Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Women’s descriptive and substantive political representation : the role of political institutions. Lore, Grace Alexandra 2016

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

Item Metadata

Download

Media
24-ubc_2016_september_lore_grace.pdf [ 3.6MB ]
Metadata
JSON: 24-1.0307353.json
JSON-LD: 24-1.0307353-ld.json
RDF/XML (Pretty): 24-1.0307353-rdf.xml
RDF/JSON: 24-1.0307353-rdf.json
Turtle: 24-1.0307353-turtle.txt
N-Triples: 24-1.0307353-rdf-ntriples.txt
Original Record: 24-1.0307353-source.json
Full Text
24-1.0307353-fulltext.txt
Citation
24-1.0307353.ris

Full Text

i     Women’s Descriptive and Substantive Political Representation: The Role of Political Institutions.   by   Grace Alexandra Lore  B.A. (Honours), The University of British Columbia MSc (Distinction), The London School of Economics   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)   July 2016   © Grace Alexandra Lore, 2016  ii Abstract This dissertation explores how political institutions shape the behaviour of women in politics. It asks whether the relationship between the gender of representatives (descriptive representation) and the representation of women’s interests (substantive representation) depends on the institutional context.    Using prominent institutional perspectives a theory is offered for how the division or fusion of powers and electoral systems affect substantive representation in both individual behaviour and policy outcomes. When party ties are weakened and internal party competition increased, women do more to substantively represent women, while men focus on other unincorporated interests.  Institutions affect policy outcomes, not only by affecting individual behaviour but also by determining how those actions are aggregated.   The theory is tested using a mixed methodological approach.  Two datasets capture individual behaviour - an international survey of representatives and an original dataset examining representative’s tweets. The results demonstrate that the gap between substantive representation by women and that by men is larger under the conditions of a division of powers and when electoral systems incentives the representation of unincorporated interests.  Interviews and surveys with more than 90 legislators from 7 countries provide evidence of the causal mechanisms. The aggregate effect of the number of women on gendered policy outcomes is tested using data combined from a range of sources.  The findings results are equivocal: institutions that facilitate individual action can also make policy change more difficult.   In short, institutions moderate the relationship between descriptive and substantive representation at the individual level and, to a lesser extent, at the aggregate level. This ‘institutionalist turn’ improves understanding of how, when, and why women act to represent women.  The ‘gender turn’ in the study of institutions demonstrates the flexibility of the theories and the broad and consequential impact of institutions. There are implications that extend beyond gender to include other issues and identities not incorporated into the party system, such as ethnicity.     iii Preface  This dissertation is original, unpublished, and independent work by the author, Grace Lore.    The fieldwork conducted in Chapters Five and Six are covered by the UBC Behaviour and Research Ethics Board for the completion of this dissertation (H14-02832).     iv Table of Contents Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iii Table of Contents ......................................................................................................................... iv List of Tables .............................................................................................................................. viii List of Figures ............................................................................................................................... xi Acknowledgments ....................................................................................................................... xii Dedication ................................................................................................................................... xiv Chapter 1: Introduction ............................................................................................................... 1 Women’s Political Representation & Women’s issues .......................................................... 6 Defining ‘Women’ and ‘Women’s Interests’ .......................................................................... 7 Testing the Link Between Gender and Representation ......................................................... 9 Preferences and Priorities ...................................................................................................... 10 Discourse and Behaviour ....................................................................................................... 11 Policy Outcomes .................................................................................................................... 11 Women’s Representation & Institutions: The State of the Literature ............................... 14 Clues in the Gaps: Where do we Look for Substantive Representation? .............................. 18 From Self-Selection to Policy Outcomes: Substantive Representation & the Career Paths of Women Politicians .............................................................................................................. 21 Using a Principal-Agent Framework .................................................................................... 25 An Alternative to Personal Votes: Unincorporated Representation .................................. 27 Women’s Issues as Unincorporated Issues ............................................................................ 30 Institutions and The Representation of Women – An Interactive Effect .......................... 33 Outcomes .................................................................................................................................. 36 Looking Forward: A Road Map ............................................................................................ 37 A Note on Methods and Sources ........................................................................................... 39 Conclusions .............................................................................................................................. 42 Chapter 2: Theorizing the Moderating Effect of Institutions ................................................. 45 Moderating Behaviour: The Division of Powers .................................................................. 47 Discourse & Behaviour: Observable Implications ................................................................ 62 Women’s Experiences of Political Institutions: Observable Implications ............................ 63 Moderating Behaviour: The Electoral System ..................................................................... 63 Proportional Representation versus Single-Member-Districts ........................................... 65 Discourse & Behaviour: Observable Implications ................................................................ 70 Women’s Experiences of Political Institutions: Observable Implications ............................ 70 Unincorporated Representation............................................................................................. 73 Discourse & Behaviour: Observable Implications ................................................................ 79  v Women’s Experiences of Political Institutions: Observable Implications ............................ 80 Substantive Representation & Descriptive Representation: In Conflict? .............................. 81 Combining Constitutional Choices: Division of Powers and Electoral Systems ............... 82 Moderating Outcomes: Presidential and Parliamentary Systems ...................................... 87 Parliamentary Systems: Unfulfilled Potential for Substantive Representation ..................... 87 Presidential Systems: Veto Players and Indecisiveness ........................................................ 88 Moderating Outcomes: Electoral Systems ............................................................................ 90 Policy and Legislative Changes: Efficiency .......................................................................... 90 Parties and Party Systems ...................................................................................................... 91 Institutions in the Real World ................................................................................................ 91 Belgium ................................................................................................................................. 92 Canada ................................................................................................................................... 93 Germany ................................................................................................................................ 93 The Netherlands ..................................................................................................................... 94 Switzerland ............................................................................................................................ 95 The United Kingdom ............................................................................................................. 96 The United States................................................................................................................... 96 Chapter 3: Testing the Theory - Substantive Representation in Presidential and Parliamentary Systems ............................................................................................................. 105 The Division of Powers and Individual Behaviour – The PARTIREP Data ................... 107 Capturing Women’s Substantive Political Representation .................................................. 108 Women Representing Women ............................................................................................. 110 Hypotheses........................................................................................................................... 111 Results ................................................................................................................................. 113 The Representation of Women by Women, Party Affiliation, and Moderating Effect of a Division of Powers .............................................................................................................. 114 Institutions as Filters, Incentives, or Both ........................................................................... 115 Discussion ............................................................................................................................ 118 Twitter – The Impact of Gender and the Division of Powers on Public Discourse in 140 Characters .............................................................................................................................. 119 The Data .............................................................................................................................. 121 Analysis ............................................................................................................................... 124 Results ................................................................................................................................. 125 Beyond the Statistical Analysis ........................................................................................... 129 Discussion ............................................................................................................................... 131 Chapter 4: Testing the Theory - Electoral System and Women’s Substantive Political Representation........................................................................................................................... 144 Results ................................................................................................................................. 150 District Magnitude: Does it Moderate the Moderating Effect? ........................................... 153 Targeting the Causal Process ............................................................................................... 155  vi The Representation of Women by Women, Party Affiliation, and Moderating Effect of Incentives for Unincorporated Representation .................................................................... 157 Institutions: Incentives, Filtering Systems, or Both. ........................................................... 159 Looking In-depth: Understanding the Effect of Different Institutional Features ................ 161 Revisiting Hypotheses 2-3: Exploring the Null Finding ..................................................... 166 Discussion: Reconsidering the Role of the Division of Powers and the Electoral System on Substantive Representation. ................................................................................................. 168 Chapter 5: In Their Own Words ............................................................................................. 184 Methods ............................................................................................................................... 184 The Interviewees .................................................................................................................. 190 Representing Women .......................................................................................................... 193 The Representation of Women by Men ............................................................................... 198 Why Women Represent Women ......................................................................................... 200 Women’s Issues ................................................................................................................... 204 Intersectionality: Representing Diversity among Women. .................................................. 205 Representing women: Challenging and Problematic ........................................................... 206 Discussion and Looking Forward ........................................................................................ 207 Chapter 6: Political Institutions in Women’s own Words .................................................... 213 The Division or Fusion of Powers: Women’s Experiences in their own Words .............. 213 Parliamentary Systems –Backbench Constraints ................................................................ 215 Parliamentary Systems –Backbench Disincentives ............................................................. 218 Parliamentary Systems - Backbench Opportunities ............................................................ 219 Parliamentary Systems - Opportunities and Influence in Government ............................... 232 Presidential Systems - Individual Opportunities, Collective Constraints ............................ 234 Electoral Systems................................................................................................................... 239 Unincorporated Representation vs. Party Votes ................................................................ 240 The Vote Dimension ............................................................................................................ 240 Ballot Dimension ................................................................................................................. 245 Vote Pooling ........................................................................................................................ 252 The Tension Between Descriptive and Substantive Representation ................................... 254 Personal Electoral Systems and Division of Powers – A Contingent Relationship ......... 255 Revisiting the Null Finding: Proportional Representation and Single Member Plurality ................................................................................................................................................. 260 The Geographic Tie of Representation in Single Member Plurality Systems ..................... 260 Proportional Representation: Geography still matters......................................................... 265 The Electoral Threshold ...................................................................................................... 272 Multiple Representatives ..................................................................................................... 274 Conclusions ............................................................................................................................ 278 Chapter 7: Women’s Descriptive Representation, Institutions, and Policy Outcomes ...... 281  vii The Comparative Agendas Project: Legislation, Gender, and the Division of Powers .. 282 Analysis ............................................................................................................................... 288 Results ................................................................................................................................. 289 The Role of Individual Women ........................................................................................... 290 Analysis without the US ...................................................................................................... 291 The Role of Political Institutions: Insights in the Form of Limited Data ............................ 292 Gender, Institutions, and Policy Outcomes ........................................................................ 295 The Data .............................................................................................................................. 296 Limitations of the Analysis.................................................................................................. 304 Interpreting the Results ........................................................................................................ 305 Women’s Political Representation, Division of Powers, and Policy Outcomes ................. 306 Parliamentary Systems - Women in Cabinet ....................................................................... 308 Women’s Political Representation, Electoral Systems, and Policy Outcomes ................... 309 Women in Cabinet – The Role of Electoral Systems .......................................................... 313 Conclusions and Looking Forward ..................................................................................... 315 Chapter 8: Conclusions ............................................................................................................ 337 Implications for the Study of Women’s Political Representation..................................... 342 Implications for the Study of Political Representation ...................................................... 345 Proportional Representation vs. Single Member Plurality Systems .................................... 346 Extending and Adapting Carey and Shugart’s Incentives for Personal Votes .................... 348 Combining Constitutional Choices: The Contingent Relationship Between the Division of Powers and Electoral Systems ............................................................................................. 355 Implications for Activists and Advocates ............................................................................ 356 Final Words ........................................................................................................................... 359 References .................................................................................................................................. 361 Appendix A: Supplementary Content for Chapter 3 ............................................................ 383 Appendix B: Supplementary Material for Chapter 4 ........................................................... 395 Appendix C: Supplementary Material for Chapter 5 ........................................................... 404 Appendix D: Supplementary Material for Chapter 7 ........................................................... 419      viii List of Tables Table 1.1. Survey of the Literature ............................................................................................... 44 Table 2.1. Division & Fusion of Powers  - The Moderating Effect on Discourse & Behaviour .. 98 Table 2.2. SMP vs. PR: The Moderating Effect on Discourse & Behaviour ............................... 99 Table 2.3. Unincorporated Representation & Party Votes: The Moderating Effect on Discourse & Behaviour............................................................................................................................. 100 Table 2.4. Division and Fusion of Powers: The Moderating Effect on Outcomes ..................... 101 Table 2.5. Electoral Systems: The Moderating Effect on Outcomes .......................................... 102 Table 2.6: Institutions in the Real World .................................................................................... 103 Table 3.1. The Effects of Gender and Division of Powers ......................................................... 135 Table 3.2. Predicted Importance of Representing Women by Gender and Institution ............... 136 Table 3.3. Effect of Gender & Division of Powers on Behaviour Controlling for Preferences . 137 Table 3.4. The Representation of Women via Twitter ................................................................ 138 Table 3.5. Proportion of Tweets Addressing Women’s Issues ................................................... 139 Table 4.1. Distribution of Cases on the Ballot Dimension ......................................................... 170 Table 4.2. The Effects of Gender and Electoral Institutions on Behaviour ................................ 171 Table 4.3. Testing the Effect of District Magnitude ................................................................... 172 Table 4.4. The Effects of Gender and Campaign Preferences on Behaviour ............................. 173 Table 4.5. Predicted Importance of Representing Women by Gender and Institution ............... 174 Table 4.6. The Effect of Institutions & Gender on Behaviour Controlling for Preferences ....... 175 Table 4.7. The Effects of Gender and Components of the Ballot Dimension ............................ 176 Table 4.8. Testing the Features of the Nomination Procedure ................................................... 177 Table 4.9. The Effects of Gender and Components of the Vote Dimension .............................. 178 Table 4.10. The Effects of Gender and the Ballot and Vote Dimensions ................................... 179 Table 4.11. The Effect of Gender and PR on Forms of Substantive Representation. ................ 180 Table 5.1. Qualitative Research in National & Subnational Legislatures .................................. 210  ix Table 5.2. Qualitative Research in the European Parliament ..................................................... 210 Table 5.3. Summary of Institutions............................................................................................. 211 Table 7.1. Cases: The Comparative Agendas Project Database ................................................. 317 Table 7.2. Percentage of Bills/Laws Addressing Issues by Country .......................................... 317 Table 7.3. The Effects of Women’s Representation and Division of Powers on Legislation .... 318 Table 7.4. Bills Passed In House of Representatives Introduced by Women in the United States ............................................................................................................................................. 318 Table 7.5. Dropping the U.S. - The Effects of Women’s Representation and Division of Powers on Legislation ...................................................................................................................... 319 Table 7.6. Summary of Variables ............................................................................................... 320 Table 7.7 Cases By Model .......................................................................................................... 322 Table 7.8: Women’s Representation, Presidential Systems, and the Impact on Policy Outcomes ............................................................................................................................................. 323 Table 7.9. Women’s Representation, Executive Dissolution of Legislature, and the Impact on Policy Outcomes .................................................................................................................. 324 Table 7.10. Summary of Results for Presidential/Parliamentary Systems ................................. 325 Table 7.11. The Effect of Women in Cabinet on Policy Outcomes ........................................... 326 Table 7.12. Women’s Representation, Proportional Representation, and the Impact on Policy Outcomes ............................................................................................................................. 327 Table 7.13. Women’s Representation, the Ballot Dimension, and the Impact on Policy Outcomes ............................................................................................................................................. 328 Table 7.14. Women’s Representation, the Vote Dimension, and the Impact on Policy Outcomes ............................................................................................................................................. 329 Table 7.16. Summary of Results for Electoral Systems ............................................................. 331 Table 7.17. Women’s Representation in Cabinet, Incentives of Unincorporated Representation, and the Impact on Policy Outcomes .................................................................................... 332 Table A.1. PARTIREP Data: Representatives by Country ......................................................... 383 Table A.2. PARTIREP Data: Representatives by Institution ..................................................... 384  x Table A.3. PARTIREP Data: Self-Reported Representation of Women by Gender .................. 385 Table A.4. PARTIREP Data: Self-Reported Representation of Women by Gender .................. 386 Table A.5. Testing Cut Points ..................................................................................................... 387 Table A.6. Testing Cut Points ..................................................................................................... 388 Table A.7. Robustness Tests ....................................................................................................... 389 Table A.8. The Representation of Women: Interacting party affiliation .................................... 390 Table A.9. The Effects of Institutions & Gender on the Importance of Representing Women .. 391 Table A.10. Twitter Analysis Key Words .................................................................................. 392 Table A.11. Twitter Analysis Robustness Tests ......................................................................... 393 Table A.12. Twitter Analysis Robustness Test2 ......................................................................... 394 Table B.1. Robustness Tests: Multilevel regression analysis grouping individuals by party. .... 395 Table B.2. Robustness Tests: Multilevel regression analysis grouping individuals hierarchically by party, parliament, country. .............................................................................................. 396 Table B.3. Robustness Tests: OLS Regression, Errors Clustered by Parliament. ...................... 397 Table B.4. Robustness Tests: OLS Regression, Errors Clustered by Country. .......................... 398 Table B.5. Robustness Test: OLS Regression, Errors Clustered by Institution. ........................ 399 Table B.6. Robustness Tests: Multilevel Ordered Logistic Regression Analysis Grouped by Party ............................................................................................................................................. 400 Table B.9. The Representation of Women: Interacting Party Affiliation ................................... 402 Table B.10. The Effects of Institutions & Gender on the Importance of Representing Women 403 Table C.1. Interview Methods Table .......................................................................................... 408 Table C.2. Response Rates by Country and Party ...................................................................... 416     xi List of Figures Figure 3.1. Average Representation of Women by Gender & Self-Placement .......................... 140 Figure 3.2. Predicted Behaviour by Gender and Institutions ...................................................... 141 Figure 3.3. Predicted Behaviour by Gender, Party Orientation, and Institutions ....................... 142 Figure 3.4. Women’s Issue Tweets by Gender and Presidential System.................................... 143 Figure 4.1 Predicted Behaviour by Gender and Institutional Features  Testing Hypothesis 4 ... 181 Figure 4.3. Predicted Behaviour by Gender, Institution, and Party Orientation ......................... 183 Figure 7.1. Women’s Representation, Division of Powers, & Family Legal Issues .................. 333 Figure 7.2. Women’s Representation, Division of Powers, & Strategic Gender Issues ............. 334 Figure 7.3. Effects of Women in Legislasture and Electoral System on Policy Outcomes ........ 335 Figure 7.4. Effects of Women in Cabinet and Electoral System on Policy Outcomes ............... 336  xii xii Acknowledgments This dissertation is the product of years of vigorous discussions, brainstorming, and ah-ha moments made possible by the curious, enthusiastic, and energetic people in my life. I’d like to thank, first and foremost, my supervisor, Richard Johnston, for joining me on this intellectual journey.  Without your guidance, wisdom, and expertise, this dissertation would not be what it is and the process would have been much less fun and far less rewarding.  Thank you also for facilitating this project while I grew my family – your support and encouragement made it possible to do both.   I would also like to thank my committee –Anjali Bohlken and Antje Ellerman.  Both contributed significantly to the rigour of the methods and the quality of the theoretical contribution. Incorporating their comments and suggestions was gratifying and I feel like I am a better scholar for their involvement in this work.   This research was supported by the Social Sciences and Humanities Research Council of Canada with a Joseph-Armand Bombardier CGS Doctoral Scholarship and a Michael Smith Foreign Study Supplement, which enabled the fieldwork.   My colleagues contributed to this work through their feedback and thought-provoking questions provided during workshops through the UBC Institute for European Studies and Political Science Department Can-Comp series.  In particular, I’d like to thank Alan Jacobs, Andrea Nuesser Dominik Stecula, Alexander Hemmingway, Edana Beauvais, and Daniel Westlake. Not only did Dr. Ken Carty provide informal feedback on this project, but as my undergraduate honours thesis advisor, he, in many ways, started me down the path of academia.  I thank Dr. Carty for inspiring a love of research over the last eight years. A number of other professors at UBC need to be thanked for their informal feedback, willingness to help with data and analysis, and support along the way – Dr. Andrew Owen, Dr. Fred Cutler, Dr. Barbara Arneil.   Dr. Emilie Van Haute was generous with the PARTIREP data before it was publicly available and provided me with a home at The Université Libre de Bruxelles while I conducted my fieldwork in Europe.  Dr. Van Haute, Dr. Silvia Erzeel and her colleagues at CEVIPOL also provided assistance with the PARTIREP data and very productive feedback on Chapter 4.  Several scholars enabled this research by graciously sharing their personal datasets – Pippa Norris, Amy Atchinson, Shaun Bevan, and Roy Gava.   Finally, none of this would have been possible without my friends and family.  My father Stan Lore and father-in-law Bill Parker both provided constant feedback and helped me with both the big picture behind this research and the small picture (someday I will figure out affect vs. effect). My dad, in particular, has been editing papers for me for ten years during my post-secondary journey and I will be forever grateful to him for instilling in me a love of thinking and for saving me from many embarrassing typos.  My mom inspired in me a love of politics and news in my earliest years.  She also encouraged my commitment to giving back and to community involvement – both of which continue to enrich and add meaning to my academic work.  My  xiii sisters have offered undying love and support for both my little family and me.  How does one do anything without that?   Several friends contributed to this project by sharing in and contributing to my enthusiasm, feistiness, and curiosity – Ursula Kerr, Kaytee Stewart, Julie Bero, Kelsey Wrightson, Alison James, Katie Robb, Nancy Peckford, Kiel Giddens, Judy Kitts, Kristin Price, Justin Alger, Rebecca Kemp, Lindsay Walton, and Lisa Beca. I owe much to the inspiring conversations that opened my mind, to their passion and energy, and to their commitment to a better world.  Thank you for fuelling my fire!   My last thank you goes to the people who sacrificed the most to make this happen. My partner, Rob Lore, was the biggest cheerleader, sounding board, big-thinker, brain-stormer, stay-at-home dad, comprehensive exam-widower, and general super-spouse. He not only made this possible but fun.  My little love Eve put up with her mom on the computer a lot and figured out how to cuddle while I typed.  Finally, Asher, who waited till I handed this in before joining our family and then let me cuddle my way through defense prep; welcome to this world, we’re working on making it a better place.     xiv Dedication      To my family, by chance and by choice.   And to Evie and Asher-  I hope by the time you care to read this, the work will be done.    1 Chapter 1: Introduction “The majority of feminist political science is broadly institutional in focus, yet does not use new institutional theory.” – Kenny, 2007, 95.  This dissertation explores how political institutions shape the behaviour of women in politics. While the effect of political institutions on women’s descriptive representation—the number of women—is well understood, the study of the substantive representation—the representation of women’s issues—has occurred largely in an institutional vacuum. As a result, it has failed to benefit from institutionalist perspectives that clearly outline how political institutions shape behaviour and influence policy. Many of the field’s most prominent scholars have observed (and bemoaned) the gap between the study of women’s representation and institutionalism. Little has been done, however, to make use of existing institutional theories to explain how and when women act to represent women.  Moreover, although the inaugural work written on the topic (Dahlerup, 1988) explored the experience of women in Scandinavian countries, the empirical research has been Anglo-centric and astonishingly little comparative work has been undertaken. Related to the narrow scope of cases is the significant and problematic under-theorizing of the relationship between gender and the substantive representation of women. The individual women and men themselves and their agentic, strategic behaviour have not been given due consideration; analyses have failed to consider their conflicting and diverse roles and have focused too broadly on representatives in the aggregate. As a result, despite decades of research, conclusive statements and generalizations about the nature and effects of women’s political representation remain elusive.   2 This dissertation takes seriously the call by Kenny, Krook, Mackay and others for an ‘institutional turn’ in feminist political science (Kenny 2007; Krook and Mackay 2011). It is an response to Celis’s (2008) demand for a study of the relationship between women’s descriptive and substantive representation that “take[s] context seriously” (116) and an answer to Driscol, Krook, Kenny, and Mackay appeal for a feminist institutional analysis “attentive to gender, strategy, institutions, power and change” (Driscol and Krook, 2009, 3; Franceschet, 2011; Kenny, 2007; Krook and Mackay; 2011).  To enable an institutional turn in the study of women’s substantive political representation, a ‘gender turn’ in institutionalist approaches is also required. The institutional theories themselves are inattentive to gender and built around a universal, ‘neutral’ actor that is implicitly masculine. The autonomous, self-interested actor remains central, and market dynamics and calculations leave little consideration to care roles and responsibilities, unpaid labour, and dependence on the state (Geortz & Mazur, 2008; Krook & Driscoll, 2012). Yet existing institutionalist theories have clear implications for the representation of gender and other identities besides. Insisting on a ‘neutral’, ‘universal’ actor masks the applicability and power of the theories. A ‘gendered turn’, on the other hand, demonstrates the flexibility of the theories and the broad and consequential impact of institutions.  Incorporating gender also addresses an abiding concern with institutionalism: the endogeneity of the very institutions under study. Most political institutions were established before women were allowed to enter politics. To the extent that women’s political participation was a matter of “add women and stir”, institutions can be taken as exogenous. Contrast this with the relationship between institutions and, say, the distribution of political resources or the number and  3 significance of political cleavages; which causes which? Incorporating gender enables strong tests of theories of institutional impact, with implications beyond the case.  To bridge these two literatures, I situate the substantive representation of women by women squarely in the institutions within which it occurs. I theorize how women, as representatives, navigate institutional incentives and opportunities to meet diverse career and policy objectives. By structuring the context within which female politicians make choices over who and what to represent and how to do so with limited resources (time, money, political capital), institutions may encourage or discourage the representation of women by women. In doing so, institutions moderate the relationship between the gender of a representative and the attention to and action on gendered political issues. Simply put, the substantive representation of women by women may vary with the institutional context.  Incorporating a consideration of the institutional context into other theories has proved fruitful – first challenging and then improving existing concepts (for example see Duch and Stevenson, 2008; Huber, Kernell, and Leoni, 2005; Kedar, 2009). In Voting for Policy, Not Parties, for example, Orit Kedar (2009) argues that voting behaviour models - whether voters choose the party or candidate closest to their ideal opinion (proximity voting) or the most extreme candidate or party on their “side” of the left-right spectrum (directional voting) – must be attentive to the institutional context. Institutions, she argues, determine the relative costs and benefits of different voting strategies. Combining the two literatures thus improves voting models and increases our understanding of ways institutions affect political behaviour and outcomes. Given the well-established importance of both gender and institutions on political preferences and behaviour, linking these two fields offers significant opportunities to increase understanding of both women’s political representation and institutions.   4 A principal-agent approach structures the analysis. While feminist academics have tended to see their methods as antithetical to or at least at odds with rational choice, others have recently called for a reconsideration of the objections to the methodological tool (Krook and Driscol, 2012). By remaining self-reflective of the masculine assumptions that under-pin the ‘neutral’ actors and by incorporating an explicitly gendered agenda, there exists an untapped opportunity for a conversation between rational choice and feminist political science (see also Goertz and Mazur, 2008). Krook and Driscol (2012) argue that rational choice improves on feminist explanations “by highlighting the micro-level interactions leading up to… decisions … and the ways in which political institutions – both formal and informal – shape the strategies and constraints of different actors.” (19). By demonstrating the ways in which social identities matter and are neither antithetical to nor incompatible with rational choice, the approach used in this dissertation allows for a consideration of women as strategic political agents and challenges the gender-blindness of existing institutional work (Swers, 2002).  In what follows, I provide the framework for my research. I begin with a discussion of the key ideas, including descriptive and substantive representation, a note on the concept of critical mass, and what is meant by ‘women’ and by ‘women’s interests’. These sections grapple with difficult concepts; where possible I provide working definitions and in other instances simply make clear what has had to be bracketed in order to continue with the research. I then discuss the state of the literature - when institutions are and are not considered and the clues inherent in the gaps in the literature. To turn attention to women as strategic actors whose behaviour must be understood as institutionally based, I outline the paths through which women end up in political office and engage in representation. The theorized career pathway is explicitly institutional and allows for a consideration of who becomes a representative and how and what  5 they do when they get there. A general argument about how institutions moderate substantive representation by structuring principal-agent relationships follows. As part of this, I introduce a concept called unincorporated representation, which parallels but expands on the idea of personal as opposed to party representation. Chapter 2 offers a more specific theory on the anticipated effect of two institutions - the separation or fusion of powers and the electoral system. Both institutions are well known for structuring principal-agent relationships, generating incentives and opportunities, and affecting political outcomes (Carey & Shugart, 1992,1995; Cox, 1997; Dewan and Spirling, 2011; Duverger, 1954; Haggard and McCubbins, 2001; Hammond and Butler, 2003; Immergut, 1998, 2010 Strom, 1997, 2002; Tsebelis, 1995; Uslaner & Zittel; 2006; Weingast and North, 1989). For example, the fusion of the executive and legislative branches and the practice of confidence votes in Westminster Parliamentary systems incentivize cohesive voting and party control. As a result, individual representatives have little autonomy over their vote and even the topics they address in speeches or introduce in legislation (Cox, 2005; Dewan & Spiring; 2011; Kam, 2001). Electoral systems can also structure incentives to follow party lines by determining the extent to which representatives rely on party or personal votes (Carey & Shugart, 1995). As I argue in the remainder of this chapter and more thoroughly in Chapter 2, these institutional features have clear implications for the representation of women’s interests.  The remainder of this dissertation tests the theory using a mixed methodological approach. Chapters 3 and 4 use quantitative analysis and employ multiple datasets to test the moderating effect of the division of powers and electoral systems on political behaviour respectively. In Chapters 5 and 6 nearly 90 qualitative surveys and in-depth interviews with women in seven countries allow for a further test of the theory. Finally, Chapter 7 returns to quantitative analysis  6 to test the moderating effect of institutions on policy outcomes. Chapter 8 offers conclusions and directions for future research.   Women’s Political Representation & Women’s issues This research uses the framework offered by Pitkin (1967) in The Concept of Representation, which distinguishes between descriptive representation and substantive representation.1 Descriptive representation means standing for and refers to the make-up of a legislative body. In the aggregate, it means the number or proportion of women. Substantive representation indicates acting for, referring to the actions of individual representatives and collective decisions in legislative bodies. Although Pitkin did not view these two types of representation as necessarily mutually exclusive, she favoured a focus on substantive representation. The notion that descriptive and substantive representation may be linked is not new and is in fact much older than Pitkin’s work. John Stuart Mill argued, “in the absence of its natural defenders, the interest of the omitted is always in danger of being overlooked; and when looked at, is seen with very different eyes from those of the persons whom it directly concerns” (1967, 22). Since Pitkin’s influential work, however, the assertion that a candidate or politician’s gender is or at least can be relevant to what they do has been common, both among activists and academics (Phillips, 1995).                                                   1 The term descriptive representation was originally coined by Griffiths and Wolheim (1960) before it was adopted by Pitkin (1967) (Mansbridge, 1999). In The Concept of Representation, Pitkin also discusses the formalistic uses of ‘representation’, beginning with Thomas Hobbes who uses the term to mean formal authorization or authority to act. She then discusses representation as accountability, wherein representation is understood as the arrangements or institutions which follow (terminate) it. Pitkin also explores symbolic representation which, related to descriptive representation, suggests an ‘identification’ and ‘alignment of wills between ruler and ruled’ (106, 108). I confine my discussions here to descriptive and substantive representation.   7  Defining ‘Women’ and ‘Women’s Interests’ The claim that ‘women represent women’ implies both that there is such a thing as ‘women’ and that there are ‘women’s issues’ to represent. While the notion that women are under-represented is hardly a controversial idea, the idea that there is such a thing as women is not. Women are not a homogenous group but are further defined by race, ethnicity, class, and sexual orientation. The suggestion that ‘women’ constitute a group, masks these intersecting differences and inequalities and can operate in a hegemonic manner, favouring a definition of women that reflects the already privileged/dominant women (Phillips, 2010, Mansbridge, 1999).  Women of colour, those facing poverty, and women with low levels of education, for example, are among the most under-represented in politics (Black, 2006; Crenshaw, 1991; McCall, 2005). Beyond an intersectional argument, since Simone Beauvoir (1953, 249) espoused that ‘one is not born a woman, but rather, becomes one’, there has been an effort to understand gender as something people do rather than something they ‘have’. Gender has been conceptualized, for example, as a performance (Butler, 1990) or system (Rubien, 1975).  The theory and analysis offered in this dissertation, however is not possible unless one is able to speak about a group called ‘women’.  Iris Marion Young (1994) describes this problem in the following way:  “On the one hand, without some sense in which "woman" is the name of a social collective, there is nothing specific to feminist politics. On the other hand, any effort to identify the attributes of that collective appears to under- mine feminist politics by leaving out some women whom feminists ought to include” (714). To undertake this research I have had to bracket the question of how to define, or rather if one can define, women.  In doing so I am not claiming that it is either possible or easy to  8 conceptualize women as a group, but rather that it is a necessary, if problematic, part of the research agenda.  Without a clear definition of ‘women’, how can one claim that there are ‘women’s interests’ in politics? To be sure, systematizing what exactly is in ‘women’s interests’ is as complicated and controversial as defining ‘women’ (for example, see Goertz & Mazur, 2008). As Phillips (1995) argues, this is not to say that there are not some issues that are gendered. While recognizing the heterogeneity of women and their experiences and the intersectional nature of oppression, women’s position in the unequal gender order, defined by a sexual division of paid and unpaid labour and relative exclusion from centres of economic and political power, does create distinct gendered interests.   Molyneux (1985, 1998) separates practical and strategic gendered interests; the former serving women’s needs generated within the current gender order while the latter seeks to permanently reposition women in society. This distinction differentiates, for example, between maternity and paternity leave - the former reinforces women as caregiver while also facilitating paid employment and the latter challenges the private/public division of labour by enabling caregiving by men. Other scholars distinguish between feminist policies such as affirmative action legislation and those policies that address more traditional areas of women’s interests such as childcare, maternity leave or family allowance (see Saint-Germain; 1989). These categories are not entirely clear-cut, and the boundary between the two depends, to a certain extent, on the timeline of the perspective taken; still the distinction is analytically useful. Whether practical or strategic, feminist or traditional, ‘gendered issues’ are systematized as that arising out of women’s unequal position in the sexual division of labour or unequal access to  9 economic and political power; this includes feminist interests (see Saint-Germain, 1989). In particular, I follow Wängnerud (2000) who identifies women’s issues in politics as those that 1) recognize women as a social category, 2) acknowledge the unequal balance of power between the sexes, and 3) aim to increase the autonomy of female citizens (see also Lovenduski and Norris, 2003; Grey; 2006).  The specific issues and polices considered throughout this dissertation fall within this definition but do not, of course, cover the range of issues that might.  The decisions made over what to include and what to exclude are in some cases somewhat arbitrary and in other instances heavily influenced by available data. These decisions are, however, necessary part of the empirical research. Fortunately, much of the data – both qualitative and quantitative – rests on self-defined substantive representation where legislators have the opportunity to fill in the content of what it means to act for women.  Moreover, some of the other quantitative analysis is informed by the insights from women themselves, using the policies legislators identified as important to women during interviews as the measure of substantive representation.  This approach still does not cover the potential universe of women’s issues and may bias the working definition and operationalization towards those important to women more likely to be represented in politics – white, educated, straight, middle/upper class, and cis-gendered.  Testing the Link Between Gender and Representation The research exploring the link between Pitkin’s concepts of descriptive and substantive representation can broadly be grouped into three broad categories: those that examine preferences and priorities, those that look at political behaviour of individuals, and those that consider policy outcomes; I discuss each in turn.  10 Preferences and Priorities The idea that women represent women depends on a difference of priorities and preferences among men and women (Norris, 2001). If women and men do not differ in their political preferences or objectives, then it can hardly be said to matter whether it is men or women making political decisions. The political preferences of men and women are expected to differ as a result of a social context defined by patriarchy, unequal division of paid and unpaid labour and inequalities in access to power - economic, social, and political.  The empirical research supports the notion that women’s and men’s preferences and priorities differ. Historically, women were more conservative than their male counterparts but are now more likely to express views to the left of men (Studlar and McAllister, 1992; Inglehart and Norris, 2010).2 When it comes to candidates and representatives, party affiliation is (unsurprisingly) the strongest determinant of opinions and priorities on general issues such as the level of free market economy, international relations, crime, and environmentalism. Differences do exist, however, within parties, with women being more ‘left’ than their male counterparts (Gidengil et al., 2003; Norris, 2001; Poggione, 2004; Vega and Firestone, 1995).  More importantly, significant differences have been observed between men and women (candidates, politicians, and voters) on gendered issues, such as affirmative action in politics, support for children and families, and equal opportunities in paid work (Bashevkin, 1993; Childs, 2002; Dodson, 2001; Erickson, 1997; Gidengil et al., 2004; Lore, 2008; Lovenduski, 1997; Lovenduski and Norris, 2003; Meier, 2008; Norris; 2001; Studlar and McAllister, 1992; Thomas,                                                  2 McAllister (1992) suggests the change occurred in the 1960s in Australia while Inglehart and Norris (2010) estimate the timing closer to the 1990s in most industrialized societies. Inglehart and Norris suggest that women in developing countries are still more conservative than their male counterparts.   11 1991; Thomas and Welch, 1991; Tremblay, 1998; Young, 2002). Other empirical investigations have found that women feel that they represent women (and have a responsibility to do so) based on “a shared sense of identity and a concern for women’s issues” (Childs, 2002, 150; see also Carroll, 2002; Childs, 2003; Celis and Childs, 2007; Freeman, 2002; Reingold, 1992, 2000; Tremblay, 2006; Wängnerud, 2000; Whip, 1991).  Discourse and Behaviour While women may place a higher priority on women’s issues, substantive representation explicitly requires acting for. Numerous studies have tested the relationship between descriptive and substantive representation by looking at behaviour, for example voting, the introduction of legislation, speeches, and statements. In general, these studies have found that women expand the political agenda to include women’s perspectives, make speeches on women’s issues, meet more frequently with women’s organization, introduce more legislation related to women’s interests, and are more likely to vote in favour of policies which support women (Bratton, 2005; Bratton and Haynie, 1999; Celis, 2006; Dodson, 2001, 2005, 2006; Kathlene, 1994, 2005; Gerrity, Osborn, and Mendez, 2007; McDonald & O’Brien, 2010; Reingold, 1996; Rosenthal, 2000; Saint-Germain, 1989; Swers and Larson, 2005; Taylor-Robinson and Heath, 2003; Thomas, 1991; Tremblay, 1998; Wängnerud, 2000; Wolbrecht, 2002). There are, however, several studies which have found no or very minimal relations between gender and behaviour (Barnello, 1990; Reingold, 2000; Vega and Firestone, 1995).  Policy Outcomes Finally, a third way to explore the link between descriptive and substantive representation is to explore changes to policy and legislation. Policy changes, Dahlerup (1988) argues, are the “most crucial element” of substantive representation (283). Several studies find that women are more  12 successful than men at passing legislation on women’s issues and that certain policy provisions are correlated with the presence of women (Bratton, 2005; Bratton and Haynie, 1999; Saint-Germain, 1989), including those on abortion regulation (Berkman and O’Connor, 1993), public funding for child care (Bratton and Ray, 2002), child support legislation (Case, 1998; Crowley, 2004), availability of maternity and childcare leave, family and old age services, health transfers to families and individuals, and total government expenditure (Bolzendahl, 2008, 2011; Kittilson, 2008). A Note on Critical Mass A focus on the number of women raises the possibility that there might be a critical mass effect, that a certain level of women’s political representation (usually 30%) has to be achieved before changes in politics and policy will follow. While many activists and organizations committed to women’s representation (for example see Equal Voice Canada; Global Database of Quotas for Women [GDQW], 2011; IDEA Network, 2005; United Nations, 1997) continue to focus on the necessity of a critical mass of 30%, the empirical research is at best ambiguous on this point.  Several studies find support for the existence of a critical mass effect (McDonald & O’Brien, 2010; Saint-Germain, 1989; Sawer, 2012; Thomas, 1991), while others observe a non-linear relationship or no relationship at all (Beckwith, 2007; Berkman and O’Connor, 1993; Bratton, 2005; Bratton and Ray, 2005; Celis, 2006; Crowley 2004; Riengold, 2000; Towns; 2003; Thomas and Welch, 1991; Thomas, 1991).  Most scholars have moved beyond the notion of a single critical mass at which change is observable (Celis and Childs, 2008; Childs, 2004, 2006; Childs and Krook, 2008; Dahlerup, 2006; Grey, 2002, 2006; Norris, 2001). As Dahlerup (1988) argued in her original contribution, identifying a critical turning point or a critical mass is not possible because, put simply, “politics is not physics’” (290).  13 Moreover there are two significant and related theoretical problems with the concept of critical mass. The first is that it ignores the process through which the individual women that collectively comprise the 30% come to be in politics and the second is that it does not theorize the incentives and opportunities they face and their strategic decision-making once they get there. As Childs and Krooks (2008) have argued: “despite the focus on actors in the critical mass theory, the broader context may limit or enhance opportunities for individuals to translate priorities into policy initiatives” (129).  Childs (2006) goes as far as to “give up” on critical mass because “it … abstracts women representatives from the context in which they act as if numbers were everything” (162).  This dissertation does not, therefore, use the concept of critical mass, but instead focuses on the micro-level of individual behaviour.  But this is not to say that numbers don’t matter – indeed, the literature in this research demonstrates that they do, but it does not require a so-called ‘turning’ or ‘tipping point’. As Bratton (2005) notes, “even if the distinctive behaviour and legislative success of female representatives do not depend on increasing the percentage of women within the legislature, an increase in female legislators who focus on policies relevant to women may change policy outcomes” (120; see also Curtin, 2006).  In other words, the absence of a critical mass or turning point does not preclude a relationship between descriptive and substantive representation.   It is possible, however, that moving away from the concept of critical mass as a significant theoretical component of research may actually help us better understand when a threshold might matter. Celis, Childs Kantola, and Krook (2008) argue “a mere increase in the numbers of women elected – a critical mass – does not always translate automatically into policy gains for women, given various constraints related to party affiliation, institutional norms, legislative  14 inexperience, and the external political environment” (102) and it is these factors that this dissertation focuses on.  In doing so, it may illuminate the conditions under which an increase in the number of women, or even a critical mass, results in substantive representation and when does it not.   Women’s Representation & Institutions: The State of the Literature Research on women’s political representation cannot entirely avoid institutional context. For example, the relationship between proportional representation and women’s descriptive representation is well known (Lijphart, 1991; IDEA, 2011; Krook, 2010; McIvor, 2003; Norris, 1997; Tremblay, 2006). Other analyses focus on how institutions are inherently gendered - systematically favouring typically masculine styles of behaviour and decision-making because women did not and were not expected to participate in politics at the time institutions were designed and adopted. (Carroll, 2001; Duerst-Lahti and Kelly; 1995; Duerst-Lahti, 2002, 2005; Hawkesworth, 2003; Kathlene, 1994, 1995; Kenny, 1996; Rosenthall, 2000; Sawer, 2004). Rosenthall (2000) argues that masculine behaviour of confrontational and aggressive politics has become conflated with institutional norms. Duerst-Lahti (2002) similarly claims that, to be effective, women have had to “perform” politics like their male colleagues. As a result “the masculine character then persist despite the female presence” (44).  In a few cases, the constraints and incentives imposed on individual women by institutions have been illuminated or used as an explanation of the behaviour of representatives (Childs, 2003; Gleb, 2002; Poggione, 2004; Rosenthall, 2002; Sawer, Tremblay, and Trimble, 2006; Schwindt-Bayer and Taylor-Robinson, 2005). In most cases, however, the institutional context remains implied or assumed (Berkman and O’Connor, 1997). Where institutional considerations are explicit, the research, with a few exceptions, is not comparative nor are the comparative  15 implications for the link between the descriptive and the substantive representation of women considered (Rosenthal, 2002; Sawer, 2012; Swers, 2009). Those that are comparative are limited to two-case comparisons (Franceschet, 2011; Gelb, 2002; Tremblay, 2003; Young, 2000) or are oversimplified. In Sawer, Tremblay and Trimble (2006), for example, a four-country analysis consists largely of individual chapters exploring a single case.  In a field of under-theorized and single-case studies, three comparative and institutionally attentive pieces stand out. First, in Feminists and Party Politics (2000), Young explores the partisan strategies of the women’s movements in Canada and the US and finds that the differing institutional contexts matter. In both countries, Young argues, the strategy can be described as a function of an interaction between the characteristics of the movement itself and the political opportunity structure. Canada’s political system is defined by high party discipline and unity imposed on its legislators while the American system is “characterized by the autonomy of individual legislators to pursue their policy interests or agendas with little constraint from party leaders or organizations” (Young, 2000, 185). In response, the women’s movement targets individual representatives in the United States while directing their efforts towards parties and party leadership in Canada. Young’s work focuses on the behaviour of and strategies adopted by the women’s movements, but her insights fundamentally say something about the incentives and opportunity structures facing women politicians. Pure presidential and Westminster Parliamentary systems generate divergent opportunities and incentives for women legislators to represent women’s interest and this is what the women’s movement is responding to.  Second, Gleb (2002) explores the differing political opportunity structures in the United Kingdom and the United Sates. Both countries experienced a notable increase in the number of women representatives during the 1990s; 1997 saw the election of ‘Blair’s Babes’ in the UK and  16 1992 was dubbed the Year of the Women in the US. The increase in descriptive representation in the UK was more significant and has been attributed in large part to powerful and centralized parties. The Labour Party in particular used all-women shortlists (AWS) to significantly increase the number of women in their caucus. After the election, however, little was observed in the way of an increase in substantive representation (see also Childs, 2002, 2004). In the United States, descriptive representation was harder to come by because candidate-centred elections made powerful incumbents harder to replace. The smaller increase in women’s representation in the U.S., however, may have been as good or better than the bigger change in the UK because “institutional realities of congressional seniority and relatively lax party control” (441) increased opportunities for the women elected to have a greater impact. While her insights are both unique and intuitive, Gleb’s analysis is limited in that it is based solely on existing literature, comparing and combining studies of women in the United States and women in the United Kingdom.  Third, Franceschet (2011) looks at two presidential systems - Argentina and Chile - and argues that institutional differences between the two countries lead to divergent representation of women by women. Formal rules in the two countries, namely the degree of legislative power exercised by the executive and the “presence and location of policy gatekeepers who control the legislative agenda” result in differences in the link between descriptive and substantive representation (Aleman in Franceschet, 69). In Chile, women introduce less legislation and report feeling a lack of power due to an inability to introduce bills that affect the national budget - a constraint. They are, however, more successful in terms of passing their legislation if they “are willing to ally with the executive women’s agency” (69), which can lobby the president and “help to move women’s rights bills through the committee process and on to chamber debates” (70) – an opportunity. On the other hand, Argentine legislators did not report feeling the  17 constraint that Chilean women did and were more active in the introduction of gender rights because the executive has less dominance in the area of agenda-setting - an opportunity for women in the legislature. The president and leaders of legislative blocks, however, control the extent to which bills come before committees and thus there are constraints on the ability of women to affect final policy outcomes. In other words, in Argentina descriptive representation is more closely tied to substantive representation as policy outcomes rather than behaviour, while the opposite is true in Chile. There are also several of the studies that use quantitative comparative analyses to explore the relationship between the number of women and policy outcomes have incorporated institutional variables. Bolzendahl (2011) includes an indicator for the number of veto points, including federalism, presidential system, bicameralism, and the strength of the judiciary. Kittilson’s (2008) findings demonstrate that federalism hinders the expansion of maternity benefits but not childcare leave, a proportional electoral system reduces the benefits associated with childcare leave, but has little effect on maternity benefits or length of childcare leave. These analyses remain limited. Bolzendahl incorporates institutions as an alternative explanation rather than an important context that can increase or decrease the impact of women on policy outcomes. Kittilson argues that the “opportunity structure” needs to be considered because “political institutions mediate the relationship between women’s parliamentary presence and policy outcomes” (326). The institutional indicators are not, however, interacted with the representation of women and so the results do not, in fact, illuminate whether there is a mediated effect between the number of women and the policy outcome.   18 Clues in the Gaps: Where do we Look for Substantive Representation? Exceptions aside, a lack of comparative analysis plagues the study of descriptive and substantive representation. Interestingly, this is a function of scholars recognizing, if sometimes implicitly, that institutional variation can make a direct comparison misguided. In the absence of explicit comparative research, preliminary insights can be gained by comparing single or two-case studies. In fact, a simple comparison of where scholars look for evidence of substantive representation provides some clues to institutional differences in substantive representation. Table 1.1 takes a survey of the literature, grouping research by the division of powers and form of substantive representation. It is important to note, as already stated, the literature (at least in English), is extremely Anglo-centric; most, though not all, of the work studies Canada, the United States, and the United Kingdom. Please see Table 1.1 at the end of this Chapter. The first point worth noticing is that nearly half (48%) of the studies on parliamentary systems explore gendered differences in preferences, priorities, or thoughts on representation. The same is true for only 16% of studies looking at presidential systems. Poggione (2004) argues that looking at opinions to assess the effect of women’s descriptive representation in the United States is more appropriate because “the policy impact of women legislators is mediated by legislative institutions and women’s positions within them” (313). As a result, votes, proposed legislation, and policy outcomes are not reliable indicators of gender differences. It is not that Poggione is wrong – indeed, the representation of women in the United States may be mediated by women’s seniority, re-election prospects, and their responsibility to their district or state. What is clear, however, is that this is a much greater concern in parliamentary systems. The second important difference is the number of studies assessing substantive representation by looking at public behaviour, including the content of speeches, votes, and introduction of  19 legislation. These studies comprise 33% of those looking at parliamentary systems and a full 65% of those examining presidential systems. The differences are even more significant when broken down by behaviour. Of the studies on presidential systems, 24% look at the substantive representation through roll call and another 38% (a plurality) assess the introduction of legislation. On the other hand, the vast majority of the 33% of studies examining behaviour in parliamentary systems looking at discourse – what is said and by whom, including via early day motions (Childs and Withey, 2004) and oral questions (Bird, 2005, Erzeel, 2012).  Of the two studies that look at voting by representatives in parliamentary systems, one relies on surveys rather than roll call data (Skjeie, 2012) and the other is an attempt to explain why the new Labour women elected in the 1997 British election did not vote to support women’s interests (Cowley and Childs, 2003). The failure to look for substantive representation in the votes cast by MPs is not an oversight, but rather an acknowledgment of the institutional constraints. Tremblay (2006) observes that “it is not possible to evaluate a representative’s commitment to articulating women’s various interests based on her voting record” in parliamentary systems such as Canada and Australia (127; see also Tremblay, 1998). Similarly, Lovenduski and Norris (2003) argue that voting records are not an appropriate tool for exploring preference differences between men and women in the United Kingdom; the high party discipline in the Westminster system is likely to overstate agreement between men and women. Roll call analysis, they argue, is more appropriate to the study of substantive representation and preference differences in the US Congress (see also Dolan, 1997; Swers, 1998). While this is passed over as a methodological question, it is, in fact, an observation on different incentives and opportunities.  20 Finally, few studies of either parliamentary or presidential systems look at the relationship between women’s descriptive representation and policy outcomes, just 11% and 6% respectively. This is somewhat surprising given that Dahlerup’s (1989) original contribution to the study of women elevated policy change as the most important component of substantive representation. Most of the research that looks at the introduction of legislation by women in presidential systems also assesses the success of women in passing legislation and finds that they are more likely to be successful (Berkman and O’Connor, 1993; Bratton and Haynie, 1999; Crowley, 2004 Saint-Germain, 1989). This is different, however, than exploring the relationship between the number of women and policy outcomes in an aggregate sense. Moreover, instead of focusing on the proportion of women over all, Berkman and O’Connor (1993) focus on the over-representation of women on committees dealing with abortion legislation and find this is one way women exert some influence over the final policy outcome. Similarly, Atchison and Down (2009), test the relationship between policy outcomes and the proportion of women in cabinet rather than the parliamentary legislatures. The limited and less direct attention to policy outcomes may indicate that it is not only the most important, but it may also be the most challenging aspect of substantive representation.  Collectively, these single country studies offer an indication that substantive representation may vary across political institutions. What is needed is an explicitly comparative and institutional approach that uses existing literature to theorize when and how women should be expected to act for women. Women as political actors with political objectives operating within and navigating through institutions need to be at the centre of the analysis. To do this, I outline the career paths and motivations of women and men in politics and the ways in which these are inherently  21 institutionally embedded as well as the ways in which they lead to gendered differences in the representation of women’s issues. From Self-Selection to Policy Outcomes: Substantive Representation & the Career Paths of Women Politicians  Women are, in a very simple sense, central to the study of the relationship between descriptive and substantive representation – the question in its most basic form is whether women do or do not act in a way to represent women. Women themselves are, however, also notably absent. Few studies consider the career path of politically ambitious women, take into account their agency, or understand women politicians as strategic actors. In other words, little thought is given to how women come to be in politics and the process through which their presence leads to the substantive representation of women. A well-theorized investigation which includes a conceptualization of the policy-making process, on a macro-level, and an understanding of the ways in which women themselves come to participate in politics and why they do what they do, on a micro-level, addresses this problem. This approach does not take women in politics as a given but rather asks how and why they come to be a part of the political process. Doing so draws explicit attention to how and why women affect political discourse, expand the political agenda, or affect policy outcomes, rather than simply testing whether they do or do not.  In her study of the ways in which the electoral geography affects political representation of the poor, Karen Jusko (2014) outlines five simplified steps in the democratic policy-making process: Stage 1. Preference Formation. Individuals form preferences on one or more policy dimensions.  Stage 2. Mobilization. Parties and candidates opt into the political process, and the electorate turns out and supports the party or candidate of their choice.   22 Stage 3. Elections. Votes are tallied, and seats are allocated according to the current electoral rules. Stage 4. Government Formation. Governments are formed, usually by the party that won a plurality of seats and potential allies.  Stage 5. Legislation. Governments implement policy more or less in accordance with the preferences of those they represent.  Jusko argues that differences in anti-poverty policies may be attributed to any one of these stages, though her work focuses particularly on stage 3. I build on her approach here, by focusing on the individuals at each stage. I incorporate the micro-level experiences and actions of women at each of these stages and draw attention to the ways in which these stages are fundamentally gendered. This place women at the centre of the analysis and explores why they behave in particular ways and not others and why they pursue some objectives over others.   In my adaptation, five stages of career development are bookended by two stages, Jusko’s first stage of preference formation, which sets the context and a final stage, Jusko’s fifth stage of legislation becomes stage six, where the behaviour and preferences of legislators are aggregated into outcomes. These stages are fundamentally gendered. I have broken down stage 2 into two components – self-selection and successful nomination – and added a stage at which individuals make decisions about who and what to represent and how.  Stage 1: Preference Formation. Individuals form preferences on one or more policy dimensions. Preferences include those on gendered issues and are expected to differ between men and women as a result of the unequal division of paid and unpaid labour and inequalities in access to power - economic, social, and political.  Stage 2a. Self-selection. Individuals opt into the political process, choosing to run for nomination. Self-selection is more likely when parties or elites ask individuals to run. Women are less likely to self-select into politics than men and less likely to be actively recruited to do so by other political actors (Fox and Lawless, 2004; 2010).  Stage 2b. Nomination. After women self-select, they must win the nomination for their chosen party. Research has shown that women often face more competition in primary  23 races and so “have to be better in order to fare equally well” (Lawless and Pearson, 2008). Stage 3. Elections. Once a candidate, women must win election to office. In some contexts, once candidates, women are less likely to be elected than their male counterpart (Stambough and O’Regan, Studlar and Matland, 1999). In other instances, electoral quotas that require a minimum representation or other, less formal, affirmative action strategies result in distinct pathways to candidacy and perhaps election for women. Stage 4. Representation and Agenda Setting. After election, women make choices on what issues to represent with scarce resources, including political capital and time. Stage 5. Promotion. Women may seek and be offered promotions or other particularistic benefits, including cabinet and committee positions, or party leadership positions. In an effort to achieve greater equality, women have, in some cases, had greater success at promotion to higher posts than men. In 2015, of the countries in which women make up more than 40% of ministers, all have more women in the executive than they do in the legislature, meaning that a greater proportion of women are promoted to cabinet than are their male colleagues.  Stage 6. Legislation and Policy outcomes. The agenda setting of individual legislators and the votes they cast is aggregated into policy and legislative outcomes.   These steps depict the process through which women come to be in politics and may substantively represent women. It moves away from a simple approach that takes women in office as a given starting point. These career stages allow for an explanation of the steps women must navigate before they achieve office and what they must do to obtain promotion or realize other desired outcomes once elected. It gives due credit to women’s political agency and draws attention to the incentives and opportunities that women have to represent women’s issues. The approach proceeds from the assumption that women are strategic political actors (Gertzog, 2002). I assume that women in politics have three distinct motivations (Muller and Strøm, 1997; Strøm, 1999).  1) Women are vote seeking: They seek votes first for nomination, and then election.  2) Women are office seeking: Women seek the benefits of office, including more rewarding and powerful appointments to cabinets or committees.   24 3) Women are policy seeking: They pursue policy, including policy addressing women’s issues.   Women, like their male colleagues, act strategically to secure nomination, election, promotion, and re-election (objectives one and two) as goods in and of themselves and as necessary conditions for achieving their policy objectives (objective three). Importantly, this approach is explicitly institutionalist and gendered. Political institutions establish how representatives can or must pursue their political objectives and structure all five stages of representatives’ career paths. Electoral institutions determine how one seeks and wins nomination, what one must do to get elected, whom politicians represent, and to whom they are accountable (Stage 2b and 3 and Motivations 1 and 2). To the extent that electoral institutions structure the likelihood of winning nomination and election, they will also affect the stage of self-selection. The rules of the institution establish who can do what and when, and thus affect the choices women make to represent particular issues over others and how they do so (Stage 4). Stage 5, promotions and other benefits, is similarly a product of the institutional context; institutions determine what rewards are available, the comparative worth they hold, and who has the power to dole them out. By determining how individual behaviour, such as voting, is aggregated into collective decisions on policy and legislation, institutions provide the rules for the final stage and affect how representatives pursue their policy objectives.  At the same time, gendered differences affect the resources prospective legislators have to pursue their vote-seeking and office-seeking objectives and the systemic barriers or opportunities they face while navigating the stages of the career path. Moreover, gendered differences in lived experiences manifest at stage 1 of preference formation and affect the content of the policy-seeking objectives.  25 In short, as representatives progress through the gendered stages of a political career in pursuit of their objectives, they must navigate institutions that incentivize some behaviour over others, enable some representational activities while explicitly prohibiting others, and require certain actions before or at the expense of others.  Using a Principal-Agent Framework Across and within each of the steps of a political career, legislators must navigate a multiplicity of principal-agent relationships determined by the institutional context. A basic principal-agent model of representative democracy takes representatives to be the agents of a larger body of citizens (the principal), addressing their issues and acting on their behalf (Mezy, 2008). The principal authorizes the agent to do this by electing them and holding them accountable through re-election. In between this, the agent decides what to do, what to prioritize, and how to act. While Mezy argues that constituency control is at the heart of principal-agent approach to representation (25), the relationship can be much more complicated than that.  The evidence from the existing literature establishes that women often think of themselves as representatives of women; in other words, women representatives are, or at least can be, the agents of women. The presence of multiple principals in conflict is, however, not only possible but also probable. High party discipline and an obligation to vote with their party to pursue promotion or the passage of legislation, whatever the views of their constituents, is a clear indication that sometimes parties are the primary principals. Local party and constituency organizations also act as additional principals when they control nomination. An elected president operating alongside legislative party leadership can double the party principals to which legislators must respond (Samuels and Shugart, 2010). If large donations or campaign  26 assistance made election possible, organizations such as unions may serve as a competing principal.  Put simply, the extent to which women representatives are able to act on behalf of women as an agent depends on these other principals to which they have an obligation. Political resources are scarce resources and representatives must make choices about which principal to prioritize and which issues and interests to pursue. Where the tie to another principal is particularly strong or there are several competing principals, there will be limited space, time, and resources to dedicate to women’s representation. When parties are functionally the primary principal for which representatives as agents are responsible, women will be unable to pursue interests independently and, as a result, the behaviour of women and men will be very similar. On the other hand, if women have greater independence from their political parties, women will take the opportunity to do more to represent women. Such ability to depart from a party line can also create incentives if women legislators are able to establish themselves as a principal for women and differentiate themselves from their competitors on this basis to cultivate electoral support. Party leaders remain predominantly male, but this is neither assumed nor necessary. Women, perhaps most notably Margaret Thatcher and Angela Merkel, have ascended to leadership positions. The ‘party’, however, includes not just the formal leader, but also those behind the scenes, those in party, but not legislative or executive, leadership positions and other powerful members. Moreover, women who are successful in leadership are likely to reflect the culture of the party; when parties do not focus on women’s issues, women who prioritize these issues are unlikely to ascend to leadership.   27 An Alternative to Personal Votes: Unincorporated Representation  That institutions structure principal-agent relationships and thus affect behaviour is no way a new idea. As Carey (2009) argues, “institutional factors shape whether, and to what degree, legislators are also subject to pressures from other principals whose demands may conflict with those of party leaders” (92). What is new is that this has implications not just for the trade-off between the representation of constituency or particularistic interests, on the one hand, and party interests, on the other, but also for the representation of women’s interests. While this research focuses on gender, the argument is much broader and includes a range of issues that are not fully incorporated into the party system – what I call “unincorporated representation”.  To be clear, the notion of ‘unincorporated representation’ is a direct parallel to Carey and Shugart’s (1995) ‘personal representation’. When individual candidates do not or cannot rely on their party’s success for election and re-election, they have incentives to cultivate personal support via, for example, pork barrelling or celebrity status. One can, however, also conceive of more general incentives for candidates and legislators to represent interests or issues other than that which their party dictates – what I term ‘incentives for unincorporated representation’ (IUR). Legislators may seek personal support, not by fostering traditional clientelistic relationships, but by distinguishing themselves as an agent for some other issue or identity not already absorbed into the party system.  The idea of incorporated issues is analogous to Lijphart’s (2012) ‘issue dimensions’, which include socioeconomic issues, religious and linguistic, cultural-ethnic, urban-rural, as well as support for the regime, foreign policy, and postmaterialist issues (76). Which issues are ‘incorporated’ vary significantly from one context to another and depend on the number and extent of social cleavages and the political institutions. For example, the socioeconomic  28 dimension is critical in the United Kingdom and New Zealand, while religious and linguistic dimensions are salient in Switzerland and Belgium. Moreover, party systems in some countries are defined by one dimension, while others include multiple dimensions. Lipset and Rokkan (1967) explain the party system using a primarily sociological approach. Whether a new interest is incorporated into a political system, they argue, depends on the answer to the following question: “must the new movement join larger and older movements to ensure access to representative organs or can it gain representation on its own?” (27). In other words, the relative cost and benefits of mergers, alliances, coalitions, or going at it alone are key determinants of the translation of cleavages into the party system. It is the “intensity of the inherited hostilities and the openness of cleavages” (32) that, according to Lipset and Rokkan, determine the possibility of alliances.  An institutional perspective can be incorporated into Lipset and Rokkan’s argument, since electoral institutions provide the context within which such mergers, alliances, competitive ‘old’ parties, and the possibility a new party will exist. Cox and Neto (1997) hypothesize that the effective number of parties can be explained by an interaction between institutional structure and the number of social cleavages. The logic is extended in Cox’s (1997) Making Votes Count, where he argues that increasing permissiveness of an electoral system will not increase the number of parties unless the system had been acting to artificially limit the number of parties.  In other words, unless there are a number of social cleavages for parties to exploit in the new electoral system, a change to a more permissive electoral system will not necessarily lead to a greater number of parties. Taagepera and Shugart made a similar argument a decade earlier – when there are more issues, more parties should exist to represent them. On the other hand, if the electoral system suppresses the number of parties, then the number of issues might also decrease  29 and one party may come to represent multiple interests from multiple issue dimensions: “one party may come to represent low-income and racial minority and urban interests, while another may present a package of high income and racial majority and rural interests” (66).3 The point here, however, is that the issue dimensions in the party system do not represent all the possible issue dimensions that could be incorporated into the party system. It is true that the electoral system may suppress the number of parties and, potentially, the number of issues incorporated, but there are additional factors that determine which issues remain unincorporated. First, if some groups that have been historically and systematically excluded from the political process, their interests may be what Mansbridge (1999) calls “uncrystallized in that they “have not been on the political agenda for long, candidates have not taken public positions on them, and political parties are not organized around them” (643). Second, many groups that have been formally integrated into the political system continue to face ongoing systematic and structural barriers to political participation, including resources, expertise, networks, and more. As such they are neither able to join nor influence an existing movement in any meaningful way, nor are they able to create a new political party however permissive the electoral system. Third, to the extent that membership in a group whose interests are not incorporated cuts across other social cleavages that are established and incorporated into the party system, opportunities to organize for incorporation is likely to be extremely limited.  In short, the dimensions that are incorporated do not represent the universe of potential issues that could be incorporated into the party system. Historic and present inequality and the entrenchment of a party system based on other issues means that not all issues will be                                                  3 There is a very large and established body of literature on this topic and most are not cited here including, for example, Taagepera and Grofman (1985). Others focus on how the issues incorporated into the party system play a critical role in determining the adoption of the electoral system (Boix, 1999; Cusack, Iversen and Soskice, 2007).   30 incorporated, regardless of the electoral system. Of those issues that are unincorporated, not all will depend on descriptive representation. Because the historical exclusion and ongoing systemic barriers to participation are often based on personal characteristics rather than ideas or ideology, many unincorporated interests are be related to identity politics and descriptive representation.  Mansbridge (1999) argues descriptive representation is important under conditions of inter-group distrust and where political interests are uncrystallized. The general approach is consistent with Mansbridge’s argument that descriptive representation can and does matter in certain circumstances, but it moves away from normative argumentation. It shifts the focus from a question of the contexts within which blacks should represent blacks or women should represent women and toward a discussion of their capacity for substantive representation and the institutional context in which they are more or less able to do so.  In the words of John Stuart Mill, the ‘natural defenders’ are necessary if unincorporated issues, including women’s issues, are not to be overlooked. But the presence of the ‘natural defenders’ is not sufficient – they must also have the capacity and opportunities to act. When institutions structure strong parties and representatives must act first and foremost as their agent, their behaviour will reflect the preferences and priorities of their party. When institutions facilitate what Carey and Shugart term ‘personal’ representation, representatives may instead focus on issues outside the party system and sell themselves as an agent for unincorporated interest.  Women’s Issues as Unincorporated Issues  What makes women’s issues unincorporated is that parties are neither organized around them nor have parties taken consistent and opposing positions on gendered issues: at elections, voters do not face alternative and opposing menus on gendered issues from which to choose. For example,  31 while they are politically salient issues, Mansbridge observes that parties have not taken distinctive and opposing views on women’s interests like sexual harassment and men’s violence against women. Mansbridge’s focus is on the United States, but the observation holds true in other political contexts. As a result, it is women who are “more likely…to act as their descriptive constituents would like them to act” (207). Even in the case of abortion, there are Republican equivalents to the Democratic EMILY’s List, like Wish List, which raises money and funds pro-choice candidates. It is true that Wish List is smaller and less influential than EMILY’s List, and it is easier to find a pro-choice Democratic woman than a pro-choice Republican woman. But the more central point is that there are pro-choice Republican women and Republican organizations in a way that there are not socialist Republicans or Republican organizations.  There are several reasons women’s issues remain unincorporated. First, the exclusion of women from the political process has been near universal until relatively recently and underrepresentation remains a worldwide phenomenon. Women and women’s issues were relegated to the private sphere until the expansion of the franchise and so views on them are “developing in a relatively ad hoc way to meet a rapidly evolving situation” (Mansbridge, 1999, 646). Second, parties, legislatures, and positions of leadership continue to be dominated by men. Third, even after women themselves have been incorporated, incorporating gendered issues remains extremely difficult because of the ways in which they cut across existing cleavages. Women are not geographically concentrated, they are racially, religiously, and ethnically diverse, and, while women are more vulnerable to poverty, women and women’s issues also span class divides. In the absence of incorporation, the representation of women relies on a form of personal representation. Shugart (1994) himself suggests that the same institutions that facilitate personal  32 representation might be relevant to the political representation of women and minorities. While he does not use the terms and conflates the concepts, his argument is that, by facilitating women’s substantive representation, opportunities for personal votes should increase women’s descriptive representation: “The ability of members within a party to cultivate personal ties might facilitate the representation of women and minorities because, for example, a female parliamentarian can seek a personal tie to women within the district by devoting special attention to their concerns.” (38). Once the personal tie is established, Shugart argues, women voters will support women and descriptive representation will increase. The concept of unincorporated interests turns Shugart’s intuition into a clear alternative to ‘personal’ votes.  One important distinction between personal representation and unincorporated representation is worth highlighting. Women are expected to want to represent women more than men. I have argued that, because of differences at the preference formation stage, the content of policy-seeking objectives of men and women are expected to differ on gendered issues. In other words, unincorporated representation is not simply about vote or office seeking, but rather includes an element of policy-seeking. Purely personal or particularistic representation, on the other hand, is fundamentally vote and office seeking behaviour. It is true that a representative may engage in personal representation to achieve votes and office to pursue some other policy objective, and the actions undertaken are not done as ends in and of themselves. Unincorporated representation, on the other hand, may be both at the same time; in systems that allow it, women may address women’s issues as a way to increase their personal votes and chances at electoral success and to pursue raise women’s issues to the agenda as a good in and of itself.  While unincorporated, women’s issues are still on a dimension on which there are conflicting positions. As a result, parties, candidates, and legislators do sometimes take positions on the  33 issue. The concept of unincorporated representation does not require that parties always or even often act in opposition to women’s interests. Nor does it require that all parties oppose and support women’s issues equally. Clearly, parties do differ and some ideologies, namely those on the left, are more compatible with some gendered issues, including anti-poverty policies, decommodification, proactive government strategies to address inequality.  To be clear, it is not that party does not matter - opportunities for ideological congruence are greater for parties on the left. Individuals self-select into parties because they have values that align with the party’s policy positions and we should expect differences across the political spectrum in how often legislators (both men and women) act to represent women. To be explicit: women on the left are expected to do more to represent women than women on the right. But gender matters and women on the left, centre, and right will all want to do more than their party encourages its members to do. This is because important gendered differences exist within parties and can surpass partisan differences (Dolan 1997; Gidengil et al., 2003; Norris 2001; Poggione 2004; Vega and Firestone 1995). For example, women on the right report that representing women is more important to them than do men on the left (see Chapter 3 for more). As such, given the chance, when the bonds to their party are loosened, women will do more to represent women whatever part of the political spectrum they are aligned.  Institutions and The Representation of Women – An Interactive Effect “Behaviour occurs in the context of institutions and can only be so understood.”   - Immergut, 1998, 6. The argument made here is that the effects of gender and institutions are interactive, not additive. Institutions are not concurrent or alternative influences on substantive representation, they are the context within which representation is embedded and as such interact with the gender of  34 representatives. To make this point clear, it is worth framing the effect of institutions as incentives/disincentives on the one hand and opportunities on the other.  Incentives Institutions structure the incentives facing political actors by structuring how one is nominated and elected, to whom one is accountable (stages 2 and 3), and how promotion or power is achieved (stage 5). In doing so, they influence self-selection and shape behaviour and representation (stages 1 and 4). Institutions may incentivize the representation of women by facilitating a principal-agent relationship between women and their legislators or may disincentivize it by requiring legislators to follow the direction of party leadership as a primary principal.   These incentives for substantive representation still rely on descriptive representation. Because of gendered lived experiences, women in politics are more likely than men to see and respond to gendered issues in politics while men are expected to respond to the same institutional incentives by addressing other unincorporated interests. Further, the incentives to represent women are less for men simply because they are men.  Voters often use the personal characteristics of candidates as informational shortcuts and voters may attribute a greater willingness and capacity to represent women’s issues to women in politics (Matland, 1994; Matland and King, 2003; Sanbonmatsu and Dolan, 2009). As a result, even if men do respond to incentives for unincorporated representation by focusing on women’s issues as candidates or once in office, the benefits will be smaller.  Opportunities While incentives induce certain types of behaviours and disincentives reduce behaviour, opportunities allow actions and constraints prevent actions. Put differently, opportunities and  35 constraints are not about vote-seeking or office-seeking but are strictly about policy-seeking and acts of representation. Opportunities are important because institutions vary in the degree to which they encourage strategic behaviour or are strategy-proof, meaning that not all behaviour can be reduced to strategy. The unequal division of paid and unpaid work, prevalence of gender-based violence, and inequalities in access to political and economic power lead to differences at the first stage of preference formation. As a result, there are differences in men and women’s policy-seeking objectives - women are expected more than men to want to add women’s issues to the political agenda, increase the attention they receive, and pursue policy on these issues. In short, regardless of existing (dis)incentives, women are expected to want to represent women more than men and institutions determine the opportunities they have to do so.  In short, the anticipated difference in the substantive representation of women by women and that by men as moderated by the political institution reflects the divergent ways in which legislators may cultivate personal support. When institutions make space for representatives to engage in unincorporated representation, women are expected to do so (at least in potentially or in part) by representing women. Men, on the other hand, will respond to the same institutions by focusing on other unincorporated interests or perhaps by fostering more traditional clientelistic relationships.  As a result, the effect of institutions should be visible in two ways: 1) The behaviour of women will differ across institutions. Women in institutions that incentivize and enable unincorporated representation will do more to represent women than women in institutions where unincorporated representation is constrained or actively disincentivized.  36 2)  The difference between the representation of women by women and the representation of women by men will vary across institutions. The gap between the action undertaken by women and that taken by men will be larger in institutions that incentivize and enable unincorporated representation; where they are disincentivized or constrained, the behaviour of women and men will be more similar.  Outcomes The focus thus far has been on individual legislators and how institutions determine what they can, should, or must do. But legislative outputs are an important component of substantive representation. Looking at legislation moves from the micro-level of individual behaviour to the macro-level where preferences and behaviours are aggregated to reach legislative, policy, and budgetary decisions.  It reflects the difference between what Childs and Krook (2006) call the feminization of the political agenda, where individual legislators raise women’s issues and concerns, and feminization of legislation, where the actual political output is changed. Institutions affect outcomes in two distinct ways. First, the impact of institutions on individual behaviour has clear implications for the aggregate. Where the gap between women and men is small, increasing the number of women will have a more limited impact.  This is not to say that women won’t matter – even under conditions of a small gap in the behaviour of men and women, the increased presence of women will shift discourse in a woman-friendly direction. When the difference is large, however, and women do much more than men to represent women, increasing the number of women (and by default decreasing the number of men) will result in a much more noticeable change in the substantive representation in votes, speeches, and legislative proposals. In other words, all else equal, institutions that facilitate and incentivize the representation of women’s issues will lead to outcomes more attentive to women’s needs.   37 All else, however, is not equal and speaking on women’s issues or proposing or voting in favour of legislation does not necessarily lead to congruent outcomes. The second way in which institutions matter is the translation of individual behaviour into outcomes. As the “rules used to aggregate the preferences of individuals in the system into a choice of a policy” (Hammond and Butler, 2003, 147), institutions moderate the effect that the aggregate number of women has on policy outcomes (stage 5) at a given level of behavioural representation (stage 4). Institutions, for example, affect how easy or hard it is to depart from the status quo (is a simple majority sufficient?) and the feasibility of reaching policy decisions in the first place (how many individuals or groups must agree?) (Besley and Case, 2003; Carey, 2000; Haggard and McCubbins, 2001; Hammond and Butler, 2003; Tsebelis, 1995). If institutions make policy changes difficult, an increase in women’s representation may not be accompanied by an increase in policy attention to women’s issues, not because descriptive and substantive representation are not related but because institutions mute the potential effect.  Looking Forward: A Road Map The remainder of this dissertation develops and tests a theory of moderating institutional effects on the relationship between women’s descriptive and substantive representation.  It explores the relationship between gender and the behaviour of individual representatives, on the one hand, and between the percentage of women in a legislature and policy outcomes, on the other. Chapter 2 uses the basic argument and framework to theorize about the moderating effects of a division of powers and the features of the electoral system. In particular, it theorizes how division of powers systems, proportional representation systems, and features of the electoral system which incentivize unincorporated representation will all lead to greater substantive representation by women, but not by men.  Conversely, parliamentary systems, a plurality formula, and features of  38 the electoral system that disincentivize unincorporated representation are theorized to suppress substantive representation by women. The anticipated effect on outcomes as the aggregation of individual behaviour is considered but is much less straightforward. Finally, the Chapter outlines a potential contingent relationship between the two institutions wherein, under the conditions of an SMP system, a division of powers is a necessary but insufficient condition for a ‘personal’ vote based electoral system. Chapters 3 and 4 then test the impact of these institutions on individual behaviour using quantitative methods.  Chapter 3 finds that the gap between the substantive representation of women by women and that by men is much larger under the condition of a division of powers than when powers are fused.  Chapter 4 finds no evidence for a difference between PR and SMP systems but does substantiate the theory that features that incentivize unincorporated representation increase the amount that women, but not men, do to represent women.  Chapters 5 and 6 use qualitative methods to test the causal process.  Interviews and surveys with more than 90 representatives and staff in six countries explore how women experience and navigate institutions.  Chapter 5 establishes commonalities and shared sentiments that extend across countries, institutions, and parties. The findings presented in Chapter 6 demonstrate how these common sentiments and perspectives filter through the institutional context that leads to more or less action depending on the incentives and opportunities present. The results observed in Chapters 3 and 4 are confirmed; a division of powers and incentives for unincorporated representation magnify substantive representation of women by women. The null finding on a difference between PR and SMP is similarly confirmed; the theorized mechanisms do not operate as expected.   39 Chapter 7, the last substantive chapter, looks at policy outcomes.  To do so, I use two unique datasets, both of which combine data from a variety of sources, including the Quality of Government dataset, the OECD, and the Comparative Agendas Project. The results indicate that women’s overall representation increases the likelihood that a bill will address strategic gender issues, like family law or gender discrimination, but this is especially, and in some cases only, true under the conditions of a division of powers or an electoral system which incentivizes unincorporated representation.  A division of powers does not have the same moderating effect on spending on family expenditure or parental leave wage replacement, indicating the difficulty in aggregating individual behaviour into departures from the status quo in these systems.  On the other hand, electoral systems that amplify the amount women do to represent women are also found to increase the overall impact of women’s representation on the same outcomes. Moreover, these electoral institutions enhance the impact of the number of women in cabinet in parliamentary systems.  The dissertation concludes in Chapter 8 with a review of the chapters and their key results. It also spells out several implications for the study of women in politics, the study of political institutions, and for advocates and activists concerned with women’s representation.  A Note on Methods and Sources The research in this dissertation uses a mixed methodological approach. Multiple quantitative analyses tests for the size and importance of the relationship theorized and each offer significant for the theory: the behaviour of individuals and the gap in substantive representation between women and men varies with the institutional context.  The qualitative analysis, however, targets the mechanisms behind these differences and explains why women do more and men do less to represent women in some institutions than others. This section introduces briefly the data used  40 and the advantages of the combination of qualitative and quantitative analysis. Because of the number of datasets, varied sources, and the mixed-methods approach, the in-depth methodological discussion is contained within the Chapters. Two datasets are used in Chapters 3 and 4 to test the importance and magnitude of the affect of gender on individual behaviour and the modifying affect of institutions. The first, the PARTIREP comparative MP survey, conducted between spring 2009 and summer 2010, reached representatives in 15 national and 58 regional parliaments. The survey, based on written questionnaires, captures self-reported behaviour and measures the extent to which representatives speak to or make proposals on women’s interests, collaborate with others to pursue these issues, and meet with women’s organizations.  The dataset also contains a wide range of institutional variables based on publicly available sources.  The data allow for a test of the moderating affect of both electoral systems and the division or fusion of powers.  The test of the division of powers, however, is limit because few cases of a division of powers are included and none is a pure presidential system. To compensate, I also use a second dataset based on just three cases, but includes a pure presidential system (the U.S.) and two Westminster Parliamentary systems (Canada and the U.K.) and so allows for a more explicit test of the theory. This dataset includes content analysis of the tweets of almost all representatives in these institutions over a two-year period.  The data is unique and was compiled especially for this research. Neither the Twitter analysis or the PARTIREP data offer a complete test of the theory, but taken together they support the theory.  The theory offered in this Chapter and expanded in Chapter 2, however, is not about mechanical institutional affects, but about the incentives and opportunities institutions generate for representatives. A qualitative analysis based on over 90 interviews and surveys with women  41 from 29 parties in seven countries complements the quantitative analysis by targeting of the causal process. The interviews allow for an assessment of whether legislators experience the institutional incentives and opportunities in the way theorized, by directly posing the question to the actors themselves.  It gives women legislators the chance to articulate how they can go about representing women and whether they think it is in their interest to do so.   Interviews in different institutional contexts also test for, rather than assert, causal homogeneity (Bennett and George, 2005). The opportunities created by multiple votes or the constraints of a fusion of powers may be experienced by representatives to a greater or lesser degree depending on the other features of the institutional, political, or cultural context. Indeed, the interviews enable a deeper understanding of the role of institutions and, in doing so, contribute to a refining and further development of the theory.  The qualitative work also informs the final quantitative test.  Chapter 7 returns to quantitative analysis to test the extent to which women’s representation in the aggregate, that is, the proportion of representatives who are women, affects policy outcomes.  The insights from the interviews and surveys presented in Chapter 5 allow me to operationalize ‘women’s policy interests’ in a way that accurately reflects the issues that women representatives report are important to them. Two datasets are used, both of which are unique compilation of existing data from a variety of sources.  The first used to test the moderating affect of a division or fusion of powers is based primarily on the Comparative Agendas Project. It includes information on the introduction and passage of legislation in six countries (Canada, France, Spain, Switzerland, the United Kingdom, and the U.S.) and is supplemented with data on women’s representation. The analysis tests the extent to which the presence of women is associated with an increase in the likelihood that a law will address civil rights, gender discrimination, social welfare, family legal  42 issues, and strategic gendered issues. The analysis is somewhat limited because of different data collection practices in the seven countries included, but as a qualitative discussion of the data demonstrates, the problems with the data offer additional support for the theory offered.   The final dataset uses a number of sources contained in the Quality of Government data supplemented with private data on women’s representation in legislatures from Pippa Norris.  The cross-sectional time series data contains information on OCED countries over a 20-year time period and allows for a test of the effect of institutions, and women’s representation on spending on family allowances, social exclusion, and parental leave. Additional data on women in cabinets in parliamentary systems provided personally by Amy Atchison enables a test of the effect of women in executives on outcomes and the moderating effect of electoral institutions.  Conclusions This introductory chapter sets the stage for this dissertation by outlining a significant lacuna between the research on women’s political representation and that on political institutions and the opportunities for bridging the gap. Moreover, it describes stages of political career and provides a framework for thinking through the anticipated effect of institutions on behaviour and outcomes. Taking institutions into account may bring clarity to some of the ambiguities observed in the literature. During the 1990s, for example, when both the United States and the United Kingdom saw a notable increase in women’s representation, the former also saw a increase in attention to and action on women’s issues while the latter saw little achieved by the newcomers (Gelb, 2002). What explains this difference? Many studies find women representatives prioritize issues important to women and express a desire to represent women, but many others report a gap  43 between the reported feminist or pro-women attitude and actual behaviour (Carroll, 1984; Reingold, 2000; Tremblay, 1998; Trimble, 2006; Wängnerud, 2000). Why do women representatives’ behaviour sometimes fail to match their expressed interests and intentions?  In addition to clarifying the existing literature, a focus on institutions enables a comparative approach and ensures a consideration of women as political actors. How does substantive representation of women and its tie to descriptive representation vary across countries? How do women navigate institutional structures and what must they do to become representatives? Once they are legislators, how can women represent women? Attention to these questions adds depth to the theory underlying the proposed tie between descriptive and substantive representation and brings attention to the mechanisms linking numbers to representation.  44                                                  4 Find “little at all difference” between men and women except for on issues of women’s advancement. Since this is exactly the type of issues this research is concerned about, I do not consider this a null finding. 5 Discusses why the new women were the least likely to rebel.  6 Vague discussion of whether they would make a difference.  * - no difference observed, + Uses surveys or interviews. Table 1.1. Survey of the Literature   Preferences and Priorities Discourse and Behaviour Policy Outcomes   Preferences and Priorities Thoughts on Representation  Contents of Speeches Votes Introduction of Legislation Other   Parliamentary  28 Bochel & Briggs (2000) Campbell et al. (2010) Erickson (1997) Lloren and Rosset (2015) Lovenduski & Norris (2003) Meier (2008) Studlar and McAlister (1992)4 Tremblay (1993) Tremblay & Pelletier (2000) Wängnerud (2000) 10 - 37%  Childs (2002) Tremblay (2003) Sawer (2002)         3 – 11% Bird (2005) Celis (2006) Childs &  Withey (2004) Erzeel (2012) Grey (2006) Tremblay (1998) Lore (2008)   7 – 26%  Skjeie (2012)+ Cowley and Childs (2003)5        2 – 7%    Tremblay (1998)        1 – 4% Constituency work: Childs (2001)+ General: Childs (2003)+6 2 – 7% Atchison and Down (2009) Martinez-Hernandez & Elizondo(1997)* Towns (2003)    3 – 11% Presidential    37 Carroll (1984) Carroll (2002) Poggione (2004) Reingold (1992)          4 – 11% Tremblay (2003) Whip (1991)           2 – 5%  Cramer Walsh (2002)           1 – 3%  Barnello (1999)* Hill (1983)* Dolan (1997) Lloren (2015) Reingold (2000) Senti (1999) Swers (1998) Vega and Firestone (1995) Fredrick (2010)   9 – 24% Ballmer-Cao & Schultz (1991)  Bratton (2005) Bratton & Haynie (1999) Fredrick (2010) Gerrity et al. (2007) Kathlene (2005) McDonald & O’Brien(2011) Saint-Germain (1989) Swers (2002) Swers and Larson (2005) Tamerius (1995) Taylor-Robinson & Heath (2003) Thomas (1991) Wolbrecht (2002) 14 – 38% Campaigning: Dolan (2005) Herrnson, Lay, & Stokes, (2003). Committee work: Norton (2002) Constituency: Richardson and Freeman (1995)+ Thomas (1990)+   5 – 14% Berkman & O’Connor (1993) Crowley (2004)        5 – 6%   45 Chapter 2: Theorizing the Moderating Effect of Institutions This chapter takes this basic argument and the framework offered in Chapter 1 to theorize the moderating effect of two institutions – the division of powers and the electoral system. Both institutions have been widely studied, and the impact on political behaviour, decision making, and outcomes has been extensively explored (Ames, 1995; Bowler and Fowler, 1993; Carey, 2007; Carey & Shugart, 1992; Cox, 1997; Cox & McCubbins, 2001; Downs, 1957; Duverger, 1959; Gallagher and Mitchell, 2005; Haggard & McCubbins, 2001; Hix, 2004; Iversen and Soskice, 2006; Kedar, 2009;Laver; 2006; Lijphart, 1994; Linz, 1990; Norris, 2004; Sartori, 1976; Strom, 2000; Sieberer, 2010; Tsebelis, 199;). These institutional analyses, however, have been blind to gender, employing the ‘neutral, universal’ actor. At the same time, the study of women within these political systems has largely ignored the impact of the institutional context. When aspects of the division of powers have been considered, research has been confined to those that are “gender-focused” (Sawer, 2012), such as women’s issues committees or intra and inter-party women’s caucuses (Gonzalez and Sample, 2010; Grace, 2011; Steel, 2002). These gender-focused institutions, Sawer argues, “provide leverage beyond that of individual parliamentarians, who may have conflicting accountabilities” (322). What this limited institutional perspective fails to acknowledge, however, are the ways in which whether a system is presidential or parliamentary also structures the number and nature of legislators’ various, and perhaps contradicting responsibilities. These conflicting roles and responsibilities and the implications for behaviour and policy have been widely theorized and tested. An extension of existing theories to include a consideration of  46 gender demonstrate that these seemingly gender-neutral institutional features have clear consequences for gendered representation.  Similarly, the effect of electoral systems on women’s descriptive representation has been extensively studied, but as Tremblay (2006) notes “there is a conspicuous dearth of studies on their influence over the substantive representation of women” (502, see also Tremblay, 2003).7 Electoral systems determine who is elected and how, who legislators represent, and to whom they are accountable. As such, structuring the incentives and disincentives facing representatives affects the behaviour of voters, candidates, parties, elected representatives, and governments. In doing so, the electoral system may also modify, that is suppress or enhance, the relationship between women’s descriptive and substantive representation within parliamentary systems. Exploring how “electoral systems do or do not generate incentives for women to act in certain ways once elected” (Tremblay, 2006, 509) offers an opportunity to improve understanding of the relationship between gender and behaviour.  To fill these existing gaps, an explicitly institutionalist and comparative approach is needed to imagine and analyze a potential moderating effect of electoral systems and the division of powers on the relationship between descriptive and substantive representation. I begin with an exploration of the division of powers, outlining why presidential systems do more than parliamentary systems to incentivize and enable the representation of women by women. I then turn to a discussion of electoral systems, distinguishing first                                                  7 Proportional representation systems are a necessary, but insufficient condition for achieving a (comparatively) high level of women in politics (Lijphart, 1991; IDEA, 2011; Krook, 2010; McIvor, 2003; Norris, 1997; Tremblay, 2006). Party control over the nominations and elections is also found to increase women’s descriptive representation (Thames and Williams, 2010; Valdini, 2012).   47 between single-member plurality (SMP) and PR systems and then using Carey and Shugart’s (1995) classification of systems based on the incentives they generate for personal, as opposed to party votes. After discussing the division of powers and electoral systems independently, I consider the ways in which they interact and discuss how the nature of the electoral system is, in some respects, contingent on the division of powers. I then turn to an exploration of the effect of institutions on policy outcomes. The implications of these institutions for the relationship between descriptive representation and policy outcomes is much less clear than they are for behaviour.  The final section of this chapter discusses the institutional combinations in the seven countries included in the qualitative portion of this dissertation (Chapter 5). Doing so contextualizes the institutions as concrete systems in use rather than abstract and disjointed concepts.  Moderating Behaviour: The Division of Powers  In terms of the stages of the political career path, presidential and parliamentary institutions provide the venue within which behaviour occurs (stage 4), structure opportunities for promotion (stage 5), and determine how individual behaviour and preferences are aggregated into outcomes (stage 6). At these stages, institutional differences between presidential and parliamentary systems differently incentivize the representation of women. Because women operate within these institutions, the substantive representation of women by women is expected to vary with the divergent incentives and opportunities they create. As a result, research that has explored the relationship between women’s descriptive and substantive representation in presidential  48 systems8 is not directly comparable to those studying parliamentary systems.9 The insights from one cannot simply be generalized to the other as if the women operate in an institutional vacuum.  In many ways, an understanding of the extent to which parliamentary systems differ from presidential systems and, in particular, how the former constrains the representation of women, is acknowledged in the literature. The acknowledgments are made, however, mostly in passing and without drawing out the appropriate comparative implications. In a book exploring women in parliamentary systems, Mackay (2006) argues “the relationship between descriptive representation and substantive representation is by no means straightforward and is mediated by other factors, particularly in strong parliamentary systems by party identity and partisan loyalty” (177). In her assessment of women in politics in English Canada, Bashevkin (1993) acknowledges “independent feminism in a parliamentary setting cannot ‘win’ without moving its policy demands through party-controlled pipelines” (3). In research using in-depth interviews with Labour’s so-called “Blair’s Babes”, Childs (2003) provides clues on the importance of the parliamentary context to the behaviour of and representation by a large influx of new women to the UK’s House of Commons in 1997. “Any comprehensive account of the relationship between women’s presence and the substantive representative representation of women in Westminster must also recognize that women MPs are party representatives” (197).                                                   8 Including Barnello, 1999; Bratton, 2005; Kathlene, 1994; Reingold, 1992, 2000; Richardson & Freeman, 1995; Saint-Germain, 1989; Thomas, 1990, 1991; Thomas and Welch, 1991; Thomas & Wilcox, 2005; Vega and Firestone, 1995.  9 Such as Beckwith, 2007; Bratton and Ray, 2002; Celis, 2006; Childs, 2003; Erickson, 1993; Grey, 2002; Lovenduski, 1997, 2001; Norris, 2001; O’Brien, 2012; Sawer, 2000, 2003, 2012; Sawer, Tremblay, & Trimble, 2006; Skjeie, 2012; Studlar and McAllister, 1992; Tremblay, 1998, 2003; Trimble, 1998; Tremblay & Trimble, 2003; Wangnerud, 2000; Young, 2002  49 “In practice, the nature of the substantive representation is influenced by the representatives’ party (at least where parties are dominant) as well as their gender identity.” (197)  Most to the point, Childs states: “Whether women representatives seek to act for women is one question; determining whether they can, in practice, act for women, is another.” (127).  Despite these passing acknowledgments and tentative assertions what is missing is an explicit theory that articulates why women are or are not able to act for women in different systems. There has been no systematic attempt to describe the conditions under which the descriptive representation of women should be expected to matter for substantive representation. In the absence of such a theory, the role of the institutional context is often ignored or obscured. Celis (2008), for example, calls for greater attention to context but speaks of female Members of Parliament and Congresswomen somewhat interchangeably, without due consideration to the importance of the contextual differences these women face. Moreover, the lack of comparative analysis has meant that the actual impact of a division or fusion of powers has not been tested. Gleb (2002) offers one exception. In one of the few pieces of research that take the role of institutions seriously, she explores the differing political opportunity structures in the United Kingdom and the United Sates. Using existing literature in the UK, on the one hand, and the US, on the other, she provides some evidence for the impact of the division of powers on the relationship between women’s descriptive and substantive representation. This section develops on Gleb’s insights and offers an explicit theory; it lays out the key differences between presidential and parliamentary systems and the  50 expected effect on the relationship between women’s descriptive and substantive representation  As outlined by Carey and Shugart (1992, 19), pure presidentialism is defined by four criteria: 1. Popular election of the chief executive.  2. The terms of the chief executive and assembly are fixed and are not contingent on mutual confidence.  3. The elected executive names and directs the composition of the government.  4. The president has some constitutionally granted lawmaking authority (such as a veto). Based on this definition, the fundamental distinction between presidential and parliamentary systems is the relationship between the executive and legislative branches. This distinction has important implications for political behaviour, regime stability, and policy outcomes (Carey & Shugart, 1992; Cheibub, 2007; Cox, 1987; Haggard & McCubbins, 2001; Lijphart, 2012; Laver and Shepsle, 1996; Shugart, 2006; Samuels, 2007; Strom, 1997, 2000; Tsebelis, 2002) and has been conceptualized and theorized in a number of ways in the literature. As Carey and Shugart (1992) put it, the origins and survival of the legislature and the executive are fused in the parliamentary systems and separate when powers are divided. Developing on the ideas of Bagehot and Cox, they differentiate between (and within) the two systems on the basis of efficiency and representation. Cox and McCubbins (2001) focus on merged purposes and powers of the parliamentary system. Laver and Shepsle (1996) argue that because of this fusion, the most important role of legislatures in parliamentary systems is to make and break governments. Legislatures in presidential systems, on the other hand, maintain legislating as their primary purpose.   51 It must be noted that presidential systems vary significantly and the term ‘a division of powers’ is used to describe a heterogeneous group of political systems. Carey and Shugart (1992) offer a typology to distinguish between types of division of power systems based on the extent to which the survival of the assembly and the cabinet are absolute and the president’s authority over the cabinet (Carey and Shugart, 1992). In what follows my theory focuses on pure presidentialism (as defined above), allowing for an explicit theorization of the various mechanisms at play, including a vote of confidence or lack thereof, tied or separated electoral fates, the power to legislate retained by legislative branch versus concentrated in the executive, and opportunities for promotion to cabinet. The quantitative and qualitative analysis in subsequent chapters, however, does consider variation across systems.  France, for example, has a popularly elected president, but parliament can still deliver a vote of non-confidence. In Switzerland, the executive is elected from within the legislative branch, but once in place does not depend on ongoing confidence. The ability of legislatures to form governments is much less powerful or meaningful in Switzerland than it is in parliamentary systems. The use of the ‘magic formula’, wherein the partisan breakdown of the executive is consistent regardless of electoral outcomes, and the practice of returning the same individual ministers to the government until they retire mean that the legislature’s decision is, for the most part, predetermined.  Efficiency and Representativeness  Carey and Shugart (1992) argue that, because the majority composition of the legislature determines the executive, parliamentary systems optimize ‘efficiency’- “the ability of  52 elections to serve as a means for voters to identify and choose among the competing government options” (7).  This is particularly true under the conditions of an SMP electoral system. Just as they maximize efficiency, parliamentary systems also minimize representativeness - “the ability of elections to articulate and provide voice in the assembly for diverse interests” (8). As a consequence of the high identifiability of responsibility and clear government alternatives, parties are programmatic and use discipline to ensure compliance with the broader program and, as a consequence, minimize a focus on local concerns and the provision of particularistic goods. While Carey and Shugart focus on the constraint of localism and particularism, including pork and patronage, the analysis can be extended to consider the provision of other, non-national, ‘goods’; the discipline and dictation by the party applies not just to local goods but to anything outside the party’s agenda. In short, the pursuit of women’s issues, like the provision of particularistic goods or logrolling, is constrained by the need for national, programmatic, and dominant parties.  On the other hand, presidential systems can deliver both efficiency and representation because voters elect two distinct agents: the legislative body can be representative while the executive ensures efficiency. As a result, Carey and Shugart (1992) argue, congressional elections can “turn more on the provision of particularistic services to constituents than on policy” (8). Again, the delivery of particularist goods is only one possible outcome. The broader consequence is that representatives have the freedom to determine the issues they prioritize and the interests they pursue because of their separation from the executive. While this may mean particularistic services, it may also include other issues, interests, and identities, including women’s issues.   53 Fusion or Separation of Origins Samuels and Shugart (2010) argue that in “every pure parliamentary system a vote for any particular legislator – or for the party’s list – is indirectly a vote for that party’s leader as candidate for prime minister. In a sense, a perfect correlation exists between the party’s votes for executive and legislative candidates” (1). As a result, they argue, it is “relatively unproblematic to think of the party as the principal and the candidate as the agent” (124). I argue that the primacy of this principal-agent relationship limits the opportunity for women to depart from their party to pursue women’s issues. Further, the political incentives for (re)election for the executive and legislative branches are the fused and ‘electoral separation of purpose’ is, by definition, not possible in parliamentary systems. In these contexts, women are unable to distinguish themselves from each other and their male colleagues by acting as a representative for or an agent of women.  In presidential systems, Samuels and Shugart (2010) argue that separate origin means “politicians in the same party can face different and even conflicting incentives in the electoral arena” (9) and incentives are enhanced “for politicians in different branches of the same party to go their own way” (9). Because representatives and candidates running for a single party are not tied together as a single (if highly imperfect) agent, women are more able and, perhaps incentivized, to depict themselves as potential agents of women.   Fusion or Separation of Survival Presidents, Samuels and Shugart argue, have little to fear from their party, but under a pure presidential system, the same is clearly true for individual legislators. In parliamentary systems, leaders can dissolve parliament, resulting in a costly and potentially risky election for individual legislators and use this as one tool to ensure  54 compliance with the national program; disincentives for dissent are clear. In presidential systems, on the other hand, dissent, noncompliance, or off-party representation does not necessarily carry the same risks. With this freedom, representatives can provide patronage, deliver local or particularistic goods, or may choose to dedicate their time and resources to women’s issues. Using these and other theories on the division of powers the following sections outline the expected moderating effect on women’s substantive representation. Table 2.1 summarizes these arguments.  Speeches, Statements, and Other Forms of Discourse  Because of the fusion of their legislative and executive branches, governing parties in parliamentary systems have important carrots to induce certain behaviour from their members, including the issues they address in their speeches and statements. Opportunities for promotion (stage 5) can be critical to meeting both policy and office objectives; cabinet positions offer significant benefits to individuals including financial rewards, power and influence, and opportunities to pursue policy goals. Controlled by the governing party leader but available to members of the legislature, these posts can be used to motivate party members to use their speaking time in debates, statements, or oral questions to communicate the party line. Individual members who do so may be offered cabinet positions while those that do not may never receive such benefits (Strøm, 1997). An ambitious representative then, must “be willing to devote time and energy to party objectives, even if that means neglecting one’s local constituency or supporting causes that have little local support” (Strøm, 1997, 169).   55 Women may use up their political resources securing lower-order goals of election and office before they can devote any to representing women’s issues and pursuing those policy priorities. Even if women want to secure cabinet positions to address policy issues important to women, the incentives inherent in the system will mean their speaking time, voting record, and central political messaging will be dictated by the party. Moreover, because the government control the political agenda and necessary resources, if women want to pursue the representation of women, abiding by the party line is simply the most efficient (and in many cases only) way to successfully achieve changes to policies or legislation.  These mechanisms do not just affect the behaviour of members of the governing party. Instead, there are degrees of derogation from this government ideal on the opposition side. Opposition parties need to portray themselves as alternative governments, stable and strong. While the potential rewards are not as imminent, individual members need to consider their future opportunities for cabinet and similar positions. Using speaking time or public discourse opportunities for something other than party priorities may put such prospects in jeopardy should the representative find themselves part of the governing party in the future.  In addition to these incentives, parties also control opportunities to speak in parliamentary systems. If backbench members wish to speak on a debate, they must express their interest to their party officials (Parliamentary Education Service, 2007; Parliament of Australia, 2012). Parties act as gatekeepers not only by controlling access to higher office but also by literally controlling the parliamentary agenda. Simply put, in many cases parties determine who can speak on what and when and individual  56 backbenchers are correspondingly constrained in their ability to speak on issues of their choosing. Party control plays a role not only when the issue is in contradiction to the party position, but also if the issue is simply outside the party’s key issue priorities and central messaging.  In presidential systems, the separation of powers means that ambitious women who seek executive positions do so outside the legislative arena (Strøm, 1997). As a result, these posts do not serve as incentives to induce particular party-friendly behaviour among legislators. Moreover, because the success of individual legislators does not determine the executive’s electoral success, there is less imperative for all party members to be speaking from the same messaging and on the same priority issues. In the absence of such significant incentives to follow party leadership and parrot the party line, alternative incentives may encourage substantive representation of women.  In short, the prospect for promotion and additional benefits of office (stage 5) and the tied electoral fates (stage 3) in parliamentary systems induces particular behaviours (stage 4). As a result, parliamentary systems incentivize behaviour directed by and for the benefit of the party, disincentivizing and constraining the use of political resources for the representation of women. These mechanism do not operating in presidential systems and, as such, opportunities and incentives for unincorporated representation are greater.  Voting The fusion of the executive and legislative branches in parliamentary systems leads to cohesive voting arising from party discipline and not just ideological congruence (Dewan and Spirling; Kam, 2001; Laver and Shepsle, 1996). As the survival of legislatures and governments is tied, backbench members of the government party are incentivized to  57 vote alongside their party to avoid a vote of non-confidence. Failing to do so could trigger an election, which is costly for both the individual and the party and may result in the loss of one’s job. Opposition parties also have incentives for cohesive voting. Dewan and Spirling (2011) argue that by committing to vote cohesively against government proposals, opposition parties will force governments to appeal to their own party’s median member which will, overall, move policy toward the opposition’s ideal point. Kam (2001) argues that party discipline also offers a way for individuals to deal with the collective action problem of party reputation (though he does not use the term): discipline or the threat of it and the need to create stability and avoid elections offer members an excuse for voting for an unpopular policy.  Further, whether opposition or government, the ‘carrots’ of promotion discussed in the proceeding section also serve as a tool for parties to control voting. Those who dissent by, for example, voting against their party but in support of a policy that benefits women, are unlikely to receive benefits of higher office from either a current or potential future government. Cohesive voting is not just consistent with theory or the perceived institutional incentives; it is an empirical fact (Carey, 2007, 2009). Occasionally party leadership offers members a “vote of conscience” or “free vote”, meaning that they can vote as they choose and need not follow the party line. These votes sometimes overlap with gendered issues, notably abortion. On the other hand, they are rare, and their very existence demonstrates the power parties have in determining the vote cast by their members the vast majority of the time (Franks, 1997). In the United Kingdom, a three-line whip system signals to MPs when their attendance and vote is mandatory. Defying a three-line whip comes at the cost of expulsion from caucus. In  58 other cases, MPs have greater freedom to depart from the party line and may take the opportunity to either express the preferences of their constituency or their own moral conscious. In New Zealand, a vote of conscience has not been held for several years and in Australia, only 33 conscience votes have been held since 1949 (Green, 2012).10 Bálint and Moire (2013) find that while there is not a marked gender difference in most conscience votes in Australia, there is in the case of abortion - most women voted for removing it from the criminal code while most men voted against it. Even when the notable disincentives for dissenting votes are weakened, however, incentives to vote with the party, such as the opportunity for promotion to cabinet, still exist (McGee, 1994; Maughan & Scully, 1997; Overby et al., 1998).  Importantly, this theory does not rest on the assumption that parties will always, or even often take positions against women’s interests. Women are expected to support women’s issues more than their party, whatever part of the political spectrum they are aligned. Even when parties take positive positions on women’s issues, however, party discipline and partisan loyalties matter. Men are as compelled as women to support their party’s position and so both men and women will vote in support of the women’s issue. In other words, whatever the issue and whatever side of the issue the party falls, men’s party-directed behaviour is not distinguishable from women’s party-directed behaviour. The result is a very muted gendered difference in voting.  In contrast, Samuels and Shugart (2010) argue that in presidential systems “the separation of powers imposes electoral incentives that affect party organization in particular ways”                                                  10 The last vote of conscious was in 2012 and was on alcohol reform (New Zealand Parliament, n.d.).  59 (13). Put simply, parties are weaker as evident in the less cohesive on roll-call votes (see also Carey, 2009). Leaders do not have the same political resources to distribute to members to ensure compliance. Moreover, a direct election of the president also undermines party unity simply because it doubles the party principals trying to exert influence over individual legislators. If legislative and executive leadership disagree, legislators are pulled in more than one direction and the contradicting influence of principals further weakens cohesion (Carey, 2009). Moreover, in the absence of shared fate, the executive and the legislative branches can have “incentive incompatibility”. While Samuels and Shugart focus on the implication for executive and legislative branches of a party, this incentive incapability also matters for individual legislators whose fates are far less tightly tied than in parliamentary systems. Individual legislators need not behave in concert and may even have incentives to differentiate themselves from each other. The dissolution of such tight, cohesive voting leaves room for women and men to vote differently on women’s issues. In summary, the institutional differences clearly structure the costs and benefits of voting in a legislature. Put in terms of the stages of a women’s political career, voting with or against one’s party has significantly different consequences for navigating the political system to achieve election and re-election (Stage 3) and promotion (Stage 5) in presidential systems than it does in parliamentary systems. The substantive representation of women by voting is constrained in parliamentary systems and facilitated in presidential system. In parliamentary systems, the absence of a gendered difference in voting habits means that descriptive representation is simply not that important to this form of substantive representation. On the other hand, given the freedom to depart from party  60 lines, a relationship between descriptive representation and substantive representation as voting should be observable in presidential systems. Introduction of Legislation In presidential systems, the division of power means that legislatures have the responsibility and power to legislate (Carey and Shugart, 1992). Individual legislators can introduce and pass legislation if they can coordinate support from other members, including, of course, legislation addressing women’s issues. Reingold (2000) argues that women in American state legislators have control over their own policy priorities, determining largely for themselves what kind of policy to propose or co-sponsor. Women, she argues, are not constrained by partisan demands or discriminatory coercion in making decisions on their policy activity, including where they dedicate their energy, time, and resources.  In parliamentary systems opportunities for women in opposition or on the backbench to substantively represent women by introducing legislation are significantly limited. The fusion of the executive and legislative branches means that the government is empowered to legislate as well as enforce and implement the law while the legislature’s most important role is to make and break these governments (Laver and Shepsle, 1996). Governments control the legislative agenda, determining what issues are presented to parliament to consider and introducing legislation that is likely to pass into law. Moreover serious restrictions exist on when and on what topics non-government members can introduce legislation. For example, backbench members cannot introduce legislation that involves the use of public funds. Further, the “confrontational game”, particularly prevelant in Westminster systems, makes cross floor-cooperation on the  61 drafting and introduction of legislation on women’s issues very difficult (Sawer et al., 2006, 4).  The same institutional features that limit opportunities for backbench women mean that women on the government frontbench, that is women in cabinet, have significant influence. Individual cabinet ministers in parliamentary systems have significant control over their portfolios and can pursue their policy priorities within their department (Atchison and Down, 2009; Laver and Shepsle, 1994). Moreover, because cabinet ministers, unlike backbenchers, can introduce legislation that does have spending implications, they have influence over a much broader spectrum of women’s issues (for example, publicly funded childcare). The practices of party discipline and party loyalty mean that once a woman in cabinet opts to pursue a policy and the rest of government supports the policy, the likelihood it receives support from the legislative branch is exceptionally high and indeed much higher than for legislation introduced by women in presidential systems. Indeed, the proportion of women in cabinet in parliamentary systems has been found to have an effect on parental and maternity leave entitlements independent of the percentage of women in the legislature (Atchison and Down, 2009).  Presidential and parliamentary systems also offer differing opportunities and incentives for making amendments to proposed legislation so that they may more adequately address women’s interests or needs. Decentralizing decision-making in the American presidential system increases opportunities for women to interject in the policy-making process. In Westminster parliamentary systems, women on committees may be confined in their ability to make changes at the committee stage in much the same way as they are  62 in expressing their opinions through votes or speeches. This is expected to be particularly true when party leadership acts as the gatekeeper for their delegates on committees.  In short, women on the backbench in parliamentary systems face significant disincentives for the representation of women. Further, their opportunities for influence are minimal. Women on the frontbench, however, have very real opportunities to represent women in a way that can translate into policy outcomes that improve the lives of women. As Atchison and Down (2009) put it - “individual legislators are likely to be relatively weak proponents of female-friendly social policy relative to their counterparts in cabinet” (Atchison and Down, 2009). Individual women in presidential systems lack the influence of cabinet ministers in parliamentary systems but have incentives and opportunities that women on the front bench lack. Table 2.1 at the end of this chapter summarize these arguments. The theory outlined generates a number of observable implications. Chapters 3 provides more specific hypotheses based on the data available and tests the theory using a mixed methodological approach. Discourse & Behaviour: Observable Implications  1. Women elected in parliamentary systems will address women’s issues less frequently in representation acts (including oral questions, committees, debates, statements, voting, and the introduction of legislation) than their counterparts in presidential systems. 2. Because men will not respond to institutional (dis)incentives in the same way as women, the gap between women and men on these representational acts will be larger in presidential systems.   63 Women’s Experiences of Political Institutions: Observable Implications  1. Women in presidential systems should feel that women are an important constituent, both in terms of their (re)election prospects and their obligations for representation. 2. Women in parliamentary systems are expected to feel limited in their ability to vote in favour of women-friendly policy or legislation if their party is voting against it.  3. Women on the backbench in parliamentary systems should feel constrained in their opportunities for proposing legislation addressing women’s issues. Women in cabinet should perceive themselves having significant opportunities to raise the profile of and implement policies that address women’s issues.   Moderating Behaviour: The Electoral System The constitutional design is not the only institution that can influence the nature of representation or affect the number and strength of competing principals. As the set of rules that determine how citizens choose their government electoral institutions are “the most fundamental element of representative democracy” (Lijphart, 1996, 1). They are not, however, equally ‘representative’. Instead, electoral systems may be “scaled to the degree to which they represent diversity” (Carey and Shugart, 1992, 9), which can include localism, in the form of policy or particularistic services, or group interests, based on ideology, ethnicity or religion. An extension of this logic to a consideration of gender demonstrates the ways in which electoral systems can also do more or less to facilitate the substantive representation of women.  In terms of the career paths and political motivations of representatives, electoral systems most explicitly determine how representatives win nomination, seek votes, and if and how they win elections (stages 2b and 3 and motivations 1 and 2). By affecting what one has to do to be elected and the chances of doing so successfully, they are also relevant to  64 stage 2a - whether women and men self-select into politics. Moreover, because they structure the relationship between legislators and voters, they determine who legislators represent and to whom they are accountable and thus impact behaviour at stage 4. Electoral systems also affect the number and type of parties and thus have an important influence on the formation of governments and the final stage - aggregation of individual preferences into policy outcomes.  Electoral systems are diverse and complex and conceptualizing the differences among them requires a consideration of several different dimensions. For example, electoral systems are often differentiated on the basis of the electoral formula (how votes become seats), the district magnitude (the number of representatives per district), and on the structure of the ballot (how voters express their preferences) (Blais and Massicotte, 1996). Some scholars offer alternative dimensions, including further dividing ballot structure into the number of votes a voter can cast and how much choice voters have between individual candidates (Gallagher and Mitchell, 2005). Others look earlier in the process, including candidate selection rules as a critical dimension for understanding electoral systems and their effects (Carey and Shugart 1995; Hix, 2004).  I use two key conceptual distinctions to theorize the expected effect of electoral institutions on behaviour of individual representatives. First, I broadly compare proportional representation and single-member-plurality (SMP), focusing in particular on the strong geographic basis of the latter and higher district magnitude of the former. While saying nothing explicitly about mixed systems, the logic of the theory leads to divergent expectations for the behaviour of members elected under different electoral systems in mixed systems. Second, I use Carey and Shugart’s (1997) framework for  65 differentiating systems that incentivize personal votes from those that incentivize party votes. This allows for a more complex consideration of electoral systems, discriminating between different proportional systems, for example.  Proportional Representation versus Single-Member-Districts Carey and Shugart argue when parliamentary systems use single-member plurality, efficiency is at a maximum and representativeness at a minimal. Parliamentary systems that use proportional representation, on the other hand, are more inclusive of broad interests and can facilitate a greater representation of group interests, including ideologies, ethnicity, or religion (9). I identify two differences between single member plurality and proportional representation systems that are relevant to women’s substantive representation – the strength of the geographic tie and district magnitude. A larger district magnitude has two separate implications – a lower electoral threshold (percentage of votes needed to win a seat) and a greater number of representatives per constituency. These differences affect the incentives and opportunities that legislators have to represent women.  Geographic tie of representation As Carroll (2011) observes, the notion of ‘women representing women’ fails to take into account geography. Which women are represented? Unlike many other groups seeking representation (based on race or class, for example), women are in no way geographically defined. The extent to which representation is geographically defined will thus affect the incentives and opportunities to dedicate political resources towards representing women. SMP, in particular, lends itself to a conceptualization of representation in geographic terms (Mezey, 2008), that is, representatives are elected within and responsible to their  66 own clearly defined electoral district. A woman elected under SMP must respond to the particular interests of the geographic community rather than broader collective needs (Sawer, 1998). Due to limited resources (time for speeches, legislation to propose, etc.), she will be constrained in the acts of representation or resources she can dedicate to explicit women’s issues (Erzeel et al., 2014; Tremblay, 2006).  While SMP systems can and do process broader conflicts of interest, such as class and race, they do so when these political issues have a geographic base. Neighbourhoods often contain a certain homogeneity of occupation and income, and as a result, a geographically based politics is class politics (see, for example, Rodden, 2010). In the United States, racial gerrymandering means that sometimes a geographically based politics is also a racialized politics. Women, on the other hand, have no geographical concentration - there is simply “no gender equivalent to a rural district or an African American district” (Reingold, 2000, 6). The time and resources necessary to represent the issues of the constituency, whether local, class, or racial, will leave little opportunity to focus explicitly on women. Several studies find that women, much more than men, feel that they do (and should) represent women even in the absence of an electoral connection, a phenomenon Mansbridge (2003) calls surrogate representation (Carroll, 2002; Childs, 2002; Swers, 1998; Waring et al., 2000). Reingold (2000) argues that the tie women feel to their female constituents indicates that this institutional constraint is “rather weak” (220). Given the shared (if diverse and varied) experience and perspective and significant numerical under-representation, it should not be surprising that women representatives continue to feel connected to their female constituents, including those outside their district This is a  67 distinct argument from the claim the institutional constraints don’t change behaviour and the fact that surrogate representation exists does not mean that women would not do more to represent women if greater opportunities or incentives were present. The expressed intention or desire to represent women may translate into greater action and more substantial policy changes when the electoral system allows for it. In fact, this may help explain the often-observed difference between what women say they represent and what they actually do or, as Carroll (1984) puts it, the gap between attitudinal and behavioural feminism (Reingold, 2000; Tremblay, 1998; Trimble, 2006; Wangnerud, 2000). Carroll explains this gap as follows: “too much emphasis on women’s issues in a campaign might lead to speculation that the candidate is narrow in her interests and would not adequately represent all the people” (319).  In PR systems, on the other hand, the activity of representation is less geographically defined. The weakening of the conventional link between a representative and an electoral district can enable ties based on shared experiences and interests “defined by identity rather than geography” (Tremblay, 2006, 505). With less time and effort consumed by local concerns, there is a greater opportunity for representatives to address cross-boundary issues, including those of interest to women (Sawer, 2002). In short, the disincentive identified by Carroll is simply not as strong in PR systems because the institutional context is different. District Magnitude District magnitude is a key determinant of the proportionality of an electoral system and has been found to be a key determinant of women’s descriptive representation (ACE Network, 2014; Rule, 1987; Norris, 2006; Taagepera and Shugart, 1989). It also has two  68 distinct implications relevant to substantive representation. First, the electoral threshold determines how broad or narrow appeals to voters must or can be. Second, the greater the number of representatives responsible for one constituency, the greater the likelihood that constituents have a woman representative, which in turn increases the number of times they raise gendered issues to the legislators who have a responsibility and incentive to address them. Electoral threshold The high percentage of votes needed to win a seat in SMP systems requires legislators to reach beyond the electorate interested in women’s issues and into the male electorate (so to speak) if they hope to be re-elected. The result is a clear disincentive to focus on women in particular or women’s issues as such. On the other hand, under PR political parties may be more likely to support their candidates in advocating for women because of the electoral incentives to directly target specific electoral clienteles (Tremblay, 2006).  A recent example provides some support for the notion that district magnitude matters. In 2014, the Swedish Feministiskt Initiativ (Feminist Initiative) Party elected their first Member of European Parliament. The party received 4.3% of the vote, just surpassing the 4% necessary to win one of the country’s 20 seats in the European Parliament. In a plurality system, the party would not have achieved any representation and, given the dim prospects for success, would have been unlikely to form in the first place. Representatives Per Riding Simply having more than one representative may affect incentives and opportunities for substantive representation. With multiple legislators looking out for the constituency’s interests, an MP’s burden to represent all interests of the geographic area and its residents  69 is decreased, creating an opportunity to dedicate resources to more specific and cross-constituency rather the geographic interests. Further, on gendered issues, such as sexualized violence or childcare needs, women may be more comfortable approaching a female representative (Childs, 2002; Tremblay, 2006). In a multimember district, the odds of having a female representative to approach are greater, thus increasing the options available to women voters and with it the number of times such issues are brought to the attention of representatives. If a woman represented by a man in a single member plurality system wishes to speak to or be represented by someone with whom they believe they share a common experience, they may approach a woman representative in a different electoral district (Mansbridge, 2003). This representative is not, however, their agent and her primary obligations are to her constituents. In a context of limited resources, providing assistance to a woman outside her constituency may come at the cost of representing an issue important to her constituency.  As a result, she has few electoral incentives for the representative to offer assistance. Moreover, there are strict conventions in SMP systems that MPs do not assist the constituents of their colleagues so as not to interfere with the relationships between representatives and districts. At the same time, the women’s local representative with whom she does have a principal-agent tie will not be made aware of the issue. In other words, in SMP systems, constituents will raise women’s issues to representatives less often and, when they are raised to representatives in other districts, there are limited incentives for them to be addressed.  In sum, two women, one elected under SMD and the other elected under PR, both of whom prioritize women’s issues more than their male colleagues, are nonetheless expected to behave differently because they operate within distinct incentive and  70 opportunity structures. Further, the gap between the representation of women’s issues by women and that by men should be larger in proportional representation systems than in SMD systems. While the theory offered here focuses on a PR/SMD dichotomy, the implications for a Mixed Member Plurality (MMP) system are clear: representatives elected off the list will face different incentives and opportunities than their colleagues elected directly in a constituency. Please see Table 2.2 at the end of this chapter.  The theory outlined generates a number of observable implications.  Discourse & Behaviour: Observable Implications 1. Women elected under SMP will address women’s issues less frequently in representation acts (including oral questions, committees, debates, statements, voting, and the introduction of legislation) than their counterparts in proportional representation systems. 2. Because men will not respond to institutional incentives in the same way as women, the gap between women and men on these representational acts will be larger in PR systems than in SMP systems.  Of course, the theory offered is not about the institutions per se, but is fundamentally about perceived incentives and legislators’ experiences of political institutions. Women’s Experiences of Political Institutions: Observable Implications  1. Women in SMP systems should feel a tension between time spent representing women and representing their geographic constituency.  2. Women in SMP systems will perceive a disincentive to campaign on or speak to women’s issues in speeches or other forms of discourse. 3. Women in PR systems should perceive themselves as having the opportunity to address women’s issues in speeches, campaigns, and other forms of discourse.   Preliminary Evidence: Substantive Representation in PR and SMP Systems Though largely theoretical, the proposed differential effects of SMP and PR electoral systems find some support in previously conducted research, both on representation  71 generally and on women’s representation specifically. Bowler and Farrell (1993) find that in the European Parliament representatives elected under a candidate-ballot system (plurality) kept regular contact with their geographic electorate while those elected under party-ballot system (PR) focused more on groups of voters. Research on Germany’s mixed member electoral system has found notable differences between MPs elected in districts and those elected via lists. Sieberer (2010) finds that district MPs are more likely to defect from the party vote (though this is only partially due to their territorial obligations). Others have observed that these MPs pursue more individualistic electoral campaigns (Zittel and Gschwend, 2008), place more emphasis on representing local interests (Klingemann and Wessels, 2001), and are more likely to sit on parliamentary committees dealing with local issues (Stratmann and Baur, 2002).  The tension between representing women and representing one’s geographic constituency has been observed in several studies looking at SMP systems. Several of the UK MPs interviewed by Childs (2002:148-149) had been contacted by women constituents of male colleagues about gendered issues and, more importantly, she observed “a perceived tension between representing one’s constituency and substantively representing women” (151). Carroll (2002) found that many Congresswomen believed women outside their district still looked to them as “their congresswomen”. Most of the representatives also felt a responsibility to represent those women in addition to their districts, what one woman called “a pretty big burden” (53). In her analysis of United States Congress, Swers (2002) identifies and explores the tension women feel between their responsibility to represent women and the necessity to represent their constituency and “maintain the support of their partisan colleagues” (2006). In Australia, a candidate for the Australian  72 Women’s Party argued that “we tend to arrange our representative democracy around location… but not much around gender” (Kelly, 1997) even though gender, much more than location, has a profound effect on life (in Sawer, 2000).  Interestingly, in SMP systems, women often undertake more representational acts than their male counterparts, not just on women’s issues but rather overall. Women have been observed to do more constituent work and express their desire to do even more (Richardson and Freeman, 1995; Thomas, 1990) and to make more speeches and statements than their male colleagues (per representative) on any issue (Lore, 2008). This gendered difference may reflect the responsibility and electoral incentive to represent one’s geographic constituency paired with women’s additional desire to represent women’s issues.  Sawer (2000) observes that issues of domestic violence were raised three times as frequently in the Australian Senate as in the House of Representatives from 1981-1993, a difference she attributes to the higher proportion of women representatives. In 2005, women in the Senate worked across party lines to co-sponsor and pass a private member’s bill, which lifted a ministerial veto on importing the RU486 “abortion” pill (Sawer, 2012). Only after it had passed in the Senate did the Prime Minister allow it to pass through the lower house (elected through AV) without division. Sawer again attributes this, in part, to the critical mass of women in the Senate. The theory offered here suggests, that in both, cases the proportional system used in the Senate may have exerted its own effect, quite separate from its effect on the number of women elected.   73 Unincorporated Representation Carey and Shugart (1995) offer an additional way to think about electoral systems. Their analysis focuses specifically on the incentives generated for individuals to rely on either their own or their party’s reputation for (re)election. A party’s reputation, they argue, constitutes a collective action problem - all involved have an incentive to maintain a positive party reputation and a cohesive, strong front, but individuals also have an incentive to speak out against an unpopular position taken by their party to try to maximize their own support. Electoral systems, Carey and Shugart argue, structure the incentives that politicians face in navigating the tension between maximizing their party’s votes and their own votes.  Four aspects of the electoral system are relevant: the degree leadership controls access to and rank on the ballot, vote pooling, and the number and type of votes to be cast by the individual voter. Carey and Shugart assign three possible values for each of the dimensions, with higher values indicating greater incentives for personal votes. First, incentives to cultivate a party vote are maximized when party leadership has control over the ballot. If candidates can access the ballot without approval from national leadership and voters can alter their rank on the list, the incentives and opportunities to cultivate a personal vote are much greater. The second characteristic, vote pooling, refers to the extent to which a vote is counted as support only for a candidate or whether votes are pooled to benefit a faction of candidates or the party as a whole. When votes are not pooled, individual candidates are elected purely on the votes they themselves receive. When votes are pooled, candidates’ success depends on the support their party receives. Finally, the incentives for personal votes are the smallest in systems where voters have  74 one vote that they must cast for a party. Incentives to seek personal votes are greater when voters have multiple votes and are at their maximum when voters have a single vote that they must cast for one individual candidate. Combining these three characteristics, there are 13 (feasible though not necessary existing) combinations ranked from the lowest incentives for personal votes to the highest. A fourth characteristic, district magnitude, has the effect of amplifying the incentives created by other three other features. By structuring the nature of representation - party or personal - electoral systems affect the relationship between constituents and their representatives, representatives and their parties. Legislators elected under party based systems must rely on their party for their re-election prospects and, as a result, need to follow their party’s lead and direct their efforts to furthering their party’s interests. With the bonds to party loosened, representatives in a personal vote system can build personal relationships. Politicians in these systems are more able to speak to issues other than those dictated by their party and have greater flexibility in voting, proposing legislation, and other behaviours.  Carey and Shugart mention national celebrity, particularism, and pork barrelling as potential pathways to personal support, but do state that their focus is the value of personal support, not how it is most effectively developed (419). The concept of ‘unincorporated representation’, including women’s issues, offers an alternative. Given the freedom from their party, legislators can take the opportunity to pursue women’s interests and in doing so distinguish themselves from their competitors by as an agent for women  75 Personal votes and unincorporated interest votes are not direct equivalent and several changes are made to the ranking of electoral systems. First, Carey and Shugart focus on party versus personal reputations at the district level, while I consider party reputation more broadly and do not limit the analysis to the district level. As outlined above, women are not geographically defined and so a consideration of the implications of unincorporated representation for women’s substantive representation requires a national scope. Moreover, to reflect the difference between incentives for unincorporated representation and incentives for purely personal votes I make two adjustments to the scoring of the original framework: the first applies to the ballot dimension and the second to the vote dimension.  Ballot Dimension The first adjustment arises out of a critique of the framework itself. The classification of ‘ballot’ as offered by Carey and Shugart does not capture the full range of combinations that might affect incentives for personal votes. The ballot dimension includes two components – the control leaders have over a ballot and the extent to which voters can disrupt a list. The intersection of these two dimensions yields four, not three, possible combinations. What is missing in Carey and Shugart’s classification is a system where the list cannot be disrupted but where leaders do not control the ballot. If control over access to and rank on the ballot is local as opposed to national, there are clearly greater incentives for cultivating a personal vote even if voters cannot change the list once it is compiled. In other words, a system should score one if voters can disrupt the list or leaders do not have complete control over access to and rank on the ballot. Carey and Shugart’s original framework ranks such systems as zero on this dimension. Countries like Canada and the United Kingdom where voters cannot disrupt the list of candidate (a  76 list of 1) but where local party members do have some control over nomination is scored ‘1’ instead of ‘0’ under this new classification.  Access to the ballot is not as simple as leader control versus individual politicians as entrepreneurs as Carey and Shugart (1995, 421) suggest. Rahat and Hazan (2001) take instead a much more multi-dimensional approach to evaluate access to a ballot. Their framework includes a consideration of the rules around who can be a candidate, who makes up the selectorate, and how territorially decentralized the process is. The former two range from exclusive to inclusive and the latter from national to local. Selection processes can also be determined by vote or by appointment. The features of the process have importance consequences for party control over candidates. When nominations are highly democratic, there is “a tendency [for candidates] to act in a manner that largely disregards the group(s) with which the candidate is associated (i.e. the party, the coalition, opposition, etc.); …[and] a drastic increase of the rather basic political trend toward individualist and populist politics” (313). Again, however, candidates might focus on unincorporated representation instead of individualistic politics. Indeed, Rahat and Hazan acknowledge that decentralization, for example, might be used to ensure the representation of women (or minorities or trade unions).  Rahat and Hazan paint a much more complete picture of the ways in which selection processes can vary and, as with Carey and Shugart’s framework, there are clear implications for the representation of women as an alternative to personal politics. For simplicity’s sake, however, I continue to use Carey and Shugart’s simple categories. That said, in testing the theory offered I am attentive to the complexity provided by Rahat and Hazan where possible (Chapter 4 and 6).  77 0. Leaders have control over nomination and voters cannot disrupt the list 1. Voters can disrupt the list or leaders do not have complete control over access to and rank on the ballot 2. Leaders do have limited or no control over the ballot and voters can disrupt the list Vote Dimension While the theory can broadly be extended beyond the personal-party dichotomy, the representation of ideas or identities is simply not the same as securing political support on a strictly personal relationship. The second adjustment is a reflection of the way in which the representation of unincorporated interests is not an exact equivalent to a purely personal politics. Carey and Shugart classify systems where voters have multiple votes as having lower incentives for personal votes than those where voters have a single vote to cast for a candidate and “candidates are competing simultaneously for the same indivisible support of each voter” (422). In the case of women’s representation or other unincorporated interests, it seems reasonable to expect that incentives might be greater when voters have multiple votes. Women voters don’t vote on the basis of gender or women’s issues alone, but if they can indicate several preferences, they may offer support for candidates who represent their policy preferences and who promise to further women’s interests in the political arena. Women in these systems will have incentives and opportunities to substantively represent women. In other words, incentives for unincorporated interests in general and women in particular are greatest when voters have multiple votes to cast below the party level, lower when they have a single vote for a candidate, and the lowest when voters have a single vote for a party.  0. Voters have one vote to cast at the party level. 1. Voters have one vote to cast for an individual candidate.  78 2. Voters have multiple votes to distribute to candidates.   Polling Dimension The polling dimension, which “measures whether votes cast for one candidate of a given party also contribute to the number of seats won in the district by the party as a whole” (421) remains the same as that offered by Carey and Shugart.  0. Votes are pooled at the party level. 1. Votes are pooled below the party level.  2. Votes are not pooled.   Incentives for Unincorporated Representation These three dimensions are combined to create a classification of systems based on the incentives they generate for unincorporated representation. Incentives are highest when votes are not pooled, leaders do not control nominations, and when voters have multiple votes with which they can disrupt the presented lists of candidates. Incentives for unincorporated representation are at their lowest and incentives for party-based representation at their highest when voters have one vote to cast at the party level, votes are pooled, leaders control access to the ballot, and voters cannot disrupt the list.  District Magnitude In the previous section, I argued district magnitude would increase opportunities and incentives for substantive representation, both because it lowers the electoral threshold and simply because more representatives per district increase the chances for a woman representative and enable specialization. Carey and Shugart, however, argue that the effect of district magnitude depends on the other institutional features: “In all systems  79 where there is intraparty competition, as M grows, so does the value of personal reputation. Conversely, in systems where there is no intraparty competition, as M grows, the value of personal reputation shrinks” (418). In party-based systems, as district magnitude increases, the party, rather than voters, increasingly becomes the primary principal to which representatives are responsible. On the other hand, when there are incentives for personal votes, “as the number of other copartisans from which a given candidate must distinguish herself grows, the importance of establishing a unique personal reputation, distinct from that of the party, also grows” (430).  Although incentives for unincorporated representation may grow alongside incentives for more traditionalistic personal representation, it is not at all clear that the number of representatives per district will have the same differential effect on the representation of women. Given the diverse ways a representative may cultivate a ‘personal’ vote, the competition between copartisans may manifest in other, non-party based approaches. If greater competition is the result of a larger district magnitude, it is more likely to take the form of divergent forms of representation, including the representation of women, rather than simply increasing the representation of women.  For summary, see Table 2.3 at the end of this chapter. Discourse & Behaviour: Observable Implications 1. Women in systems that do not generate incentives for unincorporated representation (IUR) will address women’s issues less frequently in representation acts (including oral questions, committees, debates, statements, voting, and the introduction of legislation) than their counterparts in systems that do. 2. Because men will not respond to IUR in the same way as women, the gap between women and men on these representational acts will be larger in when these incentives are at their maximum.  80  Women’s Experiences of Political Institutions: Observable Implications  1. Women in systems that score high on IUR will perceive that it is in their electoral interest to represent women.  2. Women in systems that score low on IUR will feel constrained or limited in the amount of energy they can dedicate to represent women.  3. Women in systems that score high on IUR will seek preference votes or support in nomination as a candidate on the basis of representing women.   The classification adapted from Carey and Shugart enables thinking about differences among proportional and SMD systems beyond the simple dichotomy between the two. One could compare, for example, a system that includes some incentives for personal votes (like Congressional elections in the United States, which use a primary system) and a proportional representation system that encourages party vote (like the Spain where voters can only indicate their party of choice). For the sake of simplicity of theory development and testing, I keep the consideration of SMP vs. PR distinct from the consideration of incentives for unincorporated representation. In the cases of some of the empirical analysis in Chapter 4, I explore variation along the party-personal vote dimension within proportional systems because of certain features of the available data. The qualitative research in Chapter 6, does, however, enable a consideration of the implications for an interaction between the SMD/PR distinction and the party-personal vote distinction.  Preliminary Evidence – Substantive Representation in Personal and Party Systems. As with the SMP-PR comparison, there is some preliminary evidence that incentives toward personal votes lead to a greater sense among candidates and legislators of being a   81 representative for women. In two American states, Reingold (2000) found that 34% of women representatives spontaneously mentioned women in a discussion of group representation. Two other studies, exploring countries where party control over access to the ballot is much higher (Canada and the United Kingdom), find that far fewer representatives mention women when asked about the meaning of representation or the nature of their mandate (13.6% and 11.8% respectively) (Childs, 2001, 2002; Tremblay, 2003). Further, Tremblay (2003) found that substantially more Senators than MPs in Australia cited the party system as a barrier to representing women. This difference, Tremblay argues, may be a consequence of party control - senators, more than MPs are “likely to view themselves as completely at the mercy of the party faction” (230). Where there is less voter capacity to influence the ballot, MPs see themselves more beholden to the party.  Substantive Representation & Descriptive Representation: In Conflict? The effect of electoral systems on descriptive representation has been extensively studied, and so the focus here has been and will remain exclusively on the effect on substantive representation. It is worth, however, taking the time to explicitly compare the expected effect on descriptive representation with the expected effect on substantive representation. It has been well established that PR systems increase the number of women elected (Lijphart, 1991; IDEA, 2011; Krook, 2010; McIvor, 2003; Norris, 1997; Tremblay, 2006). I have offered a theory that suggests PR will also increase substantive representation by these women. On the other hand, research has established that descriptive representation is decreased when institutional incentivize personal votes and increased when there is greater party control (Thames and Williams, 2010; Valdini,  82 2012). In this case, then, institutional features oppositely affect descriptive representation and substantive representation. Combining Constitutional Choices: Division of Powers and Electoral Systems While the division of powers and the electoral system have clear and independent implications for the representation of women, they do not exist or operate independently in practice. The representativeness and efficiency of a political system depends not just on basic constitutional choice between a parliamentary or presidential system but also on its interaction with electoral systems (Carey and Shugart, 1992). Under a division of powers, systems that also use SMP are extremely efficient because of the smaller number of parties. For the same reason, representativeness is at a minimum. Carey and Shugart argue that despite the geographic tie of representation “politics in the districts from which members of parliament are elected can no longer focus so much as in the absence of party discipline on the stuff of legislative ‘logrolling’ and compromise… instead it must focus much more on nationally divisive but often not divisible concerns” (10). While the consequence of interest for Carey and Shugart is the diminution of local representation, the paramount importance of national agendas and corresponding national leader control and extreme party discipline also reduce opportunities for representation of women by women. For the major cases included in this research, disaggregating the differential effect of the two institutions is very difficult. Canada and the United Kingdom are the major parliamentary systems used to compare a fusion of powers with the division of powers in the United States. While all three countries use SMP systems, there are substantial incentives to pursue personal votes in the latter that are absent in the two parliamentary  83 systems. A more ideal comparison would be possible if parties in the United States had greater control over nomination or if the electoral systems in Canada and the UK did more to incentivize personal votes by allowing voters to rank preferences (for example). Such a thought experiment, however, runs into a theoretical problem – a personal vote based SMP system is in contradiction with a parliamentary system. Carey does not make this argument and Shugart in either their 1992 or 1995 work, but a simple extension of their arguments makes the conclusion clear. Parliamentary systems that use SMP are defined by high party discipline to avoid votes of non-confidence and early elections, with the result that opportunities to engage in logrolling or provide particularistic services are extremely limited. A vote for the local candidate is, for all intents and purposes, a vote for the executive. Put differently, even before taking the particular features of the electoral system into account representatives have minimal (to no) opportunities to cultivate a personal, clientelistic basis of support. The extension to women’s representation as one particular manifestation of a ‘personal’ base of support is simple. Individual legislators cannot cultivate votes based on the representation of women when their primary principal is the party and they have little capacity to deliver on any such representational promise.  In short, the party versus personal nature of the SMP system is contingent on the division or fusion of powers. In parliamentary systems, SMP systems are party-focused to match the institutional constraints on personal votes generated by the fusion of powers. In other words, a division of powers is a necessary condition for personal votes. The institutional features of a presidential system enable representatives to depart from the party line and to pursue an independent representational agenda, making the division of powers  84 compatible with a personal-vote based systems. The primary system in the United States gives voters an extra vote to cast and votes are not pooled. Fundamentally, though, if representatives campaign on a promise to provide particularistic or local goods or, importantly, policies important to women, they can credibly pursue these objectives (though the outcome is not guaranteed). Of course, a division of powers is not a sufficient condition for a plurality system to be based on personal votes. For example, the French  semi-presidential system is much more party-vote based system than the United States. The point is, however, that presidential systems are not inconsistent with personal-vote based electoral system whereas parliamentary systems that use SMP are.  These sharp differences between presidential and parliamentary systems based on the trade-off between efficiency and representativeness draw attention to more subtle differences between parliamentary systems. Parliamentary systems which use proportional representation, are already more representative and less efficient than their SMP counterparts. The higher district magnitude increases the number of parties and with it the range of interests represented. At the same time, these parliamentary systems lose some of the efficiency as they frequently produce minority, multiparty, and coalition governments that reduce voters’ ability to assess accountability and choose between governing alternatives.   The altered balance between efficiency and representativeness facilitates greater flexibility in the extent to which personal votes are incentivized. To be sure, Carey and Shugart’s conceptualization of personal votes came at a later date and so the interaction  85 with parliamentary and presidential systems is not considered in their 1992 book.11 Instead, they focus exclusively on the broad interests brought into the legislature by the greater number and diversity of parties. Parliamentary systems that use PR systems where parties maintain significant control over nomination and order on the ballot will indeed be representative in a way that reflects party composition. Parliamentary electoral frameworks that give voters multiple votes that they can use to change the order on the ballot can increase representativeness above and beyond a greater number and diversity of parties. Included in this broader sense of representativeness is, of course, the increased incentive and opportunity for women to represent women.  The impact of presidential systems also depends on the timing of elections between the executive and legislative branches. To the extent that executive and legislative elections are concurrent, parties may be more “presidentialized” if executive dynamics rather than the legislative dynamics drive elections (Samuels and Shugart, 2010, 127). If elections are separated, executive and legislative branches can campaign on distinctive issues, but if they are held concurrently, there is an expectation of ‘fusion of purposes’. What is of interest in relation to unincorporated representation, however, is the intra-party competition – even if elections are held concurrently, the absence of a confidence motion still enables individuals to depart from the party line. Moreover, the extent to which concurrent elections result in a more cohesive party depends on the features of the electoral system, including whether ballots are fused. Finally, the timing of elections,                                                  11 In Carey and Shugart (1995), it is argued “that, ceteris paribus, personal reputation will be more important in a presidential than in a parliamentary system” (432). The model developed, however, “focuses exclusively on electoral rules, leaving the systematic incorporation of other variables, like constitutional system, to subsequent research.” In Samuels and Shugart (2010), the Carey and Shugart (1995) classification of personal and party votes is used only to argue that the “separate origin of the executive and legislative branches has an analogous impact on political parties” (127).   86 Samuels and Shugart (2010) matter to the extent to which presidential rather than legislative candidates make appeals on the basis of personal or party votes. In short, the timing of electoral systems in presidential systems matters less to unincorporated representation than the difference between parliamentary and presidential systems and the nature of the electoral system itself. As a result, I do not include this aspect of variation in the analysis.  In summary, presidential systems that use proportional representation systems that facilitate unincorporated representation will result in the highest representation of women by women and the largest gap between women and men. Parliamentary systems that use SMP systems which (necessarily) incentivize party votes will do the most to disincentivize and constrain substantive representation of women and, as a result, will have the lowest levels of substantive representation by women and the smallest gap between women and men.  In the analyses that follow it is not always possible to disaggregate whether it is the features of the electoral system or the division of powers that leads to women in presidential systems to do more to represent women, but the point made here is that, in many ways, they are fundamentally linked. In the qualitative analysis, effort is made to consider in a more complex and integrated fashion the institutional features that incentivize or enable representation of women by women. The interviews allow for a discussion with, rather than about, legislators themselves and target their experiences of institutional incentives and opportunities to represent women. Representatives do not  87 experience the electoral system and the division of powers as separate entities but instead pursue election, promotion, and policy objectives in a single institutional context.  Moderating Outcomes: Presidential and Parliamentary Systems There are two distinct ways that institutions can moderate the relationship between women’s descriptive and substantive representation – through their effect on the behaviour of representatives and, independently, through their effect on the aggregation of that behaviour into outcomes. This section turns to the latter and asks how a division and fusion of powers differently translate individual actions into policy outcomes. Parliamentary Systems: Unfulfilled Potential for Substantive Representation  While they are likely to minimize the substantive representation of women by women, Westminster-style parliaments, in theory, provide important opportunities for substantive representation as policy outcomes (Trimble, 2006). Because they are defined by government domination, if feminist policy objectives make it to the agenda of the governing party, parliamentary systems will result in more policy addressing women’s issues. Parliamentary systems are decisive and efficient - once a proposal has been made by government party discipline means it is very likely to be enacted (Carey and Shugart, 1992; Chiebub et al., 2004; Cox and McCubbins, 2001; Strom, 2000; Tsebelis, 2002).  The lower decision-making costs could mean that an increase in descriptive representation and a preference for women-friendly policy among women legislators is quickly and easily translated into policy outcomes. Thus, if women’s issues do make it to the agenda, government is both able and obliged to implement programs (Swer et al., 2006). As argued in Chapter 1, however, institutions affect outcomes in two ways – by  88 influencing behaviour and by aggregating that behaviour into outcomes. It is the former that poses the initial problems in the case of parliamentary systems. While parliamentary systems might provide efficient and predictable aggregation of behaviour into outcomes, their effect on behaviour means that these outcomes are likely not to include women’s issues. As Peckford (2002) argues of the Canadian case, “the promotion of outside interests, particularly feminist in nature, enjoys little room in most parties, especially when the majority of those elected have been white, male and middle class” (10).  In short, while parliamentary systems may hold opportunities for substantive representation in the form of policy outcomes, these remain largely unfulfilled because of the ways in which the system affects behaviour. To the extent that a relationship exists between the numbers and substance of women’s representation in parliamentary systems, it may be the number of women in cabinet that matter. These women, having satisfied lower-order objectives (nomination, election, promotion), may be more able to pursue their policy objectives and will have more institutional power to do so.  Presidential Systems: Veto Players and Indecisiveness  According to Lijphart (2012), consensus systems, defined in part by the balance of power between the executive and legislative branches, lead to “kinder, gentler” democracy. Consensus systems, Lijphart finds, spend a greater proportion of GDP on social expenditure than their majoritarian counterparts, and so we might expect that presidential systems lead to substantive representation in the form of policy. As argued, presidential systems, more than parliamentary systems, do enable and encourage the representation of women by women through votes, discourse, and the introduction of legislation.   89 Institutions, however, also structure the aggregation of these individual actors. Using Tsebelis’s (1995) veto player framework, Haggard and McCubbins argue that the extent to which policy decisions can be reached (decisiveness) and the extent to which those decisions are definitive (resoluteness) is a function of the number of effective veto players involved in decision making. A division of powers means there are multiple veto players and, as a result, decisions are harder to reach and the status quo more persistent. This logic can be extended to consider the relationship between women’s descriptive and substantive representation and the moderating effect becomes evident: even if the number of women increases dramatically in a presidential system where decisiveness is low but resoluteness is high, an increase in the provision of women-friendly policies may not follow. In other words, although the behaviour of individual women may include the substantive representation of women, the institutional structure may minimize the extent to which this is translated into policy outcomes.  Tables 2.4 at the end of this chapter summarize these arguments. It is difficult to make percise hypotheses on the affect of these two institutions on policy and legislative outcomes as the two ways in which the institution affects outcomes – by influencing behaviour and by determining the aggregation of behaviour into outcomes – do not operate in the same direction. For example, while Westminster parliamentary systems generate disincentives for representing women, if there are women within the executive intent on pursuing women-friendly policies, the parliamentary system offers an efficient means of doing so. On the other hand, while incentives and opportunities do exist for women in the American presidential system, the number of veto players can result in a status quo that is hard to change. In other words, there are significant reasons  90 to expect a limited relationship between the proportion of women in the legislature and policy outcomes in both presidential and parliamentary systems, though the exact institutional mechanisms differ. Given these contradictory expectations, the theory does not lend itself to clear hypotheses like those about the expected effect of institutions on individual behaviour. In fact, even if institutions moderate behaviour and even if they affect the aggregation of that behaviour, no relationship may be observable between the number of women and policy outcomes. This complex relationship may explain why so few studies look for a correlation between the number of women and policy outcomes in either presidential or parliamentary systems (see Chapter 1). As a result, the empirical work in later chapters is much more exploratory when testing the impact of institutions on policy than when testing behaviour. Table 2.5 at the end of this chapter summarizes these arguments. Moderating Outcomes: Electoral Systems Policy and Legislative Changes: Efficiency  Electoral systems affect the stability and efficiency of governments. While not without exception, PR tends to create minority or coalition governments while SMP systems produce single party majority governments (Gallagher, 2005). Similarly, systems that incentivize party votes lead to more cohesive voting than those that incentivize personal votes. In both cases, this leads to a greater number of active players who must agree on proposed legislation, making decision-making less efficient and the status quo more  91 persistent (Tsebelis, 2002). As a result, an increase in the number of women may not translate into policy outcomes, regardless of their individual levels of behaviour. Parties and Party Systems According to Duverger’s law (1959), the psychological and mechanical effects of SMP systems should result in elections and governments dominated by two political parties, and while there do exist notable counter cases, systems using SMP do tend to have fewer effective parties in legislatures (Gallagher, 2005). Downsian logic (1957) suggests that where two parties exist, they converge on the median voter, towards the centre of the political spectrum. Proportional representation systems, on the other hand, create opportunities for parties on the left (Cusack & Soskice, 2007). Because of the affiliation with between the left and policies related to women’s issues, proportional representation systems may lead to more substantive policy outcomes because of the parties elected rather than individual behaviour within the parties.  Institutions in the Real World The institutions as introduced are somewhat abstract and disjointed. This section adds flesh to that skeleton and describes the institutions used in seven countries – Belgium, Canada, Germany, the Netherlands, Switzerland, the United Kingdom, and the United States. These are the countries in which I conducted in-depth interviews with women representatives to test the theories. They do not, of course, represent all possible combinations of institutions, but there is important variation in the key independent variables – the electoral system, division of powers, and women’s political representation.   92 The ranking on the vote and ballot dimensions are based on the incentives for unincorporated representation and reflect the adjustments made to Carey and Shugart’s original concepts. Unless otherwise specified, information is based on the Interparliamentary Union Database (IPU) (2015), the Global Database of Quotas (2015), and the PARTIREP Dataset (2014). 12 Table 2.6 provides a summary. Belgium  Belgium is a parliamentary system that uses a proportional representation system to elect 150 members to the federal parliament. There are 11 multi-member districts that have between 4 and 24 seats. Ten of the districts coincide with the country’s ten provinces and the eleventh is made up of the Brussels Capital Region. Political parties do not run nationally but rather run either in Wallonia, Flanders, or the German-speaking region. Only in the Brussels Capital Region do voters have the choice between voting on a Walloon and Flemish ballot. Belgium is a bicameral system and since 2014 senators are no longer directly elected but rather are members of local and regional government or elected by their peers.  Incentives for unincorporated representation are moderate. Lists are compiled mostly at a regional or local level. Voters may choose between voting for a party and indicating a preference for one or more candidates within the same list. Preference votes are counted first as a vote for the party as a whole and position on party list matters more than the number of preference votes received. As a result, it is very difficult to disrupt party lists.                                                  12 The IPU was used as a source for world ranking of the representation of women, but the ranking themselves were reinterpreted. The IPU (2015) ranks countries that tie as taking up one spot. For example, Canada (as of September, 2015) was ranked 50th internationally but there are actually 61 countries in which women make up a higher proportion of representatives. To reflect this, I count Canada as 62nd.   93 Women make up 39.3% of the national parliament in Belgium, which ranks 18th internationally for women’s political representation. Since 1999 a quota has been used to increase the number of women elected. Specifically, every list must reflect gender parity and the first two candidates on any list must not be of the same gender. Lists that fail to meet these requirements are not accepted by the electoral authorities.  Canada Canada is a bicameral, parliamentary system and its 308 Members of Parliament are elected through a simple SMP system.13 The upper chamber is unelected and is meant to represent the provinces.  Incentives for unincorporated representation are minimal. Voters have one vote to cast and so cannot indicate preferences among candidates. Moreover, because voters face a ‘list’ of one candidate, votes are pooled for the party. Local riding associations do have some control over nominations, though it is not absolute.  At the time, Canada ranked 62nd in the world for women’s political representation with women comprising just 25% of MPs.14 There are no legislated quotas, although some parties, notably the NDP, have internal goals for representation and procedures for encouraging the nomination of women and minorities. Germany Germany is a parliamentary republic that uses a mixed-member proportional system (MMP). There are at least 598 elected members, 299 elected from single-member                                                  13 In the election following the interviews (held in October of the same year, 2015) the number of MPs increased to 338. 14 After the election, the representation of women increased marginally to 26%, putting Canada 60th internationally.   94 plurality constituencies and another 299 elected through list PR from 16 multi-member constituencies that correspond to the Länder. Lists for the election of PR MPs are closed and votes are cast at the party level. Seats are distributed using the Sainte-Laguë/Schepers method. To ensure that all parties receive the minimal number of seats that they are entitled to over-hang or balance-seats are possible. The upper house, the Bundesrat, is comprised of delegates from the Länder.  Incentives for unincorporated representation for those elected directly are similar to those in Canada – voters have a single vote to cast, and all votes are pooled, but there is some local control over nomination. Incentives for unincorporated representation are even less for those elected off the party lists – voters are pooled, lists are closed and established at the Land level, and the votes themselves are cast for the party and not for candidates. Germany ranks 25th internationally for women’s political representation and 36.5% of MPs in the Bundestag (lower house) are women. There are no legislated quotas, but most parties have voluntary quotas for the composition of the party lists. The Netherlands  The Netherlands is a parliamentary system that uses a PR system to elect the 150 members of the Tweede Kamer.  Incentives for unincorporated representation are minimal. Although there are technically 18 multi-member districts, the lists are nearly identical and established nationally making the whole country effectively a single electoral district. Voters have a single vote to cast for an individual candidate, but all votes are counted first as a vote for the party. Seats are allocated among candidates according to the order in which they appear on the list unless  95 an individual candidate receives enough preference votes to meet 25% of the Hare quota. These ‘preference votes’ rarely change which candidates are elected - between 2002 and 2010 just six MPs were elected solely because of the personal votes they received (Andeweg & van Holsteyn, 2011). The upper house, known as the First House or Eerste Kamer, is comprised of representatives elected by the 12 states-provinces.  The Netherlands ranks 22nd internationally for women’s political representation with women making up just over 37% of all representatives in the country’s lower house. There is no legislated quota, but both the Labour and the Green parties use internal party quotas.  Switzerland  The 200 members of Switzerland’s House of Representatives are elected from the country’s 26 Cantons. Five cantons use SMP and elections are conducted using proportional representation in the remaining 21. When proportional representation is used, incentives for unincorporated representation are high; multiple votes can be cast for lists and can also be used to modify lists by crossing out or repeating names or voting for members of multiple parties, a practice known as panachage.  The government of Switzerland is composed of seven members representing five different parties, with the president chosen on an annual, rotating basis. There are no votes of confidence and rejection of a proposal from the government either by the Parliament or through a referendum does not constitute a government crisis nor does it necessitate ministerial resignations. The upper house, the Council of the States, represents the cantons.   96 Just over 30.5% of legislators in the House of Representatives are women, putting Switzerland 44th internationally. There are no legislated quotas and only the Social Democratic Party has internal quotas.  The United Kingdom  The United Kingdom is a Westminster parliamentary system that uses an SMP system to elect its 650 members. Like Canada, incentives for unincorporated representation in the SMP systems are minimal.  The unelected upper house, The House of Lords, is mostly appointed by the Queen on the advice of the Prime Minister.  At the time of interviews, women comprised 22.6% of all MPs, putting the United Kingdom 71st internally. There are no legislated quotas aimed at increasing women’s political representation, but the Labour party employs all women shortlists (AWLS) in 50% of winnable ridings.  The United States  The United States is a presidential system that uses single member to elect the lower house – the House of Representatives. The upper house, the Senate, acts to represent the interests of States. In contrast to the SMP systems in Canada and the UK, there are significant incentives for unincorporated representation. The use of the primary system means that control over nominations is local, voters have more than one vote to cast at different times, and there is no pooling in the first round.   97 Of the countries included in this study, the United States has the lowest representation of women (less than 20%) and ranks 96th internationally. There are no legislated or voluntary quotas.    98   Table 2.1. Division & Fusion of Powers  - The Moderating Effect on Discourse & Behaviour  Incentives & Opportunities  Disincentives & Constraints Division of Powers  More freedom from party discipline and absence of incentives to follow party direction creates opportunities to address women’s issues, including in speeches and votes.   Women elected via support through EMILY’s list or similar organizations will have incentives to address women’s issues through speeches & statements.  Increases opportunities to propose legislation and legislative changes to reflect women’s interests.   Opportunities for cross party cooperation on women’s issues  Fusion of Powers Occasional “votes of conscience” may provide opportunities to represent women.  Significant opportunities for women in cabinet to introduce and pass legislation. To remain in caucus and access promotions to government, representatives incentivized to represent party interests, potentially at the expense of women’s issue. Re-election ambitions similarly incentivize focusing communication on party priorities. All else equal, party reputation will be more important than personal reputation.  Dissenting votes are very costly. Substantial disincentives exist for women to represent women or support policy through votes if doing so requires dissenting. More confrontational system disincentives cross-party cooperation on legislative proposals.  Party leadership control over agenda, including who makes a speech or asks an oral question, constrains opportunity for discourse address women’s issues. Fused executive and legislative branches mean government controls most introduction of legislation. This constrains individual women’s ability to introduce legislation.  99   Table 2.2. SMP vs. PR: The Moderating Effect on Discourse & Behaviour  Incentives & Opportunities  Disincentives & Constraints Single Member Plurality   “Too much emphasis on women’s issues in a campaign might lead to speculation that the candidate is narrow in her interests and would not adequately represent all the people” (Carroll, 1984, 319).   Limited issues and identities can be represented in limited opportunities for communication. Focus on geographic area constraints time and space dedicated to women’s issues.  Being seen to represent women’s interest at expense of geographic interests in votes cast may harm re-election.   No incentive to deal with issues brought to attention of representatives by constituents in other districts.  Proportional Representation  Lower electoral threshold mean women are more able to seek votes on the basis of representing women.   Opportunity to create and appeal to tie based on common experience & identity rather than geography. More able to dedicate time to addressing in speeches and other forms of communities women’s issues because responsibility to represent geographic area shared with other representatives.   With higher district magnitude & lower threshold, women may be able to make their votes, legislative proposals and speeches based on the opinions of their women constituents, rather than all constituents.    100    Table 2.3. Unincorporated Representation & Party Votes: The Moderating Effect on Discourse & Behaviour  Incentives & Opportunities  Disincentives & Constraints Unincorporated Representation / Personal Vote Systems Greater freedom from party direction on speeches and statements creates opportunities to focus on issues of choice, including women’s issues. Freedom from party enables legislators propose and to vote in favour of women’s interest.  Incentive to differentiate themselves from other candidates (in and outside their party) and women can do so on the basis of gender representation.    Party Vote Systems    Incentives largely about maintain party reputation through speeches and statements.  Tie to party for election and re-election creates disincentives to make legislative proposals on topics other than those dictated by parties. Similarly, legislators need to vote with party, regardless of position, to ensure access to ballot and re-election.  Dedicating time and resources to party agenda constraints those available to representing women.  Party controls access to agenda and legislative priorities to maintain party reputation.  101    Table 2.4. Division and Fusion of Powers: The Moderating Effect on Outcomes  Behaviour  Aggregation of behaviour  Expected relationship between descriptive representation & outcomes Division of Powers Higher levels of representation of women’s issues by individual women. High number of veto players leads to persistent status quo.  Moderate  Fusion of Powers Lower levels of representation of women’s issues by individual women. Government control over agenda, tight party discipline, ensures most government proposed legislation is passed. Minimal  102   Table 2.5. Electoral Systems: The Moderating Effect on Outcomes  Behaviour  Aggregation of behaviour  Expected relationship between descriptive representation & outcomes SMP  Minimal opportunities for the representation of women’s issues. Notable disincentives to represent women’s issues.  SMP likely to lead to majority governments able to efficiently change policy and legislation. Low.   PR Incentives and opportunities to represent women.  Minority governments create more veto players and preferences the status quo Moderate.  Personal / Unincorporated Representation Vote Based Systems High incentives and opportunities for representation of women’s issues More veto players due to more independent behaviour. Harder to depart from status quo and reach new decisions. Moderate. Party Vote Based Systems Disincentives and constraints for the representation of women’s issues  More cohesive voting and proposing of legislation, fewer veto players. Low  103 Table 2.6: Institutions in the Real World Division of Powers PR Vote Dimension Ballot Dimension Vote Pooling Incentives for Unincorporated Representation Belgium Parliamentary system.   Coalition Government Yes Multiple votes to indicate preferences among candidates  2/2    List very difficult to disrupt.   Most lists established at local or region level.   1/2 Maximum vote pooling.  All votes counted first as votes for the party  0/2  3/6 Canada Parliamentary system.   Majority government.  No While voters’ cast their single vote for a candidate, single they are voting for a party ‘list’ of one it is considered a single vote at the party level.  0/2  No ‘list’ disruption.  Significant local control over the nomination of candidates.   1/2 Maximum vote pooling.  All votes functionally counted first as votes for the party  0/2  1/6 Germany Parliamentary system.   Grand Coalition government.  Mixed  299 SMP seats and at least 299 PR seats Single vote cast at party level for PR list.  SMP seats: Votes cast for a party ‘list’ of one and thus is considered a single vote at the party level.  0/2  No ‘list’ disruption for PR seats  Some local control over the nomination of candidates.   1/2 – for SMP seats. 0/2 – for PR seats Maximum vote pooling.  All votes functionally counted first as votes for the party  0/2  1/6 – SMP seats  0/6 – PR seats        104 Division of Powers PR Vote Dimension Ballot Dimension Vote Pooling Incentives for Unincorporated Representation The Netherlands Parliamentary system.   Coalition Government PR Single vote cast for an individual candidate.  1/2    List very difficult to disrupt.   List established nationally.  0/2 Maximum vote pooling.  All votes counted first as votes for the party 0/2  1/6 Switzerland Unique political system. Division of Powers Mostly PR  5 SMP seats Multiple votes can be cast within and across parties.   2/2    Panachage enables list disruption   List established nationally, though require some support locally. 1/2  Maximum vote pooling.  All votes counted first as votes for the party  0/2  3/6 The United Kingdom Parliamentary system.    No While voters’ cast their single vote for a candidate, single they are voting for a party ‘list’ of one it is considered a single vote at the party level.  0/2  No ‘list’ disruption.  Significant local control over the nomination of candidates.   1/2 Maximum vote pooling.  All votes counted first as votes for the party  0/2  1/6 The United States Presidential System No   Multiple votes to be cast below the party level over time.   2/2 No leadership control over ballot.  List determined by primary. 2/2 Votes are not pooled during primaries or for MMD election.   2/2  6/6  105 Chapter 3: Testing the Theory - Substantive Representation in Presidential and Parliamentary Systems The separation of purposes, powers, origins, and survival generate greater opportunities for women to introduce legislation and collaborate with others to raise the profile of women’s issues and incentivize such behaviour as a way to seek electoral and financial support. By contrast, in parliamentary systems, the carrots of promotion, the stick of dissolution of parliament, and the overlapping electoral incentives of the executive and legislative branches of parties limit the opportunities for women to act independently to represent women.  This chapter tests these propositions using quantitative methods and two datasets that capture individual behaviour. The first dataset based on the PARTIREP comparative MP survey, includes self-reported behaviour from over 2000 representatives across over 70 regional and national parliaments. This dataset allows for several tests of the theory. First, I assess whether the representation of women by women and by men varies across presidential and parliamentary systems. In both institutions, women do more than men to represent women, but the gap is larger when there is a division of powers. Second, I assess the extent to which the institutions differently affect legislators in left and right parties. The results indicate that it is men’s behaviour that is institutionally dependent; in particularly men in the left do more to represent women when they are bound to their party in parliamentary systems. On the other hand, men in left parties take the freedom of presidential systems to do less to represent women. Finally, I explore whether the observed effect is due to institutions acting as filters – encourage some types of  106 individuals to opt into or out of politics – as incentive structures or both. The findings suggest institutions affect behaviour in multiple ways.  The second is a new dataset created for this research and analyzes the use of Twitter by representatives in Canada, the United Kingdom, and the United States. The analyses demonstrate that women, more than men, use Twitter to talk about women’s issues, but women in presidential system do so much more frequently. Moving beyond the statistical analysis demonstrates the influence of women’s organizations in the United States and formal, parliamentary roles in Canada and the U.K. Women who Tweet the most in the U.S. are heavily funded by groups like EMILY’s lists, while the most prolific Tweeters on women’s issues in Canada and the U.K. are government minister or opposition critics with official responsibilities for women. The analysis in this chapter is a matter of trade-offs in the face of imperfect data. The PARTIREP data includes a large number of countries and depth in terms of the measures of substantive representation, but is limited in that there are few examples of systems that use a division of powers, and is pure presidential.  The Twitter data, on the other hand, enables an explicit comparison between a pure presidential system and two Westminster parliaments, but is limited to only three cases. Despite the limitations in the data however, the results are consistently supportive of the theory. Chapters 5 and 6 take a qualitative approach to the question and allow for further triangulation of the findings. Throughout the chapter, however, I remain aware of the limitations of the data and discuss the implications of these limitations for the findings.       107 The Division of Powers and Individual Behaviour – The PARTIREP Data The PARTIREP survey, conducted between spring 2009 and summer 2010, used written questionnaires and reached representatives in 15 national and 58 regional parliaments.  The survey asked legislators about their thoughts on representation, their behaviour in parliament, and attitudes on democracy and governance. The dataset also contains a wide range of institutional variables based on publicly available sources. Details on the breakdown of respondents by country and institution are available in Appendix A, Tables A1 and A2. Except Israel, all countries involved were located in Europe and variation of division of powers is somewhat limited. In total, 37% of the representatives in the PARTIREP survey are classified as elected in a system with some level of division of powers. Although the division of powers variable does not correspond exactly with the presidential/parliamentary division as theorized in Chapter 2, it does still allow for an initial test of the theory.   The survey does not include any pure presidential systems and there are only three semi-presidential systems – France, Portugal, Poland, Austria. Austria, however, is an almost pure parliamentary system by convention (Muller, 1999; Shugart, 2005) and so is grouped with the parliamentary