportional representation before the national elections four years later. This led to the entrance of 34 representatives of the FN in the Assemblée Nationale (Mayer, 1998: 21).

5 Perlmutter also takes another Italian party into account, the National Alliance (AN). Whether this party can be seen as anti-immigration at the relevant time-points is questionable, however. It did not have an anti-immigration stance by 2000 (Lubbers, 2001) or by 2004 (Van Spanje et al., 2006). Nor did the party attach much importance to the issue (Benoit and Laver, 2006; Carter, 2005: 33—4). Therefore, the party was not included among the cases selected for this study.

6 Recently, Williams has contributed to the debate with a comprehensive cross-national study on the impact of anti-immigration parties, including contagion effects on the issue of immigration in 17 Western European countries (2006). She did not address the question of how the party positions of mainstream parties are affected by the electoral performance of anti-immigration parties, however. Instead, she examined the position shifts of mainstream parties on this issue as a response to the shifts of the radical right parties. Not much empirical evidence was found on this point, and Williams concludes that ‘the other parties do not adapt their positions on immigration directly because of the position shifts on the issue by radical right-wing parties’ (p. 70).

7 This presupposes that the mainstream parties actually have a position on the immigration policy dimension. As Meguid rightly points out, this is not a given when there is no relevant anti-immigration party in the system (2005: 349). However, I selected countries in which significant anti-immigration parties exist only. Moreover, in each of the political contexts that I deal with in this article, contemporary Western European countries, the immigration issue enjoys high degrees of salience (see, e.g., Benoit and Laver, 2006). It can therefore be assumed that parties in these contexts have a position on the immigration issue.

8 If an effect were found in accordance with Hypothesis 4, an alternative explanation would be that a hard line on immigration sits uneasily with the ideologies of far left and green parties. Most notably, the notion of the universal brotherhood might be incompatible with very restrictive immigration policies.

9 The left—right dimension includes more issues than immigration, of course. In addition, party positions in terms of left and right are not only determined by a party’s immigration position, but also by the salience of the immigration issue. However, it is consistent with the line of reasoning of Adams et al. (2006) to expect that the other niche parties stick to their key issues. Moreover, having very left-wing profiles, they are not expected to co-opt policy positions on an issue that is predominantly owned by parties of the right in the countries of study. After all, this would be the same as moderating their ideological positions in terms of left and right, of which Adams et al. (2006) have shown that it presents considerable electoral costs to these parties.

10 Hypothesis 4 may seem difficult to separate from Hypothesis 2. After all, (ex-) communist and green parties are all left-wing. However, the theoretical basis on which Hypothesis 2 is based differs from that of Hypothesis 4. Note that Hypothesis 2 is not about right-wing parties only, but states that the more right-wing its ideologies, the more vulnerable a party is to contagion effect, and green parties are not necessarily far-left parties. The green parties in the sample were all coded as having moderate positions, between 3.46 and 4.24 on average on a 1—10 left—right scale. They can therefore be expected (on the basis of Hypothesis 2) to be vulnerable, and more so than the (ex-)communist parties, that were all placed to the left of the greens. Thus, assessing Hypothesis 2 does not render the assessment of Hypothesis 4 superfluous.

11 Note that this conceptualization is a major departure from the usual classification of parties in the existing literature. Instead of classifying a party according to its origins in society or its ideological background (see Mair and Mudde, 1998), this article introduces a different basis of categorizing parties, which will not be discussed at length, however, as this does not have a large impact on case selection or results. As mentioned in the text, the group of selected parties is similar to other studies on anti-immigration parties.

12 For a few small parties that were not included in either of these two expert surveys, the author relied on descriptions of party ideology regarding immigration given by Carter (2005).

13 Established parties are defined as parties that already existed before the emergence of the anti-immigration parties in Western Europe at the end of the 1970s. Admittedly, the FPÖ was founded as early as 1955. However, before Haider’s take over in 1986, the FPÖ can be considered as a completely different party from the FPÖ afterwards (e.g. Luther, 2000).

14 Belgium is considered to contain two separate political systems, Flanders and Wallonia. Thus, this adds up to 13 political systems in 12 countries.

15 Even if the question asked in 2004 is identical to the one asked four years earlier, it is questionable whether the immigration restriction scale of 2004 is comparable to that of 2000. If the entire perception of immigration restriction changed in the minds of the experts in these four years, a 4.0 score in 2004 does not mean the same as a 4.0 score in 2000. To the extent that this influences the results, it will have a dampening effect conducive to type-II errors. If empirical evidence is found in support of the hypotheses, it is therefore likely that the impact is even larger than predicted. Because hypotheses 1 and 3 are confirmed on the basis of the available data (see below), this strengthens the findings of the article, however.

16 The reason underlying the choice for change as the major independent variable, rather than the absolute levels of anti-immigration party success, is that the other parties are expected to adjust their positions mainly as a result of ‘electoral shocks’, and not as a result of the mere presence (and possible growth stagnation or vote decrease) of the anti-immigration parties. This methodological choice did not matter much for the findings of this article, as both the change and the average anti-immigration party success by country have significant positive impacts significant at the p = 0.05 level.

17 National-level election results are used for this variable. If no national-level elections were held in one or either of these years, then the result was estimated, assuming a linear relation from one election to the next. For example, the combined electoral performance of the anti-immigration parties in Germany in 2000 was estimated by averaging these parties’ results in the national elections of 1998 and 2002. The combined result of the German People’s Union (DVU), National Democratic Party of Germany (NPD) and Republicans (REP) was 3.3 percent in the 1998 national elections and 1.0 percent in 2002, which produces a result of (3.3 + 1.0)/2 = 2.2 for the year 2000. Because these parties obtained 2.4 percent of the vote in the 1990 national elections together, the change between 1990 and 2000 is estimated at 2.2 — 2.4 = —0.2. See Table 1 for descriptive analyses of the variable computed in this way.

18 Estimations by experts derived from three surveys conducted at different points in time by Marks and Steenbergen (1999), Lubbers (2001) and Benoit and Laver (2006) are all correlated for more than r = 0.90 with the voter perceptions in terms of ‘left’ and ‘right’ derived from the EES at the corresponding time-points (significant at p = 0.01, one-tailed).

19 Carter also measured a similar institutional variable, the proportionality of each national election, but this is almost identical to the effective threshold variable (r = 0.95).

20 Data are available for virtually every significant party that is not labelled as ‘antiimmigration’. The average vote-share obtained by the parties included in the analyses together is 95 percent, varying from 68 percent in Italy in 2004 to 98 percent in Sweden (also in 2004).

21 The government status dummy is also included in Model 4, because it is a lower-order effect without which the cross-level interaction would be difficult to interpret (see Brambor et al., 2006).