The Gap between Public Preferences and Policies on Immigration: A Comparative Examination of the Effect of Politicisation on Policy Congruence

The existence of a gap between public preferences for more restrictive immigration policies and relatively expansive immigration policy in Western democracies has received considerable attention. Sometimes, this gap has been explained by the nature of immigration policies: dominated by elites while the public remained uninterested. In many countries, however, immigration has gained considerable salience among the public. There are competing expectations and accounts relating to whether policy-makers ignore or follow public demands on immigration. In this article we examine the potential drivers of variations in the opinion–policy gap on immigration in seven countries (1995–2010). We analyse the effect of the politicisation of immigration on this opinion–policy gap. The strength of anti-immigrant parties is unrelated to the opinion–policy gap on immigration. The salience of the issue and the intensity of the public debate are associated with the opinion–policy gap, and the combination of negative attitudes with extensive media coverage seems particularly conducive to policy congruence.


Introduction
Representative government requires at least some degree of responsiveness to the will of the public. A significant body of scholarship on policy responsiveness suggests that governments and legislators indeed respond to the preferences and demands of the public by delineating policies that are consistent with public preferences in order to ensure re-election (Soroka and Wlezien 2010;Klemmensen 2005, 2008;Arnold and Franklin 2012). Policy-makers will, however, pay more attention to the issues citizens care about, and they will be more responsive in those policy domains.
For a long time, scholars have been arguing that this general pattern was not applicable to immigration policies, which were characterised by a large discrepancy between the preferences of the public and the policies in place (but see a discussion of cases with different opinion configurations in Freeman, Hansen, and Leal 2013). Gary Freeman's (1995) pioneering piece argued that whereas citizens in Western democracies were holding restrictive views on immigration, demanding less immigration, immigration policies were-for the most part-expansive or liberal. 1 Several scholars, including Freeman himself, have reaffirmed more recently the existence of an opinion-policy gap on immigration (McLaren 2001;Beck and Camarota 2002;Thomassen 2012;Freeman, Hansen, and Leal 2013). One central element in these explanations is that immigration policies are adopted out of public view; immigration policies are better understood looking at the role of organised interests (trade unions, business lobbies, pro-migrant social movements) or at the institutional constraints imposed to policy-makers (i.e. European and international law and courts) than by looking at the diffuse demands of the public (Lahav and Guiraudon 2006).
Freeman's original argument described a situation that is almost 20 years old, when immigration was much less salient on the political agenda, in voters' minds but also in the media (Boomgaarden and Vliegenthart 2009). More recently, some researchers have questioned whether the gap between public preferences and policies is as large as suggested by Freeman (Statham 2003;Lahav 2004;Zapata-Barrero 2009), or whether policy-makers are constrained by pro-immigration lobbies (Statham and Geddes 2006). In particular, it has been argued that there is significant variation across countries in how much politicians have responded to public demands for reforming immigration policies. This article builds on this body of scholarship and analyses the connection between public preferences and policies on immigration.
The central argument is that the varying levels of policy congruence that are observed across countries may find their source in the degree of politicisation of the issue. By politicisation we mean, first, the emergence of a public debate, which then signals demands for policy change (Birkland 1997). In line with theories of party competition, politicisation involves the existence of diverging views on the topic, and actors that will challenge the status quo and polarise the debate (Downs 1972). Building on this definition we look at the salience of immigration in the public debate, as well as at the actions of anti-immigrant parties and civil society actors to see whether these factors affect how responsive policy-makers are to public demands.
We study the dynamics in public attitudes and immigration policies in seven countries across 15 years. The choice of countries is based on dissimilar case selection (Przeworski and Teune 1970;Lijphart 1971), which facilitates examining patterns linking immigration, public opinion and policy responses. The seven countries differ in their immigration histories, with Britain, Belgium, the Netherlands, and to some extent Switzerland and Austria, having received mass immigration for a much longer period, and Ireland and Spain having recently become destinations of immigration. They also show varying levels of politicisation of immigration. In some, immigration has hardly become politicised (Spain, Ireland); in others, new or established parties successfully mobilise support against immigration (Switzerland, Belgium and Austria); in the Netherlands such parties have been around for a long time, but only recently with some electoral success; whereas in Britain the issue has become strongly contested without giving rise (yet) to successful nation-wide anti-immigration parties (van der Brug et al. 2015).

Linking Public Attitudes and Immigration Policies
The Opinion-Policy Gap Since Freeman (1995), scholars have debated the gap between public preferences and the reality of immigration policies in Western democracies (Lahav and Guiraudon 2006). In most European countries, restrictive views about immigration have increased (see Ceobanu and Escandell 2010). Yet, a uniform pattern of policy change towards restriction is not apparent. For example, the Migrant Integration Policy Index (MIPEX) reports (Huddleston et al. 2011) show that, of 31 countries studied, half of them had enacted policy changes in a direction more favourable to immigrants and only four had shifted towards more restrictive policies (cf. also Bale 2003;Bale et al. 2010). Case studies and comparisons between two or three cases also suggest that immigration policies tend to be disconnected from public preferences. Public attitudes are almost invariably restrictive-or favouring the status quo-whereas legislation and policy-making is more often expansive (Breunig and Luedtke 2008;Green-Pedersen and Krogstrup 2008).
Other research suggests that this gap is not always as wide or omnipresent as generally argued (Lahav 2004). For example, some scholars show that British immigration and asylum policies have become more restrictive and in line with public preferences over the years (Statham and Geddes 2006;Jennings 2009), while Hobolt and Klemmensen (2005) suggest that governments are responsive to public opinion moods on immigration in Denmark but not in Britain (cf. also Ruedin 2013).
Considering the scholarship on policy responsiveness, the persistence of an opinion-policy gap on immigration in most western democracies is puzzling. The expectation is that political elites respond to the preferences and demands of the public by delineating policies that are consistent with them (Soroka and Wlezien 2010;Hobolt and Klemmensen 2008;Arnold and Franklin 2012). Anticipating electoral penalties and rewards, governments and legislators will pay more attention to the issues citizens care most about and will try to follow the position most preferred by the public in these policy domains (Wlezien 1995). Research on agendasetting goes in the same direction (Klingemann, Hofferbert, and Budge 1994), though it shows that policy-makers are often slow to react, and that when they do, they often overreact (Baumgartner et al. 2009).
However, to date, these general theories have not been comprehensively assessed for immigration policy. This is partly due to the fact that responsiveness is often measured using budgetary expenditure (e.g. Page and Shapiro 1992;Stimson, MacKuen, and Erikson 1995;Soroka and Wlezien 2010) and immigration is, for the most part, a regulatory policy area. Thus, both in the fields of migration studies and of public opinion and policy-making, we still have limited comparative evidence of the existence of an opinion-policy gap in the area of immigration.
We examine whether there is indeed a predominant opinion-policy gap in the area of immigration across Europe. As we have seen, the expectations that can be derived from the literature are contradictory: Following Freeman's work we should expect pervasive opinion-policy gaps in all or most countries. By contrast, the responsiveness and agenda-setting scholarship leads us to expect opinion-policy congruence, at least when immigration is salient among the public. In prior scholarship, we find a wide variety of both opinion-policy congruence and gaps. Hence, we turn to examining the factors that may account for such cross-national and over-time variations.
Following Freeman (1995), the dominant explanation for the opinion-policy gap is that the formulation of immigration policies is dominated by organised groups with a direct interest-primarily business lobbies interested in reducing labour costs-while the demands of the public are less articulated because of the diffuse costs of immigration for most citizens. Immigration policies are thus often shaped 'out of public view and with little outside interference' (Freeman 1995, 886). This is often referred to as the 'control gap' thesis. Other scholars have disputed this account and argue that many actors intervene in immigration policies, including the courts, bureaucracies, trade unions and the 'organised public'-like ethnic groups, NGOs, religious organisations and local actors (Cornelius and Rosenblum 2005;Statham and Geddes 2006;Lahav and Guiraudon 2006). The general idea, however, remains that these actors are constraining the extent to which government is steered by the public mood.
Both explanations regard the demands of the public as not too influential. With the rise of anti-immigration parties and a more prominent discourse on immigration by many mainstream parties, the situation has changed considerably in recent years and immigration has become a highly contested and politicised issue in many Western democracies. As Lahav and Guiraudon (2006, 212) put it, '[t]he construction of immigration as a public problem-highly salient in public opinion and partisan politics-requires us to revisit Gary Freeman's (2002) persuasive client politics model'. However, there is no study to date that has looked comparatively at whether the degree of politicisation of the issue can account for the degree of congruence between public opinion and immigration policy-making. Against this backdrop, this article examines seven countries to assess whether the degree of politicisation of immigration can account for variations in opinion-policy congruence.

The Effect of Politicisation on the Opinion-Policy Gap
We examine three key aspects of the politicisation of the issue: (i) the intensity of the public debate on immigration in the media, (ii) the strength of anti-immigration parties and (iii) the level of mobilisation for more restrictive immigration policies. The general expectation is that all three factors are strongly associated with the politicisation of immigration, and should thus induce policy-makers to be more attentive to and more congruent with the attitudes of the public. Moving in the opposite direction to what the mobilised public wants on a salient issue bears serious electoral risks for governments (Stimson, MacKuen, and Erikson 1995).
Information is an essential component of theoretical models of government responsiveness (e.g. Soroka and Wlezien 2010). For the public to notice what politicians do, they need to be informed at least of the broad lines of policy-making and policy outcomes. For politicians to evaluate the opinion mood, they need to be informed of what issues are salient in society. In both cases, the media play a central role in providing (some of) the information citizens and governments need. Obviously, with the growing use and sophistication of polling techniques, policymakers have other indicators of what the public think and want (Geer 1996;Jacobs and Shapiro 1996). Yet, the media have remained key actors in making citizens' views public, in shaping and formulating them, sometimes by forcing the debate to occur (Walgrave and Van Aelst 2006).
Previous scholarship has shown that media attention is central to understanding when the public reacts to increasing levels of immigration, and that increasing media saliency coupled with a negative rhetoric reinforces feelings of threat and negative views about immigration (Schuck 2007;Lahav 2013). We provide a first attempt at examining, in a comparative study, the contribution of media coverage to opinionpolicy congruence on immigration. The expectation is that the more often the media report on immigration-be it about facts and figures or about claims made by actors on the issue-the more congruent public mood and policy-making will be.
The second aspect of politicisation we examine is the presence and success of antiimmigrant parties. Schain (2006) hypothesises that when these parties are stronger, policy-makers are more likely to adapt immigration policies in more restrictive directions (cf. also Howard 2010). Like other new parties, anti-immigration parties alter the political competition by focusing on a new issue that has been ignored by mainstream parties. They will often make visible attitudes that were kept silent (Meguid 2005). In reaction, governing parties and policy-makers will try to address the new issue and to adopt policies in line with the demands put forward by the new parties. In the case of immigration, these new parties tend to be extreme-right, populist and anti-immigration (Green-Pedersen and Odmalm 2008; van Spanje 2010). Our expectation is that the presence of these parties and their relative strength (in terms of parliamentary seats and influence over government formation) can help understand differences in opinion-policy congruence.
Finally, following the work on social movements as mediators in the politicisation of new issues (Della Porta and Diani 1999;Giugni and Passy 2004), we look at the role of civil society/non-governmental actors and the level of mobilisation by antiimmigration movements. The expectation is that higher levels of mobilisation will push policy-makers to react and to change immigration policies in restrictive directions. Previous research has shown that the amount of extra-parliamentary and non-party-led extreme-right and anti-immigrant mobilisation can vary considerably across countries (Koopmans et al. 2005, 190-195). The results indicate that civil society mobilisation is particularly important in countries where parties are not discussing immigration to a great extent, and where no anti-immigrant or extremeright party politicises the issue. There are, however, reasons to think that mobilisation by civil society actors on its own is less effective in making political elites respond to public pressure, as the absence of an electoral threat will diminish their incentives to change course.

Data, Indicators and Methods
In this article we compare seven countries (Austria, Belgium, Ireland, the Netherlands, Spain, Switzerland and the UK) over 15 years (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) using data collected in the context of the Support and Opposition to Migration (SOM) project. 2 The data were collected using the same definitions, sources and protocols across all countries, through a tightly coordinated research collaboration, drawing on a multiplicity of sources. Public opinion is captured in two dimensions: attention to the issue and attitudes towards immigration. We measure attention with the proportion of respondents citing immigration as one of the three most important political problems in public opinion surveys (MIP3). 3 This is a relative measure, and any increase or decrease of this indicator can simply mean that other issues have gained or lost in importance (Jennings and Wlezien 2011). However, voters tend to focus on a limited set of issues when they decide to cast their vote, and if immigration is not among the three most important issues, the likelihood that they will punish electorally a party that does not pay enough attention to immigration is low (Bélanger and Meguid 2008). In this situation, we would not expect much responsiveness or congruence.
We measure the direction of voters' opinion towards immigration with the proportion of respondents declaring that they do not think that immigration has been beneficial for the economic development of the host country. 4 More fine-grained measurements of how voters perceive the various aspects of immigration would have been preferable, but existing surveys do not include better indicators that are available for the entire period in all seven countries. We also considered questions included in multiple cross-national surveys (WVS, EVS, ESS, ISSP, Eurobarometer, etc.) on the perception that there are 'too many' immigrants, and whether government should limit the entry of immigrants, but these resulted in even poorer time series. Moreover, beliefs on the impact of immigration on the national economy are among the strongest determinants of individual attitudes and preferences regarding immigration policies (Citrin et al. 1997).
Regarding policy-making, we look at two dimensions: policy activity and the direction of policy. For both, we use data from the MIPEX research project that has been extended backwards by the SOM project team to cover the whole period of our study. MIPEX is a project that measures immigration and integration policies in many Western countries (Huddleston et al. 2011). It uses 145 indicators to capture differences in the laws and policies related to the integration of immigrant populations. A value of 0, 50 or 100 is assigned to each of these indicators depending on the responses to a series of questions, where 100 indicates a more 'expansive' or 'liberal' policy position. The indicators are usually grouped into six strands of immigration policies: labour market access, family reunion, long-term residence, political participation, access to nationality and anti-discrimination. MIPEX allows comparing changes over time, with data for 2004, 2007 and 2010, and it scores well in terms of reliability and internal consistency (Ruedin 2011). To cover the full period analysed, the data were extended for 1995 and 2000 and by adding five new indicators to cover policies related to asylum. In the following we use the shorthand MIPEX to refer to these data.
Using MIPEX data, we measure the degree of policy activity and the direction of policy over time. Policy activity is measured as the number of indicator changes between two adjacent waves per country. For the direction of policy, we use the MIPEX scores to examine whether the policies in any given year are more or less favourable to immigrants in the four policy areas generally politicised: access to labour market, family reunion, long-term residence and asylum. In all figures, the MIPEX scores are reversed so that higher values denote more restrictive policies. The MIPEX data reduce complex immigration policies to scores, and subtle changes may be concealed. Nevertheless, MIPEX data remain appropriate, as they allow differentiating immigration policies across different domains, and therefore avoid overestimating a highly publicised policy change that will hide other reforms in the opposite direction (Ruedin, Alberti, and D'Amato 2015). Moreover, they are consistent with Freeman's advice to avoid treating immigration policies as a cohesive whole.
Politicisation is measured with indicators on media coverage, anti-immigrant party success and anti-immigrant civil society mobilisation. The indicators on media coverage and anti-immigrant civil society mobilisation use data from an extensive analysis of the claims on immigration made in the media between 1995 and 2010. A random sample of 796 days for each of the seven countries was drawn. For each sampled day, all articles related to immigration and integration were coded for two newspapers per country, usually one broadsheet and one tabloid. The content of the claim, its tone, the identity of the claimant and of the addressee of the claim were coded. Media coverage is measured as the average number of claims per day per country/year obtained with these data. Anti-immigrant civil society mobilisation is measured as the yearly average number of negative claims made by any non-party and non-state actor per day. Finally, anti-immigrant party success is measured as the share of seats of anti-immigration parties in the lower chamber of the national parliament.
The data available do not allow us to use statistical time series analyses; our analyses are thus based on the graphical displays of the trends observed in the indicators described before. To aid our interpretation of the figures and go beyond mere 'eye-balling' we have produced AJUS plots, computed Pearson and Kendall tau correlations and plotted cross-correlograms. AJUS is a system to classify and reduce the complexity of distributions according to shape introduced by Galtung (1969). We use a slightly modified version distinguishing six types of distributions: A (unimodal distribution with peak in the middle), J and L (unimodal with peaks on the right and left respectively), U (bimodal with peak at both ends), S (bi-or multi-modal with multiple peaks) or F (flat, no peak). Using an implementation in R, we can systematically determine the shape of a distribution once a threshold parameter is established, and two distributions of the same shape are treated as being associated. We will refer to these distributions and correlations in the footnotes when reporting on the results. 5 Is There an Opinion-Policy Gap?
We first examine whether an opinion-policy gap exists on immigration policies in the seven countries studied, both in relation to the attention given to the issue and to the position with regard to policy direction. Figure 1 displays the salience of the issue of immigration among the public and the degree of policy activity. The expectation is that policy activity increases when the public is more concerned about it. The first observation is that the two lines are not parallel in all countries. In three countries (Austria, Spain and, to a lesser extent, the Netherlands), the level of policy activity remains flat and relatively low, though concerns for immigration among the public are on the rise. In Belgium, policy activity grew significantly since 2001, while concern for immigration remained stable and low. In Switzerland, attention to the issue has moved up and down over the period, while policy activity rose almost linearly. Finally, in Ireland and the UK, policy-makers started legislating on immigration in the early 2000s, preceding the growing concern for immigration among the public that is observable from the mid-2000s. 6 These patterns suggest that, in terms of policy activity on immigration, policymakers are not responding to growing concerns among the public with reforms. Actually, if there is a relation at all, it appears that public concern follows policy activity. In most of our cases-the UK, Ireland, Belgium, and to a lesser extent Austria and Spain-levels of policy activity have increased before changes in public opinion (see similar findings in Morales et al. 2012). Only in Switzerland, and maybe in the Netherlands, did policy activity rise after changes in public opinion. These findings seem to contradict Baumgartner and colleagues (2009) who, looking at policy responsiveness in general, argue that politicians often react with some delays to shifts in public opinion.
Congruence and responsiveness are not only about adopting new policies on issues citizens are concerned about but also about the direction of policies. We look at citizens' attitudes on immigration and the policy direction in four dimensions of immigration policies (Figure 2): access to labour market, family reunion, long-term residence and asylum. We show the results for the average MIPEX indicator (solid black line) and for those policy dimensions that depart from the average policy direction pattern in each country. When the policy lines go up, policy has become more restrictive; when the attitudinal line goes up (line with circles) public attitudes about immigration are becoming more negative. There is congruence or responsiveness when both lines are going in the same direction. The first thing that stands out is that there is no uniform trend towards more restrictive immigration policies. 7 In some policy domains the legislation has become more restrictive, while in others it has become more favourable to immigrants. A good example is the UK, where legislation on asylum has become much more restrictive (consistent with Jennings 2009) while areas related to labour market access are nowadays more favourable to immigrants than in 1995. The same discrepancy in the evolution of policies between policy domains is found in the seven countries, except perhaps in the Netherlands and Switzerland. In these two countries, policies have been more stable throughout the period. When and where they have changed, however, like on labour market access in Switzerland or asylum in the Netherlands, it is towards more favourable positions. Overall, one cannot conclude that policymakers are completely disconnected from the attitudes of citizens. In each country, there are policy dimensions where policies have evolved in parallel to public opinion. The data available for these seven countries do not support Freeman's thesis of the existence of a pervasive opinion-policy gap. In another piece, Freeman points to only the UK as an exception to his general model (Freeman 1994). Here, we already find more variation in the link between public attitudes and immigration policies. In Austria, the Netherlands and Spain we do find a gap in the attention to and direction of immigration policies relative to the concerns and preferences expressed by the public, though at least in Austria asylum policy seems to have followed the public's wishes to a certain extent. In contrast, in Belgium, Britain, Ireland and Switzerland there is no meaningful opinion-policy gap either in attention to the issue or policymaking direction. Where we find that policy activity does not match public concernas in Austria, the Netherlands and Spain-we also see that the direction of policies does not follow the more restrictive course citizens prefer.

Examining Variation in the Opinion-Policy Gap: The Role of the politicisation of Immigration
By not observing a uniform trend towards more open immigration policies, we can easily exclude often-mentioned explanations about European integration that assume policy convergence (Koopmans, Michalowski, and Waibel 2012). Instead, our results lend credence to arguments that immigration policies are not nearly as determined by EU-level policies as many suggest, and that domestic factors are still the main driving force of policy-making in this area (Vink 2005).
In this section we examine whether domestic political dynamics can account for the variations in the public-policy gap we have found. We examine to what extent opinion-policy congruence is related to three aspects of politicisation: (i) the intensity of the public debate on immigration in the media, (ii) the strength of antiimmigration parties and (iii) the level of anti-immigration mobilisation by civil society actors. The expectation is that all these three factors contribute to the politicisation of the issue, and their presence or absence should help us understand when policy-makers respond to public opinion moods. Figure 3 reports the intensity of media coverage of immigration for each country (the average number of claims per day each year), the attitudes of the public on the issue and the trends in immigration policies. The first observation is that media coverage of immigration varies considerably, despite averaging over a year, and there is no single cross-national pattern over time towards either less or more coverage. Similarly, there is no year that stands out in all the countries, as one might have expected of 2001 due to the 9/11 attacks. The other observation is that trends in media coverage of immigration seem to match considerably the sentiment on the issue for a number of countries. Starting with the countries where we found an opinion-policy gap, in Austria and the Netherlands media reporting of immigration and negative attitudes towards immigration go in parallel. This is, however, not the case for Spain, where attitudes are monotonically becoming more negative and media reporting shows a fluctuating pattern with two peaks in the 2000s and a drop in attention since 2007. The British and Irish patterns are exemplary of situations where media coverage, public sentiment and policy-making move all in the same direction. By contrast, Belgium and Switzerland show patterns of oscillating moods and media reporting that do not seem to correspond. Overall, media attention and the politicisation of the issue that it brings does not account well for the policy-making direction across the countries under study. 8 In some countries media attention amplifies the negative sentiment of public attitudes while political elites design policies in the opposite direction (Austria and the Netherlands), whereas in others this consistency of the politicisation of the issue leads to policy congruence (Britain and Ireland). In the remaining countries, media reporting was either not consistently concerned about immigration (Spain) or media coverage is so erratic that it is hardly possible to expect any clear correlation with the trends in immigration policies and public attitudes (Belgium and Switzerland). Overall, thus, we only see a clear role for the media in Ireland and the UK, and very limited evidence of the media setting the agenda for policy changes in the other countries.

The Role of Anti-immigration Parties
We now consider the role of successful anti-immigration parties in shaping the opinion-policy gap. Figure 4 depicts the trends in the share of seats of antiimmigration parties alongside those of public attitudes towards immigration and policy direction. This line of explanation is even less successful than media coverage in providing a satisfactory account of the opinion-policy link. In all countries, the level and trends in the success of anti-immigrant parties seem completely unrelated to the attitudes towards immigration in the population. Of course, the mediating role of the electoral systems in Britain, Ireland and Spain-which make it more difficult for new challenger parties to gain representation as compared to the other four countries studied-is an important factor that should not be disregarded. Yet, we find no correspondence either in the countries where the electoral system does not impose a considerable barrier for electoral success.
The success of anti-immigrant parties on its own cannot account for the presence and absence of an opinion-policy gap. In some countries we find an opinion-policy gap despite the periodic success of anti-immigrant parties (Austria and the Netherlands), whereas in others we find no opinion-policy gap despite the absence of a successful anti-immigrant party (Britain and Ireland). The success of antiimmigrant parties does not even seem to propel policies in a more restrictive direction. In none of the countries where anti-immigration parties have been relatively successful in the last 15 years (Austria, Belgium, the Netherlands and Switzerland) do we see that immigration policies have become consistently more restrictive. In fact, the opposite seems to be the case. 9 This result is even more surprising given that in three of these countries radical right parties have also been in government. The parliamentary strength of anti-immigrant parties does not seem to foster policy responsiveness of governments, partly because there is no clear evidence that attitudes are more negative where and when these parties are more successful.

The Mobilisation of Anti-immigration Civil Society Actors
Previous scholarship has indicated that where anti-immigrant parties are not successful in accessing national parliamentary representation, anti-immigration mobilisation is channelled through civil society actors (Freeman 1995). It might be that mainstream political elites discount or ignore the pressure of anti-immigrant parties-as we have shown above-even if this runs counter to the expectations in the electoral competition scholarship, but that they are sensitive to anti-immigrant mobilisation by the wider civil society. Examining the number of claims by anti-immigration civil society actors next to the trends in public attitudes and policies, we find that this aspect of politicisation does not account for the patterns we find in most of the cases. 10 Patterns in mobilisation are quite erratic in a number of cases (Austria, Belgium, Spain and Switzerland). Moreover, anti-immigrant mobilisation and anti-immigrant attitudes do not seem to move in parallel in many countries (Austria, Belgium, the Netherlands, Spain and Switzerland). Indeed, anti-immigrant mobilisation is high and increasing in countries where we find an opinion-policy gap (Spain) as well as in countries where we find congruence (Britain).
Nevertheless, this aspect of politicisation might shed some additional light on the patterns for Ireland and the UK, especially when considered jointly with the patterns of media attention. Although in these two countries there is no nation-wide antiimmigration party with parliamentary representation that can play a major role in the politicisation of immigration, it seems that anti-immigration movements might have taken up this role by mobilising consistently the negative mood of the public. In both cases, we can observe that a shift towards more negative attitudes in the 1990s and a shift towards more restrictive immigration policies was accompanied by a higher number of negative claims by non-party and non-state actors. When jointly considered with the pattern of media coverage, the result is an amplification effect of the negative public mood that puts additional pressure on political elites to respond (Agnone 2007).

Conclusion
This article addresses a major puzzle in the study of immigration politics: the gap between public attitudes towards immigration and the reality of migration policies. Some recent literature has examined this gap by focusing on the growing concern among the public for the issue of immigration. In such circumstances, general theories of policy responsiveness suggest that policy-makers would face strong incentives to reform immigration policies in the direction of public demands. Our findings show that, although there are some indications that this logic applies to some countries, it is not a universal pattern. In the seven countries covered, we find no evidence of a systematic opinion-policy gap (Table 1). Whereas in some countries there is a clear disconnection between the concern and preferences of the public and the policies implemented, in others there is no obvious gap, or we find policy congruence. Moreover, not all changes in immigration policy are in the same direction: while it is true that some areas of legislation have changed towards more restrictive policies, there are areas of immigration policy that have become significantly more expansive. These findings confirm that the study of opinionpolicy gaps and policy responsiveness is far from straightforward when it comes to immigration policies.
Given these contrasting results, we have investigated possible explanations for cross-national differences in the existence of an opinion-policy gap on immigration. The dominant hypothesis in the literature is that policy-makers are constrained by a variety of actors (primarily lobbies) and international legal norms that reduce their capacity to comply with public demands. However, this account can only work if and when the issue of immigration remains elite-dominated. If the issue becomes politicised and the public cares about it-particularly if the public debate becomes intense-ignoring public preferences will bear greater electoral risks for elected politicians. In line with this last expectation derived from general models of policy responsiveness, we have examined the link between the politicisation of immigration and policy congruence. The expectation was that immigration policies and public attitudes would be less distant when politicisation is stronger.
We have considered three elements of the politicisation of immigration and compared them to both public attitudes and the evolution of immigration-related policies: media attention, the strength of anti-immigration parties and the degree of collective social mobilisation by anti-immigration groups. For each, the expectation was that the greater the politicisation and the more consistent with public opinion, the more likely policy-makers are to respond to public demands. Perhaps unsurprisingly, no single factor can explain cross-national variation in the existence of an opinion-policy gap on immigration across the seven countries and for the period under study (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010). In none of the seven countries is the strength of antiimmigration parties directly related to more restrictive immigration policies. For the other two factors, we observe some connection with policy congruence in some countries but not in others.
One interesting difference emerges between Ireland and the UK, and the five other countries. In Ireland and the UK policies have evolved in a direction congruent with public preferences. In the other countries, policies in some domains have moved in the opposite direction to the public preferences (Austria, the Netherlands and Spain) while in others they are neither obviously congruent nor obviously at odds with the Of the three countries where we found a considerable opinion-policy gap, in Austria and the Netherlands media coverage amplifies the negative views of the general public, while in Spain immigration is not consistently politicised to amplify the public mood. Of the four countries where we found no opinion-policy gap, in Belgium and Switzerland the lack of policy congruence seems consistent with a pattern of erratic and possibly uninformed public opinion mood that political elites might feel free to ignore, especially in the absence of consistent politicisation in the public arena.
By contrast, in the two cases where we do find policy congruence (Britain and Ireland) we observe the same combination of a concerned and negatively disposed general public coupled with extensive media coverage of immigration and a strong mobilisation of anti-immigration movements in the debate. The fact that the media coverage and anti-immigrant mobilisation are consistent in the timing and direction of preferences with the negative views held by the general public serves to amplify demands for more restrictive policies. It is the joint pressure of multiple forms of politicisation that seems to induce policy congruence in the area of immigration; a policy field where political elites might not a priori be very willing to attend to the growing concerns of the public, as suggested by Freeman.
Even if only suggestive due to the limitations in the number of cases and the paucity of the data, our results have implications for future research. The role of various elements of the politicisation of immigration in the public arena needs to be considered alongside the trends in public opinion in future studies of immigration policies. Our study suggests that immigration is not necessarily a 'blind corner' of democratic political representation (Thomassen 2012), and there seems to be no reason why immigration politics should fall beyond the theoretical models of policy responsiveness. Our findings indicate that political elites actually respond to public pressures in the field of immigration-at least sometimes and for some domains of immigration policy. It just seems that they need more, and consistent, pressure in this field to respond in a responsive way. The policy 'thermostat' (Wlezien 1995) might just be a bit faulty and less sensitive than in other policy fields.

Disclosure statement
No potential conflict of interest was reported by the authors.

Notes
[1] We follow Givens and Luedtke (2004) by referring to restrictive vs. expansive immigration policies. Restrictive policies refer to strict entry control mechanisms as well as to limited rights and social benefits legally guaranteed to migrants already settled in the country. Expansive immigration policies refer to legal frameworks with few entry barriers as well as to legal norms guaranteeing multiple rights to migrants and giving them access to welfare benefits under the same conditions than country nationals.
The SOM project has received funding from the European Commission's Seventh Framework Programme (FP7/2007(FP7/ -2013 under grant agreement number 225522. Full information is available on http://www.som-project.eu/, including links to data files. [3] The surveys used are: for Austria, Election surveys (SW9409, SW9902, SW2006_01, AUTNES 2009); for Belgium, Belgian National Elections Studies (1991,1995,1999,2003); for Ireland, IMS data extracted from the Irish Political Studies journal and the ISSDA at UCD (2001UCD ( , 2003UCD ( , 2004UCD ( , 2007; for the Netherlands, the Dutch Parliamentary Elections Study/NKO joint data-set (1995,1998,2002,2003,2006,2010); for Spain, the CIS September barometer (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011); for Switzerland, the annual Sorgenbarometer (every August, 1996(every August, -2011; and for the UK, the IPSOS-Mori Issues Index with yearly data (every June, 1997(every June, -2011. Despite undertaking a thorough search of all relevant national and crossnational data sources, these were the only surveys that covered a reasonable number of time points for the three decades. Correlations are moderate: 0.28 (Pearson), 0.21 (Kendall's tau). The cross-correlograms suggest a temporal positive 'lead' of public opinion concern on policy activity only in the UK, and a negative one (more concern results in less activity) in the Netherlands. ; with family and residence policy indicators is moderate and negative (i.e. the more restrictive the attitudes the less restrictive the policies) at -0.11 and -0.33 respectively; and it is more substantial and positive with the asylum policy indicator (0.40). The cross-correlograms confirm that increases in negative attitudes only lead to more restrictive policies in Ireland and the UK, and suggest that they lead to more expansive policies in Austria and Switzerland. The cross-correlograms confirm that negative claims only leads policy direction in the expected way in Ireland and the UK and, to a certain extent, Switzerland.