UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Multiculturalism and political parties : exploring the positions that parties take and their influence… Westlake, Daniel Jesse 2017

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2017_september_westlake_daniel.pdf [ 4.33MB ]
Metadata
JSON: 24-1.0348691.json
JSON-LD: 24-1.0348691-ld.json
RDF/XML (Pretty): 24-1.0348691-rdf.xml
RDF/JSON: 24-1.0348691-rdf.json
Turtle: 24-1.0348691-turtle.txt
N-Triples: 24-1.0348691-rdf-ntriples.txt
Original Record: 24-1.0348691-source.json
Full Text
24-1.0348691-fulltext.txt
Citation
24-1.0348691.ris

Full Text

MULTICULTURALISM AND POLITICAL PARTIES: EXPLORING THE POSITIONS THAT PARTIES TAKE AND THEIR INFLUENCE ON POLICY  by  Daniel Jesse Westlake  B.A., Carleton University, 2009  M.A., The University of British Columbia, 2010   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF   DOCTOR OF PHILOSOPHY  in    The Faculty of Graduate and Postdoctoral Studies  (Political Science)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  June 2017  © Daniel Jesse Westlake, 2017      ii Abstract Global migration and the rise of far-right parties have increased the importance of political debates surround multiculturalism. These two forces pull parties in competing directions. On one hand, migration increases the number of ethnic minority voters in countries, increasing the pressure on parties to support multiculturalism. As ethnic minorities become a larger portion of the electorate, parties have a greater incentive to respond to their interests. On the other hand, the emergence of far-right parties places pressure on parties to oppose multiculturalism as parties try to prevent anti-multicultural voters from defecting.  This dissertation maps the development of multiculturalism over time, examines parties’ influence over policy adoption, and ethnic minorities’ and far-right parties’ influence over mainstream party positions.  It includes three sets of findings.  First, multiculturalism policies are subject to path dependence. Second, parties influence policy adoption but only when there is cross-party support for policy adoption, with mainstream right parties having a particularly important effect on adoption. Third, parties respond to the competing pressures of ethnic minorities and far-right parties. Increases in ethnic minority electoral strength increase mainstream parties’ support of multiculturalism, but only in single member district electoral systems, while the emergence of far-right parties decreases it. Both of these factors have a particularly powerful affect on mainstream right parties.   The dissertation employs a mix of cross-country and single country analysis. Cross-country analysis establishes trends in policy development, parties’ impact on policy, and ethnic minorities’ and far-right parties’ influence over mainstream parties. It analyzes data from 21 industrialized countries using descriptive statistics, time-series  iii cross-section regression, and Cox proportional hazard models in order to establish broadly generalizable trends in policy development and parties’ influence over policy. The dissertation then conducts detailed case analysis focusing on Canada and the Netherlands. Analysis of Canada demonstrates the impact of ethnic minority pressure and electoral systems on party positions. Analysis of the Netherlands demonstrates the impact that far-right parties are having on mainstream parties.                   iv Lay Summary This dissertation examines the role of political parties in the development of multiculturalism policy.  It looks at the conditions under which parties influence policy and the way ethnic minorities and far-right parties change mainstream party positions.  The dissertation finds that the likelihood of policy adoption increases only when both government and opposition parties support multiculturalism.  In particular mainstream right support has a strong impact on policy adoption.  The dissertation further finds that increases in the electoral power of ethnic minorities increases parties’ support of multiculturalism in single member district (often called first past the post) electoral systems but not in proportional systems.  In contrast, the emergence of far-right anti-immigrant parties decreases mainstream party support for multiculturalism.  Finally, the dissertation shows that mainstream parties are more responsive to both increases in ethnic minorities’ electoral power and the emergence of far-right parties.             v Preface This dissertation is the product of an extensive examination of parties’ role in the development of multiculturalism policies. My research involved the use of data from a variety sources as well as the creation of two data sets.  I expanded the the Banting and Kymlicka Multiculturalism Policy Index (Multiculturalism Policy Index, 2016) creating a time-series cross-section data set from an index that scored countries in three years, 1980, 2000, and 2010.  I also used Manifesto Project Data (Volkens et al, 2013b) to create data sets the measured the development of multiculturalism policies and party positions on multiculturalism from 1960-2011. I drew on data on the size of countries’ foreign-born population (United Nations, Department of Economic and Social Affairs, Population Division, 2008) and access to citizenship (Janoski, 2010) to create a measure of ethnic minority electoral strength for 19 of the 21 countries in the Banting and Kymlicka Index.  Finally, I coded party manifestos for Canadian and Dutch parties obtained from three sources (Benoit et al., 2009; Birch et al., 2016; Volkens et al., 2013a) to obtained more detailed measures of Canadian and Dutch party positions.  This data formed the basis of a number of different tests of the theories put forward in the dissertation. I used data on policy adoption to examine the impact of particular multiculturalism policies on further policy development and parties influence on policy adoption using both descriptive statistics and hazard modeling. I used data on ethnic minority electoral strength as well as data on the presence of far-right parties to test ethnic minorities, electoral systems, and far-right parties’ impact on mainstream positions. For this analysis, I conducted tests using time-series cross-section analysis. Finally, I conducted additional single case analysis on Canada and the Netherlands using  vi Canadian census data (Computing in the Humanities and the Social Sciences, 2014) and my own coding of Canadian and Dutch manifestos. All analyses in this dissertation are my own original work.  A version of chapter 6 has been published as a single author academic journal article. Westlake, D. (2016) Multiculturalism, Political Parties, and the Conflicting Pressures of Ethnic Minorities and Far-Right Parties. Party Politics. Online First. All of the analysis and writing in the article (as well as in the dissertation chapter) is my own original work. In addition to this, some of the analysis in chapter 7, on Ukrainian-Canadian lobbying of the Royal Commission on Bilingualism and Biculturalism was included in my Masters thesis, Building Multiculturalism: The Contribution of the Ukrainian-Canadian Community to a Re-Thinking of Canadian Identity. All analysis in that chapter is my own. Finally, scores for the year 1990 that I created as part of the expansion of the Multiculturalism Policy Index have been added to the online version of the index and can be found at: http://www.queensu.ca/mcp/home.              vii Table of Contents Abstract .................................................................................................................................. ii Lay Summary ....................................................................................................................... iv Preface .................................................................................................................................... v Table of Contents ............................................................................................................... vii List of Tables ........................................................................................................................ ix List of Figures ....................................................................................................................xiii Acknowledgements ........................................................................................................... xv Dedication ......................................................................................................................... xvii Chapter 1 Introduction ...................................................................................................... 1 Defining Multiculturalism .......................................................................................................... 6 Situating the Dissertation within the Literature ............................................................... 11 Dissertation Structure .............................................................................................................. 18 Chapter 2 Theory: Explaining Policy Development and Party Positions ........... 21 Multiculturalism in the Context of Broader Migration Politics ..................................... 22 The Development of Multiculturalism Policies ................................................................. 45 Parties and Policy Adoption .................................................................................................... 54 Ethnic Minorities, Far-Right Parties, and Party Positions .............................................. 65 Summary....................................................................................................................................... 73 Chapter 3 Methods and Measurement ......................................................................... 76 Multi-Country Analysis- Variables and Measurement ..................................................... 76 Multi-Country Analysis- Methods Used .............................................................................. 107 Single Case Analysis- Overall Approach ............................................................................. 115 Qualitative Methods- Case Selection ................................................................................... 116 Single Case Analysis- Methods Used ................................................................................... 121 Chapter 4 The Development of Multiculturalism Policies .................................. 126 The Distribution of Multiculturalism Policies ................................................................. 127 Multiculturalism and Path Dependence ............................................................................ 141 What Counts as Multiculturalism? ...................................................................................... 155 Summary..................................................................................................................................... 160 Chapter 5 Political Parties and the Adoption of Multiculturalism Policies .... 163 Left and Right Parties and Support for Multiculturalism ............................................. 164 Parties and Overall Policy Adoption ................................................................................... 171 Parties and the Adoption of Particular Policies .............................................................. 192 Summary..................................................................................................................................... 197 Chapter 6 Ethnic Minorities, Far-Right Parties, Electoral Systems, and Party Positions............................................................................................................................ 200 Ethnic Minority Advocacy in Favour of Multiculturalism ............................................. 200 Ethnic Minorities, Electoral Systems, and Support for Multiculturalism ................. 204 Far-Right Parties and Left and Right Support for Multiculturalism .......................... 221  viii Summary..................................................................................................................................... 225 Chapter 7 Ethnic Minorities, Electoral Systems, Far-Right Parties and Multiculturalism in Canada and the Netherlands.................................................. 227 The Development of Canadian Multiculturalism ............................................................ 228 Parties’ Appeals to Ethnic Minorities ................................................................................. 233 Electoral Institutions and the Power of Ethnic Minorities in Canada........................ 240 Mainstream Left and Right Support for Multiculturalism in the Netherlands ........ 251 Far-Right Parties and Mainstream Support for Multiculturalism in the Netherlands ....................................................................................................................................................... 258 Summary..................................................................................................................................... 273 Chapter 8 Conclusion ..................................................................................................... 275 Summary of Findings and Implications ............................................................................. 275 Limitations of the Research................................................................................................... 280 Areas for Future Research ..................................................................................................... 283 Conclusion .................................................................................................................................. 286 Bibliography .................................................................................................................... 287 Appendices ....................................................................................................................... 304 Appendix A Party Categorizations ....................................................................................... 304 Appendix B Effect on Total Policy Adoption for Models Starting in 1970, 1975, and 1980 ............................................................................................................................................. 311 Appendix C Effects on Individual Policy Adoption for Models Starting in 1975 and 1980 ............................................................................................................................................. 313 Appendix D Table 4.2 Values with No Controls................................................................ 316 Appendix E Controls for Table 4.2 and Associated Appendix ...................................... 319 Appendix F Table 4.3 Values With No Controls ............................................................... 331 Appendix G Controls for Table 4.3 and Associated Appendices .................................. 334 Appendix H Table 5.1 Without Ethnic Minority Electoral Strength Controls .......... 357 Appendix I Table 5.2 Without Ethnic Minority Electoral Strength Controls............ 358 Appendix J The Effects of Cross-Party Positions on Specific Policy Adoption ......... 359 Appendix K Right and Left Party Effects on Individual Policies .................................. 362 Appendix L Effects of Ethnic Minority Electoral Strength, Electoral Systems, and Far-Right Parties on Party Positioning Using Party Positions in the Most Recent Election ....................................................................................................................................... 368 Appendix M Controls for Table 7.1 ...................................................................................... 371 Appendix N Controls for Table 7.2 ...................................................................................... 379     ix List of Tables Table 3.1: The Banting and Kymlicka Index by Policy Types................................... 79 Table 3.2: Decision Rules for the Banting and Kymlicka Index ................................ 84 Table 4.1: Instances of Policy Retrenchment ........................................................... 144 Table 4.2: Effects of Individual Policies on Overall Policy Adoption (1970 Start Date) ........................................................................................................................... 146 Table 4.3: The Effect of Individual Policies on Individual Policy Adoption (1970 Start Date) ................................................................................................................. 150 Table 5.1: Summary of Party Positions Across Countries ....................................... 168 Table 5.2: Cross-Party Positions' Impact on Policy Adoption with Ethnic Minority Electoral Strength Controls ...................................................................................... 174 Table 5.3: Government Party Positions' Impact on Policy Adoption with Ethnic Minority Electoral Strength Controls ...................................................................... 178 Table 5.4: Right Party Positions' Impact in Policy Adoption with Ethnic Minority Electoral Strength Controls ...................................................................................... 184 Table 5.5: Left Party Positions' Impact on Policy Adoption with Ethnic Minority Electoral Strength Controls ...................................................................................... 185 Table 5.6: Right Party Position's Impact on Policy When Left Parties are in Government ............................................................................................................... 188 Table 5.7: Left Party Positions' Impact on Policy Adoption when Right Parties are in Government ........................................................................................................... 190 Table 6.1: Determinants of Cross-Party Positions on Multiculturalism ................. 206 Table 6.2: Determinants of Mainstream Right Party Positions in Multiculturalism .................................................................................................................................... 215 Table 6.3: Determinants of Left Party Positions on Multiculturalism .................... 220 Table 7.1: The Effect of the Number of Immigrants in a District on Party Vote Share Outside of Quebec by Election ....................................................................... 242 Table 7.2: The Effect of Immigrants in a District on Party Vote Share in Quebec by Election ...................................................................................................................... 243 List of Tables (Appendix) Table A.1: Party Categorizations (taken from Manifesto Project data) ................. 304 Table B.2: Effects of Individual Policies on Overall Policy Adoption (1975 Start Date) ........................................................................................................................... 311 Table B.3: Effects of Individual Policies on Overall Policy Adoption (1980 Start Date) ........................................................................................................................... 312 Table C.1: The Effect of Individual Policies on Individual Policy Adoption (1975 Start Date) ................................................................................................................. 313 Table C.2: The Effect of Individual Policies on Individual Policy Adoption (1980 Start Date) ................................................................................................................. 315 Table D.1: Effects of Individual Policies on Overall Policy Adoption (1960 Start Date) ........................................................................................................................... 316 Table D.2: Effects of Individual Policies on Overall Policy Adoption (1975 Start Date) ........................................................................................................................... 317 Table D.3: Effects of Individual Policies on Overall Policy Adoption (1980 Start Date) ........................................................................................................................... 318  x Table E.1: Affirmation (1970 Start Date) ................................................................. 319 Table E.2: Affirmation (1975 Start Date) ................................................................. 319 Table E.3: Affirmation (1980 Start Date) ................................................................. 320 Table E.4: MC School (1970 Start Date) .................................................................. 320 Table E.5: MC School (1975 Start Date) .................................................................. 321 Table E.6: MC School (1980 Start Date) .................................................................. 321 Table E.7: Media (1970 Start Date) .......................................................................... 322 Table E.8: Media (1975 Start Date) .......................................................................... 322 Table E.9: Media (1980 Start Date) .......................................................................... 323 Table E.10: Dual Citizenship (1970 Start Date) ....................................................... 323 Table E.11: Dual Citizenship (1975 Start Date) ....................................................... 324 Table E.12: Dual Citizenship (1980 Start Date) ....................................................... 324 Table E.13: Exemptions (1970 Start Date) ............................................................... 325 Table E.14: Exemptions (1975 Start Date) ............................................................... 325 Table E.15: Exemptions (1980 Start Date) ............................................................... 326 Table E.16: Funding (1970 Start Date) ..................................................................... 326 Table E.17: Funding (1975 Start Date) ..................................................................... 327 Table E.18: Funding (1980 Start Date) ..................................................................... 327 Table E.19: Bilingual Education (1970 Start Date) .................................................. 328 Table E.20: Bilingual Education (1975 Start Date) .................................................. 328 Table E.21: Bilingual Education (1980 Start Date) .................................................. 329 Table E.22: Affirmative Action (1970 Start Date) ................................................... 329 Table E.23: Affirmative Action (1975 Start Date) ................................................... 330 Table E.24: Affirmative Action (1980 Start Date) ................................................... 330 Table F.1: The Effect of Individual Policies on Individual Policy Adoption (1960 Start Date) ................................................................................................................. 331 Table F.2: The Effect of Individual Policies on Individual Policy Adoption (1975 Start Date) ................................................................................................................. 332 Table F.3: The Effect of Individual Policies on Individual Policy Adoption (1980 Start Date) ................................................................................................................. 333 Table G.1: Affirmation (1970 Start Date) ................................................................ 334 Table G.2: Affirmation (1975 Start Date) ................................................................ 335 Table G.3: Affirmation (1980 Start Date) ................................................................ 336 Table G.4: MC School (1970 Start Date) .................................................................. 337 Table G.5: MC School (1975 Start Date) .................................................................. 338 Table G.6: MC School (1980 Start Date) .................................................................. 339 Table G.7: Media (1970 Start Date) .......................................................................... 340 Table G.8: Media (1975 Start Date) .......................................................................... 341 Table G.9: Media (1980 Start Date) .......................................................................... 342 Table G.10: Dual Citizenship (1970 Start Date) ....................................................... 343 Table G.11: Dual Citizenship (1975 Start Date) ....................................................... 344 Table G.12: Dual Citizenship (1980 Start Date) ....................................................... 345 Table G.13: Exemptions (1970 Start Date) ............................................................... 346 Table G.14: Exemptions (1975 Start Date) ............................................................... 347 Table G.15: Exemptions (1980 Start Date) ............................................................... 347 Table G.16: Funding (1970 Start Date) .................................................................... 348  xi Table G.17: Funding (1975 Start Date) .................................................................... 349 Table G.18: Funding (1980 Start Date) .................................................................... 350 Table G.19: Bilingual Education (1970 Start Date) ................................................. 351 Table G.20: Bilingual Education (1975 Start Date) ................................................. 352 Table G.21: Bilingual Education (1980 Start Date) ................................................. 353 Table G.22: Affirmative Action (1970 Start Date) ................................................... 354 Table G.23: Affirmative Action (1975 Start Date) ................................................... 355 Table G.24: Affirmative Action (1980 Start Date) ................................................... 356 Table H.1: Cross-Party Positions' Impact on Policy Without Ethnic Minority Electoral Strength Controls ...................................................................................... 357 Table I.1: Government Party Positions' Impact on Policy Without Ethnic Minority Electoral Strength Controls ...................................................................................... 358 Table J.1: Cross Party Positions' Impact on Different Policies (1970 Start Date) .. 359 Table J.2: Cross Party Positions' Impact on Different Policies (1975 Start Date) .. 360 Table J.3: Cross Party Positions' Impact on Different Policies (1980 Start Date) .. 361 Table K.1: Right Party's Positions' Impact on Different Policies (1970 Start Date) .................................................................................................................................... 362 Table K.2: Right Party's Positions Impact on Different Policies (1975 Start Dates) .................................................................................................................................... 363 Table K.3: Right Party's Impact on Different Policies (1980 Start Date) ............... 364 Table K.4: Left Party's Impact on Different Policies (1970 Start Date) ................. 365 Table K.5: Left Party's Impact on Different Policies (1975 Start Date) ................. 366 Table K.6: Left Parties' Impact on Different Policies (1980 Start Date) ................. 367 Table L.1: Determinants of Mainstream Party Positions ........................................ 368 Table L.2: Determinants of Right Party Positions ................................................... 369 Table L.3: Determinants of Left Party Positions ..................................................... 370 Table M.1: Immigrant Effects on Vote Share by District in 1993 Outside of Quebec .................................................................................................................................... 371 Table M.2: Immigrant Effects on Vote Share by District in 1997 Outside of Quebec .................................................................................................................................... 372 Table M.3: Immigrant Effects on Vote Share by District in 2000 Outside of Quebec .................................................................................................................................... 373 Table M.4: Immigrant Effects on Vote Share in 2004 Outside of Quebec .............. 374 Table M.5: Immigrant Effects on Vote Share by District in 2006 Outside of Quebec .................................................................................................................................... 375 Table M.6: Immigrant Effects on Vote Share by District in 2008 Outside of Quebec .................................................................................................................................... 376 Table M.7: Immigrant Effects on Vote Share by District in 2011 Outside of Quebec .................................................................................................................................... 377 Table M.8: Immigrant Effects on Vote Share by District in 2015 Outside of Quebec .................................................................................................................................... 378 Table N.1: Immigrant Effects on Vote Share by District in 1993 in Quebec .......... 379 Table N.2: Immigrant Effect on Vote Share by District in 1997 in Quebec ............ 379 Table N.3: Immigrant Effect on Vote Share by District in 2000 in Quebec ............ 379 Table N.4: Immigrant Effect on Vote Share by District in 2004 in Quebec ............ 380 Table N.5: Immigrant Effects on Vote Share by District in 2006 in Quebec .......... 380  xii Table N.6: Immigrant Effects on Vote Share by District in 2008 in Quebec .......... 380 Table N.7: Immigrant Effects on Vote Share by District in 2011 in Quebec .......... 381 Table N.8: Immigrant Effects on Vote Share by District in 2016 in Quebec .......... 381      xiii List of Figures Figure 2.1: MIPEX Access To Nationality Scores and Multiculturalism Policies .... 26 Figure 2.2: MIPEX Anti-Discrimination and Multiculturalism Scores .................... 31 Figure 2.3: CIVIX and Multiculturalism Scores ........................................................ 32 Figure 3.1: The Distribution of Party Scores for Support for Multiculturalism ...... 90 Figure 3.2: The Correlation between Recoded Scores and Manifesto Project Multiculturalism Scores .............................................................................................. 92 Figure 3.3: The Correlation between the Inverted Barriers to Naturalization Index and Jus Soli Laws ........................................................................................................ 97 Figure 3.4: Mean Ethnic Minority Electoral Strength by Country ........................... 99 Figure 3.5: The Development of Multiculturalism Policies Over Time ................... 108 Figure 3.6: Smoothed Hazard Rates for Total Policy Adoption in Settler and Non-Settler Countries ........................................................................................................ 111 Figure 3.7: The Strength of Multiculturalism Prior to the Emergence of the Far-Right Compared to the Drop in Support for Multiculturalism Caused by the Far-Right........................................................................................................................... 121 Figure 4.1: Multiculturalism Policies Adopted by 2011 ........................................... 128 Figure 4.2: Multicultural Policies Adopted by 1990 ................................................ 129 Figure 4.3: Multicultural Policies Adopted by 1970 ................................................ 131 Figure 4.4: Recognition Policy Adoption by 2011 .................................................... 134 Figure 4.5: Recognition Policy Adoption by 1990 .................................................... 134 Figure 4.6: Participation Policy Adoption by 2011 .................................................. 136 Figure 4.7: Participation Policy Adoption by 1990 .................................................. 136 Figure 4.8: Support Policy Adoption by 2011 .......................................................... 137 Figure 4.9: Support Policy Adoption by 1990 .......................................................... 138 Figure 4.10: Average Policy Adoption by MCP Strength in 1990 ........................... 140 Figure 4.11: The Adoption of Multiculturalism over Time ..................................... 142 Figure 4.12: The Adoption of Recognition Policies over Time ................................ 143 Figure 4.13: Average Policies Adopted Before, After, and Without Affirmation Policies ....................................................................................................................... 148 Figure 4.14: Percentage of Years in which Policy Adoption Occurs by Presence of Affirmation Policy ..................................................................................................... 153 Figure 5.1: Left-Right Positions Compared to Support for Multiculturalism ........ 165 Figure 5.2: Left-Right Position Compared to Support for Multiculturalism Amongst Parties that Make At Least One Statement on Multiculturalism ............................ 166 Figure 5.3: Support for Welfare Compared to Support for Multiculturalism ....... 167 Figure 5.4: Average Cross-Party, Left, and Right Positions by Country ................ 169 Figure 5.5: Cross-Party Positions When Policies Are Adopted ............................... 172 Figure 5.6: The Effect of Cross-Party Positions on Total Policy Adoption (1980 Start Date) ........................................................................................................................... 175 Figure 5.7: Government Positions When Policies Adopted ..................................... 177 Figure 5.8: Left and Right Consensus and Policy Adoption .................................... 180 Figure 5.9: Right and Left Effects on Policy Adoption (1980 Start Date) ............... 186 Figure 5.10: Right and Left Effects on Policy Adoption when the Opposite Party is in Government ........................................................................................................... 189  xiv Figure 5.11: Cross Party Effects on Different Policy Adoption (1980 Start Date) .. 193 Figure 5.12: Right Party Effects on Different Policy Adoption (1980 Start Date) .. 195 Figure 5.13: Left Party Effects on Different Policy Adoption (1980 Start Date) .... 196 Figure 6.1: Ethnic Minority Influence Over Cross-Party Positions in SMD Systems over Time ................................................................................................................... 207 Figure 6.2: Ethnic Minority Influence Over Cross-Party Positions in SMD and Proportional Systems ................................................................................................ 209 Figure 6.3: Predicted and Average Support for Multiculturalism in SMD Countries .................................................................................................................................... 211 Figure 6.4: Predicted and Average Support for Multiculturalism in Proportional Countries ................................................................................................................... 213 Figure 6.5: Ethnic Minority Influence Over Mainstream Right party Positions in SMD systems over Time ............................................................................................ 216 Figure 6.6: Ethnic Minority Influence over Mainstream Right Positions in SMD and Proportional Systems ................................................................................................ 218 Figure 6.7: Party Positions When Far-Right Parties are Absent and Present ........ 222 Figure 6.8: Change in Party Support for Multiculturalism Across Diffent Countries When a Far-Right Party is Present........................................................................... 224 Figure 7.1: Seat Changes with No Immigrant Effect ............................................... 246 Figure 7.2: Liberal Seat Changes with No Immigrant Effect .................................. 247 Figure 7.3: Reform and Conservative Seat Changes with No Immigrant Effect .... 248 Figure 7.4: Reform and Conservative Support and Opposition to Multiculturalism .................................................................................................................................... 250 Figure 7.5: Recoded Total Support for Multiculturalism ........................................ 252 Figure 7.6: Recoded Positive Statements on Multiculturalism ................................ 259 Figure 7.7: Recoded Negative Statements on Multiculturalism .............................. 260 Figure 7.8: Anti-Immigrant and Anti-Multicultural Appeals by the LPF and PVV .................................................................................................................................... 269        xv Acknowledgements  I am grateful to my dissertation supervisor Richard Johnston and committee members Antje Ellermann and Andrew Owen.  They have provided invaluable advice throughout the research and writing processes and have endured unpolished writing in several early drafts. I have learned a great deal from each of my committee members about migration, multiculturalism, parties, electoral systems, and quantitative methods.  I am deeply indebted to them for this.  I am also grateful to Anjali Bohlken and Christopher Kam for their advice with the hazard modeling in the dissertation.  Early versions of chapters in the dissertation were presented at different American Political Science Association and Canadian Political Science Association Meetings.  Julie Croskill, Peter Graefe, Edward Koning, Clare McGovern and Phil Triadafilopoulos all provided important feedback based on those presentations.  I also presented parts of this dissertation at the UBC Institute for European Studies’ Doctoral Fellows meeting, and received valuable feedback from Augstín Goenaga, Alexander Hemingway, Kurt Hübner, Conrad King, Grace Lore, Julia Partheymüller, Andrea Nuesser, and Serbulent Turan.  As a PhD student in the UBC department of political science I have been the beneficiary of a vibrant scholarly community.  My thinking about multiculturalism, Canadian politics, and comparative politics has been greatly influenced by a number of professors outside of my dissertation committee.  I am indebted to Barbara Arneil for her supervision of Masters thesis on Ukrainian-Canadians influence over multiculturalism.  That work served as the starting part for my research into the party politics and multiculturalism.  I am also grateful to Gerald Baier, who I have done extensive work for  xvi as a teaching assistant and who has greatly influenced my approach to teaching Canadian politics.  I am finally thankful to all of the faculty in the department that through course work, the comprehensive exam process, workshops, or other interaction have influenced my thinking.  I have also been lucky to be part of a vibrant graduate student community that has shaped my thinking about political science, particularly about parties and multiculturalism. In particular, Jennifer Allan, Nathan Allen, Edana Beauvais, Yoel Kornreich, Yana Gorokhovskaia, Alex Held, Grace Lore, Clare McGovern, David Moscrop, Andrea Nusser, Elizabeth Schwartz, and Carla Winston have made important contributions to my thinking through discussions or course work.  I finally am grateful to my parents, Wendy Caplan and Don Westlake, and my brother Joshua Westlake, for all of their support and encouragement throughout my PhD work.           xvii Dedication    To my Bubba, Rose Caplan 1 Chapter 1 Introduction The high levels of migration brought on by globalization make understanding immigrant integration, particularly the development of multiculturalism policies, important. As migration increases, countries have to find ways to accommodate diverse populations. In particular, increases in a countries' immigrant population should place pressure on countries to adopt multiculturalism as immigrants seek ways to integrate without giving up their cultural and religious practices. Despite this, policy adoption in a country and the size of its immigrant population are only loosely related (Multiculturalism Policy Index, 2016). While there are countries with large immigrant populations and strong multiculturalism policies such as Australia and Canada, as well as countries with few immigrants and multiculturalism policies such as Japan, there are also several outliers. Switzerland has an immigrant population similar to New Zealand but much weaker multiculturalism policies. The level of multiculturalism in a country is not simply a function of its immigrant population. This dissertation considers the role of parties in the development of multiculturalism, focusing on two major questions.  First, it asks what impact parties have on policy adoption?  If parties are translating citizens’ views into policy, changes in parties’ positions should have a significant impact on the likelihood of policy adoption.  This dissertation finds that party positions matter to policy adoption, but do so in a complicated way.  Support from parties across the political spectrum, not just from the governing party, is necessary to increase the likelihood of policy adoption.  Particularly important is support from mainstream right parties, even if they are in opposition.    2 Second, the dissertation asks why parties take the positions that they do?  In doing so, it looks at the competing pressures that parties face from ethnic minorities and far-right parties.  Because ethnic minorities benefit from multiculturalism, increases in their electoral strength should increase the incentives for parties to support such policies.  This increase in incentives should be particularly strong in single member district electoral systems, as the geographic concentration of minorities in particular electoral districts amplifies their influence over election results in these systems.  Far-right challengers, in contrast, should push mainstream parties to decrease their support of multiculturalism.  Political parties, especially mainstream right parties, are a key missing link in explaining the different levels of policy adoption across industrialized countries. As actors that control both the legislative and executive policy making processes, parties have a great deal of influence over policy adoption. Parties are crucial gatekeepers in the policy process. Government parties have the ability to pass the legislation and provide the finances needed for policy change. Partisan ministers that control departments implement policy changes.  Cross-party support for multiculturalism is critical to long-term policy expansion because it ensures that governments will continue to support multiculturalism regardless of the party that wins election, and because it ensures policies easy passage through legislatures. Finally, parties are supposed to be a link between the public and the policy making process (Budge et al., 2012; Kedar, 2009). Because parties are the main competitors in elections, they serve as the primary vehicle through which the public indicates policy preferences. When democracies are functioning well, parties should offer voters different policy platforms, voters should select the platforms that they prefer, and successful parties should implement the policies that they promised to the electorate. It is  3 important to understand parties' roles in policy development both because parties control the institutions central to policy change and because of the extent to which parties' influence over policy has meaningful consequences democratic representation. On multiculturalism, as with other issues related to immigration, parties have to be particularly careful. On one hand, parties have to respond to changes in the demographics of the electorate. As countries become more diverse, pressure on parties to adopt positions that speak to the needs of immigrants and other ethnic minorities increases. Parties' incentives to respond to immigrants and other ethnic minorities, however, are not just a function of the number of immigrants in a country. They also depend on immigrants' and ethnic minorities' electoral power. This is influenced by rules regarding access to citizenship, and the way that electoral systems translate minorities' votes into seats. On the other hand, parties also have to respond to challenges from the far-right. As multiculturalism and immigration issues have grown in salience, far-right parties have emerged in countries across Europe to advocate for more restrictive immigration policies and assimilationist integration policies. Parties have to balance these competing pressures when taking positions on multiculturalism.  Parties are likely to face challenges translating their positions into policy. The first of these challenges lies in the degree to which path dependence affects multiculturalism. These policies grant benefits to ethnic minorities that range from funding for organizations to programs such as mother-tongue education to affirmative action to dress code accommodations that make it easier for minorities to find employment. Removing these policies imposes concentrated costs on a group of voters, and makes it likely that they will punish the party responsible for such retrenchment. This is likely to place  4 limitations on the extent to which a party opposed to multiculturalism can influence policy change after policies have been adopted.  Second, parties' actions on multiculturalism are likely to influence the salience of multiculturalism and immigration issues. Individuals opposed to multiculturalism and immigration are more likely to vote based on that opposition when a political party raises the salience of such issues (Green-Pedersen and Krogstrup, 2008; Koopmans et al., 2012; Odmalm, 2012; Perlmutter, 1996). Governing parties, therefore, have to pay careful attention to opposition parties when deciding whether to follow through on commitments to adopt multiculturalism. If a government party tries to introduce multiculturalism in the face of strong opposition, they could face backlash that hurts them in the next election. If, however, the opposition is supportive of multiculturalism, governments can introduce policy without much difficulty. It is therefore important not to simply assume that parties matter to policy adoption because they control crucial policy levers. It is important to understand the influence of both government and opposition parties on policy.  This dissertation takes as its departure point much of the existing work on parties and immigration policy. It builds on existing work such as Koopmans et al. (2012), that shows that increases in the size of a country's immigrant population increases its multiculturalism policy commitment, by examining the extent to which parties and electoral systems affect ethnic minorities influence over policy. It builds on work examining the relationship between parties and policy adoption, such as Bruening and Luedtke (2008), by looking specifically at multiculturalism and by considering the importance of both government and opposition parties on policy adoption as well as distinguishing between the different effects that changes in left party positions and right  5 party positions have on policy adoption. The dissertation thus presents a sophisticated analysis of how ethnic minorities and far-right parties can work through mainstream parties to influence policy adoption. Two main theories guide the dissertation. One looks at the conditions under which parties are likely to influence policy, and the other at the way that mainstream parties respond to pressures from ethnic minorities and the far-right. The dissertation finds that partisan support for multiculturalism increases the likelihood of policy adoption, but only if both government and opposition parties support policy adoption. Furthermore, changes in mainstream right positions have a particularly strong influence on policy adoption. The dissertation additionally finds that increases in ethnic minority electoral strength increase partisan support for multiculturalism, but only in single member district electoral systems. In contrast, and consistent with existing literature (Arzheimer, 2009; Givens, 2005; Ivarsflaten, 2008; Oesch, 2008; van Spanje, 2010), the emergence of a far-right party reduces mainstream parties' support for multicultural policy. Both changes in ethnic minority electoral strength and the emergence of far-right parties have particular strong effects on mainstream right positions. This introductory chapter will provide a brief overview of the dissertation. It will first provide a definition of multiculturalism that will guide the theory and analysis through the rest of the work. It will then place the dissertation of the context of existing work on parties, multiculturalism, and immigration. The final section of the chapter will provide a brief chapter outline discussing how each chapter fits into the broader argument.  6 Defining Multiculturalism  Any comparative work must address the different, and often contradictory, understandings of multiculturalism. The term can be used to refer to the presence of national minorities (such as French, Italians, Germans, and Roma in Switzerland), policies that deal with the concerns of aboriginals and national minorities, the presence of a large immigrant population in a country, or policies that speak to ethnic minorities' interests. It can also describe the development of separate institutions and social organizations for national minorities, aboriginals, or immigrant communities (Bloemraad and Wright, 2014). Any definition of multiculturalism thus has to respond to three questions: Does the definition speak to demography or to policy? To what groups does policy respond? What kinds of policies ought to be considered multicultural? This dissertation considers multiculturalism from a policy perspective. A country may have a high level of demographic diversity without having many multiculturalism policies. Countries respond to diversity differently. Some push back against it with assimilationist policies, some ignore it by leaving culture to the private sphere, and some pursue multicultural policies that accommodate minorities' cultural practices. Canada, Switzerland, and the United States all have substantial immigrant populations but different approaches to immigrant integration. Canada has a strong multiculturalism policy designed to accommodate cultural differences, the United States has a moderate multiculturalism policy matched by liberal institutions that tend to place culture within the private sphere, while Switzerland has very few policies that accommodate minorities (Banting and Kymlicka, 2006; Bloemraad, 2006).   7 This work focuses on policies that respond to the interests of ethnic minorities. Kymlicka (1995) draws distinctions between multiculturalism that addresses the needs of national minorities, aboriginals, and immigrant and similar ethnic minorities.1 Each of these groups has very different needs and makes different demands of the state. National minorities and aboriginals often seek different levels of self-determination in order to ensure some level of political as well as social autonomy. Immigrants and similar ethnic minorities, in contrast, are often looking for policies that allow them to integrate into existing political and social communities while maintaining their own cultural practices. These demands require very different policy responses and as a result produce very different politics. As a result, they should be studied separately. Drawing a distinction between multiculturalism aimed at ethnic minorities and multiculturalism aimed at national minorities is particularly important because of the extent to which these types of policies can come into conflict. Tensions can develop between the claims of ethnic minorities and the policies pursued by national minorities seeking to protect their culture. National minorities that feel threatened by the majority group in a country may pursue highly assimilationist policies towards immigrants. These policies often place pressure on ethnic minorities to assimilate into the national minority community. Quebec's Charter of Values for example, which prohibited public employees from wearing conspicuous religious symbols such as hijabs or kippahs, would limit the ability of religious minorities to maintain their practices while engaging in public life.2                                                1 Kymlicka refers to these groups as polyethnic groups. 2 It is important to note that the Charter of Values did not pass the legislature in Quebec. There will often be disagreement amongst members of a national minority community as to what as necessary to protect and preserve the national minority's culture. In Quebec, some nationalists argued that Charter Values was important to preservation of Quebec's identity while many other Quebec nationalists would have argued that it was not necessary.  8 National minorities may also see multiculturalism as a way to avoid recognizing national minorities' distinct status. In Canada, many Quebecers criticized multiculturalism because it replaced biculturalism (McRoberts, 1997) and in New Zealand, many Maori are wary of the extent to which multiculturalism can undermine their own recognition claims (Spoonley, 2005). It is possible for recognition of national minorities and multiculturalism to coexist, but the tensions that sometimes exist between these two types of policies make distinguishing between them important. Several theoretical approaches can justify multiculturalism. Charles Taylor (1994) makes an argument for multiculturalism that rests on the importance of recognition. For Taylor the way that groups are recognized has important implications for the way that they see themselves in society. Recognizing that people from different backgrounds belong in society is important to ensuring that they can be equal participants in it. For Carens (2000) and Parekh (2006) multiculturalism is important to ensuring people from different backgrounds can participate in society with equal dignity. They make the case that, because laws and practices do not develop in culturally neutral environments, accommodations may be required to ensure that people from different cultures have an equal ability to participate in society. For example, a country might need to grant exemptions from dress code requirements to Sikhs to allow them to join state institutions like the police and military. Carens' and Parekh's arguments for multiculturalism in this case would highlight the fact that many Western countries' designs of police uniforms took place in a cultural context where the majority of the population is not Sikh. As a result, dress codes have not taken into account Sikh religious practices to the same extent  9 as they would have taken into account other cultural practices.3 Finally, Kymlicka (1995; 1989) makes the argument that multiculturalism is necessary to allow individuals to exercise their liberal rights. Because a person's cultural background plays an important role in the way that they understand the world, the ability to maintain one’s culture is essential to the way that people exercise their rights. Under this theoretical framework, multiculturalism policies ensure that people can maintain their cultural backgrounds as a way to inform the way that they exercise the rights a liberal state grants to them. A right to freedom of religion, for example, might be of little value if societal practices (such as dress codes) made it difficult for one to practice a minority religion. Distinguishing between different theoretical justifications of multiculturalism is not necessary to create a policy definition. Each theoretical approach has in common the idea that the government has a role to play in supporting and protecting cultural diversity. Regardless of whether an appeal to the politics of recognition, equal dignity, or the exercise of equal rights justifies multiculturalism, it requires that the government intervene to ensure that minorities can integrate into society without giving up their cultural practices. Such intervention, be it symbolic or through the provision of resources                                                3 This is not to portray these arguments as defences of all cultural practices. There are limits to which an equal dignity approach to multiculturalism can justify forcing individuals to adhere to particular values.  Such an approach could over-ride a law that protects a particular value but limits the practice of a minority culture. For example, a society might enact a law requiring motorcycle riders to wear helmets because it values public safety. That society might create an exemption to the law for Sikhs because such a law limits their ability to wear religious headdress. For the majority in the community, there is no conflict between the motorcycle helmet law and their religious or cultural values and therefore the law imposes a limited burden. That burden however, is greater for many Sikhs because the law comes into conflict with their religious values. Such an exemption would recognize that a motorcycle helmet law imposes a much greater difficulty on the ability of Sikhs to live their lives with dignity than it does on others. A law against blasphemy that limits the ability of adherents to criticize particular religious practices, however, would be difficult to justify using Carens' or Parekh's theoretical perspectives.   10 to allow immigrants and minorities to preserve their culture, will serve as the basis for the definition of multiculturalism policy that runs through the dissertation. Multiculturalism is distinct from anti-discrimination and other integration policies. Anti-discrimination policies take steps to ensure that minorities do not face discrimination in areas of society such as housing, the labour market, or educational institutions. However, they do not necessarily include any protection for minority cultural practices. Policies such as these might include that anti-racism policies adopted in Britain's 1970s Race Relations policies (Geddes, 2003). Integration policies such as civic education programs or majority language programs are also not necessarily multicultural if they do not include other programs that help to support minority cultural practices. In some cases, other integration policies can end up replacing multiculturalism. In the Netherlands the adoption of civics courses were parts of a shift away from multiculturalism (Entzinger, 2006). A country can have strong anti-discrimination policies, devote considerable resources to immigrant integration, and still not have a strong multiculturalism policy if it does not do much to protect and support minority cultural practices. None of this is to suggest that multicultural policies are mutually exclusive with anti-discrimination and other integration policies. Indeed, countries that commit to multiculturalism are likely to have strong anti-discrimination policies and to make significant efforts to integrate immigrants into society. Multiculturalism, should however, not be conflated with anti-discrimination and integration policies in general.  Multiculturalism is also distinct from citizenship and immigration policies. Immigration policies cover a wide range of areas including rules governing the reception  11 of immigrants, citizenship acquisition, and integration. Multiculturalism policies speak only to the way that governments approach cultural diversity. While there may be a strong correlation between the adoption of multiculturalism and countries that have open immigration policies, these are separate policy areas. Situating the Dissertation within the Literature  There is little work tying together ethnic minority electoral strength, electoral systems, far-right parties, mainstream party positions, and multiculturalism. The dissertation makes three broad contributions to the existing work on parties and multiculturalism. First, it maps the development of multiculturalism policies, showing how path dependence influences policy. Second, it links parties to policy adoption, distinguishing between broad cross-party support, government support, and left and right support for policy adoption. Finally, the dissertation examines the relative influence of ethnic minorities, electoral systems, and far-right parties on mainstream support.  Mapping the Development of Policy Adoption There is considerable disagreement with respect to how multiculturalism has developed and what policies count as multicultural. Brubaker (2001) and Joppke (1999) argue that the late 1990s and early 2000s saw a significant backlash against multiculturalism that led to the adoption of more assimilationist integration policies. These findings echo work on Australia (Murphy et al., 2003; Tate, 2009) and in remarks made by former British Prime Minister David Cameron (2011) and German Chancellor Angela Merkel (BBC, 2010). In contrast to this, Banting and Kymlicka (2013) find little evidence of multicultural retrenchment. These conflicting findings raise an important question as to whether the level of multiculturalism in North America and Europe has been increasing or decreasing.  12 Complicating this question is significant disagreement over what ought to count as a multiculturalism policy. A variety of indexes measure citizenship and immigration policy (see Howard, 2009; Janoski, 2010; Koopmans et al., 2005). Banting and Kymlicka's (2006) measure, however, is the most comprehensive that looks specifically at multiculturalism. The inclusion of a number of the policies in their index is controversial, particularly dual citizenship, mother tongue education, and affirmative action. Wright and Bloemraad (2012) have criticized the inclusion of dual citizenship while Duyvendak et al. (2013) have criticized the inclusion of mother-tongue education and affirmative action. Disagreement over what counts as multiculturalism can lead to disagreement over the development path of multiculturalism and over which countries count as strong multicultural countries. There is significant disagreement, for example, over whether the Netherlands had ever adopted a substantial multiculturalism policy (Banting and Kymlicka, 2006; Entzinger, 2006; Vink, 2007). Determining which policies are multicultural is important to understanding policy development. Confusion over which policies should be considered multicultural may be partly responsible for disagreements over the policy’s development pattern. Different measures of multiculturalism can lead to different conclusions regarding how widespread its adoption is and whether it has faced significant retrenchment. This dissertation tests the Banting and Kymlicka (2013) findings against narrower conceptualizations of multiculturalism. Doing so provides an assessment of how robust Banting and Kymlicka's findings are. The inclusion of a wide-range of industrialized countries allows for further careful analysis of policy development. There has been a great deal of analysis done on the  13 Netherlands as a case of multicultural retrenchment (Entzinger, 2006; Prins and Sahrso, 2009; Sunnier, 2010; Vink, 2007), showing a fair amount of policy retrenchment. It is unclear, however, how representative the Dutch case is of broader trends in policy development. Other analyses, such as the ones conducted by Brubaker (2001) and Joppke (1999) include examinations of countries such as France and Germany, which never adopted large numbers of multiculturalism policies. Cross-country analysis can provide a more complete picture of the pattern of multicultural policy development.  There is, further, a lack of analysis in the existing literature on the extent to which the adoption of one multiculturalism policy leads to additional policy adoption. There is some work on Canada by Pal (1993) that suggests that symbolic recognition of multiculturalism can open up new avenues through which advocates can influence policy. There has been little effort, however, to test the extent to which this phenomenon extends beyond Canada. The particularly positive reception that multiculturalism has received in Canada makes it problematic to generalize about policy development from the Canadian case alone. By examining the influence of different policies on additional policy adoption, the dissertation is able to provide a generalizable understanding of the broad patterns of policy development. Parties' and Policy Adoption There is a wide range of literature that shows that political parties have an impact on policy (Araki, 2000; Bartels, 2008; Garrett, 1998; and Hacker and Pierson, 2010). There has also been an effort to bring parties back into work on the development of immigration policy. This includes analyses of how parties are responding to public opinion and far-right challenges on immigration (Akkerman, 2005; Ireland, 2004; Lahav, 2004;  14 Perlmutter, 1996; van der Brug and van Spanje, 2009). It also includes analyses of how parties are influencing policy (Bruenig and Luedtke, 2008; Howard, 2009; Layton-Henry, 1992). Existing work however lacks a comprehensive analysis of the translation of party positions into policy. Bruenig and Luedtke (2008) come closest to this with an examination of how government positions on immigration affect immigration policies. This analysis, however, fails to account for the way that the positions of non-government parties affect policy adoption. It also looks widely at immigration policy, which as noted in the definitions section of this chapter, is not necessarily the same as multicultural policy.  There is a particular need for a focus on how party positions affect policy given the extent to which partisan left-right ideology is often associated with immigration policy. Both Howard (2009) and Ireland (2004) argue that left parties are more likely to adopt immigration policies. The argument here is that because left parties tend to have more support amongst immigrants, that they should be more likely to enact policies in their interest. This overstates the pressure on left parties to follow through with policy commitments and understates the anti-immigrant pressure that left parties face. When parties of the left have particularly strong support within immigrant and ethnic minority communities their incentive to adopt multiculturalism policies should actually decrease. When parties have strong support within a particular block of voters they do not need to adopt additional policies to win more voters. The lack of a danger that the party will lose support weakens the parties' incentives to adopt the policies they prefer. This is why neither the Democrats nor the Republicans are particularly responsive to African-American voters (Bartels, 1998; Leighley, 2001).   15 There are also substantial numbers of anti-multicultural and anti-immigrant left voters who oppose multiculturalism.  They fear that multiculturalism and immigration policies will take resources from other social programs and lead to greater competition over low-skilled jobs (Hinnfors et al., 2012; Kriesi et al., 2008). There is need to look more carefully at the extent to which parties' positions on multiculturalism are translated into policy. It is problematic to assume that parties of the left will necessarily be supportive of multiculturalism because of their left-right ideology. This dissertation will do careful analysis of both left and right parties' incentives to translate support for multiculturalism into policy adoption. It is further problematic to focus only on government support for multiculturalism as Bruenig and Luedtke (2008) do. Parties that are concerned with re-election need to be careful about passing policies that the opposition can use against them in future elections. Governments that adopt multiculturalism face a particular danger in that anti-multicultural sentiment tends to become salient only when parties mobilize voters around the issue (Green-Pedersen and Krogstrup, 2008; Koopmans et al., 2012; Odmalm, 2012; Perlmutter, 1996). Opposition support can reduce the likelihood that a party that introduces a multiculturalism policy faces an electoral backlash.  Ethnic Minorities, Electoral Systems, Far-Right Parties, and Party Positions A substantial amount of work shows that ethnic minorities influence multiculturalism policy. This includes case analysis of Australia (Lopez, 2000; Jupp, 1991), Britain (Adolino, 1998; Grillo, 1998), Canada (Fleras and Elliot, 2002; Pal, 1992), and the Netherlands (Ireland, 2004). There is also cross-case analysis that establishes this trend in countries across Europe (Koopmans at al., 2012). This work routinely ignores parties.  16 Political parties play an essential role in translating support for multiculturalism amongst ethnic minorities into policy changes. Because parties control the legislative and executive branches of government needed to implement multiculturalism, it is essential that ethnic minorities get parties to support multiculturalism if they want to see policy adoption. Understanding ethnic minorities’ influence over party positions is thus essential to understanding minorities’ influence over policy.  Analysis on minorities' influence over policy often ignores the way that electoral systems effect parties' incentives to respond to minorities' interests. Votes are only valuable to parties in so far as they lead the party to win more seats. Considering the way that electoral systems shape minorities' influence over the number of seats that a party wins is essential to understanding minorities' influence over parties. The importance of electoral systems to understanding minorities' electoral power is underlined by work that shows that minorities are more likely to be elected in SMD systems (Dancygier, 2014; Marschall et al., 2010; Trounstine and Valdini, 2008). These analyses, however, fail to consider how electoral systems affect minorities’ policy influence. It is one thing for minorities to be able to win party nominations and seats in elections, it is another to shape the policy positions of a national party. By considering the way electoral systems shape ethnic minority influence over party positions this dissertation connects ethnic minority electoral influence and electoral systems to policy adoption.  Far- right parties often counter-balance ethnic minority influence over parties. There is substantial work on the emergence and influence of far-right parties in France (Schain, 2006), the Netherlands (Akkerman, 2005; Irwin and van Holsteyn, 2003; van Heerden et al., 2014; and van Kersbergen and Krouwel, 2008), Sweden (Dahlström and  17 Esaiasson, 2011), and Switzerland (Husbands, 2000; Skenderovic, 2007). This literature is complemented by work that looks broadly at far-right parties (Carter, 2005; Green-Pedersen and Krogstrup, 2008; Green-Pedersen and Odmalm, 2008; Jensen and Frølund Thomsen, 2013), as well as work that looks at how mainstream parties have responded to the far-right (Arzheimer, 2009; Givens, 2005; Ivarsflaten, 2008; Oesch, 2008; van Spanje, 2010). This work often excludes countries without far-right parties such as Canada, Germany, and the United Kingdom.4 This exclusion is problematic because it makes it difficult to distinguish the influence that far-right parties have had on mainstream parties from changes that would have happened regardless of the emergence of the far-right. This dissertation's inclusion of a wider range of countries than existing analysis allows for a greater distinctions to be drawn between far-right parties' influence over mainstream parties and the general evolution in mainstream party support for multiculturalism that has developed over time.  The dissertation further adds to existing analysis by contrasting the influence of ethnic minorities with far-right parties. Neither ethnic minorities nor far-right parties operate in a vacuum. Rather, they compete with each other to influence parties and policies. Because many developed countries are seeing both increases in ethnic minority electoral strength and the emergence of far-right parties, it is important to understand these forces' influence relative to each other. Ignoring the extent to which growing ethnic minority electoral strength can mitigate the influence of far-right parties misses the extent to which far-right parties have to compete with other actors to influence party's positions.                                                 4 Both Germany and the United Kingdom saw the emergence of far-right parties after 2011.  This makes them useful as cases without far-right parties for the purposes of this work.  18 The dissertation is finally able to build on existing work on ethnic minorities, electoral systems, and far-right influence over parties by tying that influence to policy change. By linking analysis of the determinants to party positions to analysis of parties' influence over policy, the dissertation shows the extent to which ethnic minorities' and far-right parties' influence over parties translates into increased or decreased likelihoods of policy adoption. In further distinguishing between impacts that ethnic minorities and far-right parties have on parties with different left-right ideologies, the dissertation can further show how ethnic minorities' and far-right parties' particularly strong influence over mainstream parties has particularly important consequences of policy development. Dissertation Structure Chapter Two will outline three sets of theory that guide this work. First, it will discuss the development paths of multiculturalism and the dissertation’s expectations with respect to retrenchment, path dependence and policy development. Next, it will outline theory regarding party influence over policy. This portion of the chapter will distinguish between cross-party, government party, and left and right influence over policy adoption. The final section of the chapter will theorize ethnic minorities', electoral systems’, and far-right parties' influence over party positions. Chapter Three discusses methods. In outlines strategies for cross country and individual case analysis. It discusses several challenges related to analyses of policy development and multicultural party positions. It justifies the need for hazard modeling to deal with the path dependent nature of multiculturalism policies and time-series cross-sectional models to account for the influence that both time and the unique politics of each country have on parties' multicultural positions. This section further presents a new  19 data set created for this dissertation that expands on the existing Banting and Kymlicka Multiculturalism Policy index and creates a yearly score for party systems' positioning on multiculturalism. Both are significant additions to the data that already exists on multiculturalism policy and party positions. The second part of this chapter outlines methods used in the case analysis. It justifies the choice of Canada and the Netherlands as illustrative cases and outlines the methods used analyzing those cases.  The next three chapters present the results of the dissertation's cross-case analysis. Chapter Four looks at the development path of multiculturalism. It finds that multiculturalism policies are not polarized between countries with strong and weak policies. Rather, there is a wide range of levels of policy adoption across countries. It further confirms that policy retrenchment is rare and finds that the symbolic recognition of multiculturalism leads to the adoption of additional policies. The lack of retrenchment justifies the use of hazard modeling throughout the analysis on policy adoption. Chapter Five examines the impact of party positions on policy. This chapter finds that increases in support lead to increases in policy adoption but only when there is broad cross-partisan support for multiculturalism. It further finds that changes in mainstream right support have a large impact on policy adoption, but that changes in left party support have a negligible impact. This chapter demonstrates that party positions matter to policy adoption. Chapter Six outlines the influence that ethnic minorities, electoral systems, and far-right parties have on party positions. It finds that increases in ethnic minority electoral strength increase cross-party and mainstream right support of multiculturalism, but only in single member district electoral systems. In contrast, it finds that far-right parties  20 decrease cross-party and mainstream right support. Changes in ethnic minority electoral strength do not affect left parties’ positions. They do respond to the emergence of far-right parties, but they do so in significantly different ways. In some countries, mainstream left parties co-opt the anti-multicultural positions of far-right parties while in others they increase their support of multiculturalism. Chapter Seven illustrates some of the findings of the early chapters by focusing in two illustrative cases. Analysis focused on Canada highlights the pressure that ethnic minorities place on governments to adopt multiculturalism and way that single member district systems increase minorities' influence over parties. The second half of the chapter uses the Dutch case to show that mainstream right parties can be as supportive of multiculturalism as left parties and to show how parties co-opt the anti-multicultural positions of far-right parties.        21 Chapter 2 Theory: Explaining Policy Development and Party Positions Three questions are central to understanding parties' influence over multiculturalism. What does the pattern of multicultural policy development look like? Under what conditions do changes in party support for multiculturalism lead to policy adoption? And what influences party positions? Each question speaks to a different way that parties interact with multiculturalism. The first considers the impact that path dependence has on policy development. Understanding this is essential to understanding the ways in which parties can influence policy. The second question considers the influence that government and opposition parties have on policy adoption, as well as the relative importance of changes in left and right party positions. In doing this, the chapter goes beyond much of the existing literature on integration and immigration policy, which largely focuses on broad left-right ideology and parties in government to the exclusion of opposition parties. Finally, the dissertation considers how ethnic minority electoral strength, electoral systems, and the emergence of far-right parties affect party positioning.  The chapter develops set of theories that respond to these questions. First, it considers the broader immigration and integration context in which multiculturalism policies exist. Here, multiculturalism is distinguished from other approaches to immigrant integration such as assimilation and civic integration. The chapter then theorizes the development of multiculturalism policies, arguing that path dependence should have a significant effect on the policy. The third section of the chapter argues that changes in party positions should affect policy adoption, that both government and opposition party positions should matter to policy, and that the positions of mainstream right parties should have a particularly important influence over policy adoption. The final section of  22 the chapter argues that party positions should reflect three things. First, parties should respond to changes in ethnic minority electoral strength by increasing their support for multiculturalism.  Second, they should be responsive to ethnic minorities in single member district electoral systems than in proportional ones.  Third, they should respond to the emergence of far-right by decreasing their support for multiculturalism. Crucially, mainstream right parties should be particularly susceptible not only to the emergence of far-right parties but also to growth in ethnic minority electoral strength. Multiculturalism in the Context of Broader Migration Politics  Two sets of distinctions should be drawn when examining multicultural policy.  The first is between multiculturalism as an integration policy, and immigration and citizenship policies.  The second is between different types of integration policies, multiculturalism policies, civic integration policies, and assimilation policies.  While there are often relationships between these different types of policy, they should not be conflated.  In particular, it is important not to overstate the extent to which traditional classifications of countries as following multicultural, civic integration, or assimilationist policy models pre-determine policy development. A country may have policies that can be linked to multiple different integration models. Distinguishing Between Immigration and Integration Policies Integration policies should be set apart from citizenship and immigration policies. Citizenship policies speak to immigrants' access to citizenship, immigration policies to admission and exclusion of immigrants from territory, and integration policies to the incorporation of immigrants into society (Goodman, 2010; Hammar, 1990; Koopmans et al., 2005; Wright and Bloemraad, 2012). These distinctions are important because  23 countries can adopt different configurations of these policies. For example, a country might allow a significant number of foreigners to enter a country under a guest-worker program, but then limit access to citizenship and make little effort to integrate immigrants in the hopes that they return to their countries of origin. Alternatively, a country might have a liberal citizenship policy, allowing immigrants to obtain citizenship easily, but pursue aggressive assimilatory integration policies designed to encourage immigrants to adopt the cultural practices and language of the majority. Finally, a country might have liberal immigration and citizenship policies as well as multicultural integration policies that encourage immigrants to maintain their own cultural practices while integrating into society.  How many immigrants should be allowed into a country, what rights immigrants should have (and how easy it will be to obtain citizenship), and how immigrants should be integrated are all separate questions to which countries might give different assortments of answers. When understanding the distinctions between these types of policies, it is important to acknowledge that they are sometimes related. This can happen both because the motivations behind policy actors are similar for different sets of policies and because policy actors might see one policy area as influencing another. Policy actors who support the inclusion in society of individuals from diverse backgrounds are likely to favour both open immigration policies and multicultural integration policies. Actors who have a strong ethnically defined idea of who belongs in a country are likely to support restrictive immigration policies and assimilationist integration policies, while individuals with a republican understanding of national identity are likely to support open immigration policies but assimilationist integration policies. Because there is likely to be some  24 similarity in the motivations behind policy actors' views on immigration, citizenship, and integration policy, one should expect to see some connection between the three types of policies. In addition to this, policy actors are likely to see integration and immigration policies as related. For example, a policy actor that believes in a highly restrictive citizenship policy is likely to place stronger obligations on immigrants to integrate. The belief that it should hard to obtain citizenship is likely related to the belief that new immigrants should be expected to adopt many of the host societies civic and cultural values when integrating. Beliefs about integration can also be related to ideas about when immigrants should become eligible for citizenship. Policy actors can have different ideas about how much integration and immigrant should have to demonstrate before becoming eligible for citizenship. Disagreements over reforms to citizenship and integration policies in Germany in the early 2000s illustrate this. In debates over immigration and integration the Christian Democratic Union saw demonstrating integration as necessary to obtaining citizenship while the Social Democrats and Greens saw obtaining citizenship as an important part of the integration process. These different ideas about the way that citizenship policy ought to relate to integration policy had a significant amount of influence over how each party approached both integration and citizenship policy (Williams, 2014). Because one policy area can impact others, one should expect some connection between policy actors' views on integration, immigration, and citizenship policies. Figure 2.1 shows the relationship between citizenship and immigration policies and multiculturalism. This figure compares countries' Migrant Policy Index (MIPEX)  25 scores for access to nationality (Huddleston et al., 2015) to 2010 multiculturalism scores from the Banting and Kymlicka Multiculturalism Policy Index. There is a positive correlation between easier access to nationality and multiculturalism policies, but there are also significant differences in levels of multicultural policy adoption between countries with similar access to nationality rules. Australia, Belgium, Canada, France, Germany, Sweden, Ireland, and the United States have similar scores for access to nationality but differ widely in the extent to which they have adopted multiculturalism policies. The converse is also true. Countries with the same level of multiculturalism policy adoption, such as Ireland, Norway, Spain, and Portugal, can vary significantly in their scores for nationality acquisition. While there is a relationship between countries' citizenship policies and adoption of multiculturalism policies, there are also important differences in policy adoption amongst countries with similar citizenship policies. An open citizenship policy does not necessarily lead to a multicultural integration policy.   26 Figure 2.1: MIPEX Access To Nationality Scores and Multiculturalism Policies    This has important implications for understanding the development of multiculturalism. Existing works that seeks to explain citizenship policy by focusing on factors such as national identity (Brubaker, 1992; Favell, 1998), party ideology (Breunig and Luedtke, 2008; Howard, 2009), cross-national policy convergence (Joppke, 2007; Joppke and Morawska, 2003; van Houdt et al., 2011) can be useful starting points for examinations of integration policy. At the same time, this work should not be assumed to explain the adoption of multiculturalism simply because they provide convincing explanations of citizenship and immigration policy (or for that matter other integration policies). There is enough of a difference between these types of policies to justify examining multiculturalism in its own right. In particular, the literature on parties' impact  27 on citizenship policy can be useful in providing a basis for theories on how parties might approach multiculturalism. Howard (2009), for example, shows that left party governments are more likely to adopt liberal citizenship policies. Givens and Luedtke (2005) make similar claims with respect to liberal integration policies (though they do not focus specifically on multiculturalism). These works suggest that it is important to consider left-right ideology when examining multicultural policy development. At the same time, left parties may not have the same impact on the adoption of multiculturalism. Analysis on citizenship and immigration policy can serve as a basis for theories that examine parties' impact on multiculturalism.  At the same time, there should also be theorization and testing that focuses on multiculturalism specifically. Multiculturalism and National Integration Models The immigration literature tends to classify countries as having multicultural, civic, or assimilationist/republican integration models. The argument in much of this literature is that national identity through public philosophy (Schain, 2008), policy paradigms (Favell, 1998), discursive opportunity structures (Koopmans et al., 2005), or national culture (Brubaker, 1992) will influence the integration policies that develop in countries. These classifications consider Britain, the Netherlands, and Sweden to be multicultural countries, France to follow an assimilationist/republican model, and Germany to be a country driven by an ethnic understanding of citizenship (Bertossi and Duyvendak, 2012; Brubaker, 1992; Borevi, 2014; Schain, 2010). Such classifications often describe policy models that guide integration policy in particular countries. These models represent very different philosophies with respect to integration. As discussed in the introductory chapter, multicultural models seek to promote integration  28 while allowing individuals to maintain their own cultural backgrounds. Implicit in multiculturalism is the idea that cultural background is not a barrier to integration. Civic integration models do not operate on the same assumption. While they do not go as far as assimilation models, they hold that there are certain liberal and social values that are both necessary to integration and influenced by cultural background. Civic integration policies thus place stronger obligations on immigrants to demonstrate that they have adopted the liberal and social values present in society (Goodman, 2010). Assimilationist or republican models place the strongest obligations on immigrants to integrate. They expect immigrants to adopt the social and cultural values of the majority community when integrating, and that immigrants will become part of the majority cultural community. The French assimilationist model, for example, expects immigrants to adopt French social and cultural practices (Bertossi, 2012; Goodman, 2010).  National models often miss some of the differences and nuances in the development of different countries’ integration policies.  These models often assume the policy development is static, and fail to consider challenges to old policies that lead policy change.  As such, national models have come under increasing scrutiny in recent work by Bertossi (2011), Bertossi and Duyvendak (2012), and Scholten (2015).   None of this is to argue that multiculturalism, civic integration, and assimilationist models are not important to understanding policy. Rather, it is to make that case that there is a danger in assuming that a particular country will follow a particular policy model to the exclusion of all others. Most countries will have at least some advocates of multicultural, civic integration, and assimilationist models of immigrant integration. Even in Canada, a country with one of the strongest multicultural policies in the world, there  29 have been prominent advocates for civic integration and assimilationist integration policies. Conservative leadership candidate Kellie Leitch has advocated for the introduction of a “Canadian values” test for immigrants (CBC, 2016) and the Conservative party campaigned on re-introducing a ban on niqabs at citizenship ceremonies during the 2015 election (CBC, 2015c).5 National models do not always agree with indices examining civic integration policies and multiculturalism. Goodman (2010) classifies countries by the number of civic integration policies and citizenship policy. France, the Netherlands, and the United Kingdom all end up classified as countries with conditional immigration and integration policies (countries with a large number of civic integration policies and liberal citizenship policies). This occurs even though existing work on national models often considers France, the Netherlands, and the UK to have different approaches to integration. The same is true for France, Germany, and the Netherlands. Work on integration models often finds these countries follow different models, yet they have similar scores on the Banting and Kymlicka multiculturalism index. Even though the United Kingdom is often classified as multicultural, it has substantially different multiculturalism scores than two other multiculturalism countries Australia and Canada (Multiculturalism Policy Index, 2016).  Comparing multicultural policies to other integration policies can demonstrate the difficulties that exist in assigning countries to different integration models. There are clear distinctions between multiculturalism scores and two other aspects of integration                                                5 The Conservatives had placed such a ban in place before the election but had it struck down by the courts. The Conservatives' loss of the 2015 election to the Liberal prevented further consideration of such a ban.  30 policy, anti-discrimination and civic integration.6 Figure 2.2 shows that while most countries with strong multiculturalism policies have strong anti-discrimination policies, there are countries with strong anti-discrimination policies that do not have strong multiculturalism policies. These include France, the Netherlands, Portugal, and the United States. Countries with moderate or low multiculturalism policies also vary significantly in their level of anti-discrimination policy adoption. As with access to nationality, Ireland, Norway, Portugal and Spain have similar levels of multiculturalism policy but very different anti-discrimination policies. France, Germany, the Netherlands and Switzerland also have similar levels of multiculturalism but very different scores for anti-discrimination policy adoption. Anti-discrimination policies, therefore neither fit as a type of policy exclusive to multicultural countries nor as part of integration models that are different from multiculturalism.                                                6 Anti-discrimination scores come from the MIPEX index while civic integration scores come from Goodman's (2014) CIVIX index.  31 Figure 2.2: MIPEX Anti-Discrimination and Multiculturalism Scores  The correlation between the adoption of stricter integration policies and the strength of countries' multiculturalism policies is stronger but still includes some variation that needs explaining. Figure 2.3 shows some variation the levels of multiculturalism in countries with low integration requirements. Sweden, Belgium, and Finland have low CIVIX scores and strong multiculturalism policies while Ireland, Portugal, and Spain have low CIVIX scores but only moderate multiculturalism policies. The United Kingdom stands out as a country with a high CIVIX score and relatively strong multiculturalism policy. Finally, France and Germany have similar levels of multiculturalism but substantially different CIVIX scores. While the distinction between countries with multiculturalism policies and countries with civic integration policies is clearer than for anti-discrimination policies, there is still important overlap between countries with at least some multiculturalism policies and some civic integration policies.  32 While one might consider a particular country to follow a predominantly multicultural approach to integration (or a predominantly civic integration model), it can be problematic to assume these classifications preclude the adoption of policies for other integration models. A country can have a large number of civic integration policies and still and adopt a few multiculturalism policies.  Figure 2.3: CIVIX and Multiculturalism Scores   Policy developments in the Netherlands and France highlight problems with national model approaches to integration policy. Analyses of the Netherlands have identified several different policy paradigms that have shaped the development of integration policy throughout Dutch history. Duyvendak and Scholten (2012) identify four different frames that have guided the development of integration policy. The Netherlands goes from a pillarisation model (which had social services such as education and healthcare provision linked to cultural and social communities such as Catholics,  33 Protestant, and socialists)7 in the 1970s, to multicultural model in the 1980s, to a liberal model in the 1990s, to a civic integration/assimilationist model in the 2000s. Existing literature points to an integration policy shift in the Netherlands from a multicultural model to a civic integration one. This model places much greater demands on immigrants to adopt Dutch values surrounding public morality (particularly relating to acceptance of equality regarding gender and sexual orientation) and social interaction (Entzinger, 2006; Michalowski, 2011; Vink, 2007). Indeed the Netherlands has gone from a multicultural model of integration to one of the most strict civic integration approaches to policy in Europe (Jacobs and Rea, 2007).  These shifts in classifications of Dutch integration policy make it difficult to convincingly argue that there is a single Dutch integration model.  Rather, there have been several different models that, at different periods of time, have each influenced policy. France also changed it approach to integration over time. Bertossi (2012) identifies four different stages of integration policy that exist in France from 1980 to 2010. These include a focus on a national identity that runs through much of the 1980s, an anti-discrimination framework that develops in the late 1980s and early 1990s, a lacité framework that becomes dominant in the 1990s, and a public order framework that dominates policy decisions in the 2000s. While elements of each of these frameworks are present in French policy making throughout the last three decades, the extent to which a particular policy framework is dominant changes significantly over time.  Though the paradigms surrounding French integration policy go through changes that are less radical than Dutch paradigms, there is still evidence that multiple different perspectives on                                                7 There is some debate over whether a multicultural pillar developed in the Netherlands (Vink, 2007).  34 integration have influenced French policy. French integration and immigration policy should not be reduced to a single understanding of republican citizenship. The divided views of policy actors, political parties, and the electorate on integration policy can lead to the development of a mix of different policies that do not adhere strictly to a single model. A government that favours civic integration policies for example, might replace one that favoured multiculturalism. If a new government introduces civic integration policies without removing the old multicultural policies (to the extent to that path dependence affects multiculturalism, it may be easier to add new policies than remove old ones), a country can end up with a mix of integration policies that belong to different and contradictory policy models. Thus, countries rarely considered multicultural, such as France of Germany, may see the development of one or two multiculturalism policies. One might also see the co-existence of multicultural and civic integration policies in a country such as the United Kingdom or significant policy change over time as happened in the Netherlands.8  Thus, any explanation of integration policy that relies too heavily on national models can miss the extent to which multicultural policy adoption can occur in countries rarely considered multicultural. Why Consider Parties Amongst Other Explanations of Integration Policy? Examinations of multiculturalism can use several approaches common in explaining immigrant and integration policy. Approaches that focus on national identity argue that the identities create ideas of belonging that determine countries’ approaches to immigrant integration. Theories focused on cross-national policy convergence argue that prevailing ideas about human rights and global concerns about increasing migration drive policy                                                8 This is not to argue that once a country has one or two multicultural policies that it is a multicultural country. Rather, a country can have policies that fit with different models of integration.   35 decisions.  Finally, explanations that see Canada as a unique case highlight some of the unique circumstances in Canada that have increased the likelihood of strong policy development. All of these approaches to policy are valuable and contribute to explaining why some countries have strong multiculturalism policies and others do not. Each, however, has limitations. A focus on parties can explain some of the variation in policy adoption that existing theories cannot. Brubaker (1992) and Favell (1998) make strong cases that a countries' national identity guide the development of immigration policies. They argue that ideas around ethnic, republican, or more pluralistic understandings of citizenship guided the development of integration policy in Germany, France, and the United Kingdom. The way that both citizens and policy actors understand national identity influence who they see as eligible for citizenship and the obligations that they place on newcomers to integrate. In a country such as Germany, which has a long history of linking ethnicity with national identity, this has meant that citizenship has often been restricted to those who could claim a German ethnic identity (Brubaker, 1992). Only recently have reforms made German citizenship available to immigrants without German ancestry. In France, a republican idea of national identity meant that policy makers saw citizenship as something that most could attain, but also led them to attach strong obligations to assimilate into French culture and society to citizenship to integration. This led to relatively open citizenship policies, but assimilatory integration policies (Brubaker, 1992; Favell, 1998). In Britain, an understanding of national identity that is less tied to ideas of ethnicity or culture has opened up space for integration policies that allow more space for cultural diversity (Favell, 1998).  36 National identity and immigration history have also found their way into explanations of Canadian multiculturalism. Kymlicka (2008) highlights Canada's history as an immigration country as one of many explanations of why multiculturalism has been resistant to retrenchment in Canada. Kymlicka argues that Canada’s long history of immigration has forced policy makers to develop immigration policies that are more accommodating of diversity. This argument can also explain the development of multiculturalism in other settler countries such as Australia or New Zealand.  There are limits to the extent to which national identity and immigration history can explain the adoption of multiculturalism policies, however. There is significant variation in policy adoption amongst countries with similar identities and immigration histories. The United States, for example, is a settler country with a long history of immigration like Australia, Canada, and New Zealand. It however has a much weaker multiculturalism policy than any of the other settler countries. Belgium, the Netherlands, and the United Kingdom are all former colonial countries that have had to grapple with immigration from their former colonies, but the Netherlands has substantially weaker multiculturalism policies than Belgium or the United Kingdom. In Scandinavia, there is a wide range of multiculturalism policy adoption. Sweden and Finland have developed strong policies, Norway a moderate policy, and Denmark is the only country in the index other than Japan to have adopted no policies. Explanations of policy development that focus on national identity and immigration history cannot explain this variation. Explanations that rely on national identity and immigration history have further difficulty explaining changes in the level of multiculturalism commitment over time. National identity and immigration history do not change rapidly, yet there has been a  37 sharp increase in the adoption of multiculturalism policies through the 1980s, 1990s, and even into the 2000s. This means that many countries including Australia, Canada, Belgium, and Sweden saw increases in their levels of multiculturalism policy over a shorter period than one might expect if policy adoption were determined only by a countries' national identity and immigration history. While national identity can account for the development of strong policies in Australia and Canada, it cannot explain the speed of these policies' development nor can it account for why multiculturalism spread beyond settler countries with a history of diverse immigration. Explanations based on national identity can understate the extent to which national identities are contested and evolve over time.  Examinations of integration policy highlight a trend in policy convergence across countries. Several migration scholars have pointed towards two competing pressures that have shaped the development of integration policy through the 1990s and 2000s. On one hand, liberal and human rights norms have pushed countries towards more open citizenship policies (Borevi, 2012; Joppke, 2007; Soysal 1994). Countries that have made firm commitments to treating people as equals, regardless of their origin, and respecting liberal human rights have found it difficult to justify strong restrictions on immigration policy. This is particularly the case for countries that had tied citizenship to ethnicity, such as Germany. These countries have come under increasing pressure to liberalize their immigration policies in order to fall in line with norms against treating individuals from different ethnic backgrounds differently (Joppke and Morawska, 2003). The result of more liberal international citizenship norms should be a cross-national convergence towards more liberal citizenship policies.   38  On the other hand, there is a perception that many industrialized countries are having difficulty integrating immigrants. A substantial literature has emerged that argues that integration difficulties are driving cross-national convergence on integration policies that place greater demands on immigrants.  These policies include those that require citizens to demonstrate greater language and civics knowledge as part of the integration process (and in some cases as a requirement to obtain citizenship) (Etzioni, 2007; Joppke, 2007; van Houdt et al., 2011). These concerns are reflected in the growing hostility towards multiculturalism amongst political leaders and electorates. In Britain then Prime Minister David Cameron argued that multiculturalism was encouraging segregation and was not creating an idea of society that all people could belong in (Cameron, 2011). In Germany Chancellor Angela Merkel argued that multiculturalism in Germany had “utterly failed” (BBC, 2010). Electorates across Europe have indicated growing scepticism by voting in increasing numbers for far-right anti-immigrant, as have Australian voters by casting ballots for Pauline Hanson’s One Nation party.  Finally the election of Donald Trump on an anti-immigrant as President of the United States in 2016 has suggested a growing concern with immigration and integration in that country. Scepticism about immigration and multiculturalism seems to have a large, cross-national reach. There are, however, limits to the extent to which cross-national ideational trends surrounding citizenship and integration policy can explain the development or retrenchment of multicultural policy. On integration policies, there is a still a great deal of divergence across countries are (Antonsich, 2016; Goodman, 2014). This is particularly the case with respect to multiculturalism, as Figures 2.1, 2.2, and 2.3 show. Countries  39 such as Australia, Canada, New Zealand, and Sweden have strong policies, Ireland, Norway, Portugal, Spain, and the United States have moderate policies, while countries such as Austria, Denmark, Italy, and Japan have few if any policies. Theories based on cross-national convergence cannot explain significant variation in multiculturalism policy adoption.  The cross-national return to assimilation argument put forward by Joppke (2007) and Brubaker (2001) has problems when it is applied to multiculturalism policies. Goodman (2014) and Jacobs and Rea (2007) suggest that the turn towards civic and assimilationist integration policies does not apply evenly across cases. This turn toward integration has occurred to a much greater extent in the Netherlands than it has in the culturally similar Flemish region in Belgium (Jacobs and Rea 2007). Additionally, Banting and Kymlicka (2013) find little evidence of retreat from multiculturalism.  In examining why Canada has been particularly resistant to efforts at multicultural retrenchment, Kymlicka (2008) posits some factors unique to Canada. He argues that Eastern European immigration to Canada in the late 1800s and early 1990s meant that by the 1960s Canada had large numbers of immigrant communities that were culturally similar to the English and French majorities. This made it easier for the majority population to except the multicultural claims of immigrants. Because Eastern European communities were largely Christian, multicultural demands by Eastern European communities rarely involved religious accommodations such as dress code exemptions for hijabs or turbans. The more cultural diversity in a country, the more accommodations multiculturalism policies require. Triadafilopoulos (2012) also points to this as an explanation of the strength of Canada’s multiculturalism policy. Furthermore, Kymlicka  40 points to Canada’s experience accommodating diversity in Quebec as creating a precedent for accommodating diversity within immigrant communities.  Finally the Canada’s geographic isolation from the countries that its immigrants come from gives it a fair amount of control over its immigration. This has allowed the country to select immigrants that are more likely to integrate (those with skills make it likely they will find employment and who speak either English or French).  This in turn increases the likelihood of immigrant integration and makes it less likely that concerns over failures to integrate immigrants will lead to a backlash against multiculturalism. There are, however, limits to how much these arguments account for patterns elsewhere. While Canada is not unique as a country that received significant Eastern European immigration before adopting multiculturalism (Australia also received large amounts of Eastern as well as Southern European immigration after the Second World War) several countries with strong policies did not. Belgium, Finland, Sweden, and the United Kingdom all have strong to moderately strong multiculturalism despite not receiving such immigration. Several countries that are not geographically isolated from immigration sources, including Australia and Belgium (the latter as a result of Europe's proximity to Africa and the Middle East and the open border that comes with European Union membership) have strong multiculturalism policies. In addition, Sweden forgoes the kind of careful immigrant selection process that Canada uses by taking very large numbers of refugees, yet has the strongest multiculturalism policy in Europe. Finally, the lack of experience dealing with diversity in the form of a large national minority group similar to Quebec has not inhibited the development of strong multiculturalism policies in Australia, Finland, or Sweden.   41 It is remarkable that political parties, electoral systems, and electorates are absent from all of these examinations of integration policy. Political parties should play a central role in analyses of multiculturalism policy. Parties are a central institution through which citizens influence policy (Kedar, 2009). This should matter to multiculturalism. As global migration increases, electorates are becoming more diverse. This is placing pressure on parties to increase their support of multiculturalism, and to put multiculturalism policies into place. Indeed, parties are making efforts increase their support amongst ethnic minority voters (Dancygier, 2010; Triadafilopoulos, 2012). If parties are doing an effective job representing citizens in the policy making process, these efforts to appeal to minority voters, the changes in party positions caused by the increasing diversity of the electorate should in turn have a meaningful impact on policy. Analysis of parties influence over multiculturalism is essential to understanding the extent to which parties adequately representing the growing diversity of electorates.   Electoral systems influence parties’ incentives to try to win minorities’ support. Winning votes is a means through which parties win seats in legislatures.  The extent to which particular votes influence a parties’ seat share affects parties’ incentives to appeal to different groups of voters. In SMD systems small groups of geographically concentrated voters, such as ethnic minorities, can have a disproportionate influence on the number of seats a party wins. This gives parties a particular incentive to appeal to these voters. In proportional systems, small groups of geographically concentrated voters do not have the same influence on the number of seats a party wins, and thus parties have less of an incentive to try to win the support of these voters. Any account of parties and  42 multiculturalism that does not take into account the way that electoral systems change parties’ incentives is incomplete. Increases in migration have not just made electorates more diverse. It has also led to the rise of far-right parties that appeal to voters opposed to immigration and multiculturalism. There is a robust literature that looks broadly at the impact that far-right parties have had on mainstream parties (Akkerman, 2005; Ireland, 2004; Lahav, 2004; Perlmutter, 1996; van der Brug and van Spanje, 2009). Accounts of parties and multiculturalism policy should take this literature as a starting point. Parties’ positions on multiculturalism, like other immigration and integration policies, will be affected by the threats they face from the rise of far-right parties. It is important to examine the effects of ethnic minorities and far-right parties in contrast to each other. The combination of a rise in the diversity of electorates and the emergence of far-right parties create conflicting pressures for mainstream parties.  Understanding parties’ positions requires paying careful attention to the way these different pressures compete with each other. The extension of work on parties and broader integration policy is useful as a starting point for an investigation of parties and multiculturalism, but there are limits to its application. The first, as noted in the previous section, is that multiculturalism policies do not always correlate with the development of broader of immigration policies. The development of multiculturalism does not mirror cross-national convergence in citizenship policy and in the adoption of stricter integration policies. This suggests that the pressures that governments, and the parties that influence them, face to act on multiculturalism are somewhat different from the pressures they face to act on integration policy. In particular, careful attention should be paid to the extent to which ethnic  43 minorities can pressure parties to be more supportive of multiculturalism, and on the extent to which far-right parties' influence over mainstream parties is the same for multiculturalism as other immigration policies. Literature on immigration and integration policy further tends to attribute more liberal immigration and integration policies to left parties (Howard, 2009; Givens and Luedtke, 2005). Left parties are connected to liberal immigration and integration for two reasons.  First, they have strong links to social movements. An ideological commitment to protecting individuals from disadvantaged groups and advancing social equality is consistent with open immigration and less stringent integration policies.  Second, left parties often have higher levels of support amongst immigrants than mainstream right parties to do. Left parties therefore have an incentive to push for liberal immigration and integration policies as a way of advancing the interests of a contingent of their voters. It is not apparent that this divide between the left and the right holds for multiculturalism. The countries with the strongest multiculturalism policies, Australia and Canada, saw those policies introduced centrist (the Liberal Party of Canada) or right parties (the Australian Liberal Party and the Progressive Conservative Party of Canada) instead of left parties. Mainstream right ideologies that favour lower taxes and fiscal restraint do not preclude policies that grant some recognition to ethnic minorities, nor do they preclude policies that increase minorities' ability maintain their culture. Even socially conservative parties might be supportive of multiculturalism if they see it as a way to build cross-religious support for socially conservative policy positions.9 It is similarly not clear that the left should always be supportive of multiculturalism. In fact,                                                9 More socially conservative policies on religious freedom and practice, gay rights, and abortion, for example, might appeal to social conservatives from a wide range of different cultural backgrounds.  44 Barry (2001) and Gitlin (1995) argue that multiculturalism has hurt efforts to attain socio-economic equality.10  It is finally notable that work examining parties and immigration integration policy tends to focus on the positions of parties in power  (Bruenig and Luedtke, 2008; Howard, 2009). Doing so ignores the importance of opposition parties. The salience of multiculturalism with the broader electorate has a substantial impact on governments' incentives to adopt multiculturalism policies. When multiculturalism and other immigration policies are not salient, governments face pressure to adopt multiculturalism from ethnic minorities who benefit from the policies.  The lack of salience of the issue means that parties face little or no pressure from the public to oppose the adoption multiculturalism.   In contrast, when multiculturalism is salient, governing parties often face greater pressure to oppose multiculturalism (Perlmutter, 1996). The salience of multiculturalism depends, to an extent, on whether opposition parties are going to make an issue of a government's decision to adopt a multicultural policy. If the opposition does, they may create an electoral backlash that keeps a government from following through on a campaign promise to introduce multiculturalism. Opposition parties can run campaigns against multiculturalism in response to a government’s efforts to introduce policies and make multiculturalism and immigration a central aspect of their future election campaigns. An opposition party that does this raises the salience of multiculturalism, and because of that, can win the support of government voters that oppose multiculturalism. Centre-right and centre parties might be more likely than left parties to do this when in                                                10 This is not to suggest that Barry and Gitlin are correct, but rather that leftists are not necessarily supportive of multiculturalism.  45 opposition. However, it is possible that a left party, especially one with weak ethnic minority support and strong support amongst workers concerned that immigration threatens their employment would use opposition to multiculturalism to take xenophobic voters from a centre-right government. The potential for opposition parties to mobilize a backlash against policy adoption make taking their positions into account important when understanding policy development. The Development of Multiculturalism Policies Why Development Patterns are Important Understanding a set of policies' development pattern is important to understanding parties' influence over them. Three issues are important to this. The first is the distribution of policy adoption across countries. It is important to know if a policy is either fully adopted or not at all, or if there is a wide spectrum of policy adoption across countries. Second, one must understand whether retrenchment on a policy is common or rare. Third, one must understand whether particular policies are adopted before others. Knowing all of these things gives an observer a clearer understanding of what to look for when examining parties' influence over policy adoption. The distribution of policies helps determine whether explanations should reach across countries or should be unique to a particular case. If policy adoption is concentrated in one or two countries, it makes sense to look for unique circumstances in those particular countries that can explain adoption. Looking for trends that hold across many countries is necessary if policies are widely adopted. The distribution of multiculturalism therefore has important implications for the kinds of variables that one should look for when trying to explain policy development.  46 If retrenchment is common, analysis should look at factors that lead to policy adoption and at factors that lead to retrenchment. If retrenchment is rare, the primary focus of analysis must be on the factors that lead to adoption. When retrenchment is rare, explanatory variables will have a much stronger influence on adoption than they will have on maintaining policy stability. This affects the choice of model in quantitative analysis. If a policy is path dependent a model that distinguishes between observations that occur before and after policy adoption, such as hazard model, is needed to examine different variables’ impact on policy. To the extent that the adoption of certain policies increases the likelihood for the further adoption of others, explanatory variables will have both direct and indirect effects. If a particular explanatory variable has an impact on a policy that itself increases the likelihood of policy adoption, that effect needs to be taken into account when understanding the variable’s influence over policy development. The variable might have two impacts on policy, a direct effect on the adoption of one type of policy and an indirect effect on the adoption of another policy which is affected by the increased likelihood of the adoption of the first policy. Both need to be taken into account when understanding the variable’s influence., How Widespread is Policy Adoption Understanding how many countries have at least some multiculturalism policies is important. If a few countries have strong policies and others have none, it is important to understand what differentiates the countries with policies from those without. If some countries have strong policies, some moderate policies, and others weak or no policies, it  47 is important to look for explanatory factors are present across many countries but which vary in strength.  The strength of countries' multiculturalism policies should fall along a spectrum with some countries adopting strong policies, others moderate policies, and others weak policies. Policy adoption should be the product of a wide range of factors.  These include the influence of advocates, opponents of multiculturalism, the way institutions affect how policy makers respond to advocates and opponents, and how different immigration histories and national histories influence policy development.  They also include the way that existing immigration policies influence policy, the extent to which dramatic events such as terrorist attacks lead to the mobilization of anti-immigrant public opinion, and the extent to which different countries face pressure to conform to international norms developing around immigration and multiculturalism. These factors will combine in unique ways in different countries to create different, and often contradictory, pressures on policy makers. In one country, powerful ethnic minorities might be able to act largely unopposed to influence policy adoption, while in another an equally powerful minority group might face significant opposition because of a dramatic event that mobilizes anti-immigrant public opinion. In yet another country, a history of restrictive citizenship and weak multicultural advocates may lead to the adoption of limited policies. The number of factors that influence policy are likely to combine in a wide variety of different arrangements, some of which will favour strong policy adoption, others of which will favour weak policy adoption, and others of which that will fall somewhere in between. Perfect alignment of policy influences either in favour of multiculturalism or against it  48 should be rare. This should lead a significant number of countries to fall in somewhere in the middle of the spectrum between strong and weak policies. It is notable that even in countries that Brubaker (1992) outlines as typical cases of either republican or ethnically based citizenship, such as France or Germany, there are some pressures on policy makers to adopt multiculturalism. In France the 1980s saw the government grant some recognition to cultural diversity in the country (Koopmans et al., 2005; Safran, 1985). In Germany, some municipalities, such as Frankfurt, have adopted strong policies even though the national government has not. In Frankfurt, advocacy by the Green party was critical (Ekhardt, 2007). This is not to say that either country is multicultural, but rather that there are competing pressures that can affect policy adoption even in countries not considered multicultural. Multiculturalism and Path Dependence  Path dependence plays an important role in analyses of policy development. Path dependent approaches to understanding policy development argue that past decisions matter to the adoption of future policies. Certain decisions at critical points in time can trigger causal patterns of policy development that are very difficult or impossible to reverse (Mahoney, 2000). Fundamental to path dependence theories are the development of feedback loops that reinforce a particular pattern of policy development. These feedback loops can reinforce an existing policy or institution, making it difficult to remove. They can also increase the likelihood of a particular path of additional policy development, increasing the pressures that policy makers face to adopt particular sets of policies and reducing pressures on actors to adopt other sets of policies.   49 Path dependence can work through a variety of mechanisms that are highly relevant to the adoption of multiculturalism. Early policy decisions can influence the ideas and norms that guide future policy development in a country, they can shape institutions and agencies in ways that affect policy, and they can strengthen particular constituencies and policy advocates at the expense of others. Each of these dynamics can result from the adoption of at least some of the multicultural policies in the Banting and Kymlicka index, and make path dependent effects important to consider when examining multiculturalism.  One way path dependence can operate is through the development and reinforcement of ideas that lead to certain to the adoption of certain types of policies. Krasner (1988) argues that the way ideas surrounding sovereignty and citizenship develop in particular countries affects the range of policies open to them. The kinds of institutions a country adopts influence understandings of the concept of citizenship, and those understandings in turn affect the range of policy options seen as legitimate. Krasner uses as an example illegal immigration in the United States, arguing that it has been difficult for American governments to pursue strict policies against illegal immigration because of the liberal values embedded in the American legal system. The entrenchment of liberal ideas in American courts limited the policies pursued in response to illegal immigration. The impact of ideas on immigration and integration policy through path dependence is already reasonably prevalent in the immigration literature. Brubaker's (1992) and Favell's (1998) arguments about the influence that understandings of national identity have on immigration and integration policy in France, Germany, and the United  50 Kingdom are essentially path dependence arguments that focus on ideas. In each country ideas about belonging are deeply embedded within its history. In the case of France, republican ideas that date back to the French revolution guide understandings of who can be a citizen and the obligations those citizens have. These republican ideas led France to adopt relatively open citizenship policies but to place strong obligations on immigrant to assimilate in French culture. In Germany, ideas of an ethnic citizenship that have a history pre-dating the First World War influenced the development of very restrictive citizenship laws. In contrast to France and Germany, Britain's experience with more pluralistic immigration had roots in a complicated citizenship that was the product of its large pre-20th century empire. Britain's openness to immigration from its former colonies through the middle of the 20th century forced it to deal with significant diversity in immigration, and to adopt anti-discrimination measures such as its Race Relations policies. Brubaker and Favell highlight how which ideas shape the development of citizenship and immigration.  Path dependence can also work through the creation of departments and agencies tasked with developing and implementing policy reinforce the initial steps. Bureaucrats have incentives to ensure the programs that they run are seen as important, in particular to maximize their budgets (Niskanen, 1973). They often have autonomy in shaping policy (Huber, 2000). As policy entrepreneurs, bureaucrats with some autonomy from politicians can act and have a strong impact on policy (Carpenter, 2001). Because of this bureaucratic autonomy and self-interest, policy initiatives that include the creation of new agencies can create the means for their own reproduction and expansion.   51 Finally, the material benefits can be a source of path dependence. Such benefits affect policies including healthcare (Hacker, 1998) and welfare state benefits (Hacker, 2002; Pierson, 1994; Weaver, 2003). Pierson (2004) argues that government benefits are often difficult to retrench because they provide concentrated benefits to groups of voters. Because the groups are concentrated, they are easy to organize in defence of such benefits (Olson, 1965). Because voters are loss averse, they are likely to lobby governments that try to retrench benefits and electorally punish parties that pursue such a program.  Two policies in the Banting and Kymlicka index have the potential to create the positive feedback loops necessary for path dependence. These feedback loops can then influence policy development in two ways. Affirmation policies can provide symbolic that make publics and policy actor more receptive to multiculturalism and agencies that have an incentive to push policy expansion. Policies that provide funding to minority organization can strengthen their ability to lobby for policy change.  These feedback loops can affect policy by reducing the likelihood of retrenchment and by increasing the likelihood of policy expansion. Symbolic recognition of a country as multicultural shifts understandings of who ought to be included in a country and what an immigrants' obligation is with respect to integration. When the government affirms that it is multicultural, it does two things. First, it recognizes that it is a country of immigration and that citizenship ought not be restricted to people from particular cultural ethnic groups. Second, it affirms that individuals do not need to assimilate into the majority cultural community in order to integrate into society. Affirmation tells immigrants that they do not need to adopt the majority community's cultural practices in order to integrate into society. In the United  52 States, for example, inclusive education and welfare policies increased minorities’ sense of belonging in the country. Even those minorities that did not receive benefits but had neighbours that did saw improved performance. This demonstrates that even symbolic inclusiveness can have meaningful effects on minorities’ sense of belonging (Condon et al., 2015; Filindra et al., 2011). A government that symbolically recognizes the importance of minority communities through programs like multiculturalism is likely to empower minority communities to push for the expansion of such policies.  Symbolic recognition also tells its native-born citizens that immigrants can become well-integrated members of society even if they do not adopt similar cultural practices to the majority. This should make it easier for advocates of multiculturalism to lobby for policy expansion. If symbolic recognition changes public opinion, the public appeals of multiculturalism advocates will be more effective. Even if public opinion does not change, symbolic recognition can change the way minorities appeal to governments. When lobbying a government, particularly when lobbying bureaucrats, multiculturalism policy advocates can point to the government's symbolic recognition as acquiescence to the values and arguments that should underlie substantive policies. They can make similar appeals to the broader public.11  Policies that recognize a country as multicultural often involve the creation of new agencies tasked with developing and implementing multiculturalism. These agencies provide both a new access point through which advocates can influence policy, and an organized interest within government that has an incentive to protect and push for policy expansion. Because implementing multiculturalism often involves high levels of outreach                                                11 This is particularly the case if multiculturalism has a prominent in the education system. Individuals who grow up being told that multiculturalism is a core value in their country are more likely to be convinced by arguments that seek to appeal to multicultural values.  53 to ethnic minority organizations, government agencies tasked with implementing multiculturalism are likely to build strong ties with ethnic minority organizations and other advocates for multiculturalism. Advocates can use these ties as an access point to lobby governments to adopt additional policies. The agencies that they build relationships with have incentives to act as allies to advocates for multiculturalism because doing so increases their own policy prestige.  The Canadian case serves as an important example were the development of symbolic recognition policies, and the agencies associated with them, created positive feedbacks that led to additional policy adoption. The adoption of multiculturalism created new avenues through which advocates could influence policy. Paquet (2015) finds that bureaucrats played an important role shaping the development of subnational immigration and integration policy in Canada. Pal (2003) further finds that interaction between the agencies created in the aftermath of the introduction of Canada's first policy affirming multiculturalism and ethnic minority advocacy organizations shaped the further development of Canadian multiculturalism. Many of the anti-racism programs that were incorporated into Canadian multiculturalism were done so as a result of lobbying by ethnic minority organizations of the agencies tasked with implementing Canada's first multiculturalism policies.  Finally, some multiculturalism policies should also create path dependent effects by providing material benefits to minorities. Government funding for ethnic minority organizations, support for mother-tongue education, and affirmative action all provide minorities with tangible benefits. There are strong incentives for minorities to organize in support of these programs, and to vote against any party that advocates their  54 retrenchment. Conversely, the costs of these programs are small and spread out over a large population. Opponents of them have little incentive to organize to push for their retrenchment. To the extent that multicultural programs provide material benefits to ethnic minorities, they should be subject to the same path dependence dynamics as other welfare programs. Funding for ethnic minority organizations should also increase the likelihood of additional policy adoption. This policy can increase minorities’ political power. Well-funded organizations can attract more members (increasing their political clout) and invest more resources in political organization and lobbying. As a result, they can be more effective policy advocates. In addition to preventing retrenchment, funding can create positive feedback loops that increase the likelihood of policy expansion. The stronger ethnic minority organizations are, the more success they are likely to have when lobbying governments. If path dependence is affecting policy development three things should be true. Positive feedback loops generated by policy adoption make retrenchment rare. Second, if symbolic recognition is driving path dependence the affirmation of a country as multicultural should increase the likelihood of additional policy adoption. By the same logic, such policies should increase the likelihood of policy expansion. Parties and Policy Adoption  Parties play an important role turning citizens’ views into policy.  As a result it is important to understand the impact that parties have over policy change. Parties’ incentives to follow through with their policy commitments are likely to be influenced by the positions of their competitors. This is highlighted in the latter two subsections, on  55 cross party support and left-right ideology. Policy adoption is should require support amongst government and opposition parties and left and right parties. The amount of support multiculturalism needs in order for policy adoption to occur is high. Parties' Influence over Policy Adoption On a number of different issues, parties exert influence over policy. As organizations that control both the executive and legislative branches of government, parties influence the changes in law and policy needed for policy change. They also have strong incentives to implement their platforms. Parties that follow through on their commitments can campaign on them in future elections. There is a link between parties and policy change in a number of different policy areas. Left governments are more likely than right governments to protect the welfare state against retrenchment efforts (Garrett, 1998), the United States does more to redistribute wealth under Democratic Presidents than Republican ones (Bartels, 2008; Hacker and Pierson, 2010), and market-based pension reforms in Britain went further under the Thatcher Conservatives than under Labour (Araki, 2000).  Similarly, parties have strong incentives to follow through on commitments to introduce multiculturalism. These policies grant concentrated benefits to a small segment of the population. It is easier for voters to organize in support of these policies. It is also easy to identify the voters that benefit from the policy commitments parties make, and convince those voters that the benefits they gain are large. This means that parties’ are likely to be rewarded by voters by their follow though on commitments to policy adoption. It also makes it more likely that minority voters will remember parties’ failure to follow through with their commitments, and that they will vote accordingly. Even if  56 voters do not draw these conclusions on their own, organizations that support multiculturalism can place pressure on parties by mobilizing supporters. Multiculturalism’s concentrated benefits make it easier for organizations to solve collective action problems related to mobilization, and punish parties that do not deliver on their commitments. Parties that have not made the same commitments to support multiculturalism are unlikely to face as strong pressure to adopt policies. These parties may have an electoral coalition that includes few ethnic minority voters and therefore are unlikely to be concerned with losing minority votes. If they do have substantial support amongst minorities, they may have won the support by making commitments to other policies.  Parties may be less likely to follow though on commitments to multiculturalism when they face a mobilized anti-multicultural public. Harrison argues that governments are sometimes willing to impose concentrated costs on small groups of voters if a large number of people are mobilized around the issue (Harrison, 1996). Under most conditions, collective action difficulties should prevent individuals opposed to multiculturalism from organizing to influence government. When opponents overcome collective action problems, governments come under substantial pressure to abandon such policy commitments. Far-right parties (which increase the salience of immigration issues (Green-Pedersen and Krogstrup, 2008; Koopmans et al., 2012; Odmalm, 2012; Perlmutter, 1996)), referenda (Manatschal, 2015; Milic, 2015), or dramatic events such as the Theo  57 Van Gogh murder in the Netherlands12 (Vanparys et al., 2013) can help overcome collection action problems mobilizing anti-multicultural voters. All of these variables have the potential to increase the salience of multiculturalism either by organizing anti-multicultural voters or by publicizing unpopular policies. These variables have the potential to lead voters who otherwise would not pay much attention multiculturalism or immigration issue to consider it when voting. This increased attention is likely to increase parties’ concern about a backlash if they follow through which commitments to policy adoption. Party positions should finally influence policy adoption because parties influence the choices that are available to voters in elections. They play a central role aggregating public opinion and turning opinions into coherent policy (Budge et al., 2012). The ability for citizens to express opinions electorally is constrained by the options that parties offer them. When all parties favour multiculturalism, it can be difficult for opponents to find a party that expresses their viewpoint. Under such circumstances, parties face little threat of a backlash from voters and have strong incentives to follow through on their commitments. In these circumstances, voters unhappy with policy change have no vehicle through which to express their discontent. When there is broad opposition to multiculturalism amongst parties it can be difficult for proponents to find a party to support. In these conditions, parties can block access to the policy process and make it more difficult for advocates to see the policy change that they seek. Advocates and voters should have a more of an ability to work through parties to influence policy when there is partisan support for multiculturalism.                                                12 It is important to note that Vanparys et al. find a change in the salience and tone of public debates surrounding Islam but do not find an impact public support for making religious accommodations (Vanparys et al., 2013).  58 Parties matter to policy development, but not equally across different conditions.  The next two sections consider the confounding factors that affect parties’ influence over policy.  In particular they highlight the effects of cross-party support and left-right ideology on parties’ decisions to turn promises into policy. Partisan Consensus and the Adoption of Multiculturalism Policies Parties have incentives to break their commitments when following through on such commitments would alienate large numbers of voters or a party’s core constituency. Governments that introduce policies in the face of opposition are vulnerable in two ways. First, opposition parties may affect the salience of multiculturalism with the electorate. In the absence of an actor mobilizing anti-multicultural opinion, multiculturalism and immigration issues are rarely salient with the public (Lahav, 2004). This gives parties a fair amount of leeway to introduce pro-immigrant policies such as multiculturalism without fear of electoral repercussions. This logic is used by Freeman (2002) uses this logic to explain why countries across Europe and North America have more liberal immigration policies than public opinion in them would support. When the public is mobilized against multiculturalism, parties’ incentives change. Voters mobilized against multiculturalism policies can do significant damage to a government’s chances of re-election. This is the case even if the policies in question provide diffuse benefits. Because these diffuse benefits affect a large number of voters, parties can face large electoral costs if they ignore their interests. This analysis is highly relevant to multiculturalism (and broader integration policy) because voters tend not to be supportive of multiculturalism (Freeman, 2002; Howard, 2010).   59 Political parties play a central role in mobilizing this opposition. There is a great deal of existing work that shows that anti-immigrant and anti-multicultural is much more likely to have an impact on elections and on policy if it is mobilized by a far-right party (Green-Pedersen and Krogstrup, 2008; Howard, 2010; Koopmans et al., 2012; Odmalm, 2012; Perlmutter, 1996). The ability to mobilize opposition to multiculturalism should be limited to far-right parties. Other opposition parties may see a government's introduction of multiculturalism as an opportunity to mobilize opposition and hurt a governing party.  This is especially the case if the opposition party has a history of opposition to multiculturalism. Indeed, in the Netherlands this opposition was first articulated, not by the far-right, but by the centrist VVD. The VVD did so in the early 1990s; well before the far-right List Pim Fortuyn or Party For Freedom entered Dutch politics (Vink, 2007).  The potential for opposition parties to mobilize voters means that governments have to pay careful attention to the positions of opposition parties when deciding whether to follow through on commitments to policy adoption. If the opposition is generally supportive of multiculturalism, the government has a strong incentive to follow through on its policy commitment. Under these conditions, the government can deliver the promised benefits to its minorities without worrying about an opposition-mobilized backlash. When the opposition opposes multiculturalism, the governing party has to worry that such opposition will lead to a backlash against policy adoption. Such a backlash has the potential cancel out any benefits are governing party receives by demonstrating its commitments to minorities’ interests. Because opposition parties can use opposition to multiculturalism to hurt government parties, policy adoption should be more likely when there is cross-partisan consensus in support of it.   60 The second way in which governments are vulnerable to opposition lies in the extent to which opinion on multiculturalism cuts across left-right ideological divides. This makes it possible for opposition parties to use policy adoption as a wedge issue to hurt governments. Both parties of the left and parties of the right have electoral coalitions that include supporters and opponents of multiculturalism. On the left, there are social liberals that support multiculturalism out of a commitment to cultural inclusiveness and solidarity across different communities. There are also individuals on who oppose multiculturalism and immigration because of concerns that immigration will increase unemployment or that multiculturalism will take resources away from social programs. On the right, there are social conservatives that oppose multiculturalism, but also social liberals that may support it out of a commitment to treat all cultures equally or out of a commitment to pro-immigrant policy (Kriesi et al. 2008).13 Both left party and right party electoral coalitions can break down if the salience of multiculturalism increases. Two factors should make parties opposed to multiculturalism willing to raise the issue even though both left and right parties face potential electoral divides over it. The first is that in most countries opposition to multiculturalism usually exceeds support. This means that there are often more anti-multicultural votes to be gained by a party that mobilizes anti-multicultural opinion than there are pro-multicultural votes to be lost. Second, a party that has campaigned on opposition to multiculturalism will have likely already lost some of their pro-multicultural voters. Having lost many of these voters, a party has an incentive to use its opposition to try to win some of the government parties’ voters opposed to policy adoption.                                                13 Certain individuals on the right, particularly in the business community benefit from high levels of immigration because it increases the labour force that they can draw on.  61 Tensions in the United States over immigration and race demonstrate how these issues can be used as wedge issues.  Both Democrats and Republicans used issues surrounding race and immigration to break up each other's electoral coalitions. Through the late 1970s and 1980s, the Republicans used issues surrounding race and ethnicity to divide a Democratic coalition that included both Northern liberals and Southern conservatives. Through the 1990s, the Democrats used immigration as an issue to split the Reagan coalition that included both social liberals and social conservatives (Jeong et al., 2011). At different points in history, both major American parties were vulnerable to increases in salience of cultural and immigration issues and both have been hurt by the other's use of it as a wedge to break up their electoral coalition. Parties that support multiculturalism are similarly vulnerable to an opposition that decides to make their adoption of multiculturalism an issue. The best-case scenario for a governing party that wants to adopt multiculturalism is that they face an opposition that supports multiculturalism.  This allows them to adopt policy with little fear of a backlash. Under these conditions, the low salience of the issue with most voters protects the government, allowing them to deliver on their commitments. When a governing party faces opposition parties opposed to multiculturalism, their optimal strategy should be to avoid acting in order to reduce the risk of a backlash against them that splits their electoral coalition. Left and Right Parties and Policy Adoption Left and right parties have different incentives to follow through with their commitments. This stems from the parties' different levels of appeal with ethnic minority communities. Parties that have developed a strong appeal amongst a group of voters need to do less to  62 hold on to those votes than other parties. Parties that are trying to win over a new group of voters often are fighting a perception that they have historically unresponsive to those voters. As a result, it is essential that parties trying to make in-roads into a voting block deliver on their promises to those voters. Parties with long histories of support do not face the same credibility issues and therefore have some leeway to break their promises.  The extent to which a party has multiple avenues through which it can appeal to certain voters can also give it more leeway. A party that appeals to voters on a number of different issues may be able to deflect criticism over failure to follow through on some commitments by pointing to others that it has delivered. Parties that rely on one policy to win the support of a particular group of voters do not have this same ability.  The extent to which a block of voters is competitive will affect parties' responsiveness to them. This plays out in both the United States and Britain. In the United States, politicians are not highly responsive to African-Americans' interests because of the extent to which they vote for the Democratic Party. Democrats have little incentive to respond to their interests because they already have won the votes of large numbers of African-American voters. Republicans have little incentive to respond to their interests because it is unlikely the Republicans can win many votes in the African-American community (Bartels, 1998; Leighley, 2001).  A similar dynamic occurs in the United Kingdom where the Labour Party is much more responsive to immigrants and ethnic minorities when they face competition from the Liberal Democrats or other minority friendly parties (Dancygier, 2010). When parties face competition, failure to follow through on their commitments to minority voters carries with it the threat of defection to another party. When there is little competition  63 over minorities’ votes, parties can get away with paying lip-service to their concerns in election platforms and then not following through on those commitments because there are no other parties for minorities to defect. These mechanisms should all lead right party support for multiculturalism to have a greater impact on policy adoption than left party support. Left parties tend to have greater support within ethnic minority communities because of their greater support amongst low-income voters and their broad support for social solidarity (Bird et al., 2011; Dancygier, 2010; Dancygier and Saunders, 2006). Paradoxically this gives left parties less of an incentive to follow through on their commitments to be supportive of multiculturalism for each of the reasons above. Left parties may often support multiculturalism, but their willingness to follow through on this support should depend on mainstream right agreement. Right parties do not have the same minority support and therefore have to follow through on their commitments in order to demonstrate their responsiveness to minorities’ interests. Similarly, a left party can appeal to minorities on economic issues in a way that most right parties cannot. Because minorities tend to be more economically disadvantaged than the rest of the population, many left party economic policies are likely to have some appeal within minority communities. If a left party fails to follow through on a commitment to multiculturalism, it can point to success in other policy areas in order to demonstrate that it is still providing benefits to immigrant and minorities. Because right parties tend to have economic policies less closely aligned with immigrant and minorities, they do not have the same ability.   64 While left parties have some incentive to follow through on their commitments in order to solidify their support within minority communities they also may have incentives to balk at following through on commitments. Because opposition to multiculturalism often crosses left-right divides, left parties often have supporters that oppose multiculturalism. Being too aggressive in the pursuit of multiculturalism may open up opportunities for opposition parties to use multiculturalism as a wedge issue to break up its electoral coalition. In particular, anti-multicultural opposition parties might use a left party's support of multiculturalism to try to win over left working class voters who are sceptical of immigration. When opposition parties oppose multiculturalism, left parties are likely to be better off highlighting their work on economic issues on which their anti-multicultural working class and immigrant and minority voters agree than risking breaking up this coalition by pushing for the adoption and expansion of multiculturalism policies. Right party support for multiculturalism should increase the likelihood that left parties follow through on their commitments, because right party support changes party competition within minority communities. Absent a right party that is supportive of multiculturalism, left parties often face few challenges when trying to win ethnic minority votes. Left parties in this situation may include support for multiculturalism in their platform, but their incentives to follow through on these policy commitments are weak, especially when doing so may alienate their anti-multicultural supporters. This changes when a right party takes a position supportive of multiculturalism. Support for multiculturalism suggests that a right party is making at least some attempt to appeal to ethnic minority voters. This increases the potential for competition between left and right  65 parties over minority votes. Under these circumstances, left parties have strong incentives to follow through on their commitments in order to ensure that minorities do not defect to the pro-multicultural right party. Increases in right party support for multiculturalism should, thus, not only increase the likelihood of policy adoption when the right party is in government. They should also increase the likelihood of a left party introduces policy when it is in government. The ideology of the party that supports multiculturalism thus should be important to determining whether partisan support for multiculturalism translates into policy adoption. On their own, left parties face limited pressure to follow through on their commitments to adopt multiculturalism policies. Right parties, however, because of their weaker support in minority communities have more an incentive to follow through on policy commitments. Right party support further places pressure on left parties to keep their promises to minorities. Ethnic Minorities, Far-Right Parties, and Party Positions Because of parties’ influence over policy, it is important to understand ethnic minorities’ and far-right parties’ influence over mainstream parties.  Mainstream parties face competing pressures from ethnic minorities and far-right parties.  Their incentives to respond to ethnic minorities are further influenced by the way electoral systems affect minorities’ influence over election results. Ethnic Minorities and Support for Multiculturalism The size of a country's ethnic minority population should increase parties' support for multiculturalism. Ethnic minorities have strong incentives to advocate for such policies. They benefit from policy adoption in a variety of different ways. Policies such  66 as funding for ethnic minority organizations, mother-tongue education, and affirmative action confer material benefits on minority communities. Other policies, such as religious and cultural exemptions from dress codes, make it easier for minorities to integrate economically, and are likely to improve minorities' economic conditions. Finally, symbolic recognition of minorities' contributions to a country through legislative and constitutional recognition policies, as well as through the development of multicultural and inter-cultural school curricula, are likely to increase minorities' sense of belonging in a country. These policies make it easier for minorities to interact with government (Bloemraad, 2013) and for immigrants to obtain citizenship (Bloemraad, 2006). All of these benefits give minorities strong incentives to push parties to support multiculturalism.  Much of the existing literature confirms ethnic minorities' support of multiculturalism. Organizations representing ethnic minorities played an important role pushing for policy adoption in Australia and Canada (Fleras and Elliot, 2002; Lopez, 2000; Pal, 1993). In Britain, ethnic minority politicians have been strong advocates for multiculturalism (Adolino, 1998). Finally, cross-national work by Koopmans et al. (2012) shows that European countries with larger ethnic minority populations are more likely to adopt multiculturalism policies. This suggests that the benefits that minorities gain from these policies are sufficient at the very least, to push ethnic minority elites and organizations to advocate for their adoption. Parties' incentives to support multiculturalism depend on the size of the minority population in a country. The larger the size of a country's ethnic minority community, the larger the portion of the voting population that they make up, and the more ethnic  67 minority votes a party needs to win government. A party in a country with few ethnic minorities, such as Denmark, can afford to ignore minority voters. A party in Canada, provided there is some competition between parties over ethnic minority votes, cannot. As a result of this, countries with larger ethnic minority populations should have parties that are more responsive to minorities' interests and more supportive of multiculturalism. Parties' incentives to respond to minorities’ interests are not just a function of the size of a country's minority population, but also of its citizenship rules. Large numbers of ethnic minorities come from immigrant backgrounds and, as a result, their ability to fully participate in politics depends on their ability to obtain citizenship. Because countries have very different rules regarding citizenship acquisition, one should not assume that all increases in the size of a country’s ethnic minority population will have the same impact on parties incentives to respond to minorities. If citizenship acquisition in a country is easy, parties should be highly sensitive to changes in the size of its ethnic minority population. If, on the other hand, citizenship acquisition is difficult, parties may barely respond at all to changes in the size of the ethnic minority population. Parties only have a strong incentive to respond to minorities' interests if those minorities can obtain citizenship and vote in elections. This can explain why Switzerland has a substantial foreign born and ethnic minority population but weak partisan support for multiculturalism and weak policies. It is very difficult for immigrants to Switzerland to obtain citizenship; therefore, Swiss parties have weak incentives to respond to their interests. Parties also have incentives to support multiculturalism in order to gain the support of minority organizations. Ethnic minority organizations play a key role in  68 mobilizing ethnic minority voters (Berger et al., 2004; Bloemraad, 2005). As a result, gaining the support of ethnic minority organizations can be valuable to parties trying to reach minority voters. Support from such organizations can also win parties endorsements, volunteer support, and financial contributions. All of these things improve a parties' likelihood of winning votes with ethnic minority communities. If a party's support of multiculturalism wins it the support of a minority organization, it can increase its support within a minority community even if such positions have a limited impact on individual voters’ decisions. Ethnic Minorities and Electoral Systems Electoral systems are critical moderators of electoral strength. Voters' influence over election results will depend on the way that an electoral system translates their votes into their seats. The greater the influence a group of voters has over election results, the greater the incentive for parties to respond to their interests. In single-member district (SMD) systems, voters' influence over election results depends on their geographic distribution. Large geographically dispersed voting blocks will have a greater impact on elections. A large group of geographically concentrated voters will determine the outcomes in some districts but have no influence over others, and thus have less influence over election outcomes (Calvo and Rodden, 2015; Jusko, 2015; Rodden, 2010). The opposite is true for small groups of voters. Small groups of geographically dispersed voters have limited influence over election results because they are a small minority of voters in every district. When a small group of voters is geographically concentrated, however, they make up a large share of the vote in influence outcomes in particular districts (Johnston and Ballantyne, 1977). The way proportional systems translate votes  69 into seats does not depend on the geographic distribution of votes, so they are not subject to the same dynamics.  Ethnic minorities tend to be small voting blocks that are geographically concentrated (Dancygier, 2010; McGarrigle, 2016; Phillips, 2006; Triadafilopoulos, 2012). This should lead changes in ethnic minority electoral strength to have a greater impact in SMD systems than in proportional ones. In countries with SMD systems and very small ethnic minority populations, ethnic minorities make up a small proportion of the voting population in every district and have little effect on election results. As the ethnic minority population grows, however, the ethnic minority population is likely to disproportionately increase in districts with large minority populations. As a result, ethnic minorities will become pivotal voters determining the outcomes of seats faster in SMD systems than they will in proportional systems. Huber (2012) points to this mechanism when explaining why SMD systems politicize ethnicity to a greater degree than proportional ones. Triadafilopoulos (2012) also uses it to explain why Canadian parties have been more supportive of multiculturalism than German ones. In addition to this, in SMD systems ethnic minorities can play an important role determining election outcomes in districts that are close regardless of the size of the ethnic minority population in the district. If, for example, the difference between the first and second place candidates is less than 1%, the ethnic minority population in the district need only be 1% or 2% (depending how split minorities’ votes are) in order to have an important impact on the outcome in that district. SMD thus increases the influence of ethnic minorities in both districts with large ethnic minorities and in districts where the margin of victory is small enough that even a change in the voting behaviour of a small  70 ethnic minority community can affect which party wins the district.  This is a second mechanism that Huber (2012) points to in explaining the increased politicization of ethnicity in SMD systems.     Far-Right Parties and Mainstream Party Support for Multiculturalism Parties taking positions on multiculturalism face strategic dilemmas when confronted by far-right challengers. Meguid (2008) outlines three different strategies that mainstream parties can pursue in response to fringe parties like far-right parties. A mainstream party can pursue a dismissive strategy, hoping that refusing to engage with the fringe party will make it irrelevant. A mainstream party can adopt an adversarial strategy, emphasizing its opposition to the fringe party. A party employing this strategy hopes that its opposition will capture voters opposed to the fringe party while forcing its mainstream competitors into a difficult decisions over to how to engage with the issues raised by the fringe party. Finally, a party can pursue an accommodating strategy by co-opting fringe party positions in order to prevent voters from defecting to it. Each strategy has implications for how parties position themselves on multiculturalism. Parties that respond to the far-right with adversarial strategies should increase their support of multiculturalism while parties pursuing accommodating strategies should decrease their support of it.  Ideology should influence parties’ choice between an adversarial and accommodating strategy. For mainstream right parties the proximity of far-right parties on the left-right spectrum should make them more likely to pursue accommodating strategies. Kitschelt and McGann (1995) find that far-right parties often combine anti-multicultural and populist positions with free-market ones to create centre-right, anti-immigrant platforms. As a result, mainstream right voters do not have to compromise on  71 left-right issues in order to vote for a far-right party. This make these voters more likely than left voters to defect to a far-right party.  Failure to accommodate far-right positions carries a high degree of risk for mainstream right parties. Because the major multiculturalism and immigration issues separate the mainstream right from the far-right, failure on the part of the mainstream right to co-opt far-right positions leaves anti-multicultural centre-right voters with the easy decision to defect to the far-right party. They can get the centre-right positions they prefer from both the mainstream and far-right parties, but, unless the mainstream right co-opts the far-right's positions, they can only get the anti-multicultural policies they prefer from the far-right party. Mainstream right parties thus face strong pressure to adopt accommodating strategies. This is consistent with scholarship on the far-right Front National (FN) in France that links decisions by the centre-right Chirac government to adopt more restrictive immigration with the rise of the FN (Marthaler, 2008; Schain, 2008). Left parties face a much more difficult decisions when deciding how to respond to the far-right. Unlike mainstream right parties, there is usually some distance between the mainstream left and far-right parties on the left-right dimension of politics. This means that left voters have to make more of a compromise when defecting to a far-right party. An anti-multicultural left voter that defects may get the anti-multicultural policies they prefer, but lose the leftist policies in other areas that they support. At the same time, left parties have to be careful of the extent to which opposition to multiculturalism cuts across the left-right spectrum (Bruenig and Luedtke, 2008; Kriesi et al., 2008) and of the fact that far-right parties have demonstrated an ability to take votes from left parties (Oesch,  72 2008). Hinnfors et al. (2012) show that even in the absence of a far-right challenge, many working class left voters' concerns over competition from foreign labour can place pressure on left parties to favour restrictive immigration policies. Left parties are not as vulnerable to losing votes to the far-right as mainstream right parties are, but they are not immune far-right challenges either. Left parties also have incentives to adopt adversarial strategies. They can benefit from opposing far-right parties in two ways. First, left parties tend to have strong support within ethnic minority communities (Bird et al., 2011; Dancygier, 2010). Accommodating strategies that co-opt the positions of the far-right parties can cost left parties support within these communities. This is especially true when a left party faces competition from other pro-multicultural parties such as other left parties or green parties. Second, adversarial strategies can force left parties' mainstream right competitors into difficult decisions. An adversarial strategy can ensure that the anti-multicultural viewpoints of a far-right party play a central role in public debates during an election. The mainstream-right party then has to respond to these issues and, regardless of its response, risks losing votes. If the mainstream-right party co-opts the positions of the far-right party (which is the more likely outcome), the mainstream left party can use its opposition to the far-right party to win over pro-multicultural mainstream right voters. If the mainstream right party does not co-opt the positions of the far-right, it will lose support to the far-right and the left faces a weaker mainstream right competitor. Left parties have an incentive to pursue adversarial strategies because such strategies can weaken their mainstream right competitors.  73 The dilemmas that left parties face when responding to the far-right are likely responsible for the conflicting findings in the existing literature regarding left parties' responses to the far-right. Meguid (2008) and van Heerden et al. (2014) both suggest that left parties have adopted adversarial strategies in France and the Netherlands while cross-national work be Bale et al. (2010) suggest that left parties co-opt far-right parties anti-multicultural and anti-immigration positions. The way that left parties respond to the rise of a far-right party should depend on three things.  These include the extent to which the left party feels it needs pro-multicultural positions to hold on to ethnic minority votes, the extent to which it is worried about losing votes to the far-right, and the degree to which its losses can be off-set by the problems that an adversarial strategy can cause for its mainstream right competitors.  Summary This chapter builds a theory that examines the role that parties should play in the development of multiculturalism. It first outlines a number of expectations regarding multicultural policy development. It argues that there should be a range of different levels of policy adoption across countries, that policy retrenchment should be rare, and that the adoption of policies that affirm a country as multicultural and that fund ethnic minority organizations should increase the likelihood of additional policy adoption. These development patterns should shape parties' influence over policy development.  Widespread policy adoption leaves parties a great deal of room to influence policy development. It means that there is the potential in many countries to see the adoption of multiculturalism and that parties do not need to be able to ensure the development of a complete and comprehensive multiculturalism program in order to have a positive  74 influence on policy. Low levels of policy retrenchment mean that parties' influence over policy is much more important in years that take place before policy adoption than after. Finally, if recognizing a country as multicultural or funding ethnic minority organizations increases the likelihood of additional policy adoption, it is important to pay particular attention to parties' influence over these policies. The second part of the chapter builds a theory of how party positions should influence policy adoption. It highlights both partisan consensus and support from mainstream right parties as factors that should influence policy adoption. This has important implications for the development of party positions. If cross-party support for multiculturalism is necessary to increase the likelihood of policy adoption, it is important to pay careful attention to it. It may not be sufficient for advocates of multiculturalism to simply win the support of the governing party. In order to get the policy outcomes they want, advocates may need to convince parties across the political spectrum to support policy adoption. Further, if right parties’ positions have a particularly important impact on policy adoption it is important to understand the determinants of their positions.  The final section in this chapter outlines two theories regarding the influence that ethnic minorities and far-right parties should have over party positions. It argues that ethnic minorities should increase party support for multiculturalism, and their influence should be particularly strong in SMD electoral systems. In contrast, far-right parties should push mainstream parties to reduce their support, and should have a particularly strong effect on mainstream right parties. The chapter thus outlines a theory that ties ethnic minorities, electoral systems, and far-right parties to mainstream party positions. By also linking party positions to the policy development, it shows how ethnic minorities  75 and far-right parties can influence policy development through their impact on mainstream party positions.    76 Chapter 3 Methods and Measurement The dissertation uses of cross-case and within case analysis to test its theories. Chapters 4, 5, and 6 use quantitative analysis to identify trends in the development of multiculturalism, party influence over policy adoption, and the determinants of party positions that hold across different countries and across time. Chapter 7 focuses on two cases, Canada and the Netherlands, in order to illustrate the relationships that come through in the multi-country analysis. Analysis of Canada can demonstrate the impact of ethnic minorities and electoral systems over party systems.  An examination of the Netherlands can show support for multiculturalism across the left-right spectrum and the influence that far-right parties have on mainstream positions. This chapter will outline the methods used throughout the empirical analysis. The first section discusses the variables and measurement strategies used in the cross-case analysis in the first three chapters, while the second section will discuss the methods used in these chapters. The third section explains reasons for the dissertation's focus on Canada and the Netherlands, while the fourth will outline the analysis strategies used for chapter 7. Multi-Country Analysis- Variables and Measurement Measuring Multiculturalism Policies Measuring multiculturalism policies is a difficult and often controversial task. As discussed in the introduction, multiculturalism policies are integration policies that provide recognition or support to ethnic minorities. Several indexes measure citizenship and integration policy. Howard (2009) and Janoski (2010) have developed indexes to measure the openness of different countries' citizenship policies. Koopmans et al. (2005)  77 created an index to measure countries' citizenship and integration policies. Finally, Banting et al. (2006) developed an index that measures the strength of countries multiculturalism policies.  Of the existing indexes, the Banting and Kymlicka Multiculturalism Policy index fits best with this dissertation's aims. It looks at multiculturalism policies as opposed to citizenship policies, which is important given the distinctions made between integration and citizenship policies identified by Koopmans et al. (2005) and Wright and Bloemraad (2012). It also covers a wide range of countries including most of Western Europe, North America, Australia, New Zealand, and Japan. This gives it an advantage over work by Koopmans et al. (2005) that looks only at five European countries. A comparison that includes countries outside of Europe can lead to findings that are generalizable beyond Europe. The Banting and Kymlicka index also uses a much broader range of policies than Koopmans et al. do. When looking at the protection of cultural difference Koopmans et al. focuses primarily on policies that accomodate Muslims. While these policies are an important aspect of multiculturalism, they do not cover the full range of possible policies. The Banting and Kymlicka index, by including policies such as the inclusion of multiculturalism in the school curriculum and funding for ethnic minorities provides a much better coverage of multiculturalism policies than the Koopmans et al. index does. As a result, the dissertation uses the Banting and Kymlicka index to measure the strength of countries’ multiculturalism policies. The index is also valuable because it can be broken up to assess the extent to which a country’s multiculturalism policies are recognition policies, policies that ease minorities’ participation in society, or policies that provide material support to minorities.  78 The Banting and Kymlicka index scores 21 countries14 for multiculturalism policy adoption by looking at 8 policy indicators. These policies include:   the affirmation of a country as multicultural in legislation or in its constitution  the inclusion of multicultural or intercultural education in the school curriculum  policies requiring sensitivity to and representation of ethnic minorities in the media  policies providing for exemptions from dress codes for religious or cultural reasons  permission of dual citizenship  government funding for ethnic minority and multicultural organizations  support for mother-tongue education  affirmative action policies targeted at ethnic minorities For each policy indicator a country scores a 1 if they have fully implemented the policy, a 0.5 of they have a partial policy, and a 0 if they do not have a policy. A country might score a 0.5 if a government has indicated intent to bring a policy into place but only implemented certain aspects of it. For example, the Netherlands scores a 0.5 for representation of minorities in the media between 1982 and 1994 because the government had reported on the lack of minority representation in the media and developed some efforts to program to minority viewers. The country does not score a one, however, until 1995 when it allocated set amounts time to minority cultural programming on its national broadcaster. A country might also score a 0.5 if a policy only exists in some parts of the                                                14 These countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States.  79 country. For example, Belgium scores a 0.5 on affirmative action between 2002 and 2007 because such a policy was in place in Flanders but not Wallonia. Adding the scores for each policy provides a total score for the strength of a country’s multiculturalism policy.  The index can be broken down into three different categories of multiculturalism; recognition policies, policies that make it easier for ethnic minorities to participate in society, and policy that provide material support to minorities, as is shown in table 3.2. Recognition policies include affirmation of multiculturalism, the inclusion of multiculturalism in the school curriculum, and sensitivity to ethnic minorities in the media. Participation policies include dress code exemptions and the recognition of dual citizenship. Support policies include funding for ethnic minority organizations, mother-tongue education support, and affirmative action (Banting et al., 2006, 56-57). Summing scores for each of these groups of policy provides a score for the strength of a country's recognition, participation, or support policies.  Table 3.1: The Banting and Kymlicka Index by Policy Types Policy Type Policy Recognition Affirmation of a country as multicultural Inclusion of multiculturalism in school curriculum  Ethnic representation and sensitivity in the media  Participation  Exemptions from dress codes on religious or cultural grounds Acceptance of dual citizenship Support  Funding for ethnic minority organizations Provision of mother-tongue education  Affirmative action policies Classification of policies is taken from Banting et al. (2006).  80 Three indicators in the index are controversial. The use of dual citizenship as indicator of multiculturalism has been debated in the academic literature. Scholars tend to differentiate between citizenship policy and immigrant integration policy (Koopmans et al., 2005; Wright and Bloemraad, 2012). An open citizenship policy is consistent with a number of different approaches to immigrant integration, not all of which are multicultural. A country might have an open citizenship policy but still encourage immigrants to assimilate and adopt the majority culture. A dual citizenship policy can be a way of encouraging citizenship acquisition that is consistent with such an approach to immigration. The French republican model of citizenship and integration best exemplifies this approach (Brubaker, 1992; Favell, 2001). Dual citizenship policies would also be consistent with a neo-liberal approach to citizenship that would support an open citizenship policy but reject multiculturalism because of the government programs it involves. Dual citizenship could finally be part of a multicultural integration policy, allowing immigrants to maintain their culture by strengthening their ties with their countries of origin.  The inclusion of mother-tongue education in the index is also controversial. Mother-tongue education can be a multicultural integration policy that helps immigrants pass on their language and culture to their children. It can also be the opposite of a multicultural policy, if it is a way to discourage permanent immigration by ensuring immigrants will be able to return to their countries of origin. Germany used mother-tongue education for this purpose in the 1960s (Vermeulen, 1997).  Mother tongue education can finally be a catalyst for majority language learning. When used for this purpose, such programs are not multicultural but rather other integration policies  81 (Duyvendak et al., 2013). As with dual citizenship, there are arguments for and against considering mother tongue education a multiculturalism policy.  Finally, the inclusion of affirmative action policies in the index is controversial. These policies may be part of anti-discrimination or other non-multicultural integration strategies designed to increase ethnic minority work force participation (Duyvendak et al., 2013). When affirmative action policies are a way to reduce economic inequality or discrimination they do not contain the intent to protect or promote minority culture that is necessary for multiculturalism policies. At the same time, these policies might be considered multicultural policies if they create diverse work places that reduce the pressure on minorities to give up their culture in order to integrate economically. For example, workplaces that are more diverse might be more likely to grant minorities exemptions from policies that come into conflict with religious or culture practices.  The extent to which these are multicultural policies depends on the extent to which they fit with a broader government program to recognize and support cultural diversity. If all a country does is adopt a dual citizenship policy or an affirmative action policy, it is hard to believe that policy actors consider the policy part of a broader multicultural program. On the other hand, if a country adopts one of these policies along with other policies that are more clearly multicultural, it is reasonable to believe that these policies are part of a broader multicultural agenda.  The dissertation further uses a narrow definition to test the robustness of findings by Banting and Kymlicka (2014) that multiculturalism has been resistant to retrenchment.  To do this it looks specifically at policies classified as recognition policies (affirmation of a country as multicultural, inclusion of multiculturalism in the school curriculum, and  82 sensitivity to minorities in the media). It is unlikely that a government would adopt these policies if it did not at least have some interest in pursuing multiculturalism, and thus they provide a good robustness check against claims that Banting and Kymlicka (2014) do not find evidence of retrenchment because they have an over-broad measure of multiculturalism. The use of wide and narrow conceptualizations of multiculturalism in the dissertation allows it to make two important contributions to debates over what policies ought to be considered multicultural. First, it can determine the extent to which debates over what ought to be included matter in broader research into multiculturalism policy. If using wide and narrow conceptualizations of multiculturalism change the results of analysis, these debates are highly important to broader empirical work on policy adoption and the effects of policy. If, on the other hand, excluding a policy from the total measure of multiculturalism has little effect on findings, it may be less important to settle these debates in order to conduct empirical work.  Tests that look at wide and narrow conceptualizations of multiculturalism can shed some light on debates over what policies ought to be included in the index. If a policy fits with the general development path of other policies in the index, it is likely that governments are considering it a multiculturalism policy. If a policy does not fit the pattern, it is likely that approaches to integration policy other than multiculturalism are driving its adoption. Further, if a policy does not fit with the index, excluding it from analysis should strengthen the relationship between the adoption of a policy that is clearly multicultural and the total adoption of other policies in the index. Excluding indicators  83 that create noise should make the relationships between the presence of policies such as affirmation policies and other policies more coherent.  The dissertation uses total scores that include all policies and that exclude each of the controversial policies. In addition to tests of total policy adoption, it runs tests that look at individual policy adoption. These tests are in chapter 4, and determine the extent to which the adoption of one policy can increase the likelihood of other policies. Tests in chapter 5 determine whether parties have a unique ability to effect the adoption of particular policies. These tests use single policy indicator scores in place of total score.  The original index includes scores for 1980, 2000, and 2010.15 These gaps present a problem for much of the work in this dissertation. Analysis on the development of multiculturalism requires knowing the order of policy adoption and the gaps in time between the adoptions of different policies. Analysis that looks at parties' influence on policy adoption requires knowing parties' positions on multiculturalism in the years in which policies are adopted, and as a result, the exact years in which policies were adopted. In addition to this, the index's 1980 start date misses the adoption of significant numbers of policies in Australia, Canada, the Netherlands, and Sweden. This is not a problem for much of the past work done with the index (such as testing the impact multiculturalism has on the welfare state), but is a problem for an analysis of policy development.  To account for these problems, I created a unique and expanded version of the Banting and Kymlicka index. It covers every year between 1960 and 2011, and is the most comprehensive times-series cross-section index of policy adoption to my                                                15 The 2010 scores were added after the publication of the book cited in the previous paragraph. Scores for 2010 can be found online in the Immigrant Minorities section of the Multiculturalism Policies in Contemporary Democracies website at: http://www.queensu.ca/mcp/immigrant.html.   84 knowledge. The expanded index is as faithful to the original index as possible. It uses the same coding rules as the original index (found in Tolley, 2011) and shown in table 3.2. To expand the index I identified the policies used to justify the original scores (found in Tolley, 2011) and determined the specific ways in which policies were adopted. I also checked for the presence of similar policies not included in the original scores and for policy retrenchment.16  Table 3.2: Decision Rules for the Banting and Kymlicka Index Indicator Score Decision Rule Affirmation 1 Country has affirmed multiculturalism and has an implementing body. 0.5 Country has not affirmed multiculturalism explicitly, but has a relevant body; multiculturalism may also have been affirmed in some municipalities, but not nationally. 0 Country has not affirmed multiculturalism and does not have an implementing. MC School 1 Country has included multiculturalism in its curriculum. 0.5 Country has not formally or extensively adopted multiculturalism in its curriculum, but has engaged in rhetoric that supports such inclusion, implemented it in some districts, or developed intercultural or anti-racism education initiatives. 0 Multiculturalism is not included in school curriculum. Media 1 Ethnic representation, inclusion, sensitivity or diversity is included in the mandate of public broadcaster or media licensing. 0 Ethnic representation not mentioned in mandate of public broadcaster or media licensing. Exemptions 1 Country has granted exemptions or accommodations on religious grounds. 0.5 Some exemptions have been granted, but others have been explicitly denied. 0 Country does not grant exemptions or accommodations on religious grounds.                                                16 A full justification of the scores for each country is available online at: https://danielwestlakepolitics.wordpress.com/data/.  85 Indicator Score Decision Rule Dual Citizenship 1 Dual citizenship is permitted; foreign nationals may retain their original citizenship even after acquiring the citizenship of the host country. Note that some countries’ citizenship policies distinguish between the citizenship rights of foreign nationals and those of native-born émigrés. 0.5 Dual citizenship is officially prohibited, but tolerated in practice. 0 Dual citizenship is not permitted; foreign nationals must renounce or relinquish their original citizenship before acquiring the citizenship of the host country. Funding 1 Ethnic groups are provided state funding in the form of core- or project-based support. 0.5 Some ethnic groups receive state funding, but the practice is not widespread and the funding may be restricted to supporting the delivery of integration and settlement programs. 0 Ethnic groups do not receive state support. Bilingual Education 1 Country funds bilingual education or mother-tongue instruction either for children or adults. 0.5 Available in some provinces, states or areas, but not offered as a general rule. 0 Country does not fund bilingual education or mother-tongue instruction; refers also to cases where bilingual education is provided, but only as a means of facilitating the learning of the country’s official language. Affirmative Action 1 Country has an affirmative action policy that targets immigrant minorities; this may be in the public or private sector or both. Initiatives will extend beyond human rights policies and include targeted action aimed at removing barriers or more positive action measures such as quotas or preferential hiring.  0 Country has no affirmative action policy for immigrant minorities. Decision rules are taken word for word from: http://queensu.ca/mcp/immigrant-minorities/decision-rules.  Party Positioning on Multiculturalism  The dissertation uses Manifesto Project Data to measure parties' support for multiculturalism. The Manifesto Project scores parties' positions on a variety of policies  86 by measuring the percentage of mentions that a policy receives in parties' election manifestos. This data includes a score for a country's positive mentions of multiculturalism as well as a score for their negative mentions. For the dissertation, parties’ negative multiculturalism scores are subtracted from their positive scores. This is preferable to a total score that looks at the percentage of overall multiculturalism mentions that are positive. The emphasis of multiculturalism within a manifesto can be as important as the ratio of positive statements to negative statements. A party that devotes 5% of its manifesto to supporting multiculturalism makes a much greater commitment to the policy than a party that devotes 0.5%. This is the case even if both parties only make positive statements. Subtracting negative statements from positive statements captures this difference in emphasis in a way that a ratio does not. The Manifesto Project does not use mentions of the term “multiculturalism” to determine whether a particular statement if supportive or opposed to multiculturalism. Rather, statements are coded as positive if they include positive reference to cultural diversity or suggest that ethnic minorities should preserve their cultural heritage. Statements that refer to the importance of encouraging or enforcing cultural integration are coded as negative (Volkens et al., 2013b). For example, a 2003 Swiss Social Democratic Party mention of the value of including minorities in the police force is coded as a positive mention of multiculturalism. In New Zealand, a Labour manifesto arguing for the creation of a Minister of Ethnic Affairs and fostering language education for people from the pacific islands is coded as positive statement about multiculturalism. In the 2002 Swedish Social Democrat manifesto positive references to diversity, even without mentions of multiculturalism explicitly are coded as positive statements with  87 respect to multiculturalism (Volkens et al., 2013a). This approach to coding manifestos fits with the dissertation's approach to multiculturalism. The variation in the way that the term “multiculturalism” is used across countries requires coding party positions and policy statements for their substantive content and not use of particular terms.  The CMP index conflates multiculturalism with multinationalism. Statements that are favourable to the recognition and accommodation of national minorities such as the Quebecois in Canada or Catalans in Spain are coded in the CMP as positive mentions of multiculturalism. This is problematic because the dissertation excludes multinationalism from its definition of multiculturalism. Statements supportive of national minorities do not necessarily reflect a commitment on the part of political parties to support multiculturalism directed at immigrants and ethnic minorities. Indeed, in some cases national minorities see multiculturalism as mutually exclusive with their own recognition goals. Multiculturalism has experienced some opposition from Quebecois political elites in Canada and Maori elites in New Zealand concerned that multicultural recognition will replace recognition of their own distinct status17 (McRoberts, 1997; Spooney, 2005).  The conflation in the Manifesto Project coding is particularly problematic for regionalist and separatist parties such as the Bloc Quebecois in Canada or the Vlaams Belang in Belgium because it inflates their positive scores. This conflation also exaggerates far-right parties support of multiculturalism if far-right parties are strong supporters of federal institutions, as is the case in Belgium and Switzerland. To compensate for this, cross-party scores do not include separatist/regionalist parties or far-                                               17 This is not to say that all or even most national minority groups oppose multiculturalism. Rather it is to suggest that there is a tension between the recognition of ethnic minorities and the recognition of national minorities.   This makes the conflation of the two in the coding of party manifestos problematic.  88 right18 parties. This still leaves some noise in the party positions calculated for mainstream parties. Unfortunately without access to the manifestos for all of the countries in the analysis (the number of actual manifestos available in the data is limited), there is no way to compensate for this. Some noise in the multiculturalism scores is unavoidable for a data set that covers the range of time and countries that the Manifesto Project does. Comparing party support for multiculturalism across countries as it evolves over time requires combining the positions of multiple parties to create single scores for countries. An average of each party’s position, weighted by the share of seats each party won in the lower house, is used as this score. The weighting captures the different degree of influence that large and small parties have overall a countries' overall party system. The influence of a large party that forms government is not equal to a small opposition party that struggles to stay in the legislature. The dissertation uses four sets of scores, all of which are averages weighted by the parties' seats share. The first score is for cross-party support and includes all parties except for nationalist parties and regionalist/separatist parties. A second set of scores is for governing parties, which includes both the senior and junior partners of governing coalitions.19 Finally, the dissertation uses scores for both left and mainstream right parties. Parties are included in the left party score if the Manifesto Project data classifies them as either communist or social democratic.  Parties are included in the mainstream right score if they are classified as either a conservative or a Christian Democratic party. The dissertation does not use separate scores for other types of parties such as green parties, liberal parties, or agrarian                                                18 These parties are coded as nationalist in the Manifesto Project's coding scheme.  19 As with the cross-party, left, and right party scores, a party's contribution to the government average score is weighted by the percentage of all of the government seats that are held by the party.   89 parties.20 The variation in the presence of these parties across different countries and across time makes cross-country comparison of these parties' positions difficult.21  The positions of parties do not change suddenly at elections, but rather evolve over time. This is problematic given the Manifesto Project only has scores for party's positions in election years, and given that parties only generally publish policy platforms in the lead-up to elections. In order to account for how party positions change between elections, a linear trajectory captures how each party's position changes from election year to election year. This creates a score that can be used in non-election years, taking into account how party positions evolve and how that evolution affects the likelihood of policy adoption.  Figure 3.1 shows the distribution of cross-party multiculturalism positions. The narrow range of party positions is important, but not surprising. Multiculturalism policies make up a small proportion of most parties' platforms. Most parties take positions on multiculturalism; under 20% of observed party positions are zero (this is not fully clear in figure 3.1 because of the large number of cases that score between 0 and 0.5).22 10% of cases have party systems with negative total scores while slightly under 75% of cases have cross-party scores above 0. The overwhelming majority of cases fall between -0.5 and 2.5. This means that small effects for ethnic-minority and far-right party influence on party positions are important. When looking at the potential for different variables to                                                20 For a list of which parties fit into which category, see Appendix A. 21 For this reason, one should not expect the cross-party positions are not always an average of left party and right party positions. Cross party positions can be less supportive of multiculturalism than both left and right party positions if, for example, a centre party is strongly opposed to multiculturalism. 22 I consider parties that have an equal number of positive and negative mentions of multiculturalism to not have taken a position on multiculturalism. The ambiguity in these parties' platforms makes it difficult to determine whether the party is supportive or opposed to multiculturalism. It is likely that voters and other policy makers also have difficulty determining if such a party is supportive of multiculturalism.  90 change party positions, one should consider a change in multiculturalism of 0.5 or 1 to be a significant change in a party's position. Figure 3.1: The Distribution of Party Scores for Support for Multiculturalism  Bars show the percentage of all observations for cross party support for each score for cross-party support.  Separate manifesto coding is done for Canadian and Dutch parties for the analysis in chapter 7. Manifestos were coded for their support or opposition to multiculturalism using NVivo keyword searches for multiculturalism, immigration, cultural diversity, and related terms. Statements were coded as supportive or opposing multiculturalism depending on the context in which they appeared. Statements that were generally supportive of multiculturalism or cultural diversity or supportive of specific policies in the Banting and Kymlicka index were coded as supportive of multiculturalism. Statements that advocated the assimilation of immigrants or ethnic minorities, that  91 opposed cultural diversity, or that opposed policies in the Banting and Kymlicka index were coded as statements opposing multiculturalism. Manifestos going back to 1980 for all of the major parties in Canada and the Netherlands were coded.  This allows for an analysis that takes into account the deficiencies in the Manifesto Project coding of multiculturalism, particularly the conflation of multiculturalism with multinationalism. The manifestos were not available across a large enough time span nor enough countries to do this for the entire quantitative data set, but enough was available to provide data for the more limited case analysis conducted in chapter 7. Like with the Manifesto Project data, negative statements were subtracted from positive statements in order to create a total score for parties’ multiculturalism policies. Because this analysis looks at two countries there is no need to aggregate scores to create total scores for all parties in Canada and the Netherlands, nor for left or mainstream right parties.23  This coding is also a check for how closely the Manifesto Project data fits a narrower conception of multiculturalism. Figure 3.2 shows the relationship between the Manifesto Project scores and the recoded scores for both Canada and the Netherlands. The Dutch scores fit reasonably well with most positive recoded scores matching positive Manifesto Project scores and most negative recoded scores matching negative Manifesto Project scores. The Canadian data fits less well. There are a large number of strong positive or negative scores from the Manifesto Project that fall to zero after recoding. This results from the noise in the data created by debates over Quebec’s place in Canada.                                                23 The manifestos were downloaded from the Manifesto Project database (Volkens et al., 2013a), Political Documents Archive (Benoit et al., 2009), and Electronic Manifestos Canada (Birch et al., 2016). "The text(s) used come from the collection of political texts made available at www.poltext.org by Lisa Birch, Jean Crête, Louis M. Imbeau, Steve Jacob and François Pétry, with the financial support of the Fonds de recherche du Québec - Société et culture (FRQSC)."  92 Parties that are supportive of special recognition and accommodations of Quebec will have artificially high scores while parties that oppose such recognition and accommodation will have artificially low scores. This problem is unique to a couple of countries in the data set (likely Belgium, Canada, and Switzerland).  Figure 3.2: The Correlation between Recoded Scores and Manifesto Project Multiculturalism Scores  For Chapter 7 coding was also done of the manifestos of the two Dutch far-right parties that emerged in the 2000s, the List Pim Fortuyn (LPF) and the Party for Freedom (PVV). Anti-multicultural and anti-immigrant statements in these platforms were identified using an NVivo keyword search. These statements were then coded for the kinds of appeals that they made. Statements were coded as culturally chauvinistic, economic left appeals, or social left appeals. Culturally chauvinistic statements are statements that opposed multiculturalism or immigration out of concern for how  93 immigration or multiculturalism would affect Dutch culture. Economic left that opposed multiculturalism or immigration out of concern for the impact that it might have on the economy, employment, or social programs. Appeals were coded as socially left if they opposed immigration or multiculturalism because of the perception that it could threaten non-economic social equality. Examples of these claims include those that suggest multiculturalism threatens women’s equality or gay rights.  The coding of far-right parties' appeals measures the extent to which far-right parties are attempting to win the support of voters on the left and the right. After platforms are coded, percentages of anti-immigrant and anti-multicultural appeals that are cultural chauvinist, economic left, or social left appeals are calculated. Comparing these scores shows the extent to which these parties are appealing to voters on both the left and the right. Appeals that are not clearly culturally chauvinist, economically left, or socially left are not coded.  This means that the sum of chauvinist, economically left, and socially left appeals does not add up to 100%.  Ethnic Minority Electoral Strength  Ethnic minorities’ voting power is captured by a variable for ethnic minority electoral strength. This variable is designed to capture voting power independent of a countries’ electoral system. It serves as a proxy for the number of ethnic minority voters in a given country. The more responsive parties are to minorities, the greater the impact this variable should have on parties’ position. Two factors affect ethnic minority electoral strength. The first is the size of a country’s ethnic minority population. The larger the ethnic minority population in a country, the greater the share of the voting population they make up, and the greater  94 incentive parties have to respond to their interests. In addition, as the size of an ethnic minority population increases, so does the strength of ethnic minority organizations. Governments and political parties are more likely to listen to organizations that claim to speak for more voters. The more people an organization can claim to represent the stronger the argument they can make to governments that they represent the interests of a substantial portion of the population, and the more financial, volunteer, and vote support an organization can offer to a party. The dissertation uses the size of a country's foreign-born population as a proxy to measure the size of its ethnic minority population. Data on the size of a country's foreign-born population comes from the United Nations Department of Economic and Social Affairs (United Nations, 2008).  The electoral strength of a country's ethnic minority population is not only a function of its size, but also its access to citizenship. A country can have a large ethnic minority population, but that population will have little influence on parties and on policy adoption if it cannot vote. Because parties seek votes, it is necessary to consider access to citizenship when measuring ethnic minority electoral strength. The dissertation measures access to citizenship by using a reversed version Thomas Janoski's Barriers to Naturalization index.24 The Barriers to Naturalization index scores countries from 0-1 on 12 different aspects of the naturalization process: “good conduct, willingness to integrate, language skills, dual nationality, application complexity, application fees, state discretion in granting citizenship, residency requirements, jus sanguinis laws preventing jus soli naturalization of children, women allowed to maintain citizenship after marrying a                                                24 The Barriers to Naturalization index is preferable to Howard's (2009) measure of citizenship access for two reasons. The first is that it includes measures for scores in four periods (1970, 1980, 1990, and 2002) as compared to Howard's two. The second is that Janoski's score covers a wider range of countries, which is important to the large-n comparison conducted in this paper.  95 foreigner, and mothers when married to a foreigner being able to transfer citizenship to their children” (Janoski, 2010, 37). Countries scoring a one in the index have highly restrictive naturalization rules, while countries that score close to 0 have liberal immigration regimes. The index is reversed so that having liberal citizenship rules increases a country's ethnic minority electoral strength score. Scores for ethnic minority electoral strength are a product of the multiplication of Barriers to Naturalization scores and the size of a country’s foreign-born population. Ethnic Minority Electoral Strength  = Foreign-Born Pop. x (1-Barriers to Naturalization)  The size of a country's foreign-born population is not a perfect proxy for a country's ethnic minority population. It includes members of the foreign born population who do not identify as ethnic minorities. For example, the measure counts ethnic Germans who were born in Poland or other Eastern European as foreign born even if they do not identify as ethnic minorities. The experience of Germany is relatively unique in my data. Changes in its borders that resulted from two world wars and population movement that occurred during the Cold War meant that a significant number of ethnic Germans were born outside of Germany and immigrated back into Germany.  None of the other 21 countries have experienced quite the same border adjustments and so do not have ethnic majority immigrant populations of the same size as Germany. A larger concern regarding the validity of the dissertation's ethnic minority electoral strength measure is the number of ethnic minorities in a country that are not foreign born but have access to citizenship. If a country has a large number of 2nd and 3rd generation immigrants that have jus soli, or automatic, access to citizenship, the dissertation's measure of ethnic minority electoral strength will underestimate a country's  96 actual ethnic minority electoral strength. Some of this will be accounted for by the extent to with the Janoski Barriers to Naturalization index includes a measure for jus soli citizenship. Countries that grant automatic citizenship to the children and grand children of foreign-born residents will have a lower Barriers to Naturalization score, and therefore a higher a score for ethnic minority electoral strength.  There is a reasonably strong correlation between the provision of jus soli citizenship and the inverse Barriers to Naturalization score. Figure 3.3 compares Jonaski's scores with Howard's (2009) measure of jus soli in the twelve countries that are in both indexes for 1980 and for 2008. The presence of jus soli citizenship explains about 33% of the variation in a county's Barriers to Naturalization score. There is, thus some accounting for the presence of second and third generation immigrants in the dissertation's ethnic minority electoral strength score. Countries with larger foreign-born populations should have larger numbers of second and third generation immigrants. Countries that automatically grant citizenship to second and third generation immigrants at birth will have a lower Barriers to Naturalization score. This means that countries that grant large numbers of second and third generation immigrants citizenship will have higher ethnic minority electoral strength scores than those that do not.  97 Figure 3.3: The Correlation between the Inverted Barriers to Naturalization Index and Jus Soli Laws  This is admittedly not an ideal measure of the size of a country's ethnic minority population. An ideal measure would account for both the number of second and third generation immigrants in a country and their access to citizenship. Unfortunately, countries vary quite significantly in the ways in which they measure the size of their ethnic minority electoral strength. Some countries measure self-identification with ethnic groups, some measure citizenship and language, some measure race, and others do not measure ethnic identification at all. There is no perfect measure that captures the size of the ethnic minorities of near 21 countries in Europe and North America. Foreign born population, combined with a measure of access to citizenship that takes into account jus soli citizenship acquisition, is a measure that provides consistent data over a large number  98 of countries and time, and thus is the best available proxy for a country's ethnic minority population. In theory countries can score anywhere from a 0-100 for ethnic minority electoral strength. A country scores a 0 if it either scores a 1 on the Barriers to Naturalization index or if it has a foreign born population of 0%. Conversely, a country scores a 100 if it has no barriers to citizenship and its entire population is foreign born. In practice, scores for ethnic minority electoral strength vary between 0 and 20. Figure 3.4 shows the average ethnic minority electoral strength across 19 different countries in the dissertation's analysis.25 High immigration countries such as Australia, Canada, and New Zealand (the United States is an exception amongst these countries) unsurprisingly have high levels of ethnic minority electoral strength. Some of the countries that have had extensive empires in the past, such as France and the United Kingdom, have moderate scores. Ireland and Sweden also have moderate scores for ethnic minority electoral strength as well. Other countries such as Austria, Japan, and Switzerland have low scores.                                                25 No scores were calculated for Greece or Portugal because there are no Barriers to Naturalization index scores for those countries.  99 Figure 3.4: Mean Ethnic Minority Electoral Strength by Country  It is notable that there are several countries with high multiculturalism policy scores and low levels of ethnic minority electoral strength.  Finland and Sweden, in particular, stand out as cases with ethnic minority electoral strength scores below 5, yet multiculturalism policy scores that are higher than most other countries.  This suggests that, while there is an important relationship between ethnic minority electoral strength and policy adoption, there are important outliers where policies are adopted in the absence of electoral pressure.  In these cases it is important to look for other potential explanations of policy adoption. One explanation worth considering involves the role experts play in policy development.  While policy development can be influenced by the relative political power of advocates and opponents, it can also be affected by the way that experts and bureaucrats work through problems that their governments face.  Heclo (1974)  100 distinguishes between these two explanations of policy development referring to processes that involve responses to electoral or interest group pressures as “powering” and processes that involve experts and bureaucrats problem-solving as “puzzling.”  The importance of “puzzling” to the development of multiculturalism, particularly to cases that are outliers in models that look at electoral or interest group explanations of policy development, should not be discounted.  The growth in migration to industrialized countries has forced governments to deal with questions of how best to integrate immigrants from diverse cultural backgrounds.  Multiculturalism provides one potential solution to these questions.  Countries in which bureaucrats and other policy actors become convinced of the benefits of multiculturalism may adopt such policies even in the absence of electoral pressure.   Electoral Systems  The dissertation divides electoral systems into two categories, single member district26 (SMD) and proportional systems.27 A broad categorization of electoral systems is valuable because it maximizes the number of countries that fit into each category. Subdividing SMD and proportional systems by, for example, distinguishing between mixed member proportional and pure proportional systems, significantly reduces the number of countries in each electoral system category. This makes it more difficult to distinguish between the effect that the unique politics of a country is having in party positions and the effect of the electoral system. Including more countries in each category                                                26 The single non-transferable vote system used in Japan before 1994 is included in the SMD category even though it involved the election of multiple members. This is because an SNTV system should exhibit characteristics similar to SMD systems. The lack of proportionality in such a system means that a geographically concentrated group should still be able to be mobilized in a way that allows them exert a disproportionate influence over electoral outcomes. 27 The dissertation only considers electoral systems used to elect a country’s lower house.   101 reduces the likelihood that one or two outlier countries influence the electoral system effect.  Different types of SMD and proportional systems should have similar effects on parties’ incentives to respond to ethnic minorities. In single member plurality, alternative vote, and run-off systems the need for parties to win relatively small single districts gives them incentives to respond to geographically concentrated ethnic minorities. In contrast, the way that mixed member proportional systems enlarge districts and counter-balance them with list seats reduces the importance of these voters to election outcomes, and as a result, parties' incentives to respond to their interests. Like SMP and alternative vote systems, parties in mixed member proportional and pure proportional systems should behave similarly. Far-Right Parties Conceptually, far-right parties are parties that not only take strong anti-immigrant positions, but also make such positions a point of emphasis in their manifestos and in their campaigning.  These parties see anti-immigrant and ethnic nationalism as the key issue (or one of a couple of key issues) that set them apart from other parties.  This definition fits with work by Fennema (1997) and van Spanje (2010).  Parties that had some anti-immigrant or anti-multicultural positions but made other issues the focal point of their manifestos and campaigning, such as the Reform party in Canada28, are not considered far-right parties in this work. The presence of a far-right party measures their impact on a party system. Presence is preferable to strength because the strategies that mainstream parties pursue can affect the strength of far-right parties. A party that successfully pursues an                                                28 In the case of the Reform party these issues were regionalism and economic neo-liberalism.  102 accommodating strategy will both reduce the far-right parties’ support and end up with positions closer to it. Conversely, a party that successfully pursues an adversarial strategy may strengthen a far-right party even though it ends up taking positions that are farther from the far-right party. A far-right party may, as a result, have a strong impact on mainstream parties as it loses electoral strength, or have a weak impact on mainstream parties as it gains strength. Any examination of far-right parties that uses their electoral strength as an independent variable thus has to grapple with an endogeneity problem. It would not be clear whether far-right parties are affecting mainstream positions or whether mainstream positions are affecting far-right party strength.  In most cases, the Manifesto Project's coding of nationalist parties determines which parties are far-right parties. Four additional parties that are often considered anti-immigrant and anti-multicultural parties are coded as far-right parties. These include the Vlaams Belang (and Vlaams Blok)29 in Belgium, the Progress Party in Denmark, the Progress Party in Norway, and the New Democracy party in Sweden.30 A party is considered present in a country if it won at least 5% of the vote or a seat in the lower house of parliament. Once a party crosses this threshold, mainstream parties should start to see it as a threat and adjust their positions accordingly. A full list of the parties coded as far-right parties can be found in the table with party classifications in Appendix A. Control Variables A number of different control variables are included in the analysis. The first tests in the dissertation, examining the development path of multiculturalism policies, include three                                                29 The Vlaams Blok was the forerunner the Vlaams Belang. The Vlaams Belang was created after the Vlaams Blok was banned for violating Belgian anti-racism laws. 30 Tests that exclude that do not code these parties as far-right parties have also been run. Those tests tend to produce stronger confirmations of the hypotheses regarding far-right influence over the positions of other parties. The inclusion of these three parties thus serves as a conservative test of the paper's theories.  103 control variables. The electoral strength of ethnic minorities is used as a control to distinguish between increases in policy adoption that occur because of the adoption of previous policies (or changes in party positions) and policies that are adopted because of ethnic minority advocacy. It is possible that ethnic minority advocacy could be affecting both previous policies (such as affirmation or funding policies) and additional policies. Leaving ethnic minority electoral strength out of these models could lead to omitted variable biases. A second control used in these tests is economic growth. During periods of economic growth, governments have more revenue and should be willing to use government resources to develop or expand government programs like multiculturalism. Conversely, during periods of economic decline governments have fewer resources and may be less willing to introduce or expand existing government programs. Because overall economic growth is likely to place limits on what governments can spend on programs, GDP growth is used as a measure for this control. This is measured using OECD data (OECD, 2013). A final control is used in these tests for whether a country uses a federal system of government. Federal countries may have more difficulty introducing multiculturalism policies if control over the jurisdictions needed to implement such policies is divided between the federal and regional governments. For example, regional government control over education might limit the extent to which pro-multicultural federal parties can influence the inclusion of multiculturalism in the school curriculum or the develop of mother-tongue education programs. Federalism is measured using a binary variable taken from the Quality of Governance data set (Teorell et al., 2013).  104 Two additional controls are added for tests that look at parties' influence over policy adoption. The first control looks at whether the government is on the left or the right of the political spectrum. This is determined by whether the Manifesto Project left-right score for the parties in government is on the left of the Manifesto Project scale. This measure has a stronger impact on policy adoption than other measures of whether a government is left or right, making it the strongest control against which to test other independent variables. It also provides the best reflection of the way in which left/right positioning might affect policy adoption. A substantial amount of work links parties left parties with more liberal immigration policies (Howard, 2009; Givens and Luedtke, 2005). Scholars often assume that this relationship extends to multiculturalism. The argument made is that left parties are more likely to support multiculturalism because of their commitment to social solidarity. Left parties are expected to try to build solidarity between different disadvantaged groups be they economically disadvantaged, disadvantaged because of gender or sexual orientation, or because of cultural background or minority status. The argument is that the history left parties have of fighting for the greater social inclusion of disadvantaged groups such as immigrants and ethnic minorities should make them more likely to support multiculturalism. Using the Manifesto Project's left-right scale instead of their classification of party types allows for a measure that captures governments' changes in ideology. It distinguishes between the strong left ideology of the 1970s and 1980s Labour party and the centrist positions of Labour in the mid 1990s under Tony Blair.  Parties' support for nationalism is a second control variable. Parties that show a strong commitment to nationalism may be less willing to adopt multiculturalism out of a  105 concern that it threatens the countries' national identity. This is particularly the case in countries with an identity that connects closely to a single culture of ethnicity. Nationalism scores are calculated by subtracting the Manifesto Project's scores for negative mentions of the national way of life from positive mentions of the national way of life. Appeals to patriotism and nationalism are coded as positive mentions of a country's national way of life while opposition to such ideas are coded as negative mentions (Volkens et al., 2013b). The dissertation's measure of nationalism is calculated in a similar manner to its measure of multiculturalism. The tests that look at the influence of ethnic minorities and far-right parties over mainstream positions use two additional control variables. The first control is a lagged dependent variable, which captures the influence of the party’s previous position has on its current one. It separates out the impact that parties previous position has on its current position from the impact of ethnic minority electoral strength and far-right parties. This controls for cases in which a party is consistently supportive (or opposed) to multiculturalism independent of changes in ethnic minority electoral strength or the presence of far-right parties. The lagged dependent variable, however, cannot control for effects that co-vary with either of the independent variables of interest.31  A control for economic conditions is also included in these tests, because there is some literature that suggests that parties should not be as supportive of multiculturalism during economic declines. In weak economies, the majority population may see immigrants as competitors for jobs and for scarce resources (Alesina and Glaeser, 2004; Betz, 1994; Freeman, 2002; Joppke, 1999). If this is the case, declines in partisan support for multiculturalism should occur during poor economic conditions. As the majority                                                31 It should be noted that including a lag variable greatly increases R2 values.  106 population becomes more concerned about the economy, they should put more pressure on parties to adopt more restrictive immigration policies and to oppose the adoption or expansion of multiculturalism programs. Unlike tests that look at policy adoption, unemployment and not GDP should be influencing party positions. Where GDP is likely to affect the resources that a government has available to it, unemployment likely has a stronger impact on perceptions of labour market vulnerability. It makes sense to use unemployment as an economic control when looking at party positions because anti-immigrant sentiment often relates to labour market vulnerability (Makowsky and Stratmann, 2014; Mayda, 2006). It makes sense to use unemployment as an economic control when looking at party positions. Measures for unemployment come from OECD data (OECD, 2013).  There is no explicit control variable in the analysis for public opinion. The literature on immigration and integration policy suggests that public opinion has limited influence over party positions. The salience of immigration issues, such as multiculturalism, is often low (Freeman, 2002; Green-Pedersen and Krogstrup, 2008; Lahav, 2004). This means that it is rare that public opinion on its own will have a substantial impact on parties' positions on immigration. Parties have little incentive to respond to broad public opinion on an issue unless a large numbers of voters are considering the issue when deciding which party to support. When the salience of an issue with the public is low, individuals and interest groups, in this case ethnic minorities, most affected by the policy should wield a large amount of influence over parties' positions. Since these are the voters most likely to consider the issue (in this case multiculturalism) when voting, parties need to most concerned about their views. When  107 public opinion influences party positions, it often as the result of far-right party mobilization (Koopmans et al., 2012; Perlmutter, 1996).  Far-right parties are a key independent variable in the analysis.  In addition to this, it is difficult to find public opinion data on multiculturalism that is comparable across the number of countries and years included in this analysis. Eurobarometer and World Values Study data on immigration and multiculturalism before the 1990s is very limited. This would cause two problems for analysis. First, any findings would be limited in their generalizability across time and countries to the small number of observations for which there is public opinion data. Crucially this would exclude important incidences of policy adoption in Australia, Canada, the Netherlands, and Sweden. Second the sparse data would limit the extent to which tests with controls, particularly fixed effects controls, could generate results that one could have confidence in. Multi-Country Analysis- Methods Used Policy Adoption, Path Dependence, Survival Analysis, and Descriptive Statistics One of the challenges in modeling multicultural policy adoption relates to path dependence. Figure 3.5 shows that retrenchment is rare. As a result, observations that occur before policy adoption are more important to understanding independent variables' influence over policy than observations that occur after. Parties opposed to multiculturalism have a greater ability to prevent policy adoption in observations that occur before policy adoption, but limited ability to affect policy after adoption. Once policies are in place, retrenchment efforts have to contend with the positive feedback loops that help to keep policies in place.   108 Figure 3.5: The Development of Multiculturalism Policies Over Time  * Average policy adoption reflects the percentage of policies in each category that a country has adopted, with a maximum score of one. For example, a score of 0.5 for participation policies means that a country has adopted 1/2 participation policies. Percentages are used here so that policy categories with different numbers of policies can be compared.  A second challenge that affects the modeling of policy development is that policy adoption does not follow a linear trajectory. A country can adopt a policy, a partial policy, or no policy at all. Because there are only three possible scores for adoption, regressions models that assume a linear dependent variable do not make sense for this analysis. It is possible that a country with strong partisan support for multiculturalism will see no increase in policy because it has already hit the maximum score possible in the index. An ordinary least squares regression model would treat this as an observation  109 where party positions failed to influence policy adoption even though it is impossible to influence policy under such conditions.  In order to deal with these problems ordinary least squares (OLS) regression models are not used for any of the analyses that examine the development path of multiculturalism (chapter 4) or the influence parties have on policy adoption (chapter 5). Instead, Cox proportional hazard models are used to estimate the influence that particular policies or party positions have on policy adoption. These models use only the observations that occur before policy adoption in order to estimate the influence a different policy or a party has on policy adoption. Removing these observations accounts both for the inability of a party to push policy adoption past the maximum number of policies adopted as well as the limited ability parties opposed to multiculturalism have to cause retrenchment. For measures of multiculturalism that include multiple policies, each country re-enters the model immediately after a policy is adopted. It does this until it reaches the maximum possible number of policies adopted. This, in effect, creates a multiple failure32 hazard model. The Cox proportional hazard model assumes that each case in the study has the same baseline hazard, in this case likelihood of policy adoption. The baseline hazard is the estimated likelihood of policy adoption that does not account for the impacts of any independent variables. The difference in the baseline hazards shown in figure 3.6 demonstrates that this is not the case. Settler countries (Australia, Canada, New Zealand, and the United States) see the hazard of policy adoption peak in 1985 and then decline                                                32 Hazard models refer to events of interest as failures. These models are often used in engineering and medical analysis. Failures occur when a structure such as a bridge fails or when a patient in a drug trial dies. In this dissertation, however, a “failure” is actually a positive event. It occurs when a policy is adopted.   110 significantly while the hazard of policy adoption increase steadily over time. Settler countries such as Australia and Canada had large immigrant populations by the 1960s and so would have faced greater pressure to adopt multiculturalism policies through the 1970s. This can account for the earlier increase in policy adoption in these countries compared to non-settler countries. This also meant that there were fewer policies for countries to adopt in the 1990s, reducing the baseline hazard for policy adoption in that decade. Stratifying data can control for differences in baseline hazards (Therneau and Grambsch, 2000). Cox proportional hazard models throughout the dissertation therefore stratify data based on whether a country is a settler country or a non-settler country. This controls for the different baseline hazard rates of settler and non-settler countries. Failure to do so would weaken the fit of the Cox proportional hazard model.  111 Figure 3.6: Smoothed Hazard Rates for Total Policy Adoption in Settler and Non-Settler Countries  Lines show the baseline hazard (likelihood of policy adoption) without taking into account any explanatory variables.  A further challenge that exists when modeling the adoption of multiculturalism is that the likelihood of policy does not remain equal over time. In the 1960s and early 1970s multiculturalism policies were rare. Indeed, before the Canadian government’s adoption of multiculturalism in 1971, governments rarely considered adopting policies. It is unreasonable to expect governments to start adopting multiculturalism before that point. Analysis that has too early a start date can end up under-predicting a variable’s effect on policy adoption by including early years in which policy adoption was unlikely. To account for this, analysis uses four different start dates, 1960, 1970, 1975, and 1980.   112 Any models that include ethnic minority electoral strength do not use the 1960 start date because ethnic minority electoral strength data only goes back as far as 1970.  Crucially, no countries drop out of the analysis between 1970 and 1980. The exit of a country from the analysis could significantly change the relationship between independent variables and policy adoptions Because no countries drop out of the analysis, one can be fairly confident that that increasingly strong relationship between party positions and policy adoption that comes through in chapter 5 is not an artifact of a particular country dropping out of the analysis. The hazard models used in the dissertation can explain adoption, but cannot explain policy retrenchment. The low number of cases in which retrenchment occurs makes it difficult for a cross-country quantitative model to provide insight into the causes of retrenchment. There are simply not enough instances of retrenchment to allow for meaningful quantitative analysis. The vast majority of retrenchment occurs in the Netherlands, meaning that any attempt at quantitative analysis of retrenchment would be highly influenced by the Dutch case, regardless of whether that case is an outlier or fits broader trends. Because of this, analysis of policy retrenchment is better suited to small-n qualitative analysis. Hazard ratios show the change in relative likelihood of policy adoption for a change of one in a given independent variable. They are reported instead of the coefficients generally reported for OLS models. A hazard ratio of one means that a policy is as likely to be adopted after a change of one in the independent variable as it was before. A hazard ratio of 0.5 means that a policy is half as likely to be adopted after a change of one in the independent variable than before. Finally, a hazard ratio of two  113 means that a policy is twice as likely to be adopted after a one-point change in the independent variable than before.  Chapters 4 and 5 also use descriptive statistics to provide context for the hazard ratios presented in the regression analysis. These statistics show the number of policies adopted in different situations (for example, after countries have adopted affirmation policies or when there is strong cross-partisan support for multiculturalism). They show that the hazard ratios shown in the regression analysis correspond with meaningful increases policy adoption. The descriptive statistics can provide an indication of the overall likelihood of policy adoption and the extent to which policies adoption occurs in different contexts. Party Positions and Time-Series Cross-Section Analysis The constraints that require the use of hazard models in chapters 4 and 5 do not apply to analyses looking at party positions. Figure 3.4 shows that party positions go up and down over time. The fact that a party supported multiculturalism in the past does not prevent the party from opposing it in the future. There is also no practical upper-limit on a parties' support for multiculturalism. Technically a party cannot devote more than 100% of its platform to either supporting or opposing multiculturalism, which would place a limit of 100 or -100 on party scores. In practicality, no party is close to such a score. As a result, linear models are appropriate for tests on the determinants of party positions. The standard OLS regression model needs to be adjusted to account for the time-series cross-section nature of the data. Different observations in a time-series cross-section data set are not independent. Observations in the same country or that are close in time are related. Differences in a country's politics, national discourse, or national  114 identity may influence the relationship between ethnic minority electoral strength, far-right parties, and party positions in a particular country. As such, time-series cross-section regression models are used for tests on the determinants of party positions. Fixed effects and clustered standard errors, both by country, account for the relationship between observations from the same country. Using a fixed effects model is the equivalent of adding dummy variable for every country in the analysis; effectively controlling for the country an observation comes from. Clustering accounts for the wider standard errors needed when observations are related.  The lagged dependent variable in the models presented in chapter 6 can suppress some of the effect for each independent variable. The coefficient of an independent variable shows the immediate impact of the variable on party positions. An increase in the dependent variable in a given year, however, has additional impacts on party positions in future years because of the high correlation between the current and past positions. This means that an increase in a particular independent variable will have an indirect impact on future observations that will be hidden by the lagged dependent variable. A Koyck lag model compensate for this by calculating the over time effect an independent variable has on party positions.33 The model provides the asymptote that the variable’s effect would approach if it had an infinite amount of time to affect the dependent variable (Beck, 1991). The rate at which this increase occurs diminishes over time. Figures accompanying the discussion of ethnic minority electoral strength and far-right parties in chapter 6 show the over time effect for the major independent variables in the analysis.                                                33 The formula for the Koyck lag model is: total effect = instantaneous ethnic minority electoral strength effect / (1 - lagged dependent variable effect).   115 Like the regression analysis in chapters 4 and 5, descriptive statistics accompany the models in chapter 6. These help to put the effects found in the models into context. They show the extent to which the changes in the models correspond with changes in party support for multiculturalism over time. They also allow changes in party positions in different countries to be distinguished.  Single Case Analysis- Overall Approach  Chapter 7 builds on the work done in the previous chapters by looking at two cases to illustrate some of the relationships that come through in the cross-case analysis. The aim of this chapter is not to do independent theory testing but rather to provide a more in depth look at some of dynamics that come through in the cross-case analysis. A look at a single case can provide additional insight into the relationships presented in the cross-case analysis in a few different ways. Individual case analysis can allow for an examination of the relationship between ethnic minorities and advocacy for multiculturalism, demonstrating the extent to which ethnic minorities have been advocates for multiculturalism. Illustrative cases can also show the influence that ethnic minorities have over electoral results- providing insight into the number of seats that ethnic minorities influence and, as a result, parties' incentives to respond to their interests. Single-case work further allows for a more careful reading of the manifestos that parties put forward. This puts the commitments that parties make to multiculturalism in the context of a broader party platform. Corrections are made for the way that the Manifesto Project conflates multiculturalism and multinationalism.  Looking at the statements that parties make in context further allows for a deeper examination of the appeals that far-right parties make. Such an examination can demonstrate the extent to  116 which far-right parties are using anti-multicultural and anti-immigrant appeals to try to win the support of voters on the left and right of the political spectrum. This in turn provides some insight into the extent to which mainstream left and right parties have to deal with challenges from the far-right. Qualitative Methods- Case Selection Case Selection Framework Cases are selected to illustrate particular effects that come out of the cross-case analysis, not to provide generalizable testing of the dissertation’s theories. The aim of the case studies is not to establish generalized trends, but rather to illustrate some of the details that exist within the trends that come through in the larger cross-case analysis presented in the earlier chapters. The generalizability of relationships between parties and policy adoption and between ethnic minority electoral strength, far-right parties, and party positions should be established by the cross-case analysis, not by the illustrative cases examined in the single case analysis.  Two types of cases are valuable for illustrating the relationships that come through in the cross-case analysis. The first is a case with strong multiculturalism policies, strong partisan support for multiculturalism, a single member district electoral system, and strong ethnic minority electoral strength. The confluence of those four variables is valuable because it allows the case to demonstrate a number of things. The strong multiculturalism policy allows for an examination of the role of ethnic minorities in pushing for it. This is useful in providing some causal justification for the relationship between ethnic minority electoral strength and partisan support for multiculturalism that comes through in the cross-case analysis.   117 The combination of strong ethnic minority electoral strength, a single member district electoral system, and strong partisan support for multiculturalism allows for an illustration of the impact that electoral systems have on ethnic minority influence over parties. By looking at the propensity of ethnic minorities to support particular parties, one can examine the extent to which disproportionate support amongst ethnic minorities increases a party's seat share. Comparing this to the overall minority population in a country shows the extent to which single member district systems increase ethnic minorities' influence over election results. Showing the increased influence that ethnic minorities have on electoral outcomes in single member districts shows why there is a positive interaction between ethnic minority electoral strength and single member district electoral systems in the cross-case analysis. A case with strong partisan support for multiculturalism, a single member district electoral system, strong ethnic minority electoral strength, and strong multiculturalism policies is useful for showing why ethnic minority electoral strength interacts with single member district electoral systems to increase partisan support for multiculturalism. A second type of case that can be useful for illustrating the relationships in the cross-case analysis is one that has substantial partisan support for multiculturalism that declines following the emergence of a far-right party. The examination of party positions in such a country allows for an examination of how the statements on multiculturalism changed with the emergence of the far-right. A closer look at how party statements change after the emergence of the far-right allows for a more nuanced examination of the way that far-right parties appeal to different parties’ voters. The extent to which far-right parties connect their anti-multicultural appeals to issues important to mainstream right  118 voters (such as cultural conservativism) or mainstream left voters (such as the economic threat posed by immigrants) can be demonstrated by case analysis in a way that it cannot by larger cross-country analysis. Case Selection- Canada Canada is a case that can demonstrate ethnic minority influence in a single member district country. Canada fits the requirements for this case better than the other countries in the data set. Of the countries with single member district electoral districts, Japan34 and France have relatively weak ethnic minority electoral strength and partisan support for multiculturalism. The United Kingdom and the United States have slightly stronger ethnic minority electoral strength and partisan support for multiculturalism, but do not approach the same strength of Canada. The unique politics surrounding race in the United States also makes it possible that that country will be an outlier when one looks at the relationship between ethnic minority electoral strength and partisan support for multiculturalism (for a broader discussion of the unique politics surrounding race in the United States see Dawson, 1994 and Leighley, 2001). Australia has electorally strong ethnic minorities but weak partisan support for multiculturalism, making it a likely outlier when it comes to ethnic minority electoral strength and partisan support for multiculturalism. This leaves Canada and New Zealand as the best cases to illustrate the way that single member district electoral systems increase ethnic minorities' influence over party positions. New Zealand switched to a mixed member proportional electoral system in 1996, and, as a result, there are fewer examples in New Zealand that can be used an illustrative case than in Canada. As a result, Canada serves as the best illustrative                                                34 Japan also had a single non-transferable vote electoral system until 1994, when it adopted a proportional system.  The peculiarities of the single non-transferable vote system would make it a poor illustration of the way that single member district electoral systems affect ethnic minorities’ influence over parties.  119 case when one is looking at the influence ethnic minorities have on partisan support for multiculturalism in a single member district electoral system. Canada has electorally strong ethnic minorities, a single member district electoral system, strong partisan support for multiculturalism, and a strong multiculturalism policy.  The differences in ethnic minority electoral strength and party support for multiculturalism come from the data used in cross-country analysis of the dissertation. Different countries’ ethnic minority electoral strengths are shown in figure 3.3. This figure highlights the high level of ethnic minority electoral strength in Canada compared to other countries. Figure 5.4 shows partisan support for multiculturalism in different countries. It demonstrates that Australia, in spite if its high level of ethnic minority electoral strength, has relatively weak partisan support for multiculturalism. Case Selection- The Netherlands The Netherlands fits very well as a case that had moderate partisan support for multiculturalism that dropped with the emergence of a far-right party. The emergence of the far-right in the Netherlands has been one of the better studied instances of a backlash against multiculturalism. A great deal of literature examines both the emergence of the Dutch far right (The List Pim Fortuyn and later the Party for Freedom) and mainstream parties' response to it (Koopmans and Muis, 2009; Prins and Saharso, 2009; Scholten, 2012; van Heerden et al., 2014). There is a clear drop in support for multiculturalism that coincides with the emergence of far-right parties in the Netherlands. This provides a good case to illustrate the way that far-right parties push mainstream parties to weaken their support of multiculturalism. The fact that the centre-right Christian Democratic Appeal (CDA) is supportive of multiculturalism is supportive of multiculturalism before the  120 emergence of the far-right further makes the Netherlands a good case for in depth analysis. That the Dutch Labour party (PvdA) is also supportive of multiculturalism before the emergence of the far-right is also valuable.  This allows the Netherlands to be used as a case that illustrates that far-right’s impact on both the mainstream right and the mainstream left.  Figure 3.7 further illustrates the value of the Netherlands as a case for in depth analysis. The figure shows that the countries that had the largest drops in support for multiculturalism when far-right parties emerged, the Netherlands had the strongest support for multiculturalism. Only Austria and Denmark had larger drops in multiculturalism, and both had weaker pre-far-right support for multiculturalism than the Netherlands. Of the countries with strong pre-far-right support for multiculturalism, the Netherlands had the largest drop in cross-party support. Only Belgium and Norway had higher levels of pre-far-right cross-party support for multiculturalism, and the Netherlands had larger drops in support than both of those countries. This makes the Netherlands a stronger case than any of the others in the data set for illustrating the influence that far-right parties have on mainstream party support for multiculturalism.  121 Figure 3.7: The Strength of Multiculturalism Prior to the Emergence of the Far-Right Compared to the Drop in Support for Multiculturalism Caused by the Far-Right  Single Case Analysis- Methods Used Different methods are used in the Canadian and Dutch single case analysis because the cases are designed to illustrate two different relationships. In both cases, there is an emphasis on the recoded platforms discussed in the measurement of party positions section earlier in this chapter. The recoding looks at only statements that speak to multiculturalism targeted towards immigrants and similar ethnic minorities, excluding the statements addressing national minorities' and aboriginals' interests that are often included in the Manifesto Project coding. The recoded platforms, thus, provide a more careful and accurate depiction of party support for multiculturalism than the Manifesto Project does. This ensures that the single case analysis does not run into the same  122 problems regarding the conflation of multiculturalism and multinationalism that the cross-case analysis does. Most of the analysis in the two cases looks at the evolution of party positions over time. Additionally, the analysis of the Canadian includes some archival analysis of ethnic minority advocacy in favour of multiculturalism as well as some quantitative analysis of the size of immigrant populations in ridings and the impact that that had on the success of parties in different elections. Single Case Analysis Methods- Canada Analysis on Canada includes two sets of analysis. The first looks at ethnic minority influence on the development of Canadian multiculturalism. This section uses a mix of archival research and secondary literature to demonstrate that ethnic minorities advocated in favour of multiculturalism and played a central role in policy advocacy. Submissions to the Royal Commission on Bilingualism and Biculturalism, the Royal Commission's report, and secondary literature on the development of Canadian multiculturalism are used for the bulk of the analysis in this section of chapter 7. This section of the chapter builds on the positive relationship between ethnic minority electoral strength and partisan support for multiculturalism in SMD systems shown in that cross-country analysis.  It shows that ethnic minorities are indeed pushing policy actors to support multiculturalism. The second section on Canada uses quantitative analysis to demonstrate the impact that ethnic minorities have on election results. In this section, the Canadian census (Canadian Census Analyzer, 2014) is used to obtain the number of proportion of the immigrant population in every Canadian riding from 1993 to 2015. Regression analysis is conducted using the percentage of immigrants in a riding as an independent variable along with median income (taken from the same source as immigrant population) and  123 region to determine the impact that the immigrant percentage of riding had on each major party's vote share. A separate regression is run with each major party as the dependent variable. Separate analyses are conducted for Quebec and the rest of Canada because of the extent to which campaigns in Quebec differ from the rest of Canada. The size of the immigrant population in a riding serves as proxy for the size of riding's ethnic minority population so that the measure is consistent with the measure used in the cross-case analysis.  Once the regression analysis is complete, an additional calculation is done to determine how many seats a party would win if the size of a district's immigrant population had no effect on the vote share of a party in a district. This has the effect of simulating an election in which immigrant voters voted the exact same way that non-immigrant voters did. These simulations show how disproportionate immigrant support for parties such as the Liberals influences election results. A comparison of this to the overall size of the ethnic minority population can distinguish between the power of minorities in SMD and proportional systems. In a proportional system, immigrants can influence no more seats than their proportion of the population, and likely fewer because immigrants do not vote as a single block. When the number of seats influenced by the Liberals' disproportionate support in district approaches or exceeds the percentage of the immigrant population in a country, immigrants' electoral influence in a single member district exceeds what their influence would be in a proportional system. This section of the chapter can help to demonstrate why parties are more responsive to ethnic minorities in single member district systems than in proportional systems.  124 Single Case Analysis Methods- The Netherlands The single case analysis in the Dutch case analysis is done using descriptive statistics. The first section uses the recoded manifestos, as well as Manifesto Project data, to compare centre-left support for multiculturalism to centre-right support. This analysis provides an accurate contrast of mainstream left and mainstream right support for multiculturalism. This is useful in demonstrating that mainstream right parties often do support multiculturalism. The cross-case analysis that demonstrates that mainstream right parties have a greater influence over policy adoption, and that mainstream right are more responsive to ethnic minority electoral strength and the emergence of far-right parties, reflects genuine support for multiculturalism on the right of the political spectrum. It also helps to show that mainstream right parties are not inherently anti-multicultural. The opposition to multiculturalism that mainstream right parties take on after the emergence of far-right parties is a response to the far-right, not something inherent to mainstream right ideologies. The second part of the Dutch case analysis looks at how mainstream party positions changed after the emergence of the far-right. It uses the re-coded party manifestos to show the changes in mainstream party support. It also uses the far-right manifestos that were coded for the nature of their anti-immigrant appeals to show that the far-right parties made appeals targeted at voters on the mainstream left and mainstream right. This section of the dissertation illustrates the way that mainstream parties' positions change in response to the rise of far-right parties. It shows how far-right parties can appeal to parties of both the left and right, showing why there is, at least in some  125 countries, a shift towards anti-multicultural positions by both mainstream left and mainstream right parties.    126 Chapter 4 The Development of Multiculturalism Policies This chapter examines the development patterns of multiculturalism policies. Understanding the development pattern of policy adoption is an important first step in determining the influence that parties have on policy. If multiculturalism policies are limited to only a few countries, it is important to focus on what factors set those countries apart from others. If, in contrast, multiculturalism policies are widely distributed across countries, it is important to look for trends that hold over a large number of countries. Knowing the extent to which multiculturalism follows a path dependent pattern is also important to an examination of parties’ influence over policy. If multiculturalism follows a path dependent dynamic, parties will have more influence over policy development before adoption than after. In this case, observations that occur before adoption are particularly important. Furthermore, there may be particular policies whose adoption is important. If a policy, such as affirmation, increases the likelihood of additional policy adoption, it is particularly important to pay attention to variables that increase the likelihood of the adoption of that policy. Finally, this chapter sheds some empirical light on what ought to count as multicultural policy, identifying whether controversial policies fit within the index.  This chapter highlights four important findings about the development of multiculturalism policy. It first finds that in addition to a handful of countries with very strong or very weak policies, a substantial number of countries have moderate policies. This shows that multiculturalism policies can exist even in countries that have not committed to adopting a fully multicultural model of integration.  The second main finding is that retrenchment is rare. Because of this, parties’ influence over policy is more  127 important in years before policy adoption than after. Third, affirmation policies increase the likelihood of additional policy adoption. This means that it is particularly important to pay attention to affirmation policies when examining party influence over policy. It also suggests that path dependency affects multiculturalism. Not only is retrenchment difficult, but also the adoption of affirmation policies appears to create a positive feedback loop that increases the likelihood of additional policy adoption. Finally, mother-tongue education is the only policy that does not fit with the rest of the index, though the significant number of dual citizenship policies adopted before 1960 make drawing firm conclusions about the fit of dual citizenship difficult. The Distribution of Multiculturalism Policies Total Multiculturalism Policy Adoption There is a range of different levels of policy adoption across different countries. Figure 4.1 shows the level of policy adoption by country in 2011. Five countries, Australia, Canada, New Zealand, Finland, and Sweden have six or more policies out of a possible eight. Seven countries including Belgium, the United Kingdom, Ireland, Norway, Portugal, Spain, and the United States have 3.5 policies or more. Thus, twelve countries in the Banting and Kymlicka index had put in place either moderate or strong multiculturalism policies by 2011. Seven countries; Greece, France, Germany, the Netherlands, Switzerland, Austria, and Italy; have fewer than three policies but still have adopted some. Only two countries, Denmark and Japan, had no policies at all in 2011.  128 Figure 4.1: Multiculturalism Policies Adopted by 2011   Multiculturalism policies are widely distributed across Europe and North America. Policies exist in countries with a diverse set of national identities and histories of immigration. The countries with the strongest policies tend to be Anglo-settler countries, but strong policies are not exclusive to them. Australia, Canada, and New Zealand all have strong policies. Two Nordic countries, Sweden and Finland, also have strong policies. There is also a range of different countries that have moderate policies. Four of the seven countries with moderate policies, Belgium, the United Kingdom, Portugal, and Spain, had colonies and substantial immigration from those colonies at different points in their history. One country that never had colonies and only recently began to receive large numbers of immigrants, Ireland, has a moderate policy. One Anglo-settler country, the United States, and one Nordic country, Norway, have moderate policies as well. The range of total policy adoption in 2011 makes it difficulty to neatly  129 divide countries into those that are multicultural and those that are not. It also shows that countries with a variety of national identities and immigration histories have adopted policies. Figure 4.2 shows the level of total policy adoption across countries in 1990. It demonstrates similar patterns to the ones displayed in figure 4.1. In 1990, only two countries had adopted strong policies. Several, however, had moderate policies in place, including New Zealand, the United Kingdom, the Netherlands, Sweden, and the United States. Like in 2011, there were also a number of countries in 1990 that had adopted a small number of policies including France, Belgium, Ireland, Denmark, Germany, and Greece. Even before multiculturalism gained the widespread adoption present in 2011, most countries in the data set had adopted at least a few policies.  Figure 4.2: Multicultural Policies Adopted by 1990   130 Like in 2011, the countries adopting multiculturalism policies have a somewhat diverse set of national identities and immigration histories.  The group of countries with strong multiculturalism policies includes a substantial number of Anglo-settler countries as well as two countries that once had colonies, the Netherlands and the United Kingdom, and one Nordic country, Sweden. Unsurprisingly, the countries that are relatively recent immigration countries such as Greece, Ireland, Portugal, and Spain had few policies. There were, however, the beginnings of such policy development in Greece and Ireland. Ireland by 1990 had adopted one policy while Greece had a partial policy in place.  Figure 4.3 shows the level of policy adoption in 1970. At that point, multiculturalism was rare in all countries. The United States had the highest number of multiculturalism policies with three. A wide range of countries had one policy in place by 1970. These include Anglo-settler countries such as Australia and New Zealand, countries that had colonies such as Belgium, France, and the United Kingdom, and a country that never had colonies, Ireland.   131 Figure 4.3: Multicultural Policies Adopted by 1970  Little connects countries that had at least one multiculturalism policy in 1970 with the countries that developed strong policies by 1990 and 2011. The country with the strongest policy in 1970, the United States, did not adopt a single policy between 1970 and 2011. Two of the countries, Australia and New Zealand, with one policy in place in 1970 ended up with strong policies by 2011 and one country with a partial policy in 1970, Canada, ended up with a strong policy. Finally, two countries with no policies in place at all in 1970, Finland and Sweden, had strong policies in place by 2011. The most multicultural countries in 1970 were not necessarily the most multicultural countries in 2011. The adoption of multiculturalism is not simply a case of pluralistic countries adopting more and more policies. If this were happening, one would expect to see the countries that had the strongest policies in 1970 end up with the strongest polices in 2011.   132 National identity and immigration histories cannot explain the wide distribution of policy adoption. One cannot separate out a set of countries with a particular immigration history or national identity that are unique in adopting multiculturalism policies. This demonstrates some of the limitations of Brubaker’s (1992) and Favell’s (1998) approach to explaining immigration and integration policy in Britain, France and Germany. They fail to account for the extent to which elements of particular types of integration policies show up in countries where they may not be expected. In the case of multiculturalism, this comes through in the extent that non-settler countries that did not experience high levels of immigration before 1970, still adopt policies. It also comes through in the extent to which the countries that have strong multiculturalism policies in 1970 are different from the ones with strong policies in 2011. The findings finally show that multiculturalism is much more widespread than is sometimes acknowledged, particularly with respect to its proliferation through the late 1990s and 2000s. Contrary to the return to assimilation policy that is noted by Brubaker (2001) and Joppke (1999 and 2010), multiculturalism is still alive in both Europe and North America. Recognition, Participation, and Support Policies Examining the adoption of different types of multiculturalism can shed further light on patterns of policy development. Looking in particular, at policies that provide multicultural recognition can address some of the concerns raised by scholars such as Duyvendak et al. (2013) and Vink (2007) about the potential for indexes to mischaracterize countries as multicultural. Recognition policies are the policies in the Banting and Kymlicka index that are most explicitly multicultural and least likely to be  133 adopted by governments and countries that do not see themselves as at least somewhat multicultural or have no desire to support cultural diversity. These policies include affirmation, multicultural education, and media representation. As noted in table 3.1, these include affirmation of a country as multicultural, the inclusion of multiculturalism in school curriculum, and laws requiring sensitivity to and representation of ethnic minorities in the media. If these policies show the same development patterns as the total policy index, it is likely that the dynamics demonstrated in the broader index are robust to narrower conceptualizations of multiculturalism. An examination of participation and support policies can also be valuable as way to determine whether these types of policies follow similar dynamics to those found in the examination of total policy adoption. Figures 4.4 and 4.5 show that at least some recognition policies are present in most countries. Figure 4.4 shows that four countries had adopted the full range of recognition policies by 2011. These countries include two Anglo-settler countries, Australia and Canada, and two Nordic countries, Finland and Sweden. A wide range of countries had at least one recognition policy. These countries include New Zealand, Belgium, Portugal, Spain, and the United Kingdom. Figure 4.5 shows recognition policy adoption was more limited in 1990. Only one country, Australia, had adopted the full range of recognition policies by that year. The countries with more than one recognition policy in 1990 include a mix of Anglo-settler countries, Nordic countries, and countries that used to have colonies. These are New Zealand, Canada, the Netherlands, Sweden, and the United Kingdom. The adoption of recognition policies is reasonably common by 2011, but rare prior to 1990. More than half of the countries in the Index lacked any recognition policies in 1990.  134 Figure 4.4: Recognition Policy Adoption by 2011  Figure 4.5: Recognition Policy Adoption by 1990   135 The widespread adoption of recognition policies has substantive importance. This shows that that wide proliferation of multiculturalism found in the Banting and Kymlicka index reflects at least a partial commitment by many countries to multiculturalism. It is rare that a country does absolutely nothing to recognize cultural diversity. Only five countries in the index have no recognition policies. Furthermore, the extent to which these policies show up in countries other than Australia, Canada, or Sweden shows that policies have expanded beyond countries traditionally thought of as multicultural. While most countries do not have as strong recognition policies as they could, there are elements of multiculturalism in most European and North American countries. This finding emphasizes the importance of considering the adoption of multiculturalism in countries with diverse national identities and immigration histories.  Policies that remove barriers to ethnic minority participation in society, exemptions from dress codes and the acceptance of dual citizenship, show a similar range of adoption. Figure 4.6 shows that most countries in the index have at least one participation policy, either allowing citizens to hold dual citizenship or making exemptions to dress codes for religious or cultural reasons. Five countries, Australia, Canada, New Zealand, Sweden, and the United Kingdom have fully implemented both policies. Like, recognition policies, the period between 1990 and 2011 saw a large increase in the adoption of participation policies. Figure 4.7 shows that participation policies were much less common in 1990. Only eight countries had at least one participation policy in 1990 and only three had fully adopted both.  136 Figure 4.6: Participation Policy Adoption by 2011  Figure 4.7: Participation Policy Adoption by 1990   137 Finally, the same pattern of development exists for support policies. These policies include funding for ethnic minority organizations, support for mother-tongue education, and affirmative action. Figures 4.8 and 4.9 show the number of support policies adopted in 2011 and 1990 respectively. While only one country, Australia, had adopted the maximum number of support policies by 2011, 10 of the 21 countries had adopted half of the support policies in the index. Diverse countries, from Belgium to Canada to Germany, have support policies. Only Denmark, Italy, and Japan have none. As with recognition and participation policies, a great deal of policy adoption occurs in the years between 1990 and 2011. Compared to 2011, in 1990 only six countries had adopted at least half of the support policies in the index, and seven countries had no policies at all. Figure 4.8: Support Policy Adoption by 2011    138 Figure 4.9: Support Policy Adoption by 1990  The pattern of policy development is consistent across each category of policy in the index. For each category, there is substantial adoption across a wide range of countries in 2011. For each category, there is also substantially weaker policy adoption in 1990, suggesting that a lot of the growth in multiculturalism has taken place between 1990 and 2011. The similar patterns in policy development further suggest that there is no time period in which certain types of multiculturalism policies were more likely to adopted than others, nor is there a set of countries that are more likely to adopt certain types of policies.  An examination of the link between the strength of a country's multiculturalism policy and the number of additional policies adopted confirms the wide distribution of policy adoption. Figure 4.10 shows the change in policy strength in countries that had at least three policies in place by 1990 and in countries that did not have at least three  139 policies in place by 1990.35 This was not simply a ceiling effect. It was possible for the countries that had at least 3 policies in 1990 to adopt 21.5 policies. The total number of policies in these countries grew by 5 or 21.5% of possible growth.36 In contrast, it was possible for countries with fewer than 3 policies in 1990 to adopt 104 policies. Policy adoption in these countries grew by 29 or 28% of possible growth.37 After 1990 most of the growth in the strength of multiculturalism occurred in countries that did not have more than three policies in place before 1990. In only 1991, 2000, and 2010 were more policies adopted in countries that had more than 3 policies than those that had less than 3 policies.                                                 35 1990 is chosen as the start year for this analysis because by 1990 a substantial number of countries had at least three multiculturalism policies in place. 36 These countries adopted a total of 6.5 policies, but the Netherlands removed 1.5 policies, for a growth in policy adoption by 5. 37 29.5 policies were adopted, but Denmark removed 0.5 of a policy, for a total growth of 29.  140 Figure 4.10: Average Policy Adoption by MCP Strength in 1990  Lines show the number of policies adopted (or retrenched) in each year between 1990 and 2016 by countries with more than 3 policies and with fewer than three policies.  Figure 3.4 highlights the extent to which policy development is similar across the different types of policies. It shows a consistent increase in the adoption of recognition, participation, and support policies over time. Outside of participation policies, which start at a higher level because of the adoption of a substantial number of dual citizenship policies before 1960, the trends in policy development are very similar. Not only is there little evidence in retrenchment, but also there is little indication that recognition policies, the ones most difficult to misconstrue as multicultural, follow a different trend than the other types of policies in the index. This is further evidence of multiculturalism’s widespread uptake. If most of the growth in multiculturalism between 1990 and 2011 were a result of multicultural  141 countries becoming more multicultural, one would expect most of the policy adoption to occur in countries with three policies or more. Instead, the countries with fewer than three policies see the highest levels of adoption after 1990. More and more countries are including some elements of multiculturalism in their integration policies. Using a narrower measure of multiculturalism does not affect Banting and Kymlicka’s (2012) findings. Multiculturalism and Path Dependence Path Dependence and Multicultural Retrenchment Path dependence in a policy area can manifest itself in two ways, in the form of lack of policy retrenchment and in the form of increased policy development. The extent to which there has been retrenchment in multiculturalism is controversial. Both Brubaker (2001) and Joppke (1999) make claims that liberal democracies are turning away from multiculturalism and towards more assimilationist approaches to integration. Public statements by political leaders such as German Chancellor Angela Merkel and British Prime Minister David Cameron are often pointed to as evidence that political leaders across Europe are trying to distance themselves from multiculturalism (BBC, 2010; Cameron, 2011). At the same time, Banting and Kymlicka (2012) show little evidence of a retreat from multiculturalism.  The expanded version of the Banting and Kymlicka index used in this dissertation supports Banting and Kymlicka's (2012) findings in two ways. First, it shows the trend in the adoption of multiculturalism across time, confirming the lack of retrenchment that they find in their three-year (1980, 2000, and 2010) analysis. Second, it allows for a focus on recognition policies, demonstrating that the policies most closely linked to  142 multiculturalism show a similar lack of retrenchment to the overall index. Both of these findings come through in figure 4.11 and 4.12. Figure 4.11 shows a steady increase in the adoption of multiculturalism in the total policy index. Between 2000 and 2010, the period in which there has been the strongest backlash against multiculturalism, average policy adoption has gone from 3 to 3.5 policies. At no other point in time is there a significant drop in average policy strength across the countries in the index. The same is true for recognition policies. Figure 4.12 shows a similarly steady increase in the overall adoption of recognition policies. Increases in policy adoption outweigh the instances of retrenchment in the mid-1990s and early 2000s. Average recognition policy adoption goes from just under one policy in the early 2000s to close to 1.25 policies by the end of the decade. Figure 4.11: The Adoption of Multiculturalism over Time   143 Figure 4.12: The Adoption of Recognition Policies over Time  Neither the examination of total policy adoption, nor the look at recognition policies provides evidence that the backlash noted by Brubaker (2001) and Joppke (1999) has influenced policy. That this is true for both total policy adoption and recognition policies is important. It suggests that Banting and Kymlicka's (2012) findings are not a result of the inclusion of policies that may not be multicultural in their index, but rather, a genuine picture of the difficulty in retrenching multiculturalism.  Only 10 of the 62 full and 38 partial policies were repealed. Of the repealed 10 policies, one was re-adopted just two years after repeal and another was re-adopted 11 years after repeal. Table 4.1 shows that half of all retrenchment that occurs takes place in the Netherlands. Between 1994 and 2008, the Netherlands retrenched half of its multiculturalism policies. Four of the other countries that retrenched policies (Austria, Denmark, Germany, and Italy- each retracted one policy) are countries that always had  144 particularly weak policies. Given the general lack of adoption of multiculturalism in each of those countries, it is not surprising to see retrenchment happen. In two of those cases, Germany and Italy, the policies retrenched were re-adopted after retrenchment. Policy retrenchment appears to be a phenomenon that is limited to the Netherlands; only two other countries have repealed policies and not subsequently re-adopted them.  Table 4.1: Instances of Policy Retrenchment Country Year Policy Scale of Retrenchment Policy Re-Adopted? Austria 2002 Recognition in the school curriculum Partial to no policy No Denmark 2002 Mother-tongue education Partial to no policy No Germany 1972 Mother-tongue education Partial to no policy Partial policy re-adopted in 1983 Italy 2004 Recognition in the school curriculum Partial to no policy Policy re-adopted in 2006 Netherlands 2008 Media sensitivity Full to partial policy No Netherlands 2004 Affirmative action Full to no policy No Netherlands 1997 Dual citizenship Full to partial policy No Netherlands 1995 Recognition in legislation Full to no policy No Netherlands 1994 Mother-tongue education Full to no policy No United States 1984 Dress code exemptions Partial to no policy No  The low levels of retrenchment have important implications for the analysis conducted throughout the rest of this dissertation. It shows that the factors that influence policy are more likely to influence policy before adoption than after adoption. Once a policy is in place, the low levels of retrenchment make it difficult for opponents of multiculturalism to further impact the policy. This confirms that hazard models, which consider the impact of previous policies party positions on the likelihood of policy  145 adoption, are appropriate for much of the analysis conducted in this chapter and in chapter 5.  The low levels of policy retrenchment also contradict a significant body of work that sees multiculturalism policies declining in favour of more assimilatory immigrant integration policies (Brubaker, 2001; Goodman, 2014; Joppke, 2010; Joppke, 1999). It confirms that Banting and Kymlicka (2012) findings that suggest that multicultural retrenchment has been rare, even in spite of the backlash against multiculturalism that has occurred through the late 1990s and 2000s. The rise of significant opposition to multiculturalism, exemplified by the emergence of far-right parties across most of Europe, has had little impact on policy adoption. Path Dependence and Policy Adoption In addition to limiting policy retrenchment, path dependence can also influence policy adoption. Feedback loops that make it difficult for politicians to repeal policies may also give them incentives to adopt additional policies. The affirmation of a country as multicultural can create new avenues through which advocates can push for additional policy adoption. Increasing funding to ethnic minority organizations can increase their ability to lobby in favour of the adoption of additional policies. It is, thus, important not only to look at path dependence through the lens of retrenchment, but also to the test the impact of some policies on the adoption of others. Hazard models that look at the effect the adoption of individual policies has on the overall adoption of multiculturalism show that few policies have a strong impact on overall policy adoption. Table 4.2 shows that most policies in the index do not increase the likelihood of additional adoption. This is the case regardless of whether measures of  146 total policy include dual citizenship, mother-tongue education, or affirmative action. Most of the hazard ratios in the table are close to or below one, meaning that the presence of most policies leaves countries as likely or less likely to adopt additional policies than countries without those policies. Table 4.2: Effects of Individual Policies on Overall Policy Adoption (1970 Start Date) Variable Total MCP Total MCP Minus Dual Citizenship Total MCP Minus Affirmative Action Total MCP Minus Bilingual Affirmation 1.380 (0.777-2.452) 1.325 (0.721-2.435) 1.180 (0.649-2.147) 1.650 (0.882-3.085) MC School 0.505* (0.237-1.072) 0.461* (0.207-1.026) 0.509* (0.232-1.119) 0.589 (0.267-1.299) Media 0.669 (0.348-1.286) 0.636 (0.323-1.253) 0.643 (0.316-1.311) 0.740 (0.371-1.474) Dual Citizenship 0.831 (0.464-1.489) NA  0.803 (0.440-1.477) 0.861 (0.452-1.642) Exemptions 0.383* (0.135-1.083) 0.426 (0.147-1.231) 0.266** (0.076-0.938) 0.510 (0.162-1.603) Funding 0.759 (0.401-1.439) 0.743 (0.381-1.450) 0.858 (0.427-1.723) 0.865 (0.436-1.712) Bilingual Education 1.129 (0.567-2.250) 1.060 (0.521-1.159) 1.301 (0.615-2.753) NA  Affirmative Action 0.376** (0.163-0.871) 0.454* (0.196-1.053) NA 0.529 (0.220-1.276) * P-value<0.1, ** p-value<0.05, ***p-value<0.01 Values are hazard ratios with the range for the 95% confidence level in brackets Values for control variables are included in the appendix E  The exception to this is affirmation policies. While the hazard ratios for affirmation are not statistically significant, they do point towards a positive affect on policy adoption. Depending on whether dual citizenship or affirmative action policies are  147 included in the total policy measure, the presence of an affirmation policy increases the likelihood of policy adoption by 18% to 38%. When the total policy measure excludes mother-tongue education, the effect that affirmation policies have on additional policy adoption jumps to a 65% increased likelihood. Appendix B shows that this effect diminishes slightly when the analysis uses later start dates. By 1980, the measure that excludes affirmative action policies drops to an increase of only 7%. The other measures of total multicultural policy range between a 29% effect for affirmation on total policy adoption to a 60% effect. Appendix D shows that the effect of affirmation policy adoption increases when the model excludes controls. The estimates in table 4.2 likely slightly understate the overall effect that affirmation policies have on policy adoption because of the covariation that exists between the affirmation policy variable and the ethnic minority electoral strength control.  Descriptive statistics provide further evidence of a link between the presence of affirmation policies in a country and the adoption of additional policies. Countries that have adopted affirmation policies adopt more policies after adopting the affirmation policies than before. They also adopt more policies than countries that never affirm themselves as multicultural. Figure 4.13 shows that the average country adopts 3.11 additional policies after affirming itself as multicultural, but only 1.33 policies before affirming itself as multicultural. In contrast to this, countries that never affirm themselves as multicultural adopt an average of only 1.88 policies. This suggests that the adoption of affirmation policies precede the development of more extensive multiculturalism programs.   148 Figure 4.13: Average Policies Adopted Before, After, and Without Affirmation Policies  Averages are for the total number of years with or without an affirmation policy not for individual years. The average country with an affirmation policy adopts 3.11 policies over the total number of years during which It has had the affirmation policy.  Tests that look at the influence that affirmation policies have on the adoption of individual policies provide further evidence that affirmation policies lead to the expansion of multiculturalism. Table 4.3 shows a positive relationship between the presence of affirmation policies and the adoption of most of the other policies in the index. The table shows the results for multiple different regressions.  The first column lists the policies used as independent variables, while controls are included in Appendix G. When controls are included in the models, affirmation policies increase the likelihood of the adoption of multiculturalism in the school curriculum and media sensitivity by more than 5 times.  They increase the likelihood of the adoption of funding for ethnic  149 minority organizations by almost 5 times, and the likelihood of the adoption of affirmative action policies by almost 6 times.38 All of these effects are statistically significant at the 95% confidence level except for the effect on affirmative action, which is statistically significant at the 90% confidence level. In addition, the presence of affirmation policies increases the likelihood of the adoption of exemption policies by 1.4 times, but this effect is not statistically significant. These effects get weaker over time. When analysis starts in 1980, the effect of affirmation loses statistical significance for influence over the adoption of media sensitivity and funding for ethnic minorities. Models that exclude controls, presented in Appendix F, show a weakening of the effect of affirmation policy on many different policies. Overall, however, these findings suggest that the presence of a policy affirming multiculturalism does increase the likelihood of adopting at least some of the other policies in the index.                                                 38 There is also a very large increase in the likelihood of the adoption of dual citizenship policies, but the small number of cases where dual citizenship policies are adopted after a country adopts an affirmation policy makes it difficult to know how strong this effect is.  150 Table 4.3: The Effect of Individual Policies on Individual Policy Adoption (1970 Start Date) Variable Affirmation MC School Media Dual Citizenship Exemptions Funding Bilingual Education Affirmative Action Affirmation NA 5.084** (1.366-18.930) 5.850** (1.157-29.595) 23.812*** (2.273-249.464) 1.407 (0.289-6.857) 4.772** (1.007-22.614) 1.103 (0.217-5.622) 5.823* (0.975-34.760) MC School 11.803 (0.605-230.321) NA 11.236** (1.334-94.609) 31.001** (1.984-484.506) 2.066 (0.395-10.818) 0.963 (0.082-11.349) 9.695 (0.390-241.169) 0.942 (0.112-7.915) Media 2.020 (0.297-13.757) 2.787 (0.650-11.947) NA 5.339 (0.425-67.022) 1.362 (0.318-5.838) 1.088 (0.169-7.010) 1.092 (0.134-8.900) 1.265 (0.255-6.274) Dual Citizenship 1.034 (0.172-6.203) 0.757 (0.242-2.367) 2.236 (0.471-10.616) NA 1.396 (0.223-8.744) 0.823 (0.207-3.263) 0.357 (0.070-1.826) 1.850 (0.315-10.848) Exemptions 0.079 (0.000-16.485) 0.560 (0.064-4.875) 2.127 (0.173-26.125) NA NA 2.480 (0.156-39.410) 1.127 (0.056-22.655) 1.483 (0.068-32.566) Funding  4.939 (0.601-40.564) 1.912 (0.557-6.561) 4.550* (0.948-21.848) 8.386* (0.880-79.951) 1.588 (0.334-7.555) NA 0.803 (0.116-5.583) 2.005 (0.308-13.041) Bilingual Education 8.083** (1.033-63.250) 4.696** (1.050-20.999) 3.785 (0.724-19.788) 12.503* (0.994-157.239) 1.995 (0.320-12.423) 1.416 (0.303-6.611) NA 1.542 (0.169-14.074) Affirmative Action 0.181 (0.005-7.217) 0.268 (0.026-2.746) 1.853 (0.230-14.935) NA  3.075 (0.547-17.278) 0.590 (0.075-4.661) 0.138 (0.008-2.506) NA * P-value<0.1, ** p-value<0.05, *** p-value<0.01 Values are hazard ratios with the range for the 95% confidence level in brackets Values for control variables are included in the Appendix G. NA is used in cases where there are insufficient observations to draw a estimate and effect.  151  Affirmation policies, however, have little effect on is mother-tongue education. Countries with affirmation policies appear no more likely to adopt mother-tongue education policies than countries without affirmation policies. This is true for both models that include and exclude controls (as is also shown in Appendix F). Mother-tongue education policies appear to follow a somewhat different development pattern than the other policies in the index. This fits with the arguments made by Duyvendak et al. (2013) that mother-tongue education is not a multiculturalism policy. Two reasons may lead countries that are not multicultural to adopt mother-tongue education policies. First, as Duyvendak et al. (2013) argue, these policies can be a pedagogical tool to facilitate majority language acquisition. Second, countries with guest worker programs may have introduced these policies to ease immigrants' return to their home countries. In neither case is the adoption of mother-tongue education policies consistent with theoretical underpinnings of multiculturalism. This can explain why the affirmation of a country as multicultural has little affect on its adoption of mother-tongue education policies. None of the other policies, including funding for ethnic minorities, in the index has the kind of effect on additional policy adoption as affirmation policies. The lack of an effect for funding suggests that the main mechanism through which path dependence works occurs through affirmation policies. Symbolic recognition, and not material support, creates space for advocates to push for the adoption of additional policy adoption. Once that has occurred, funding for ethnic minority organizations does not appear to have an additional impact on the development of other policies.   152 The lack of a strong and consistent effect for other policies in the index is unsurprising given the wide range of strengths of policy adopted across most countries. One would expect that in a diverse group of countries with moderate policies, multiculturalism would manifest itself in many distinct ways. It is likely that the particular policies in the index that any given country adopts are highly dependent on the politics surrounding multiculturalism and immigration in that country. Because of this, one would expect to see a substantial amount of inconsistency in the links between the adoptions of different policies. Descriptive statistics comparing policy adoption when affirmation policies are present to when they are absent confirm the hazard models’ findings. Figure 4.14 shows the rate of policy adoption for each policy in countries with and without affirmation policies as a percentage of the number of observations in which affirmation policy is present and absent. For every policy, except mother-tongue education, there is a substantially higher rate of adoption when affirmation policies are present. Countries with affirmation policy range from a 3.94% likelihood of adopting an exemption policy in any given year to a 15.91% likelihood of adopting a funding policy. This compares to a range of 0.6% (for affirmative action policies) to 1.17% (for multicultural school curricula) in countries that do not have affirmation policies.39 Mother-tongue education adoption occurs at an only slightly higher rate in countries with affirmation policies. Countries with affirmation policies have a 2.86% likelihood of adopting mother-tongue education policies in any given year, while countries without affirmation policies have a 1.15% per                                                39 Included in these numbers are both policies adopted in countries that never adopted affirmation policies and policies adopted before countries adopted affirmation policies.  153 year likelihood of adoption. This provides further evidence that presence of an affirmation policy increases the likelihood of additional policy adoption. Figure 4.14: Percentage of Years in which Policy Adoption Occurs by Presence of Affirmation Policy  The bars show the percentage of policies adopted for years in which it was possible to adopt a policy in category of note.  For example, funding policies were adopted in 15.91% of years in which a country both had an affirmation policy and had not yet fully adopted a funding policy.  All of this data points to the presence of path dependence in the development of multiculturalism policies. Not only do multicultural policies appear difficult to retrench, but the adoption of affirmation policies increases the likelihood of policy expansion. This means that explanations of the development of multiculturalism should pay particular attention to the determinants of affirmation policies. If supporters of multiculturalism can get a country to recognize itself as multicultural in either legislation or in the constitution,  154 they stand a much better chance of getting the country to adopt additional policies. Conversely, if opponents of multiculturalism can keep their countries from adopting affirmation policies, they will decrease the likelihood of policy expansion. The competition between advocates and opponents over whether a country ought to call itself multicultural is, thus, not merely symbolic. It has important implications for future policy development. Any explanation of why advocates or opponents win debates over affirmation is also particularly important to explaining the broader development of multiculturalism. It is further remarkable that it is symbolic recognition, and not the provision of material support, leads to the expansion of multiculturalism. This is a surprising finding given the extent to which material benefits are often emphasized in explanations of path dependence in other policy fields (Hacker, 2002; Pierson, 1994; Pierson, 2004; Weaver, 2003). This, combined with work by Brubaker (1992) and Favell (1998) that shows ideas of belonging and citizenship have an important influence over immigration policy, highlights the relative importance of symbolic understandings of who belongs in a country to broader immigration policy. Groups' access to government support does not appear to determine their ability to influence policy. Rather, the ability of ethnic minorities to use multicultural recognition to ground arguments for policy expansion is the key mechanisms through which positive feedback loops to influence policy adoption. These findings highlight the importance of two factors pointed to in the theory chapter regarding the way that symbolic recognition can influence further policy adoption. The first is the way that symbolic recognition changes ideas about who belongs as a citizen in a country and the obligations that immigrants have when integrating.  155 Symbolic recognition sends a signal to both policy makers and to the public as a whole that citizenship should be open to people from a range of different cultural backgrounds, and that individuals ought not to have to give up their cultural identity when integrating. This provides the philosophical basis for arguments in favour of more substantive multiculturalism policies. Policy-makers and a public that see a country a symbolically multicultural should be more likely to be sympathetic to arguments in favour of policies that allow minorities to integrate while maintaining their culture, such as exemptions from dress codes or funding for ethnic minority organizations.  In addition to this, symbolic recognition should open up new avenues through which advocates for multiculturalism can influence policy. The creation of new state agencies tasked with implementing symbolic recognition policies provides advocates with potential allies within government when lobbying for policy expansion. The agencies have to build networks with ethnic minority organizations in order fulfil the mandate they have to implement symbolic recognition policies. They also have incentives to support the expansion of multiculturalism because doing so increases the prestige and the resources of their agency. When combined, these two factors should make the state agencies created by symbolic recognition powerful allies for advocates of multiculturalism. The presence of such allies within the government bureaucracy should make policy adoption more likely and help to explain the positive correlation between the presence of affirmation policies and the adoption of additional multiculturalism policies. What Counts as Multiculturalism? The tests in the previous section of this chapter can also provide some empirical evidence to support the inclusion of many of the policies in the Banting and Kymlicka index. The  156 inclusion of dual citizenship, bilingual education, and affirmative action policies in the index is controversial. Scholars tend to draw a distinction between citizenship policy and immigrant integration policies such as multiculturalism (Koopmans et al., 2005; Wright and Bloemraad, 2012). An open citizenship policy that includes dual citizenship can be consistent with non-multicultural approaches to integration. A country might allow dual citizenship in order to encourage citizenship acquisition, but still expect immigrants to assimilate into the majority culture. The French republican approach to citizenship exemplifies this (Brubaker, 1992; Favell, 1998). A country may also take a neo-liberal approach to citizenship and immigrant integration, allowing individuals to hold dual citizenship, but do little to support cultural diversity as part of a belief that government should not support either minority or majority culture. At the same time, dual citizenship might be included as part of a multicultural immigrant integration framework. Dual citizenship can support ethnic minority culture by allowing minorities to maintain ties to their country of origin while integrating into their host country. The inclusion of mother-tongue education policies in the index is also controversial. Mother-tongue education can be a multicultural integration policy if it is a way to help immigrants and other ethnic minorities pass along their language to their children. It can also be the opposite of an integration policy if it aims to ensure that guest workers maintain their language so they can return to their countries of origin. German mother-tongue education policies in the 1960s were designed in part with this goal (Vermeulen, 1997). Mother-tongue education can also be a way to help immigrants and ethnic minorities learn the language of the majority community (Duyvendak, 2013). Like  157 dual citizenship policies, mother-tongue education policies can have multicultural and non-multicultural justifications.  Finally, the inclusion of affirmative action policies in the Banting and Kymlicka index is controversial. Affirmative action policies can be part of anti-discrimination or non-multicultural economic integration programs (Duyvendak et al., 2013). They may be part of an effort to reduce discrimination against ethnic minorities or to respond to economic inequalities between minority and majority communities. In neither case would affirmative action policies necessarily be multicultural policies because they would not necessarily include measure to protect cultural diversity. At the same time, affirmative action policies may be a way to create culturally diverse workplaces. These workplaces may place less pressure on ethnic minorities to assimilate into the majority community in order integrate economically, and therefore play an important role in allowing immigrants to maintain their own culture. Ensuring that cultural diversity is not a barrier to employment can offer vital economic support to immigrant communities trying to maintain their cultural practices while integrating. Like dual citizenship and bilingual education policies, affirmative action policies can have multiple justifications. The tests from the previous sub-section of the chapter displayed in table 4.3 and figure 4.14 suggest that affirmative action policies fit within the multiculturalism index but that mother-tongue education programs do not. Like most other policies in the index, the likelihood of the adoption affirmative action policies increases with the presence of affirmation policies. Countries that consider themselves multicultural are considerably more likely to adopt affirmative action policies than countries that do not. This suggests that multiculturalism fits within the Banting and Kymlicka index. The correlation  158 between the presence of affirmation policies and affirmative action policies suggests there is something to the claim that affirmative action are often used to ensure cultural difference does not become a barrier to economic success. At the very least the presence of affirmative action policies in the index is not leading to a misrepresentation of the extent to which countries are multicultural. The opposite is the case for mother-tongue education policies. Unlike other policies, countries with affirmation policies are not more likely to adopt mother-tongue education policies. No other policy in the index increases the likelihood of mother-tongue education policy. This suggests that policy makers do not always see mother-tongue education as a multiculturalism policy. The governments that are supportive of multiculturalism are not necessarily the same governments that adopt mother-tongue education policies. It is difficult to determine the extent to which dual citizenship fits with the other policies in the index. There are strong relationships between the adoption of affirmation policies and the adoption of dual citizenship policies, suggesting a link between multiculturalism and dual citizenship. At the same time, the adoption of dual citizenship before 1960 in many countries truncates the sample for tests involving such policies. Australia, Belgium, France, Ireland, New Zealand, and the United Kingdom all have dual citizenship policies in place before 1960, and thus the adoption of dual citizenship in these countries could not be included in this chapter's analysis. It is not clear that the same positive relationship between the presence of affirmation policies and the adoption of dual citizenship policies would exist if the time period for the analysis could be extended back to include the adoption of dual citizenship policies in these countries.   159 Finally, the use of different measures for total multicultural policy adoption can shed some light on how much of an impact the inclusion or exclusion of different policies can have on analyses of policy adoption. Table 4.2 shows little difference between the effects found tests conducted on all of the policies in the original index, tests on the original index minus dual citizenship, and tests on the original index minus affirmative action. This suggests that the controversy over the inclusion of dual citizenship and affirmative action has little impact on the results of analyses of policy development. The same is not true for mother-tongue education policies. Excluding mother-tongue education policies from the analysis substantially increases both the effect of affirmation policies on total policy adoption and the effect's significance. Consistent with the analysis above suggesting that mother-tongue education policies do not fit with the rest of the index, this suggests that the inclusion of mother-tongue education policies can lead to an underestimation of the impact that different explanatory variables have on overall policy adoption. The findings here suggest that Banting and Kymlicka are mostly right in their decision over what to include in their index. Only one policy, mother-tongue education, stands out as significantly different from the rest of the index. There is empirical evidence to support the claim Duyvendak et al. (2013) make about the problems with including mother tongue education in a multicultural index. At the same time, there is little evidence to suggest that affirmative action does not fit with other policies. It is harder to make a definitive claim with respect to dual citizenship, but it does not appear that including dual citizenship in the index meaningfully changes the results of analyses.  160 Summary This chapter examines the patterns of multicultural development. There are four main findings in this chapter. The first is that the adoption of multiculturalism policies is widespread. This is true both for all of the policies included in the original Banting and Kymlicka index, and for the most explicitly multicultural recognition policies in the index. The adoption of multiculturalism is not limited to Anglo-settler states that have a history of diverse immigration stretching back over the past century. Rather, multiculturalism policies exist in a range of countries across Europe and North America. This confirms Banting and Kymlicka's (2012) findings. It also highlights the need to look beyond Brubaker's (1992) and Favell's (1998) use of national identity to explain the development of different immigration policies. National identity may be important, but it is not deterministic. Multicultural policies can develop in countries as diverse as Belgium, Canada, Sweden, and the United Kingdom.  The second and third main findings in this chapter are that multicultural policy adoption follows a path dependent pattern. This is true both in the sense that it is uncommon to see policy retrenchment and in the sense that the adoption of affirmation policies increases the likelihood of additional policy adoption. This has two important implications for the rest of the analysis in the dissertation. The first is that the observations that occur before policy adoption are more important to explaining policy development than the observations that occur after. This means hazard models, which look only at observations that occur before adoption, are necessary for analysis of different variables’ impacts on policy. The second important implication is that one ought to pay particular attention to factors that increase the likelihood of the adoption of  161 affirmation. Those variables that increase the likelihood of affirmation policy adoption are likely to have an additional, indirect, effect on the development of additional multiculturalism policies because affirmation policies increase the likelihood of the adoption of additional policies. This is an important finding as it highlights symbolic recognition of multiculturalism as something that plays a key role in policy development. It demonstrates that such recognition has meaningful consequences for the more substantive benefits that ethnic minorities can receive from the development of more comprehensive policies.  The path dependence findings also have important implications for the existing literature on immigration. Contrary to claims made by Brubaker (2001) and Joppke (1999), there is little evidence of multicultural retrenchment. This confirms similar findings by Banting and Kymlicka (2012). It also points to an interesting paradox where the adoption of stricter integration policies (Etzioni, 2007; Goodman and Wright, 2015; Joppke, 2007) has not coincided with the removal of multiculturalism policies. The proliferation of both multiculturalism and more assimilatory integration policies has the potential to lead to the development of complicated and potentially contradictory sets of immigration and integration policies across Europe and North America.  Finally, the chapter is able to provide some empirical evidence that responds to debates over what ought to be included in a multiculturalism policy index. There is evidence to support the inclusion of affirmative action as a multiculturalism policy, but mother-tongue education stands out as policy that develops apart from other multiculturalism policy indexes. This suggests that examinations of the why multicultural policies develop need to be careful about including mother-tongue education policies in  162 their analyses. It is difficult to determine whether dual citizenship fits with the rest of the index because of the large number of policies adopted before the start of analysis. This analysis is important given controversy over the inclusion of different policies in the Banting and Kymlicka index. The empirical work in this dissertations suggest that, at the very least, the inclusion of dual citizenship and affirmative action are doing little to change results of analyses of the development of multiculturalism. Critiques of the Banting and Kymlicka index for its inclusion of dual citizenship (Wright and Bloemraad, 2012) and affirmative action (Duyvendak et al., 2013) may have sound theoretical grounding, but the inclusion of those policies does not appear to create noise in the index's measure of multiculturalism. The same is not true for mother-tongue education. In line with Duyvendak et al.'s (2013) theoretical critique, mother-tongue education policies appear develop separately from other policies in the index. Overall, the index provides a good measure of multiculturalism policies across Europe and North America, but there is reason to question the inclusion of mother-tongue education.    163 Chapter 5 Political Parties and the Adoption of Multiculturalism Policies This chapter examines the relationship between political parties and policy adoption. It presents a number of important findings regarding the influence that parties have on policy adoption. It first shows that support for multiculturalism is not exclusive to left parties. There are moderate right parties that support multiculturalism and that oppose it. This highlights the importance of looking at parties across the spectrum when examining policy development. The chapter also shows that cross-party support plays a crucial role in influencing policy adoption. This highlights the importance of looking beyond government parties (which Breunig and Luedtke (2008) focus on with respect to parties and influence over immigration policy) when examining parties' influence over policy. The chapter then highlights the unique impact that that changes in right party positions have over policy adoption. Existing analyses (Howard, 2009; Ireland, 2004; Givens and Luedtke, 2005) that focuses on left parties misses the important influence mainstream right parties have over policy. It is easier for left parties to adopt multiculturalism policies when they have the support of right parties. This chapter is organized as follows. The first section compares the multiculturalism positions of left and right parties as well as party positions on multiculturalism to broader left-right ideology. The second section examines the link between parties and overall policy adoption. The fourth section looks at the impact that party positions have on specific policies in the Banting and Kymlicka index, finding little evidence that cross-party positions or left and right have parties have particularly strong  164 impact on one set of multiculturalism policies. Like the previous chapter, the analysis in this chapter includes both hazard models and descriptive statistics. Left and Right Parties and Support for Multiculturalism Comparison of parties’ left-right and multiculturalism positions shows a weak correlation. Figure 5.1 shows that parties that are more left wing on the Manifesto Project's left-right scale40 are slightly more likely to favour multiculturalism. This relationship is slight, however, and there are large numbers of outliers. The large number of outliers, particularly on the right of political spectrum, shows that left-right positioning often cannot explain a party’s multicultural positioning. Figure 5.2 shows that this holds even when parties that make no statements (positive or negative) about multiculturalism are excluded from analysis. The same is true for a measure of support41 for welfare programs (which isolates an issue related to the economy from other left right issues). Figure 5.3 again shows a slight relationship between support for welfare and support for multiculturalism. That relationship, however, is weak and has a large number of outliers.                                                40 This measure is a based on the log rile score in the Manifesto Project data. This score takes into account party’s positions on all of the policies in coded in the Manifesto Project data. 41 This measure looks at the extent to which a party favours the expansion of welfare programs such as health care child care, pensions, and social housing (Volkens et al., 2013b).  165 Figure 5.1: Left-Right Positions Compared to Support for Multiculturalism  Dots show the positions of individual parties.  The further to the right a dot is, the more right wing a party is.  The higher a dot is, the more supportive a party is of multiculturalism.  166 Figure 5.2: Left-Right Position Compared to Support for Multiculturalism Amongst Parties that Make At Least One Statement on Multiculturalism  Dots show the positions of individual parties.  The further to the right a dot is, the more right wing a party is.  The higher a do is, the more supportive a party is of multiculturalism.  167 Figure 5.3: Support for Welfare Compared to Support for Multiculturalism  Dots show the positions of individual parties.  The further to the right a dot is, the more supportive the party is of government welfare programs.  The higher a dot is, the more supportive the party is of multiculturalism.   In most countries, there is little disagreement between parties with different left/right ideologies. Figure 5.4 shows the differences in average multiculturalism positions for all, left, and mainstream right parties broken down by country. There are significant differences in average support in some countries.  In Austria, Denmark, France, and Italy left parties support multiculturalism more than right parties.42  In Belgium, the Netherlands, and Norway mainstream right parties average stronger support for multiculturalism than left parties.  Table 5.1 shows that, in spite of these differences,                                                42 There is also a significant difference in left and right party support for multiculturalism in Canada.  The high level of right support for multiculturalism, however, suggests that this more of a difference over the extent to which left and right parties emphasize their support of multiculturalism in their platforms than a difference of opinion over policy.  168 there is substantial overlap between the positions mainstream right parties take in 95% of years and the positions that left parties take. While differences in support for multiculturalism exist in some countries, it is not clear that either left or right parties are consistently more supportive of policy adoption.   Table 5.1: Summary of Party Positions Across Countries Country Parties Average Position Standard Deviation Range for 95% of Years Australia All Parties 0.156 0.240 -0.219-0.775 Right 0.215 0.505 -0.443-1.600 Left 0.188 0.303 0.000-0.979 Austria All Parties 0.156 0.492 -1.240-0.797 Right -0.055 0.645 -2.167-0.800 Left 0.390 0.526 0.000-1.711 Belgium All Parties 0.557 0.742 -0.930-1.806 Right 1.562 2.051 -1.540-6.176 Left 0.196 0.836 -2.365-1.066 Canada All Parties 1.268 1.655 -1.957-4.574 Right 0.934 2.444 -2.733-6.250 Left 2.392 1.670 0.000-6.300 Denmark All Parties -0.773 1.445 -3.919-0.546 Right -1.377 2.653 -9.051-0.549 Left -0.526 1.405 -4.642-0.883 Finland All Parties 0.240 0.434 0.000-1.687 Right 0.201 0.514 0.000-1.721 Left 0.404 0.939 -0.051-3.332 France All Parties 0.377 0.635 -1.032-1.381 Right 0.131 0.755 -1.428-1.339 Left 0.756 0.551 0.000-1.805 Germany All Parties 0.113 0.277 -0.497-0.887 Right -0.052 0.679 -1.872-0.998 Left 0.148 0.237 0.000-0.805 Greece All Parties 0.056 0.061 -0.001-0.172 Right -0.004 0.012 -0.034-0.000 Left 0.098 0.106 -0.017-0.311 Ireland All Parties 0.353 0.453 0.000-1.367 Right 0.370 0.488 0.000-1.346 Left 0.273 0.474 0.000-1.689 Italy All Parties 0.301 0.493 -0.398-1.527 Right 0.086 0.429 -1.072-1.005 Left 0.522 0.926 -0.086-3.055 Japan All Parties 0.000 0.001 0.000-0.002 Right 0.000 0.000 0.000-0.000 Left 0.002 0.008 0.000-0.035 Netherlands All Parties 0.667 0.554 -0.792-1.449 Right 1.687 0.621 0.402-2.544 Left 0.039 0.530 -1.753-0.603 New Zealand All Parties 0.995 0.726 0.139-2.979 Right 1.020 0.950 -0.413-2.897 Left 0.906 0.989 -0.511-3.220  169 Country Parties Average Position Standard Deviation Range for 95% of Years Norway All Parties 1.134 0.178 0.851-1.400 Right 1.911 0.933 0.460-3.202 Left 0.494 0.574 -0.017-1.559 Portugal All Parties -0.110 0.305 -0.813-0.143 Right -0.058 0.230 -0.599-0.237 Left -0.111 0.309 -0.828-0.131 Spain All Parties 0.390 0.285 0.047-1.409 Right 0.305 0.525 -0.347-2.081 Left 0.527 0.371 -0.042-1.167 Sweden All Parties 0.442 0.440 0.000-1.274 Right 0.271 0.391 0.000-1.200 Left 0.285 0.343 0.000-1.245 Switzerland All Parties 0.235 0.613 -1.236-1.323 Right -0.020 1.555 -4.908-1.784 Left 0.143 0.409 -0.397-1.296 United Kingdom All Parties 0.103 0.110 -0.044-0.368 Right 0.135 0.208 0.000-0.734 Left 0.115 0.189 -0.208-0.602 United States All Parties 0.333 0.474 0.000-1.930 Right 0.413 0.548 0.000-2.090 Left 0.276 0.499 0.000-1.820  Figure 5.4: Average Cross-Party, Left, and Right Positions by Country   170 The lack of a strong correlation between the left-right categorization of parties or left-right positions and support for multiculturalism suggests that party positions on the subject are orthogonal to left-right ideology. This contradicts much of the existing literature that links left parties to more liberal immigration and integration policies (Howard, 2009; Givens and Luedtke, 2005; van der Brug and van Spanje, 2009). It shows that it is important to look at multiculturalism positions independent of left-right ideology. Mainstream right parties can be important contributors to the development of multiculturalism; indeed, this has been the case in Australia and Canada. An analysis of parties' influence over policy that focuses largely on left parties will miss the extent to which right parties make important contributions to policy development.  This finding has further implications for the next chapter examining the determinants of party positions. It highlights the importance of looking at both the left and right when examining the influence of ethnic minorities and far-right parties. Consistent with work by Hinnfors et al. (2012) and Kriesi et al. (2008) there are reasons why left and right parties might support or oppose multiculturalism. Advocates for the policy might tap into ideas surrounding social solidarity to convince left parties while also appealing to liberal ideas surrounding cultural tolerance or liberal free market ideas regarding immigration in order to convince mainstream right parties. At the same time opponents of multiculturalism can use the fear that immigrants will take jobs and resources devoted to social programs in order to try to influence left parties’ positions while also appealing to social conservative and nationalist views.   171 Parties and Overall Policy Adoption Cross-Party and Government Support for Multiculturalism and Overall Policy Adoption An examination of cross-party positions and policy adoption shows a relationship between the two. Figure 5.5 shows average cross-party positions in years where policies adopted compared to years where partial policies are adopted and no policies were adopted. It demonstrates that support for multiculturalism is higher when policies are adopted. When at least one policy is adopted, support for multiculturalism is an average of 0.626 points. When at least a partial one is adopted, average cross-party support is 0.485 points. This compares to 0.369 points when only a partial policy is adopted and 0.326 points when no policy is adopted. Cross-partisan support for multiculturalism is almost twice as large in countries when policy adoption occurs than when it does not. Finally, and unsurprisingly, cross-partisan support for multiculturalism is weakest when retrenchment occurs, at -0.024 points.   172 Figure 5.5: Cross-Party Positions When Policies Are Adopted  Bars show average party positions in years when policies were adopted, partial policies were adopted, and no policies were adopted.  Hazard modeling provides some, somewhat inconclusive evidence for a link between cross-party positions and policy adoption. Table 5.2 shows a positive effect for most start dates in the analysis and most measures of multiculturalism. These effects are negligible for analysis that begins in 1970 but are substantial for the later, 1975 and 1980, start dates. The fact that parties’ effect grows over time is unsurprising. In the early 1970s, multiculturalism policies were still rare. The extent to which parties drew links between support for cultural diversity and multiculturalism in the early 1970s was likely limited by the extent to which multiculturalism was still a relatively novel policy idea. Parties may not have linked support for cultural diversity to multiculturalism at the time, because they may not have had a strong understanding or conception of what  173 multiculturalism policy was. As policies became more common, it is likely that parties would have developed a stronger understanding of the policies connected with multiculturalism.  The estimated relationship in the model is strong, but imprecisely estimated. Though the effect of one percentage point increase in parties’ positive multiculturalism statements for analysis using the 1980 start dates ranges from a 22% to a 35% increase in the likelihood of policy adoption (depending on the measure of total multiculturalism policy) the effect is only statistically significant at the 90% confidence level.  It is also only significant for measures of total multicultural policy that do not exclude either dual citizenship policies or mother-tongue education. This provides some indication that cross-party support for multiculturalism matters, but not strong proof of a relationship. This is important to consider, however, in light of findings that will be discussed later in the chapter that show that support by right parties in opposition increase the odds of policy adoption.    174 Table 5.2: Cross-Party Positions' Impact on Policy Adoption with Ethnic Minority Electoral Strength Controls MC Measure All Policies Dual Citizenship Removed  Affirmative Action Removed Mother-Tongue Education Removed Start Year 1970 1975 1980 1970 1975 1980 1970 1975 1980 1970 1975 1980 Cross-Party Positions 1.026 (0.814-1.294) 1.137 (0.883-1.464) 1.311* (0.981-1.753) 1.019 (0.806-1.290) 1.147 (0.885-1.487) 1.245 (0.934-1.660) 0.994 (0.773-1.278) 1.116 (0.844-1.476) 1.352* (0.968-1.887) 0.974 (0.764-1.242) 1.060 (0.813-1.381) 1.222 (0.895-1.669) Minority Electoral Strength 0.961 (0.893-1.033) 0.954 (0.884-1.029) 0.942 (0.866-1.024) 0.969 (0.901-1.043) 0.961 (0.889-1.039) 0.958 (0.881-1.041) 0.961 (0.892-1.035) 0.950 (0.877-1.028) 0.931 (0.852-1.017) 0.995 (0.920-1.076) 0.992 (0.914-1.077) 0.979 (0.895-1.071) Party with Left Ideology in Gov 1.751 (0.965-3.176) 1.412 (0.771-2.586) 1.331 (0.695-2.548) 2.083** (1.108-3.916) 1.681 (0.886-3.188) 1.458 (0.749-2.839) 2.013** (1.080-3.752) 1.591 ().844-2.999) 1.486 (0.748-2.951) 1.034 (0.581-1.843) 0.863 (0.479-1.556) 0.816 (0.436-1.528) Federalism 1.143 (0.667-1.959) 0.860 (0.661-1.120) 1.376 (0.760-2.489) 1.134 (0.650-1.978) 1.219 (0.686-2.166) 1.440 (0.784-2.643) 1.007 (0.571-1.775) 1.066 (0.592-1.919) 1.148 (0.604-2.182) 1.115 (0.621-2.002) 1.189 (0.654-2.164) 1.245 (0.659-2.351) Cross-Party Nationalism 0.793* (0.613-1.026) 0.860 (0.661-1.120) 0.967 (0.734-1.275) 0.785* (0.603-1.023) 0.860 (0.655-1.128) 0.948 (0.714-1.258) 0.702** (0.531-0.927) 0.771* (0.580-1.024) 0.881 (0.654-1.188) 0.755* (0.570-1.000) 0.788 (0.591-1.051) 0.869 (0.643-1.173) GDP Growth 1.018 (0.907-1.142) 1.005 (0.890-1.135) 1.005 (0.879-1.149) 0.978 (0.860-1.111) 0.962 (0.841-1.101) 0.970 (0.837-1.124) 1.020 (0.907-1.147) 1.005 (0.887-1.139) 0.997 (0.867-1.146) 0.992 (0.879-1.120) 0.983 (0.867-1.115) 0.985 (0.858-1.130) * P-value<0.1, ** p-value<0.05, ***p-value<0.01 Values are hazard ratios with ranges for the 95% confidence level in brackets  175 These results are robust to measures of policy adoption that exclude the more controversial policies from the index.  Figure 5.6 shows that the positive effect of cross-party support on policy adoption is largely consistent across different measures of total multiculturalism.  Removing dual citizenship and mother-tongue education policies slightly weakens the relationship between cross-party positions and policy adoption.  The differences between the different measures however are slight, suggesting that changing the measure of total multiculturalism does little to change the relationship between party positions and policy adoption. Figure 5.6: The Effect of Cross-Party Positions on Total Policy Adoption (1980 Start Date)  Data for this figure comes from the 1980 results in table 5.2. Dots show the hazard ratio for the effect of cross party positions on policy adoption. Lines show the range of effects that fall within the 95% confidence level for the model. Hazard ratios above 1 represent an increase in the likelihood of policy adoption while hazard ratios below one represent a decrease in the likelihood of policy adoption.   176 Analysis that looks at governing parties’ positions provides mixed evidence with respect to governing parties’ influence over policy.  Descriptive statistics that look at governing parties' support for multiculturalism and policy adoption suggest a similar connection between positions and policy. Figure 5.7 shows that governing parties average a score of 0.735 points for support of multiculturalism in years when one or more policies are adopted compared to a score of 0.382 points when no policies are adopted. The score for governing parties that adopt only partial policies, however, is lower than for governments that do not, at 0.209 points. The difference in governing party support between years in which policy occurs and years in which it does not is similar to the difference for cross-party support. Partial policy adoption however, appears only weakly influenced by governing party support. Finally, the link between opposition to multiculturalism and retrenchment is greater for government positions than for cross-party positions. Governments average a score of -0.826 points when retrenchment occurs.   177 F