Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Rethinking political thinking: gender and public opinion in Canada O’Neill, Brenda Lee 1996

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

Item Metadata


831-ubc_1996-148084.pdf [ 18.89MB ]
JSON: 831-1.0087667.json
JSON-LD: 831-1.0087667-ld.json
RDF/XML (Pretty): 831-1.0087667-rdf.xml
RDF/JSON: 831-1.0087667-rdf.json
Turtle: 831-1.0087667-turtle.txt
N-Triples: 831-1.0087667-rdf-ntriples.txt
Original Record: 831-1.0087667-source.json
Full Text

Full Text

RETHINKING POLITICAL THINKING: GENDER AND PUBLIC OPINION IN CANADA by BRENDA LEE O'NEILL B.A. (Hons.), Brock University, 1986 M.A., McMaster University, 1987 M.A., McMaster University, 1989 A THESIS SUBMITTED FN PARTIAL FULFILLMENT OF . THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Department of Political Science We accept this thesis as confoirrting to the required standard THE UNIVERSITY OF BR^ISH COLUMBIA' August 1996 © Brenda Lee O'Neill, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver, Canada Date Se.f>T>_Aer 3-^ 7 iWL DE-6 (2/88) Abstract This study argues that gender is a significant factor to consider in investigations of political opinions and presents evidence of the relevance of gender to support for various issues and in the social construction of opinion. Moreover, it argues that the patterning of women's and men's opinions, and differences in the sources of those opinions, point to a difference in political cultures: a women's political culture and a men's political culture. Using survey evidence gathered at the time of the 1988 Canadian federal election, the study follows three separate investigative paths in an attempt to uncover the existence of distinctive political cultures. The first path investigates gender gaps in opinions at the time of the election and links these findings to earlier work suggesting the existence of a women's agape ethos, their weaker hawkishness, and their weaker support of continentalism. It is shown that controls for women's lower average incomes, their lesser educational attainment, their greater support of feminism, and gender roles do not fully account for differences in women's and men's attitudes. Moreover, evidence is addressed of women's greater religious fundamentalism, which often works in such a fashion on attitudes as to attenuate gender gaps in opinions. The second path investigates the social structure of women's and men's opinions and finds that despite the similarity of opinion on a number of issues, divergence appears in the sources of opinion. The influence of economic self-interest, age cohort, region, social group memberships, religious fundamentalism and feminism are found to vary between women and men across a number of issues. The third and final path elaborates on opinion structure by the investigation of women's and men's belief systems, that is the connections between various opinions and the manner in which these connections are hierarchical. Although women's and men's belief systems are very similar, the positioning of feininist belief differs by gender. For women, regardless of their level of political sophistication, feminism is connected to the most basic ideological belief, economic liberalism. For men, however, ferninism is only connected with ideological belief among the politically sophisticated. The study links this evidence to the existence of a women's political culture and argues that it stems partially from each gender's socialization, but that it is a culture in transition. The weakening of religious belief generally is likely to result in larger gender gaps in opinion in the future. Table of Contents Abstract ii Table of Contents iii List of Tables v Acknowledgements vii Chapter 1: Introduction 1 Women and Politics in Canada 11 Gender Differences in Public Opinion 15 Organization of the Investigation 17 Chapter 2: Political Opinions, Mass Belief Systems and Gender 22 Distinct Political Worlds 31 Gender Gaps in Opinions and their Sources 35 The Social Structure of Opinions 64 Belief Systems 69 Framework for Proceeding 74 Chapter 3: Creation of the Opinion Indexes 78 Creation of Opinion Indexes 79 Chapter 4: Is there a Gender Gap in Canadian Opinion? 97 The Gender Gap: Past Evidence 97 The Gender Gap in 1988 106 A Gender Gap in Opinion 125 Chapter 5: Explaining Gender Gaps in Opinion 128 Value Pluralism and the Gender Gap Puzzle 128 Explaining the Gender Gap in Opinions 144 Comparing the Theoretical Models 167 Conclusion 181 Chapter 6: The Social Structure of Opinions Compared 184 Comparing the Social Determinants of Opinion: Results 187 Conclusion 226 -m-Chapter 7: The Belief Systems of Women and Men 235 Measuring Belief Systems 242 Results 246 Conclusion 263 Chapter 8: Gender Difference as Cultural Difference 266 A Women's Political Culture 267 Sources of a Women's Political Culture 277 A Culture in Transition 280 Future Research Directions 281 Conclusion 283 Bibliography 284 -iv-List of Tables Table 2.1 Potential Determinants of Political Opinions 75 Table 3.1 Variable Composition of Opinion Indexes 88 Table 3.2 Summary Statistics of Opinion Indexes for the Full Sample 95 Table 4.1 Gender Differences in Opinion Indexes 108 Table 4.2 Percent Agreement with Individual Issues by Gender 109 Table 4.3 Breakdown of Public Opinion on Continentalism Issues by Gender 119 Table 4.4 Association between Abortion and Other Indexes by Gender 123 Table 5.1 The Social Structure of Religious Fundamentalism: Regression Results 132 Table 5.2 The Social Structure of Support for Feminism: Regression Results 137 Table 5.3 Influence of Religious Fundamentalism and Feminism on the Gender Gap: Regression Results 141 Table 5.4 BreakDown of Part-Time Labour Force by Gender and Age, based on 1986 Census Data 147 Table 5.5 Family Income and Employment Status by Gender 151 Table 5.6 Gender Percentage Point Differences in Family Income by Marital Status 153 Table 5.7 Gender Percentage Pomt Differences m Family Income by Presence of Children in the Home 154 Table 5.8 Selected Socio-demographic Differences between Homemakers and Other Women 159 Table 5.9 Differences in Opinion Indexes between Home-makers and Other Women 161 Table 5.10 Shares in Various Socio-demographic Groups by Gender 165 Table 5.11 Explaining Gender Gaps in Opinion: Regression Results for Defence, Capitalism and Government Involvement 168 Table 5.12 Explaining Gender Gaps in Opinion: Regression Results for Extra Billing, Continentalism and British Ties 173 -v-Table 5.13 Explaimng Gender Gaps m Opinion: Regression Results for Women's issues, Civil Liberties and Energy 178 Table 6.1 Determinants of Opinion: Defence 188 Table 6.2 Determinants of Opinion: Compassion Issues 192 Table 6.3 Determinants of Opinion: Civil Liberties 199 Table 6.4 Determinants of Opinion: Continentalism 201 Table 6.5 Determinants of Opinion: Abortion Women's Issues and 205 Table 6.6 Detemiinants of Opinion: Dualism 209 Table 6.7 Determinants of Opinion: Separate School Funding British Ties and 212 Table 6.8 Determinants of Opinion: and Energy Environmental Priority 215 Table 6.9 Determinants of Opinion: Billing Taxation and Extra 218 Table 6.10 Determinants of Opinion: Union Support Immigration and 221 Table 6.11 Determinants of Opinion: Cynicism and Reform 225 Table 7.1 Factor Analysis Results: Women's and Men's Extraction Stage Results 247 Table 7.2 Factor Analysis Results: Women's Pattern Matrix 250 Table 7.3 Factor Analysis Results: Men's Pattern Matrix 251 Table 7.4 Factor Analysis Results: Women with College/ University Experience Pattern Matrix 260 Table 7.5 Factor Analysis Results: Men with College/ University Experience Pattern Matrix 261 -VI-Acknowledgements Many people helped make the writing and completion of this dissertation possible and I am extremely grateful for having received their help. I would like to thank the Political Science Department at the University of Alberta for generously providing me with a supportive work environment over the past year; a number of faculty members and staff in the department have offered advice, help, encouragement and suggestions, and for that I wish to thank them. I wish to thank the University of British Columbia for providing fellowship moneys that allowed me to work on this project for a number of years. Although a number of political scientists have provided encouragement and advice over the years, I would especially like to thank Sandra Burt at the University of Waterloo for her enthusiasm over this project and for helping me refine my tiiinking on women and politics. I also want to thank the members of my committee: Don Blake, Ken Carty (who I want to thank for among other things having made clear to me the need to complete this project), and my supervisor, Richard Johnston. This dissertation owes much to his ideas, suggestions and criticisms, and is, as a result of his input, a much better piece of work that it would have been otherwise. I am extremely grateful that he allowed me to work at my own pace, and yet provided his undivided attention when necessary. For this and more, I owe him animmense debt of gratitude. Finally, I want to thank my family for their support and encouragement over the years: first, to my second mother, Myrna Stewart, who generously flew in from Nova Scotia on several occasions to care for her granddaughter while I went off to work; to my sisters, Lynn Melna and Kim Tykolis, for being such good friends; to my Mom and Dad, Lise and Dan O'Neill, whose unending emotional and financial support I can surely never repay; and to my husband and daughter, David and Rachel Stewart, for keeping me aware of the things that really matter in life. I thank Rachel for being extremely patient with her mom while I finished my 'works'. And I thank David, not only for his ovemhelming contribution to this project as editor, sounding board and academic colleague, but for his unending love and support. Everything I am owes to your love, and I thank you. -vii-Chapter 1: Introduction "Another Mother for Peace'" Pretending that the sight of the message on this bumper sticker inspired this investigation would make for a wonderful story. Unfortunately, I saw the sticker only as this study neared completion, and must confess that five years ago its significance might have been lost on me. It nevertheless provides a useful focal point for introducing this study. This investigation is about gender and its relevance to political opinions. More specifically, the investigation centres on women's and men's thinking on a number of issues, the factors that shape opinion on those issues, and their belief systems more broadly. At its core is the examination of the assumptions made in current tiiinking on opinion formation and belief systems as they apply to women. Much of current thinking on this area of investigation is based on evidence gathered from mixed gender or on occasion from male-only samples and fails to take into account the particular forces, values and realities of women's lives and, as a result, may not accurately reflect this aspect of their political behaviour. There exists a wide range of research on the broader topic of the political values and opinions of Canadians. Some investigations address Canadian political attitudes Message on a bumper sticker seen in Edmonton on July 19, 1996. -1-directly (Johnston, 1986; Lambert et al., 1986, 1987, 1988; Simeon and Elkins, 1980; Sniderman et al., 1989; Nevitte and Gibbins, 1990), but in many cases research on public opinion has appeared as a secondary consideration in investigations of voting behaviour (Clarke et al, 1979, 1980, 1984, 1991; Frizzell et al , 1989; Kornberg and Clarke, 1983; Penniman, 1981, 1988). Only a limited number of investigations have tackled the more specific topic of gender gaps in voting and attitudes in Canada (Brodie, 1991; Everitt, 1994; Gidengil, 1995; Kopinak 1987; Terry, 1984). This investigation, by providing a comprehensive picture of the opinions of Canadian women and men at the time of the 1988 election, hopes to move towards filling this research gap in Canada and at the same time continue the process of integrating women into thinking on this subject. The working assumption of this study is that gender matters to political thinking. Women and men think about political questions in distinctive ways. Women and men bring with them to the political arena a distinctive set of core values, and different priorities, based partially on their gender. Whether these differences stem from biology or from the maimer in which boys and girls are socialized into particular gender roles is not a matter addressed in this thesis. Indeed, answering such a question is well beyond the method of study employed in this investigation. Instead, that differences exist in the values and opinions of women and men, differences that cannot be explained by social circumstances or demographic factors, is taken at face value to mean that each gender approaches politics from a different political culture. Thus the investigative framework adopted falls very much in line with mainstream evaluations of political culture in Canada. To assume that differences in opinions indicate differences in political culture is not a novel idea (Elkins and Simeon, 1979 and 1980). Applying it to gender differences in opinions and belief systems has not, however, been a preoccupation of Canadian political scientists (for some exceptions, see Vickers, 1988 and Bashevkin, 1993). The main goal, then, is to assess whether gender merits serious consideration as a culture defming factor in Canadian politics. As early as 1975 Thelma McCormack suggested just such a possibility: The alternative thesis is that women live in a different political culture from men, a culture based on differences in political socialization [...], differences in political opportunity structures, and the way in which the media of communication define each of them." (1975:25). Such an evaluative framework allows and provides a first step towards gaining a more thorough understanding of women's political behaviour as well as Canadian political culture more generally. That it has taken twenty years for the framework to be adopted in the type of investigation undertaken here suggests both the limited attention paid to this area of research generally, and the limited number of feminist researchers accepting quantitative methods as a legitimate method of inquiry for studying women. The male political culture is essentially the 'dominant' political culture. How men think about politics is reflected in the formal political arena. As the principal players in that arena, it stands to reason that the arena reflects men's values and beliefs. The female political culture, on the other hand, is less visible; the relative absence of women from formal politics, and the dominance of male norms in the political arena itself, makes it particularly difficult for this women's culture to materialize. For instance, in her reflections on the House of Commons, Sheila Copps, Deputy Prime Minister for the present Liberal Government, stated that 'from the moment you step inside, you sense that this place is foreign to women, alien to our spirit of cooperation, steeped in confrontation and simply not a place for traditional female virtues' (1986: 93). Kim Campbell, former Conservative Prime Minister, also discusses the difficulty she encountered in attempting to introduce 'inclusive justice' as a process of legal reform during her position as Justice Minister (1996).2 And Audrey McLaughlin, former leader of the Federal New Democrats, noted: When a women enters the House of Commons, she enters what in significant ways is an old-fashioned men's club. There are all sorts of reminders--some subtle, some not so subtle—that this is not a woman's place (1992: 26). Thus the feeling that the political world is not a woman's place exists across at least three of the main political parties in Canada and, moreover, among women at the highest levels of power in those parties. That women are less likely to be found in the dominant political arena and feel less comfortable in it explains in part the inability to report strong evidence that a women's political culture exists. And the lack of evidence explains in part the failure of traditional political science to account for and validate such a difference. 2 Inclusive justice is a process of four steps: 1) accept that the law does not treat everyone equally; 2) invite excluded individuals to discuss their reality; 3) invite those individuals to help in amving at a solution to the problem; and 4) implement that solution. Campbell noted the difficulty justice officers encountered in accepting the first principle, which denies the universality of justice principles, particularly in their application to women (See Campbell, 1996: 161). . 4 . Regional patterns of political cultures have been said to exist in the past given the cognitive and affective differences in flunking across the Canadian provinces.3 Such differences often dwarf any similar differences in triinking between women and men. On these grounds, the existence of 'gendered' political cultures is questioned. But regional differences in political thinking are supported by political institutions which provide channels for such differences to manifest themselves. The federal political system provides an institutional means of support for differences in political thinking. Daily political life is often characterized by inter-provincial and inter-governmental conflict. There is, on the other hand, no such institutional structure to support the existence of gendered political cultures. A government of, for, and by women does not exist. And given women's lesser involvement in many political institutions the failure to hear their distinctive political voice is understandable. I argue that it exists but that it is a whisper; you must listen closely to hear it. It is generally true that quantitative research on women's political behaviour is shamefully scarce. There has yet to be a comprehensive investigation of the manner in which women think about politics despite the recent increase in research on women and politics generally. As recently as 1983, Sapiro lamented the lack of research in this area: "Interestingly, female public opinion remains one of the most unexplored area of women's studies within political science" (150). While Sapiro might praise some of the growth in research in this area in the United States in recent years, recent studies of 3Such patterns have been less visible in recent investigations. See Stewart (1994). -5-women's public opinions in Canada are still few in number (see Everitt, 1994; Gidengil, 1995; Brodie, 1991). The lack of research in this area may be due in part to the general feniinist distrust of quantitative methods in particular, and the techniques employed in public opinion surveys more specifically. Whatever the reasons for the marginalization in women's studies, the study of women tends to be undertaken by women scholars. That many feniinist women scholars dismiss quantitative techniques as objectifying reduces the number of scholars willing to undertake such studies. But as I argue elsewhere (O'Neill, 1995), there is room for quantitative methods on the women's political behaviour research agenda and this present study hopes to take a step towards showing the merit of adopting this line of inquiry. One exception to the general mistrust of quantitative methods is the research undertaken to investigate the existence and causes of the gender gap. Research in this area employs quantitative methods and provides some understanding of women's particular opinions and attitudes and the unique forces generating them. The investigation undertaken here fits within this research agenda and implicitly accepts the benefits which can result from quantitative investigations of women's political behaviour. The limited attention paid to women's political thinking in Canadian research despite its importance may also stem from a limited availability of data sets, a problem which is quickly diminishing. As an example of the disappearance of such limitations, election surveys present an interesting opportunity for assessing a whole range of opinions. Given the scope of issues it touches upon, this investigation evaluates women's and men's political thinking using a number of the survey questions that make up a Canadian election survey undertaken in 1988. By virtue of the breadth of issues addressed, this examination of the Canadian women's political opinions will be the first of its kind. Most studies that have centred on the political attitudes of women have concentrated on the gender gap, that is on differences in the opinions held by women and men. Some scholars have extended the discussion to an attempt to account for these differences. I believe that limiting the focus to gender gaps is misplaced. Attempting to include women in this area of investigation means moving beyond simple comparisons and attempting to account for such differences. Focusing on opinion differences implies an acceptance of male behaviour as the norm and the standard of comparison (see Sapiro, 1983: Introduction). As such, women's opinions and attitudes are investigated in order to account for their deviation from men's and for no other reason. Such an inquiry does little to reinforce the belief that a women's perspective can have relevance for political questions. However, gender gap analysis should not be abandoned. The gender gap as a social phenomenon is deserving of attention for two main reasons. First, the gender gap phenomenon is linked to the larger question of whether women and men differ in their political opinions as a result of biological differences, self-interest, socially-constructed definitions of their 'proper' roles, or due to differences in the levels of social and demographic characteristics between the two groups. The question of gender difference is one that is not limited to the discipline of political science. But as it relates to political issues, research \on the gender gap presents an interesting opportunity for studying those factors that directly influence opinions about political issues and beliefs. It serves as a useful exercise for understanding the nature and source of difference between the genders. Moreover, it helps in understanding not only whether women and men differ politically, but also whether women act as a cohesive political group. Second, the gender gap phenomenon highlights the degree to which mainstream politics and political science has rested on male definitions of politics. Analysis of the gender gap has focused attention on issues of importance to women that have generally been ignored by the media and political institutions. Whether the gender gap exists or not, focusing on it provides an impetus for getting these issues on the political agenda. It also serves as a sharp reminder to political parties that over half of the country's electorate is made up of women; as of yet, this reality only seems entertained when gender gaps have the potential for guaranteeing (or denying) electoral success. If women's values differ from those of men they are nonetheless politically legitimate and deserving of political attention. Investigations into the opinions and beliefs of women must extend beyond the limited focus of gap analysis. Inquiries need to investigate the underlying values and guiding principles of women's attitudes and behaviour, as distinct from men's, in order to ensure that we move away from the tendency to discuss why women do not match men in their attitudes. Such a shift in investigative focus may well provide the basis for legitimating a differing conception on the nature of politics, a conception that redefines rationality and self-interest. Differences may exist at the level of the determination of opinions, in their sources, and in their structure. It is also true that gender gaps in opinion are normally small in size. Focusing on them necessitates concluding the difference is one of degree and not one of kind. But the underlying differences may well be a matter of kind, hidden from the view of the researcher concentrating solely on gaps in opinions. Women and men may come to hold similar opinions for very different reasons. The gender gap framework for proceeding discounts this possibility by emphasizing an investigation of small opinion differences only. This investigation attempts to move beyond gender gap analysis and towards a comprehensive evaluation of women's and men's opinions, their values and their beliefs. It also attempts to account for their opinion sources and outline the broader structure of opinion (that is belief systems). It will show the issues which reveal the greatest differences in thinking between women and men, as well as those that reveal little gender difference. It argues that the revealed differences are large enough to be considered indicative of a difference of political culture. Evidence will also be provided that these differences in thinking cannot be wholly accounted for by differences in economic circumstances or by differences in other socio-demographic factors. And importantly, it will show that women's private lives are as relevant to the detennination of their political opinions as are men's. But what makes this investigation of women's opinions different from others that have been previously undertaken is the argument that part of the reason for the lack of larger gender gaps on some issues is a competition of fundamental values. Fenrinism has provided many women with an important vehicle for speaking 'in their own voice' and its liberal influence has been recorded on many issues. It stands as a core belief for many women and thus shapes their opinions on many other issues. But investigations have rarely taken account of the fact that women are also more conservative in religious belief than men. And religious belief is no less central to the determination of opinion on some issues than feminism. Thus religious beliefs and support for feminism compete for dominance on some issues; namely, those for which religious values come into play. The end result is that an underlying division among women is masked. This is, however, only one part of this investigation. Women and men may think differently about some issues, and showing that such differences cannot be 'explained away' is a starting point. The next step is in showing that even when women and men are in agreement on issues, this agreement may nevertheless rest on very different grounds. Thus the investigation will show that gender differences extend beyond opinions themselves and are also found in the social agents that transmit political values and political culture. It will be shown that the reasons behind women's and men's political thmking vary even when women and men are in agreement. Often when they agree the construction of their opinions differs. The final part of this investigation will show that differences, although smaller in -10-size, also exist in the belief systems of women and men. Slight differences are to be found in the way that women and men relate each of their political values and beliefs. The failure to find larger differences in this element of political thinking should not be taken to mean that a women's political culture does not exist. Instead, the inability to detect greater differences in this element of thinking, it is argued, stems from the dominance of the male political culture and the relative exclusion of women from the formal political arena. Women and Politics in Canada The point has been raised that women, despite the lack of formal restrictions, have not entered the arena of politics in the same numbers as men. Their exclusion from this arena constitutes an integral element of the arguments made in this investigation. This section will sketch out the manner in which women have been politically excluded in Canada, as well as the reasons which have been advanced for explaining their lesser involvement. With the exception of voting, women in Canada do not engage in politics at the same levels or in the same numbers as men. Thus politics is dominated by men and the world of politics remains very much the domain of men.4 An understanding of the 4 That women remain informally excluded from the arena of electoral and representational politics does not mean that their involvement is limited in social organizations not traditionally understood to be engaging in politics. The study of such institutions is developing and suggests that women have been and are more political than traditionally believed. See Vickers (1988) and especially (1996). Nevertheless, their involvement in traditional political activities is limited and begs the question: is this exclusion due to -11-limited nature of Canadian women's involvement in representational politics is important to the investigation at hand, as it provides a setting for an evaluation of the political opinions of women in Canada. Gender differences in political behaviour suggest that women and men relate differently to their political worlds, although some of these gaps appear to be narrowing with the passage of time. Such differences may be accounted for by structural factors, but nevertheless highlight how the political opportunity sfructures that women in Canada face vary from men's and result in part in women's exclusion from the political arena. Differing rates of political participation and in the choice of level of political participation have been recorded and show that in the arena of Canadian politics, positions of power are held by men (Brodie, 1985,1991; Bashevkin, 1985,1991). Regardless of the reasons for their exclusion, women are likely to have reacted politically, in some manner, to it. In the first instance, Canadian women are less likely to be interested in politics than are men, and they are more likely to feel that politics is too complicated for them. Low levels of political efficacy are likely to partly explain women's lack of involvement in the system. But as Brodie (1991) points out, this weak orientation to the political system is partially the result of the political system itself. The political system fails to address many of the issues that are relevant to women's lives (such as child care and the gendered division of labour) while social and political realities limit the degree to which women can become involved in the system. That they reveal less interest in the system informal structural barriers or might it be that the culture of electoral and legislative politics is so at odds with women's political culture that many choose not to engage in it all? -12-should not be surprising. A number of differences also appear in the rates of electoral success between the genders. Despite the fact that women made up nineteen per cent of federal candidates in the 1988 election, they made up only thirteen per cent of candidates who were successful in their electoral bid (Brodie 1991: 5). Women made up a limited share of those individuals that ran for elected office, and an even smaller proportion of legislators. Their electoral success did not match that of men but this was not due to an unwillingness on the part of Canadians to vote for women candidates. It was instead partly due to the disproportionate concentration of women candidates in minor parties with an overall weaker chance of electoral success. As well, those fewer women who did secure nomination in major parties were likely to be nominated in ridings where the party had little chance of success. The major obstacle for women becoming successful electoral candidates, or electoral candidates at all for that matter, is money. Many women simply do not have the personal financial resources or the money network connections presently required to mount a successful electoral campaign at the federal level in Canada. Although the same is true of most men in Canada, it is likely to be true of more women than men. In 1988 a full 37 percent of women candidates ranked funding as the major obstacle in their campaign and slightly more than one-half indicated that money was an obstacle in gaining the nomination (Brodie 1991:44-5). If we can assume that women are at a financial disadvantage, then their limited numbers as electoral candidates may have less -13-to do with lack of interest than with realistic assessments of their chances of success. The concentration of female candidates and office holders at the municipal level provides further support for such a conclusion (Brodie, 1985; Vickers, 1978). Municipal politics involves fewer disruptions to family life, as office holders are not required to move their families or leave their families for extended periods in order to attend legislative sittings. Vickers (1978:46) attributes women's greater presence at the municipal level to two further factors: first, the reduced financial costs of mounting an electoral campaign at this level; and second, the reduced level of "power and influence" in these positions lowers both the degree of competition for them and the amount of party control exerted over them, which result in their increased accessibility to women. Thus the absence of parties in most Canadian municipal elections removes a barrier that often inhibits women from electoral success. It is also true that municipal political careers incur far less family life disruptions than do federal and provincial ones and as such provide a viable option for women with small children in the home. A glance at political party organizations similarly reveals the distance that remains before women gain equal representation within the Canadian political system. Referring to the tendency as the 'law of increasing disproportion', Bashevkin (1985,1989b, 1991) reports that women are overrepresented at the lower levels of party organizational hierarchies, and underrepresented at their higher levels. The higher up ones moves in the party hierarchy, the fewer women one is likely to find. Although women have begun to occupy positions at the higher levels of party organizations, the existence of this 'phik--14-collar ghetto' in party organizations continues despite the erosion of this female volunteer base due to the greater numbers of women entering the paid work force (Bashevkin, 1991:62). According to an individual interviewed by Bashevkin, "[there exists] far more opportunity now for women to do what used to be the men's jobs in the ridings, but little change in the willingness of men to do what have traditionally been the women's jobs" (1991:64). Women are increasingly gaining the skills required to sit among the party brass, but the opportunities are not often made available to them. The end result is that women continue to do the work that empowers them little within the party, and even less in the broader political system. Such differences in participation rates do not however include voting participation rates, where women and men prove to be quite similar in behaviour. Using CNES data from 1965 to 1988, Brodie reveals that only once did a statistically significant difference appear in gender voting levels (it appeared in 1984) and, even then, the difference was less than substantial: 3 percentage points (1991: 18). Women cannot be considered less politically motivated given their voting rates. But the political opportunities and 'choices' made available to them limit the degree to which the political sphere can be considered open to them. Perhaps somewhat symptomatic of this exclusion, their political lives, particularly their political opinions and beliefs, deviate somewhat from those of men. This exclusion may also stem from the fact that many women do not see their particular concerns reflected in the formal political arena. -15-Gender Differences in Public Opinion Research has added to the larger body of knowledge regarding women's political behaviour by investigating women's political attitudes and more specifically the gender gap. Much of this newer research is undertaken by women scholars whose explicit goal is that of "bringing women in" to the discipline. Moreover, these scholars are explicit in their desire of oveitAirning the many stereotypes and broad generalizations made about women, their political behaviour and their political opinions. Part of the stereotyping of women that occurred in traditional political science characterized them as more conservative, the guardians of morality, and idealist in their value structures; these largely unsubstantiated claims were held to account for any differences between them and men (see Vickers, 1987). Similarly, the tendency to extrapolate from women's private roles and biological roles to their broader behaviour patterns characterizes much of the early political science behaviour research. All of this took place often alongside the tendency to see women as apolitical: their behaviour and attitudes were often assumed to stem from the influences of husbands and fathers. As such women's roles as "housewives", mothers, wives and women were assumed, again with little supporting evidence, to account wholly for their behaviour and attitudes. It is understandable that many feminist scholars are preoccupied with moving away from such stereotypical portrayals of women. Others however are less willing to discount the notion that women's particular private roles may play some part in providing them with a particular point of view. And still others acknowledge and indeed elevate -16-women's biological and social roles above those of men. An argument is sometimes made that there is little reason to discount women's attitudinal differences so long as they are not then held as justifications for restricting women to the particular public and private roles most appropriate to them. Equality of the sexes need not demand an absence of difference. This investigation stems in part from an awareness of the limited degree to which women enter into many of the current investigations into public opinion in Canada. It also stems from a personal awareness that women are as complex and cross-pressured as men and that these multiple pressures are likely to shape their opinions as much as is true of men. Women are more than homemakers and mothers. And many women struggle to define and reconcile their feminism and their religious beliefs. The investigation also stems from an intuitive sense that women bring something unique into their thinking about political questions. To quote one feminist scholar at length, The difference between the content of thought for woman and man is drawn from their relation to a specific kind of body, to a specific recent past (since their birth), a specific inherited past (within their culture), a present context, and future options....We could say that the pathway to information differs according to sex. The lived experience of the body is different for a man than it is for a woman; being brought up in the Western world as female or male brings to consciousness an entirely different set of data; being born into a world of stereotypes, archetypes, and historical facts provides a different range of information for a man's reason to appropriate as a specific history than for a woman. (Allen, 1987: 11-12) The assumption of difference is to be found at every path in this investigation. Nature or nurture, despite the intriguing aspects of this question, is not the point of this investigation. The goal of this investigation is to reveal that women's political thinking is -17-distinctive and grounded in more than their private sphere roles. Moreover, it deserves full treatment in political science research. Organization of the Investigation The task of the investigation is to empirically explore the opinions and belief systems of women and men at the time of the 1988 national election in an effort to uncover the existence and potential sources of gendered ways of thinking about politics. Three distinct investigative paths will be employed in this effort: first, women's and men's opinions will be compared across a range of issues in order to determine whether gender gaps are a general characteristic which extend across the range of opinions tapped by the election survey; second, the evaluation will proceed to investigate whether traditionally evaluated sources of opinion need to include those that are of particular relevance to women and which have not as a rule entered such investigations; and third, women's and men's political belief systems will be mapped in an effort to uncover differences in those ideas or core beliefs that are central to the determination of opinions. Chapter 2 provides a framework for the investigation drawn from prior research and feminist theory that allows for the integration of women's particular public and private roles with their political cultures. It is simply not enough to undertake a traditional investigation of the sources of political opinions separately for each gender. Trying to understand the sources of women's opinion in this manner would do little more than treat them as men in drag. Rather, the investigation will include each of those -18-factors that may be of central importance to the determination of women's values but which have been omitted from traditional research in the area given their assumed irrelevance to political culture. Insight into which factors may be of particular importance for women will be drawn from feminist theory. Armed with a framework for proceeding with the empirical investigation, Chapter 3 introduces the data set to be used in the investigation. The data are drawn from a survey of Canadians undertaken immediately preceding and after the 1988 federal election. The breadth of issues addressed within the data set included in the study makes it a vehicle particularly well-suited to the investigation at hand. Chapter 3 will also outline and explain the set of issue indexes, created from the variables in the election study, that form the basis upon which the empirical evaluation rests. The creation of a smaller number of indexes from the large and unwieldy set of individual variables included in the study allows for a more intensive investigation into the opinions and beliefs of men and women. Chapters 4 through 7 form the bulk of the empirical investigation. Chapter 4 begins by detailing the existence of gender gaps in the range of issues tapped by the opinion indexes. It also shows those issues on which women and men are in agreement. Moreover, it reveals the differences that exist in women's and men's support of feminism and in their religious beliefs. Armed with this information, Chapter 5 sets about uncovering the likely sources of these differences in opinion. In doing so it reveals the relative strengths of the competing theories outlined in Chapter 2 claiming to explain such -19-gaps. Feminist beliefs, economic self-interest, and social group memberships each explain part of the gender gap puzzle. But their ability to explain gaps in opinion is limited. Evidence is also provided that on some issues religious belief competes with fenvinism for priority in the determination of opinion and thus mediates the size of the gap found on those issues. Understanding how women and men come at each of their opinions provides greater insight into the relationship between gender and the development of political cultures. Thus Chapter 6 extends the investigation on the gender gap to an evaluation of how women and men come at their opinions more generally. Every index, not simply those on which gaps appear, is examined with respect to the social structure of opinion on the issue, and women's and men's opinions are investigated separately in order to reveal any distinctions in this element of their opinions. The results of this investigation show that on many issues women and men reveal differences in the roots of their thinking on political questions. The differences are subtle but are nevertheless found to exist. There is, however, no simple pattern to these differences; distinctions in structure vary with the issue addressed. Chapter 7 moves away from studying individual opinions towards an evaluation of the connections between opinions, that is towards belief systems. The goal of this chapter is to isolate the core beliefs that underpin more specific opinions in an effort to uncover whether similar principles are at work for each gender and to compare the level of constraint and coherence that exists among these political beliefs. This investigation -20-reveals that women's and men's belief systems are similar in most respects and deviate in only minor ways. Interestingly the sharpest difference between women and men on this element of political opinion comes in the positioning of feminist belief relative to other opinions. In this one element of their belief systems, women and men are at variance. No normative assessment is made regarding which ideology or belief system is 'best'; the point is simply to determine how each gender manages within a complex political system. In the final chapter, the evidence generated on each element of political opinion will be pulled together in an effort to show that women and men are members of separate political cultures. In addition, some speculation on the sources of this particular culture will be made as well as suggestions for further research. That women must be included in such analyses is for the most part understood and accepted by many scholars; how the process of integration should occur is less well known. In the end, therefore, this investigation is not so much about women and the sources of their political opinions as it is about ensuring that every aspect of men's and women's lives, private and public, be included in investigations of political behaviour. -21-Chapter 2; Political Opinions. Mass Belief Systems and Gender This chapter will provide the theoretical background for introducing gender into research on public opinion and belief systems. Gender has appeared in some recent investigations of public opinion in Canada but is often only superficially discussed and then subsequently dismissed as statistically insignificant and thus irrelevant. This investigation contends that gender is relevant and applicable to discussions and investigations of public opinion. Women and men live in different but overlapping political worlds. Gender influences the groups that one is likely to join, the positions that one is likely to hold in those groups, the opportunity structures one faces and the resources one will bring to the political arena. The political socialization processes of women and men differ, as do the roles they take on in adult relationships. As a result the cues that women and men look to for help in determining their political opinions and the values that are brought to bear on this element of their political behaviour are likely to differ. It is possible that the manner in which the whole of their political opinions are organized and logically grouped together will also vary. If women and men live in separate political worlds, and consider different sets of values to be of fundamental importance, then the whole of their political outlooks will be shaped differently. One could consider this difference to be one of political culture. Political culture consists of 'assumptions about the political world' (Elkins and Simeon, -22-1979: 127). As a detemiinant of political attitudes and opinions, political culture restricts the range of alternatives entertained by those within its collectivity. If the assumptions that women and men make about the political world and the range of alternatives they entertain vary, then their opinions and attitudes are likely to reflect this varying perspective. Jill Vickers focuses directly on this duality in emphasizing that when women engage in political activity it is to 'get things done' while for men it is to 'get ahead' (1988). This duality, subtle and difficult to document, but nevertheless relevant, may parlay into differences in attitudes. It is imperative that an emphasis on such differences does not deny that women and men share many of the same social and political experiences. The dominant world, which equates in many respects to the male world, is nevertheless one in which women live and which affects many elements of their lives. Women and men attend the same churches, live in the same cities, towns and provinces, attend the same universities and are represented by the same Members of Parliament. Many of the cues that women and men look to in the determination of their opinions are similar. Despite one's normative assessment of the influence of this dominant world on women's views, its impact is nevertheless felt by women and must be considered a relevant determinant of their political behaviour. As such, women's political worlds,may reflect a bridging of two cultures: the dominant male culture and the less visible but nevertheless real women's culture. In order to deteimine whether it is realistic to speak of different political cultures for women and men, three individual elements of political opinion will be investigated. -23-Each provides a distinction framework for evaluating gender and public opinion and hence increases the confidence that can be placed in the finding of distinctive cultures. The three elements of public opinion to be evaluated are: •gender gaps and their sources; •the socio-demographic structure of opinion; and •belief systems. The phenomenon of attitudinal difference between the genders, commonly referred to as the gender gap, provides the investigation's introductory focal point. In a number of studies, women and men are found to differ in their opinions on issues. The first task is to assess each gender's political standing on a number of the issues at the time of the 1988 federal election in order to determine where gender gaps exist and where they do not. Gaps on proximate policy questions would not point to the existence of fundamental gaps in opinions; gaps in core political values would indicate that the difference is more radical. It is also necessary to examine as broad a scope of issues as possible for the existence of gender gaps, in order to determine the extent of difference in opinions. The number of issues examined in this investigation sets it apart from those previously undertaken. The scope of issues addressed here is unparalleled in previous gender gap research. Documenting the existence of gender differences in opinion is an important first step but charting these differences is not sufficient. An understanding of women's and men's political cultures requires an understanding of what accounts for such gaps in attitudes. As such, the second task is to uncover the sources of gender gaps in opinions. -24-A number of possibilities have been introduced in previous investigations of the gender gap in opinions. Researchers have taken direction from the literature on gender difference in their attempts to account for differences in opinions. But very little of the gender difference literature is directed specifically at examinations of political behaviour. Instead, much of it focuses on general epistemological questions and broader analyses of psychological and social behaviour. The gender difference theoretical framework is nevertheless directly applicable to an investigation of women's behaviour since it identifies and investigates the potential" sources of a women's perspective. Previous gender gap research provides some direction in applying these theories to examinations of political behaviour. Previous studies of gender gaps in opinions have also taken some direction from mainstream political science in an attempt to uncover the sources of the gender gap. Such studies have found that gender differences in socio-economic factors such as age, education and income account in part for the differences discovered in women's and men's opinions. Women's greater economic vulnerability, it is argued, accounts for their greater support of the welfare state. Others have posited that it is a difference in women's and men's core political values that may help to explain differences in attitudes. Political values have not, however, prominently figured in previous gender gap investigations despite the fact that mainstream research points to their importance in the shaping of opinions on more proximate political issues. The ability to document that gaps exist in part because of differences in women's and men's core values, net of any other independent pressures on opinions, will support the existence of gendered political -25-cultures. The second element of public opinion to be evaluated in this investigation is the socio-demographic structure of opinion. Gendered political cultures must rest on more than simply differences in the degree to which women and men support or oppose a number of political issues. A more forceful pronouncement of cultural difference depends on the existence of a difference in the manner in which women and men come at their opinions, that is, in differences in the roots of their opinions. The existence of gendered cultures depends on gendered ways of knowing. To investigate the possibility that women and men come at their opinions differently, a number of factors that might provide some insight into how women come at their opinions will be statistically evaluated for their role as opinion determinants. Those factors that have traditionally been seen as determinants of opinions—core values, life experiences, self-interest and group memberships—should not be dismissed as determinants of women's opinions. The sharing of the common dominant culture will necessarily mean that they will have some influence on women's as well as men's opinions. But the roots of women's and men's opinions may differ in the relative impact of each of these factors. That is, similar life experiences and group memberships may result in different pressures on opinions. Such an analysis is distinct from one attempting to explain gender gaps in opinion in part because it focuses on the determining factors that are at play in all opinions rather than focusing only on the sources of differences in a smaller set of opinions in which gender gaps appear. Explaining the source of the gender gap is important in and of itself, but one should not overlook cases in which the opinions -26-of women and men are congruent. This coming together of opinion may nevertheless be the result of a differing set of cues for the two groups. Women and men may agree on some issues but for very different reasons. Accounting for the structure of each gender's opinions will provide a basis for assessing the degree to which women and men think about political questions in different ways. Third, the examination of belief systems is also relevant to this investigation in that it allows for a determination of the central principles underlying each gender's attitudes. In contrast to the investigation of the social structure of opinion, belief systems research posits that ideas, rather than socio-demographic factors, are the principal shapers of more proximate opinions on issues. Ideology rather than group evaluations may account for an individual's attitudes on issues. The determination of a difference in the manner in which women's and men's opinions are structured by core ideological beliefs allows for a further testing of the proposition that women's and men's political cultures differ. Belief systems are centrally based, widely applicable, elaborate structures of attitudes (Campbell et al., 1960). These structures are organizations in that there exists an interdependence among their constituent attitudes: attitudes cluster into groups. The interdependence among attitudes in these clusters is assumed to rest on a central, or small number, of fundamental beliefs that are at play in the formation of not only political but social attitudes as well. Belief systems research thus moves beyond singular evaluations of political opinions and rather focuses on the connections between diverse sets of attitudes in an effort to uncover underlying beliefs and principles. An assumption in belief systems research is that individuals adopt an underlying -27-political philosophy with a few basic principles, and then evaluate issues and render opinions on those issues based on references back to those fundamental principles. The most commonly investigated source of constraint is the liberal-conservative ideological continuum which provides a "contextual grasp of politics that permits a wide range of more specific idea-elements to be organized into more tightly constrained wholes" (Converse, 1964: 227). Thus the greater the association among a set of opinions, the greater is the degree of ideological constraint among them. Key to this theoretical model is the two-tiered nature of the belief system, that is the movement from ideology to opinion with nothing in between. Investigations into belief systems are not new. They have a lengthy history highlighted by the changes in focus that often accompany any research path. The twist that I apply to the discussion is an attempt to introduce into belief systems research the idea that women and men may structure their attitudes around a different set of fundamental values. If women and men come at politics with different value frameworks, then these value frameworks could result in different belief systems. At the top of the two-tiered system for deteraiining opinions may lie a different set of criteria for evaluating political phenomena. These differing criteria are explained in part by differences in the political cultures of women and men. Establishing a framework for evaluating each of the three elements of Canadian women's and men's opinions necessitates an examination of both the literature on gender difference and the literature on public opinion and belief systems. As such, whenever possible the literature concerning gender difference will be introduced and discussed in -28-an effort to "bring women into" research in this area. This literature will hopefully reveal those factors that are unique to women's lives, and that should be introduced into the investigation of women's political opinions. On the other hand, gender differences in opinions may reflect notiiing more than different life experiences or differences in socio-economic factors. Thus in an effort to account for various sources of opinion, the investigation will draw not only from the gender difference literature but also from current research on public opinion in Canada. A number of socio-demographic factors have been investigated and suggested as important components in the shaping of public opinion. Gender differences in opinion may reflect differences in the levels at which women and men appear in each of these categories, and as such must be included in the investigation. It is imperative that both literatures be explored in order to properly model women's and men's opinions and belief systems. The task of the present chapter is to provide a framework for evaluating each of the three elements of public opinion in an effort to uncover whether cultural difference exists. Thus a framework for the investigation of gender gaps and their sources is the first undertaking. A number of factors have previously been examined as potential determinants of the gap and these will be outlined in order to set the stage for the empirical analysis to come. After providing a framework for explaining the gap, I will then sketch out the process for examining how each of the variables introduced as potential determinants of the gender gap structure individual attitudes across a full range of issues. Evidence of a differential social structuring of opinion will further support the conclusion of a gendered political culture, and hence necessitates this area of -29-investigation. Finally, having outlined the framework for an examination of the gender gap and the social structure of opinions, a review of the current state of research on the existence and properties of belief systems among the mass public will be undertaken to set the stage for the introduction of gender into this research area. Political opinions do not exist in individual vacuums, but rather are related and influenced by other opinions and beliefs and thus should be investigated as a group as well as individually. Undertaking separate examinations of women and men on this element of their political behaviour is a relatively new research direction. It may or may not shed light on the nature and source of difference in women's and men's approaches to political questions. It is also necessary to elaborate on some of the terms to be used throughout the course of this investigation. I follow Rokeach's lead in my use of the term opinion rather than attitude, belief or value. According to Rokeach,.an "attitude is a relatively enduring organization of beliefs around an object or situation predisposing one to respond in some preferential manner" (1968: 112). Thus the building blocks of attitudes are singular beliefs. Beliefs can be either descriptive, evaluative, or prescriptive, and will predispose individuals to particular courses of action. Values, on the other hand, are a particular type of belief, centrally located in an individual's overall belief system, that prescribe behaviour. Values are "abstract ideals, positive or negative, not tied to any specific attitude object or situation, representing a person's beliefs about ideal modes of conduct and ideal terminal goals" (124). Individuals will hold many beliefs, fewer attitudes, and even fewer values. Finally, Rokeach defines an opinion "as a verbal expression of some -30-belief, attitude or value" (125) and thus the term is more inclusive than the previous ones, and more applicable to the study at hand since all three will be included in this examination of public opinion in Canada. D I S T I N C T P O L I T I C A L W O R L D S Before proceeding with a review of the literature applicable to this investigation, it is important to provide some justification for claiming that women and men live in different political cultures. Gender is not often seriously considered a relevant factor in analyses evaluating the impact of socio-demographic characteristics in the shaping of attitudes. The argument made here is that gender is relevant to evaluations of public opinion and belief systems, and must be introduced into such evaluations in order to make them more applicable to discussions of women's political behaviour. Political culture in Canada may vary with gender, and finding evidence of gender differences in political opinions, their structure and belief systems would be an important first step toward supporting such a conclusion. The notion that women and men may differ in their political orientations has been entertained by others. There is a growing literature documenting the differences in the political behaviour and political outlooks of men and women. This literature suggests that to a limited degree, the two groups live in different political worlds. The use of the term limited is explicitly meant to underline the extent of difference that is suggested; focusing the investigation on Canada necessarily restricts the degree to which the political worlds of men and women will and can differ. Canadian women and men share a number -31-of political rights, freedoms, and benefits that alone confine the extent to which their political experiences and by extension, their political cultures, can be said to differ. But that small and consistent differences exist in light of such similarities makes the investigation of those differences all the more necessary, and all the more interesting. At the first level, there is wide agreement among feminist theorists that women and men differ in their fundamental values although there is less agreement on the source of this difference. A difference in fundamental values should translate into a difference in opinions and in belief systems. Carol Gilligan's work, In a Different Voice (1982), despite its empirical imperfections, remains an influential and intuitively appealing account of the basic difference in the fundamental values of women and men. When confronted with moral dilemmas, women and men tend to interpret them differently, 'in different voices'. In brief, Gilligan's work posits that moral reasoning in women is based on the notion of responsibility - 'the ethic of care'- rather than on the premise of rights. Thus a central underlying value among women emphasizes connection and relationships -political 'collectivism' rather than 'individualism'. In contrast, male moral reasoning is said to be imbued with values directly connected to individualism. Individual rights, rather than responsibility for others, is the basis for moral reasoning. Moral questions are interpreted as competitions between individuals and their rights, as disjointed rather than connected elements. This provides a sharp contrast to the basis of moral reasoning in women. While not directly 'political' in focus, Gilligan's work presents an important starting point for an investigation of the structure of opinion and belief systems in -32-Canada. If women and men connect their political attitudes in unique ways, the underlying value differences posited by Gilligan may present a possible explanation for such differences. Women's greater reliance on collectivism and responsibility as a basis of moral reasoning may extend to their political evaluations. If true, the structure of political attitudes would best be interpreted through such values. In contrast, the reliance on more abstract terms such as individualism and rights would be reflected in the structure of men's attitudes. Sandra Burt (1986) provides some further supportive evidence that women and men conceive of politics and representative democracy in unique ways. In her analysis of a small sample of intensive interviews, she found that women "explicitly called for a more humanitarian approach from political leaders, more caring for others and more responsiveness to citizens' requests than exists" (77). This suggests that at the most basic level there exist intersex differences in definitions of the political system and individual positions within it. The gender gap in voting and attitudes also provides reinforcing evidence that women and men are political creatures of a different stripe. The gender gap evidence suggests that women and men, based on their differing political views, have different sets of political priorities. The issues on which gaps appear include nuclear weapons and defence, government spending, welfare policies and the welfare system in general. The general trend is that women are more 'liberal' in their political outlook. They endorse programs of government spending to aid the poor and disadvantaged, and are more likely to adopt 'dovish' opinions on war and peace issues. Such differences have appeared in -33-the general public, and also among delegates to party conventions (Brodie, 1988). The use of the term 'conservative' throughout this investigation refers to a political ideology that includes strong support of military power and the free market philosophy, and an overall opposition to strong central government and a general distrust of authority. Conservatism by this definition is simply classical liberalism repackaged. By comparison, a 'liberal stance' refers to an ideology that generally endorses a strong government and welfare system, and opposes a strong military superstructure and laissez-faire capitalism. Although women are found to be less conservative on a number of issues, evidence does exist of their greater conservatism on others issues. Women score higher on religious fundamentalism scales, endorse school prayer in higher numbers, and are more likely to agree with restrictions on fundamental freedoms. Thus the direction of the gap, towards greater liberalism or greater conservatism, is dependent on the nature of the issue in question. These gender gaps in voting and attitudes should not be exaggerated. They have appeared over a relatively limited number of issues, and when such differences exist, they are usually relatively minor. However, the limited number of documented gaps may be less the result of their limited existence than of few surveys available for documenting their existence, or the result of investigators concentrating on a limited number of issues. Uncovering the sources of the gender gap, and documenting the scope of its reach across opinions, will bring us closer to an understanding of whether they constitute systemic or more temporal phenomena. While the sources of these gaps are discussed at length -34-fiiJther on, the importance of the gender gap at this point in the discussion rests on its suggestion of fundamental political differences between men and women, of varying political cultures. Some consistency over time and across sample groups lends credence to their significance. Thus mmking about women and men as inhabitants of different political worlds is not new. But the lack of overwhelming evidence of this difference may reflect the greater dominance of the male political world. Women's political world may consist of a melding of two worlds: one dominated by men and another, perhaps more central yet less politically visible, women's world. Many women and men speak in similar voices and this is to be expected given their many shared experiences and the dominance of the male political world. But the existence of difference is relevant and merits analysis in part because it suggests that women face a unique set of experiences that come to bear on their political opinions and belief systems. An exhaustive investigation into the political opinions and belief systems of women and men is fundamental to an understanding of the nature of political difference. Other differences notwithstanding, women and men are politically different, and these differences may indicate the existence of a difference in political culture. I - GENDER GAPS IN OPINIONS AND THEIR SOURCES The first task, then, is to document those issues that compose the gender gap phenomenon. As mentioned earlier, women are often more liberal in their opinions than men, but they also exhibit greater conservatism on some issues. One might assume, a -35-priori, that if gender accounts for basic differences in opinions, then the largest gender gaps should be recorded on issues that touch directly upon issues of particular concern to women. Although such "women's" issues may well be of interest to men, they can be understood to hold a less central position in their lives. The issue of abortion, as an example, might be hypothesized as one upon which sharp divisions would appear in men's and women's opinions at the very least because of the physiological relevance of the issue to women. But as will be fully documented in Chapter 4, the very questions on which one might assume large gender differences exist are those which reveal little gender differentiation. The dilemma is, then, that gender does appear to matter to the determination of opinions but mostly on issues that appear to have little direct connection to gender itself. Solving the dilemma requires fitting together the pieces that compose the puzzle of the sources of the gap. A. Understanding the Lack of Gender Gaps: A Competition of Values One particularly fruitful vein for beginning the search for the key to solving the gender gap puzzle comes from mainstream investigations of public opinion. Much of the literature on the formation of public opinion focuses on the particular values held by individuals, argued to hold a central place in the formation of opinion on particular political issues (Conover 1988). Values, "general and enduring standards" (Kinder, 1983:406), are argued to be of particular importance in the deteraiination of attitudes and beliefs. According to Feldman, individuals "need not be ideologues in order to evaluate politics on the basis of beliefs and values. To some extent, policies and actions are -36-simply judged right or wrong because of their implications for deeply held values" (1988:418). Converse (1964) discussed this psychological constraint as one of those likely to influence the adoption of political attitudes and beliefs, but subsequent research has tended to neglect this aspect of his theory focusing instead on their ideological underpinnings. Given the relevance of values to the formation of political opinion, they would seem to be an important starting point for attempting to account for the gender gap in attitudes. While values are but one type of the many opinions that individuals can hold, they are more central to belief systems than are attitudes and beliefs, and necessarily more abstract. Values reflect the particular goals and ideals that individuals consider to be of particular importance. The relevant distinction between values and opinions is mainly that values are more stable and enduring, while other opinions are more likely to fluctuate with the particular circumstances of the time. Individuals hold particular core values that work to direct the positions they adopt on various issues. These values are likely to be shaped by any number of forces including those that shape opinions more generally, such as self-interest and social group memberships, but they are more permanent than opinions. McClosky and Zaller (1984) maintain that political culture introduces such values to individuals, and that institutions, policies and the coniimtment of political elites maintain them within society. If women and men are committed to distinct sets of values this would provide evidence of the differing nature of their political cultures. American research has evaluated the degree to which such central values shape -37-political attitudes. Feldman (1988) examined the place of overall support for three core beliefs in the attitudes of Americans: equality of opportunity; economic individualism (the ability to get ahead on your own through hard work); and the free enterprise system. Of these three, support for the free enterprise system was found to have little association with political attitudes and opinions, but "beliefs in equality of opportunity and work ethic are associated with preferences on public policy issues, presidential performance evaluations, and candidate evaluations" (437). He also found some evidence that these values are differently held by women and men. Canadian research has also investigated the degree to which values shape attitudes towards certain political issues. Sniderman et al. (1989) in their examination of the Charter Project data, discover that opinions on language rights are partly grounded in "strategic calculations of group status"(283) but also in the core political value of egalitarianism which they define as "a desire that the less well-off be better-off and a willingness to accept historically unpopular and stigmatized groups as equals" (274). Moreover they assert that the effects of these two forces on language rights attitudes are interdependent. Group membership is relevant since it links self-interest with the status accorded the group, but also because group membership constitutes and reinforces the holding of certain values among its membership. According to the authors, Ultimately, it is the interplay between the core values to which citizens subscribe and their concern for the status of the groups, both linguistic and partisan, with which they identify that shapes the politics of language rights . (283). Thus the authors highlight the simultaneous effects of values and social group -38-memberships on the political opinions adopted by individuals. Fletcher and Chalmers (1991), investigating opinions on affirmative action among Canadians using the same Charter data, discovered an inconsistency between support for the principle of equality and support for affirmative action programs at both the elite and mass levels. Rather than argue this inconsistency is indicative of a lack of coherence between values and opinions, they instead suggest that rarely does only a single value come into play in the shaping of opinions. It is often the case that multiple values often compete for dominance. As they suggest, a person, in considering whether or not large companies should have quotas to ensure that a fixed percentage of women are hired, not only may reflect on the value of equality but also on how important the notion of merit is in his or her way of thinking (79-80). Thus values compete for dominance in shaping opinions, and the victor is likely to vary as individual issues are addressed. Tetlock (1986) coined this weighing of values "value pluralism". Thus values, determined in part through group memberships, have been proposed as an important source of political opinions, and as such, can be looked to for help in attempting to identify the source of the gender gap. Feldman's (1988) evidence that American women and men vary in their support of the core beliefs of equality of opportunity, economic individualism and the free enterprise system justifies an examination of values among Canadian women and men for their part in the determination of political opinions. Moreover, Gilligan's (1982) assertion that women are more collectivist than individual, and show a greater ethic of care than men, further -39-justifies this line of inquiry. The argument made in this investigation is that part of the gender gap puzzle lies in the competition of core values among women. One key element of this value pluralism is religious fundamentalism. Religious conviction will exert a conservative force on issues that have a moral dimension, and religion has been linked to political opinions in Canada (Johnston, 1986). But women are often found to be more religious than men. Bibby, for example, provides evidence from a 1985 national Canadian survey of differences in religious group membership: "more women than men say they are members [of religious groups], attend regularly and enjoy their participation" (Bibby, 1987:100). Religion and religious belief have also been argued to be of central importance in women's culture (Bernard, 1981); as such, it seems inevitable that religion will play a key role in the determination of women's values and hence political opinions. To return to the example used above, religiosity is likely to be associated with greater conservatism on the abortion issue. Among those who are religious, abortion may be understood to be a question of social morals. At the very least, it is likely that women's rights will be understood to be less pressing than those of the unborn child. Religious belief brings with it a particular view of the world: political questions may be answered in part through the lens of moralism. Women's greater religious conviction is likely to explain their greater conservatism on some issues. The other key element of women's value pluralism, it is suggested, is feminism. Feniinists view the world through the lens of sex and gender (Gibbins and Youngman, -40-1996). Moreover, the feminist revision perspective1 argues that feminism is a liberal stimulus for women's political attitudes precisely because it allows for the inclusion of women's traditional nurturance and compassion values in the public sphere (Conover, 1988). Feminism may provide the liberal impetus for the adoption of more liberal opinions in its strong endorsement of liberal democratic values, its acceptance of diversity, and its challenge to the use of force and violence as a means of conflict resolution. Thus the common socialization of women and men within the capitalist system is unlikely to result in substantial differences in their political opinions unless women as a group understand the restrictions that system has had on their life experiences, and bring this realization to bear in the determination of their political opinions. As a core value for many women, feminism is likely to bring about a liberalization of many of their attitudes, including opinions on moral questions. Thus on certain issues, religiosity and feminism exert competing pressures on opinions; the two core values will compete for dominance on more proximate issues. Returning to the example set out above, the adoption of a feminist outlook is likely to bring with it support for increased access to abortion services in direct opposition to the pressure exerted by religious belief. The lack of a gap on certain issues may thus be explained by this competition of values among women. The conservative pressure of religious belief mitigates the liberal pressure of ferninism resulting in attitudes that seem to differ little from those of men. On questions bearing a weak moral component, however, feminism is 1 This term is taken from Deitch (1988). -41-likely to account for women's greater liberalism. Both feminism and religiosity are key elements in understanding women's and men's opinions. As such, they are returned to and discussed at greater length later. B. Explaining Gender Gaps in Opinions The competition of core values goes some way towards explaining the lack of gaps on some issues but it fails in providing an answer for the existence of gender gaps. Over and above differences in women's and men's core values, life experiences and socialization vary with gender. Some feminist theorists make the claim that it is sex itself, biology, that accounts for the differences in the priorities that women and men ascribe to various political goals (e.g. Bernard, 1981; Chodorow, 1978). Many other researchers reject biological determination altogether and point to either early socialization processes or adult life experiences as responsible for gender gaps in attitudes (e.g. Sapiro, 1983). Thus women and men may approach politics from a particular perspective, or ethos, sterrmiing from either biological sources (i.e. sex itself), or from particular historical forces and life experiences, or from the socialization that is particular to each gender (i.e. political socialization). While the sex of each respondent is measured empirically with a relative degree of ease and with a relative degree of accuracy, socialization and the myriad of life experiences and historical forces important to each gender are not as easily and accurately tapped by opinion surveys of adults during election periods. Further to this, it is unlikely that the gender gap in opinion stems from only one source. The difficulty of empirically -42-disentangling competing sources of opinions complicates the investigation. Moreover, individual factors are likely to account in and of themselves for competing pressures on opinion. Nevertheless, attempting to account for the gender gap in opinion empirically, as difficult a task as it is, is important and deserving of attention if only to present an important springboard for future evaluations. The literature on gender difference and previous research on the gender gap in attitudes, as well as the more general literature on public opinion in Canada, point to several potential sources of gender gaps in opinion. Those factors that will be examined as potential sources of the gap in this investigation are: •economic self-interest; •differences in social group memberships; •gender role socialization; •feminism; and •biological determinism and early-life socialization. Each of these factors will be discussed in turn, and in the following chapters empirically investigated for their role in the existence of the gender gap in political attitudes. Economic Self-interest If women and men think differently on political issues, it may simply be a reflection of their differing financial positions. Economic self-interest may lie at the core of any difference in political opinions. A common assumption made in much social science research is that individuals act on the basis of self-interest, and as such are -43-"rational" in their behaviour. According to Kinder, "There is no more familiar presumption than that people support policies that promote their own material interests and oppose policies that threaten them." (1983:403) This "rational actor" assumption, when extended to the determination of political opinions, holds that individuals at high income levels are more likely to oppose policies that are seen as threatening to their financial positions: they are not likely to endorse redistribution policies or social welfare policies which could put a drain on their finances as they would be the principal supporters by way of taxes of such programs. Similarly, the expectation is that individuals at lower income levels will specifically favour these policies because their own economic positions would be bettered with their adoption. Actions based on self-interest as outlined here, are those that are aimed at increasing one's wealth, assets or power. In this 'economic version of political man,...,to be self-interested is to be asocial, materialistic, and oriented toward the here and now' (Citrin and Green, 1990:5). Despite the presumption of the key role of self-interest in the determination of issue positions that is often made, the notion is replete with difficulties. To begin with, there are difficulties encountered when one attempts to define precisely "self-interest". Economic motives alone do not seemingly explain all behaviour, particularly behaviour motivated by social or altruistic motives, and so definitions might do well to extend beyond such a limited focus. There is also the question of whether an individual's self-interest should be defined subjectively by the individual at hand, or more properly by an "objective" outside observer. The tjpinion of an outsider regarding what is in one's best -44-interest need not always correspond to a similar assessment by the individual in question. Defining self-interest also necessitates the selection of a reference group by the individual at hand, for when individuals attempt to determine their level of support for state-funded child care, they must primarily determine whether they consider themselves tax-payers or child-care demanders. The selection of the most salient group membership will directly determine support for the policy, since self-interest will inevitably be defined by the choice. The objective selection of a reference group is no less difficult for the researcher. But despite such difficulties, the question of self-interest has not been left out of research on political behaviour and public opinion. Rather, it has remained in spite of them. The basis for most definitions of self-interest is economic and the most often used proxies of material self-interest are income and employment status. Economic self-interest has been presented as a possible determinant of the gender gap in opinions (Kopinak, 1987; Welch and Thomas, 1988; Erie and Rein, 1988). Women's lower socio-economic status and their greater dependence on the state as employer both serve to increase the likelihood that they will endorse in greater numbers than men the redistribution of income, as well as government programs and policies designed to aid women. This investigation will examine whether economic self-interest plays some part in explaining gender gap in opinions. But economic self-interest may not be as primary a concern for women as it is for men. Taking Gilligan (1982) at face value, women's more 'compassionate' outlook and their 'ethic of care' could downplay the relevance and importance of consideration of -45-individual self-interest in the determination of individual opinions. This ethic of care and the responsibilities of women as the prime caregivers of children could extend outward toward society at large and explain to some degree their greater sympathy for the downtrodden and their greater support of welfare measures. Their concern is more likely to be directed outwards rather than towards their own individual circumstances, and their sense of responsibility will mean an incorporation of others into any calculations of 'self-interest. Support for this argument is provided by Arthur Miller (1988) who gives evidence of women's greater use of sociotropic economic orientation in their voting decisions. Unlike men, their voting decision was more likely based on the country's economic fortunes in general rather than their own personal economic fortunes. This selflessness overshadows their own personal economic self-interest, and so we would also expect women's financial situations to correspond loosely to their political opinions when compared to the same relationship among men. The definition of self-interest may have to be re-evaluated for women as a group. Changes in the size of the gender gap resulting from controls for income and employment status serve as a critical test of the relevance of economic self-interest to the gender gap. If women's lower socio-economic status and reliance on the state as employer are driving the differences in opinions, then controlling for income, employment status, and employment in the government sector should serve to reduce the relative size of the gap. If, on the other hand, women's self-interest is an extension of their private sphere role as caregiver outward into the public sphere then one would -46-expect that removing the effects of income and work status would change the size of the gap very little. Calculations of economic self-interest are not always made on basis of one's income level or employment status. Calculations can also be made on the basis of membership in a particular socio-demographic groups. At the very least, social group membership will influence an individual's perceived self-interest by providing a psychological attachment to the group and the recognition of shared interests and values. The role of social groups in the formation of political opinions can be assessed by two different concepts: group membership and group identification (Conover, 1988).2 Group membership is the objective inclusion in a social group or category, based on some observable characteristic such as sex, age and education. Differences of opinion can present themselves in groups that at first glance might not appear to have clear reasons for being opposed or supportive of particular issues. As an example, differences of opinion can appear when comparing individuals that reside in different provinces. The social influence of membership in a province can and often does result in the adoption of a particular provincial viewpoint. Membership in objective social categories, whether or not individuals feel a psychological attachment to the group, can influence opinions. Group identification, on the other hand, demands two things: self-awareness of being a member of the group, and a psychological attachment to it. This definition 2 Conover discusses a third category, group consciousness, that adds the realization by group members that their fortunes are tied to those of the group as a whole. This element of group membership will be discussed further as an element of feminist belief. -47-distinguishes between ingroup and outgroup psychological attachment: the attachments of members of the group (ingroup) and non-members (outgroup). Individuals not belonging to the group can feel close to the group, but as Conover's work demonstrates, "such attachments differ significantly in their origins and influence" and as such must be kept distinct from group identification, which demands that individuals be members of the particular group (1988:53). As an example, homemaker status has been suggested in gender gap research as a role that could possibly shape women's opinions on a number of issues if women within the category adopt the stereotypes associated with the role. In order for women to identify with homemakers as a group they must be a member of the group. Women who are not homemakers may feel close to the group, but as non-members, cannot identify with it. The relevance of this distinction becomes clear when thinking about the impact of group memberships on perceived self-interest. As an example, if the government decided to tax homemakers then one could anticipate that homemakers, perceiving the negative impact this would have on their economic self-interest, would oppose such a policy. Other women, even if they felt close to homemakers, would not consider this policy in terms of their own self-interest. Although they may well oppose the policy, it would be for reasons other than group identification. When and how particular groups will come into play in individual political reasoning depends in part on the political environment. As the association between an issue on the political agenda and the particular social group's fortunes increases, so too will the impact of group identification with that group on opinion. In addition, the -48-salience of the issue for the group's members will also determine the impact of membership on opinions. This salience may depend in part on the issue's position on the political agenda, directly determined by both the media and political actors, and on the ability of the group's elites to draw attention to the issue. Simply put, the more an individual identifies with a social group, and the more direct the connection is between a group's fortunes and one side of a particular issue, the more likely it is that his or her opinions will correspond to that group's position on issues. For example, one could anticipate that the elderly would be keenly supportive of increases to Canada Pension benefits; that women would support affirmative action policies designed to speed up their introduction into previously guarded male job environments; and that single mothers would be especially interested in seeing that the laws concerning the payments of child support be tightened. There is a certain degree of logic and rationality to the holding of such positions. The entrenchment of the Charter of Rights and Freedoms has made such distinctions all the more real in that it gives constitutional identity to social groups, and provides legal entitlement for claiming benefits based on group membership. But, in the context of economic self-interest calculations, it is not simply identification with the group that influences opinion: it is the realization that a particular policy or issue is likely to affect one's economic self-interest because one is a member of that group. Thus group identification provides an additional avenue for considering self-interest as an explanation for gender gaps in opinions. Two groups in particular potentially explain some part of the gender gap. Identification with a particular age group -49-can determine calculations of self-interest: an understanding that particular policies might differently harm certain age groups and an identification with those groups would lead to differences in support for that particular policy with age. The fact that women live longer than men, and that older age groups tend to be more economically vulnerable, suggests that age might account for some of the recorded differences in the opinions of women and men. Any impact on the gender gap due to age would be the result of the different rates at which women and men occupy older age categories. Controlling for the impact of age on opinions should serve to decrease the gap in opinions on which economic calculations are relevant. As such, age will be introduced into the analysis in order to evaluate its relevance to the gap in opinions. Education may also be connected to economic self-interest calculations in that education and income are closely and positively related. Higher education is often the key to entering higher income brackets. Thus the self-interest calculations made on the basis of income levels are also likely to be connected with education level. Differences in the rates at which women and men are found at higher education levels may be driving the gender gap in opinions. Because a smaller proportion of women have attended post-secondary institutions than is true of men, their resulting greater economic vulnerability may account in part for their greater liberalism. Although gender differences in educational attainment are dimmishing, it is still the case that women are less likely to have attended a post-secondary institution. Controlling for education would thus serve to decrease the size of the gender gap. -50-Social Group Memberships The degree to which social forces are relevant in the shaping of opinion can be inferred from the preceding discussion on the hypothesized gendered impact of socio-demographic characteristics like education and age on self-interest. Social group memberships and demographic factors are relevant in any attempt to account for the opinions of women and men and as such must be included in an attempt to account for the gender gap (see McClosky and Zaller, 1984). But the impact of social group membership can occur outside of any direct cognitive connection of membership in the group and self-interest. The impact may be more subtle than this. As sources of social influence, collectivities transmit and reinforce a set of values to their members. Social group memberships can provide members with unique or at least divergent social, political, cultural, and economic environments, and provide members with experiences particular to that group. If women and men join different social groups or if the salience of memberships in similar groups varies between women and men, then the set of values that women and men take from such memberships are likely to vary and as a result, may lead to differing political opinions. Evidence abounds on the place of social group memberships in the shaping of public opinion on a number of issues in Canada. Social identities shape our views of the political world and social groups help individuals define opinions on issues through personal contact, peer pressure or through the provision of "distinct perceptual viewpoints" (Conover, 1984: 763). It is also possible that historical political issues or past events, long off the political agenda, can linger in group assessments since people -51-store their affective reactions to social groups and transmit them across generations. Historically, Canada has been marked by highly visible groups in opposition over issues: Quebecois against non-Quebecois; French against English; have provinces against have-not provinces; and Catholics against Protestants. The attention such conflicts generated in the past increases the likelihood that individuals will have formed strong group allegiances and transmitted them to their children. This transmission must be reinforced outside the family, however, in order for cleavages to be maintained, and the reinforcement occurs chiefly when new political issues are interpreted along old cleavage lines, and as social groups reinforce a particular political perspectives (Johnston, 1985). A number of social group memberships have been found to dominate in their ability to shape opinion in Canada. Richard Johnston's research (1986) shows that the most notable group identities in the shaping of opinion are those based on region, religion, language, education, ethnicity, and class. He also found that these divisions were not usually reinforcing but rather that "issue publics" assembled themselves around individual issues, and that individuals tended to circulate around these various issue publics. When reinforcing divisions did appear, they tended to be loose ethno-religious cleavages over particular issues, most notably the questions of language, religious schooling, and to a lesser extent, abortion. On the whole, however, "the picture is one of considerable mass-level circulation among policy coalitions" (1986:115). Blake and Simeon (1980), in an analysis of the place of regional divisions in the policy preferences of Canadians, found evidence similar to Johnston's of the existence of educational, regional, and linguistic cleavages in issue preferences. Moreover, their -52-research suggests that differences based particularly on age may be widening. As they put it: The young have always been more liberal on social and moral questions, more opposed to foreign influence and more conciliatory towards French-Canada, but, at least on the last three, the gap has recently widened (90). Thus a number of socio-demographic characteristics have been linked to the structure of opinion in Canada, and gender-dependent differences in the impact of these traditionally evaluated social factors on opinion can be suggested as relevant to the determination of the gender gap. i) Socio-economic class: Most often the relationship found between class and political opinion is that the 'higher' the social class the more conservative the political attitudes. Such differences in opinion go beyond self-interest, however, and are rooted instead in class-dependent ideological viewpoints on such questions as the size of the welfare state, the place of government in the economy, and the role of labour in the business economy. Despite the difficulties in measuring the concept of social class, it has been found to be related to opinion on a number of issues. If women make up a greater share of society's lower classes, given lower levels of education and income, this might account for their greater liberalism. Several measures of social class will be introduced as controls. First, controlling for education and income should serve to reduce the size of the gap. In addition, union ties will be introduced as a proxy of socio-economic class. Labour groups implicitly represent a particular view of the relationship between management and workers. The relevance of such a view to political opinions is obvious. But women have -53-not historically had a relationship with unions that equals that of men: their jobs are often the last to be unionized and women's issues have not constituted the main focus of many labour groups (Armstrong and Armstrong 1984). Thus the degree to which the labour culture has been transmitted to women is likely to be weaker than is true of men. This difference may be somewhat mediated by the number of women employed in the public sector, characterized by a high rate of unionization, but it is unlikely to overcome the labour movement's relative weakness in welcoming women into its fold. As such, the impact of labour unions should also be considered in the examination of gender gaps: controlling for them may increase the size of the gap in opinions. ii) religion: As a core value, religious belief provides a particular vision that is brought to bear in the determination of some political issues. Social conservatism, including the protection of family values and opposition to feminism, is often based in religious conviction (Gibbins and Youngman, 1996). On such issues as abortion, pornography, and homosexual rights, one's religious beliefs are likely to play a key part in detenmriing one's opinions. Women's greater religious commitment must be taken into account when examining gender gaps in these attitudes. As such, controlling for women's greater religious fundamentalism is likely to increase the size of the gap found on such issues. On other issues, its impact on the gender gap is likely to be mmimal. iii) education: Higher education, apart from the conservative pressure exerted by self-interest considerations, is also a vehicle for the transmission of cultural change. -54-University attendance is linked to an increased tolerance and to an acceptance of new ideas. As such, those individuals that have attended university may have adopted a perspective that translates into more liberal thinking on some issues. Given that women are less likely to have attended post-secondary institutions, controlling for the impact of education is likely to increase the gender gap in attitudes. iv) age: Controlling for age might serve to increase the size of the gender gap in that age tends to be associated with increased conservatism. Quite apart from life cycle considerations, the inverse relationship between age and education might also account for the greater conservatism associated with age: older Canadians are less likely to have attended university and therefore are less open to new ideas, less exposed to them, and less tolerant of change in general. Women tend to live longer than men and thus controlling for age is necessary in order to remove its effects from the gender gap in opinions. v) membership in voluntary organizations: Apart from education, the cultural transmission of ideas is also likely to occur through membership in social organizations. Participating in some groups may mean that women's opinions come to more closely mirror the dominant opinion in the political community, which is most likely to correspond with male beliefs as men happen to dominate in many public groups. The social interaction that occurs in social groups may provide the context through which women come to adopt issue positions that are closer to those of men since such -55-involvement presumably increases interaction with men (Putnam, 1966; Cox, 1969). Political institutions and organizations, dominated by men, would seem likely vehicles for moving women's opinions closer in line with those of men. Alternatively, women's opinions may come to deviate from the political norm (i.e. men's opinions) if they are members of groups made up primarily of women who do not hold attitudes that one could consider the community norm, e.g. liberal and radical feminist groups or women's shelters, as contrasted to religious groups or groups engaged in the protection of family values. A look at patterns of participation in volunteer groups reinforces the notion that women's and men's unique perspectives may be driven in part by their unique social interactions. They also highlight the relevance of group memberships in determining interests, values and beliefs. Evidence from Statistics Canada shows that women are more likely to work for volunteer associations, and moreover, that the groups to which they volunteer their time are different. Women are more likely to volunteer their time to organizations oriented towards issues of health, education and social welfare, while men are more likely to choose leisure activity organizations (Statistics Canada, 1981). Women's membership in these organizations is likely to ensure that their attitudes continue to deviate from those of men, in that the focus of such groups, and the dominance of lfte-minded individuals within them, are not likely to move the opinions of their members closer to those of men. Moreover women's weaker participation in party politics and 'traditional' political organizations means that their opinions are less likely to be moved towards the male norm. Controlling for women's participation in non-traditional political organizations would serve to weaken -56-the gap in opinions. On the other hand, controlling for their participation in traditional political organizations and institutions should serve to increase the size of the gender gap in opinions. Gender Role Socialization In attempting to explain gender value difference, socialization theory moves away from economic calculations and group memberships and points to the importance of assumed adult roles and the expectations linked to them as strong factors in the development of political values. Gender in this instance is 'the socio-cultural manifestations of being a man or woman ' (Sapiro, 1983:36) and being socialized as a woman translates into a greater human compassion than men stemming in part from woman's maternal nurturance, the extension of familial concerns into the public sphere, and the concentration of women in the domestic rather than the public sphere. For women, the choice between home and career is one of tremendous significance. The expectations attached to each of them differ substantially and the adult role assumed by women can have a direct effect on their political beliefs. But the traditional socialization of women as women is not equally adopted by all adult women. The traditional gender role adopted by a number of women is personified by the "homemaker". The ascribed role of the homemaker has traditionally been that of the nurturing, gentle woman devoted to home and family. The "traditional" gender role ideology and its variations, to be found almost everywhere one looks, bears as its central dictum the privatization of women, their restriction to the domestic sphere (Rinehart, 1992:68). -57-The acceptance of such a role implies an acceptance of the particular place of women within it, and as such is a source of value learning. Differences in opinion may result if women, as homemakers, adopt or exhibit the stereotypes linked to that role. Moreover, the homemaking role most closely approximates cultural expectations of women and, as such, homemakers are expected to perform in the political system in stereotyped ways (Sapiro, 1983). The expectation is that homemakers will display socially-defined "feminine" characteristics: nurturance, concern for others, tolerance, cooperation rather than competition, and pacifism (Tolleson Rinehart, 1992: 11). The stereotype would also extend to women as defenders of morality, stemming again from their responsibility as guardians of children. The adoption of these stereotyped characteristics will have a direct impact on the political opinions and belief systems of homemakers. It is the case that many parochial issue stances are often linked to the homemaking role. It is likely that homemakers' political perspectives, at the very least, are mfluenced by the values centrally-ascribed to the homemaker role. Men, although they may choose to become homemakers, are not expected to conform in the same manner as women to the societal expectations that accompany that decision for women. Employment outside of the home and higher education can both serve to move women away from stereotypical images of "the feminine" and closer, it is argued, to the male standard. This possibility increases if the decision to work outside of the home is a career choice, rather than a decision based on financial need, and if employment is not in a pink collar ghetto that will only serve to reinforce private sphere role expectations in the -58-workplace. If the pursuit of a career is combined with an increased level of education, the movement away from the ascribed political perspective is very likely to be even greater. Faced with the inequalities of the workplace, and often the double burden of home and career, many career women will undergo a change in political perspective. The work of researchers has shown that on several issues, the opinions of homemakers and of women who work outside the home differ substantially (Poole and Zeigler, 1985; O'Neill, 1995) and lends credence to the notion that role choice is influential. Nevertheless, the degree to which the selection of the homemaker role is indicative of gender role socialization must be questioned. Two alternative and equally plausible scenarios for explaining the adoption of a particular viewpoint through the homemaker role can be entertained. First, it may be motherhood and not gender role socialization that accounts for the particular set of values found in homemakers (see Ruddick, 1988). Many women adopt the homemaker role because of the presence of children in the home or as a result of the birth of a child. Regardless of the reasons behind a woman's decision to remain home to care for children, the act itself may bring about a change towards 'maternal thinking'. Second, women that exhibit the characteristics of the traditionally defined role for women choose to become homemakers. Hence, these 'traditional' women select themselves into the category. What may be taken as an indication that society socializes some women into a particular role, may instead merely be a result of self-selection into the category. The confidence that can be placed in the homemaker category as a proxy for gender role socialization is somewhat weakened by these two considerations. Including a measure for the number of children in the home removes one -59-of the lirnitations in the homemaker proxy: any 'maternal thinking' influence will be tapped by this variable and removed from the relationship between homemaker status and opinions. Controlling for the impact of the adoption of homemaker status, the respondent's marital status, and for the presence of children in the home on political opinions should reduce the size of the gender gap in attitudes, as those women that have adopted the role expectations they are socialized into as children and which are reinforced by the homemaker status are removed from the relationship between gender and opinions. On the other hand, controlling for employment status and education is likely to have the opposite effect on the gender gap in attitudes, and will increase the gaps, as women whose opinions most closely resemble those of men are removed from the relationship between gender and political opinions. Feminism Drawing on gender difference theories, some researchers have looked to feminism as an explanation for the differences in women's and men's opinions. This perspective emphasizes that women's ways of knowing, whatever the particular source of the perspective, are the source of the gender gap in opinions, but that these ways of knowing will only become relevant when women gain the confidence to employ their own particular proper voice'. Feminism entails its own set of values and predispositions, and membership in the group is likely to fundamentally shape women's opinions. Gender consciousness, an understanding that one's sex is a factor in one's -60-relationship to the political world has been argued to be a necessary and sufficient condition for feminist belief (Rinehart, 1992). Gender consciousness "embodies an identification with similar others, positive affect towards them, and a feeling of interdependence with the group's fortunes" (Rinehart, 1992:14). To become gender conscious is to undergo a central redefinition of values that then translates into new issue positions and changes in the relative salience of issues. For Rinehart, this redefinition of values need not result in the liberalization of political opinions; organized women on the right of the political spectrum, as proponents of the traditional view of women and the protection of the family, are also gender conscious. Our interest in this investigation lies with those gender conscious women who espouse more egalitarian views, ones that correspond to the liberal position, and thus specifically in feminist consciousness. Feminism and feminist consciousness may provide keys for unlocking a particular set of values that can then be brought to bear on political questions. Some theorists have underlined that feminism need not be the only source of this new confidence; higher education and careers (as distinguished from part-time and/or low income employment which serves only to reinforce women's private sphere roles) are possible paths to unlocking this women's voice. A university education and a fulfilling career can each provide women with the self-confidence necessary to think "as a woman". They are also likely to increase a women's contact with feniinist groups, thus increasing their chances of adopting feniinist values and beliefs. Here then, and directly in contrast to Sapiro's assertions, marriage can be understood to restrict women's 'natural' voice in that women lose the psychological and -61-economic independence that is necessary to voice this unique woman's perspective (Carroll 1988). Marriage, the most traditional of patriarchal institutions, serves to converge the genders on the male norm. Thus controlling for marital status in this instance should increase the size of the gender gap in that this convergence would be removed from the relationship between gender and opinions. Controlling for feminist belief, on the other hand, should decrease the size of the gap in opinion if the adoption of a feminist perspective serves as the key to unlocking a particular way of knowing. Controlling for education and/or employment status should have a similar effect in attenuating the size of the gap, in that they too would serve to allow women to find their own voice. Despite some limited agreement on this differing set of values as a source of the gap however, there is little agreement on the source of these differing value frameworks. Biological Determination and Early Life Socialization The gender gap in opinions may reflect a difference of values between the genders that is grounded in biology (see Okin, 1990) or in early life socialization. As such, women's lesser support for defence measures may reflect women's natural or learned tendency to be less aggressive than men (Maccoby and Jacklin, 1974) and their greater support of welfare measures and measures designed to protect social morality may stem from their reproductive role or their early-life socialization. As the gender responsible for physically introducing children into the world, all women, not only mothers, may naturally feel a greater responsibility for protecting them. Alternatively, it may be that -62-women are socialized early in life to think 'like a woman'. If biological differences or early life socialization account for the gender gap in opinions then sex differences in opinions should persist even after having controlled for self-interest, gender role socialization, feminism, and other social factors. But declaring the remaining gaps in opinion to be one of biology or early life socialization requires that several criteria be met. First, it is necessary that the researcher be confident that all possible determinants of the gap in opinion are included in the equation. Specification errors, specifically errors in the modelling of the sources of opinion for both genders, will lead to bias in the estimated regression coefficients (Johnston, 1984). Although the model as specified in this investigation can be defended, the possibility exists that the model is underspecified. Some omissions are due to the data set available rather than faulty theory: there are no suitable measures of ideology, part-time work status nor of self-selection as a feminist identifier (see Chapter 5). Thus one could expect that the estimated remaining gap in opinion may at least be partly the result of the omission of relevant variables. Some level of specification error is unavoidable. Second, the researcher must be confident that the proxies employed in the investigation closely approximate the concepts that they intend to measure. Measurement error will also result in biased estimators of the coefficients (Johnston, 1984). As discussed in Chapter 5, the validity in employing both the homemaker and feminist proxies as measures of gender role socialization and feminist consciousness respectively is questionable but necessary given the lack of suitable variables in the data set. The proxies nevertheless allow for a preliminary investigation of the sources of the gender gap. The methodological result of -63-their weakness is, however, that the coefficient on sex that remains after having controlled for all other factors is likely to include some of the impact of the weak proxies on opinion due to their measurement error. The sex coefficient may then be artificially inflated depending on its correlation to those independent variables that are measured with error. The ability to measure with a high degree of certainty the predictive ability of biology or of early life socialization is thus limited. II - THE SOCIAL STRUCTURE OF OPINIONS The investigation will provide evidence of the existence of a difference in the political cultures of Canadian women and men. In order to do so, evidence of a difference in women and men's ways of knowing must be gathered. One method of attempting to do so is to evaluate the social structure of each gender's political opinions. While the investigations of the gender gap provides some help in understanding those forces that serve to move women's and men's opinions closer to each other, or further away from each other, it is only by separately investigating the impacts of these factors on women's and men's opinions that some insight will be gained into each gender's political minking. Much of public opinion research has overlooked the possibility that societal memberships and life experiences could translate into differential impacts on the structure of women's and men's opinions. One exception is Conover and Sapiro's (1993) analysis of the gender gap on defence questions. Another is found in Kopinak's study (1987) which found that women were more likely to espouse liberal values as their socio--64-economic status increased rather than the traditionally observed decrease in liberalism with class. It was the privilege of economic security that brought out the gender gap in attitudes, not the projection of women's family role into the public sphere. Thus Kopinak's results suggest that the relationships between various factors and political opinion should not be assumed to be identical for each gender. Some of the folly of assuming that experiences and factors will exert similar forces on opinions was made clear in the preceding discussion on the sources of the gender gap. To assume that women and men come at their opinions in a similar manner negates a priori the existence of different political cultures. A separate examination of the social structure of opinion for each gender allows for the possibility that even when the two groups agree on issues, they may do so for very different reasons. Women may differ in how they cope with large numbers of complex political issues, by looking to sources other than those traditionally evaluated in research on opinion formation, sources that may be unique to their gender. Similarly men may prove to be much more "private" in the determination of their opinions than traditional public opinion research has shown. Much of women's political participation is shaped by the unique personal and structural constraints they face (Brodie, 1991). The inclusion of feminism and homemaker status into analyses of the structure of their opinions goes some way towards accounting for such differences. But a difference in the social structuring of opinion need not come from the inclusion of factors that are only relevant to women. It need not require that women actively search out and select particular groups to join, groups whose members are unlikely to be male. Group identification is not always required for the -65-group to influence opinions; group membership can in and of itself shape opinions. Society is structured in such a way that inclusion in some groups is less likely to be a choice for women, but also in such a way that membership in the same groups need not present women with the same cues that such groups present to men. Similar experiences can translate into different values, expectations, and beliefs that are learned in these different roles. Therefore, in order to examine whether the structuring of opinion differs between Canadian women and men the set of variables introduced as factors in the determination of the gender gap will be investigated separately on each in order to allow for a comparison of their effects. The effects of both religious belief and feminism on the opinion of women and men will be evaluated as each has been introduced as a core belief in the determination of opinions. If, as suggested by Bernard (1981), religion plays a bigger role in women's culture, then it may also be true that religion matters more in the deteirnination of their political opinions. Feminism's role in the deteirnination of values is also likely to vary by gender. Feminism, it is argued brings about a liberalization of attitudes for both women and men. Feminism, in seeking to increase the opportunities available to women and their equality in society, is likely to directly influence the attitudes and opinions of many Canadians, indirectly because of its impact on political discourse as well as public policies. But feminism's influence is likely to vary with gender if only because men's adoption of the outlook comes as a member of the outgroup: men may sympathize with feminists but they are not women. Thus the impact of feminism is likely to differ between women and men. -66-As noted previously, marriage and homemaker status can serve to bring women's opinions more closely in line with men's if they restrict the freedom and confidence necessary for articulating the difference voice. Alternatively, marriage, caring for children and the homemaker role may serve to differentiate men's and women's beliefs if they serve as vehicles for the expression of an ethic of care. Neither is it immediately clear how marriage and caring for children will influence men's opinions. As a traditional and conservative institution, marriage may be linked to greater conservatism in men's opinions. Fatherhood may have a similar effect if men see the role as one of protector of children's interests. In any event, each of these factor's influence on women's and men's roles will be investigated in the chapters to come. In other experiences, post-secondary education for example, while much of the experience may be similar for each gender, some elements of the experience may impress upon one gender and not the other. Education may provide the key that unlocks the ability of women to express their 'different voice' or it may be a force which allows women to divest themselves of the societal expectations that accompany homemaking and thus to more closely resemble men in their opinions. Similarly, working outside the home could serve either to increase or decrease the gender gap in opinions. And income has been shown in one instance to increase women's liberalism on some issues, rather than lead to greater conservatism. Regional differences in public opinion have been found to exist and are often linked to regional differences in political cultures (Elkins and Simeon, 1980). For region to similarly influence the opinions of both genders, the transmission of this political -67-culture would also have to be similar for both groups. This is only likely to happen if women are members of the social and political groups, and in numbers similar to men, which are likely to transmit these political patterns of culture. One could anticipate that is not likely to be the case. Women very often do not belong to the same political or social organizations as men. Indeed the bulk of women's political activity takes place within relatively informal groups and social movements (Vickers, 1988). It may also be that women's weaker financial positions relative to men may mean that economic self-interest concerns override and take precedence over any "regional" political questions3. Women's and men's memberships in a number of voluntary organizations are included in analysis of the gender gap in order to determine their relevance of such group memberships on opinion. The degree to which feminist groups are prominent in the media and in political circles is likely to matter in the transmission of feminist belief. The transmission and legitimation of feminist thought, quite apart from women's direct participation in these groups, can influence opinions. If feminist activity is stronger in some areas women may be more likely to have adopted feminist messages than men in those same areas. For instance, feminist groups play a particularly prominent role in Quebec politics and women make up a greater share of elected representatives in that province when compared to others (Tremblay, 1997). Moreover feminist groups in Quebec have stronger ties to parties of the left than in many other provinces. Thus the ability of a 31 thank Linda Trimble for pointing out that women may simply be too overwhelmed by their own economic situations to care about regional political concerns. -68-women's political culture to counter the dominant men's culture may very well vary with region; controlling for language and region of residence will allow for the determination of whether support for various issues varies between Quebec and the remaining provinces. m - B E L I E F S Y S T E M S The preceding sections have suggested that women and men may come at politics differently, and such a difference could translate into different attitudes and into different social patterning of opinions. The assumption of different voices could extend into patterns of belief systems as well. Belief systems are the interconnections between various opinions and the degree to which such interconnections are hierarchical in any sense. A belief system, according to Converse, is "a configuration of ideas and attitudes in which the elements are bound together by some form of constraint or functional interdependence" (1964:207). The goal of belief systems research is to determine the nature of the constraint that exists on attitudes, if any exists at all. One of the basic assumptions in politics is that members of the public can be grouped into distinct and internally homogeneous groups on the basis of their political beliefs. It is often assumed that political society is made up of some combination of liberals and conservatives, the distinction resting on their common beliefs and attitudes regarding abstract principles and issues. While there are any number of elements that distinguish the two ideological camps, at the very least they differ on the role they -69-prescribe for government, the state's necessary defence stance, and the size of the welfare state. Liberals are more likely to endorse a strong role for government in society, including its role in a well-structured welfare system, and will downplay the necessity of hawkish defence policies and substantial defence weapon stocks. Conservatives on the other hand are likely to endorse strong military and defence strategies but will downplay the role of government and the welfare state in society. It is also often assumed that once one has determined the ideological stance of an individual, one can then deduce their political attitudes on other issues. Despite such presumptions, evidence suggests that ideology plays a less than direct role in the shaping of attitudes, and moreover, that a good portion of the electorate is less than clear on what constitutes the liberal and conservative camps. Such evidence suggests at the very least that the place of ideology in the structuring of opinions needs to be reconsidered, and that a change in focus may provide more insight into the sources of opinions on issues. As expressed by Kinder in an evaluation of the state of public opinion research a decade ago "Americans are not creatures of coherent, wide-ranging ideologies. [...] More generally, American public opinion is of many and diverse pieces, a mosaic of partisan attachments, social relations, values, and personality" (1983:413). There is little reason to believe that ideological positioning in and of itself accounts for the range of issue positions taken by an individual. A branch of work attempting to reverse the minimalist conclusion begins with the premise that ordinary citizens lack political information and pay little attention to politics in general, two nunimalist assertions. The distmguishing element of the work of these -70-researchers is the assumption that such realities need not correspond to conclusions of low coherence or reasonableness in political dunking. The important question to ask, according to these researchers is how, given such information shortfalls, does the public manage at all politically? Sniderman et al. respond that individuals manage politically by simplifying the choices in front of them. They do this by making use of heuristic tools, "judgmental shortcuts, efficient ways to organize and simplify political choices, efficient in the double sense of requiring relatively little information to execute, yet yielding dependable answers even to complex problems of choice" (Sniderman et al., 1991: 19). These heuristic tools may consist of affective or cognitive components, or of some combination of the two. Party identification is offered as an example of such a tool in use during the voting decision: voters can ultimately look to their preference for one party label over another, as a shorthand key for marking the ballot, in combination with, or in spite of, any campaign considerations. Sniderman et al. further hypothesize that it is incorrect to assume that political decision-making will not vary across different portions of the populations. The assumption of causal homogeneity leads to incorrect conclusions regarding political calculations, and as such, reinforces the conclusion that the public is unsophisticated in its political reasoning. By advocating the abandonment of what is at best a questionable premise, the researchers contend that a more realistic picture of political reasoning and choice will emerge. They suggest, and base a good portion of their research on the suggestion, that one element which will directly determine the type of reasoning, and the -71-particular heuristics employed in that reasoning, is the individual's level of political sophistication. Thus the conclusion of a lack of ideological coherence among the mass public may lie in the examination of mass publics as a single body. There is little reason to assume a priori that all individuals employ the same processes to reach their attitudes and opinions. Such an assumption could mask important elements of belief systems within these separate groups and may hide important differences that exist in the structuring of their opinions, in effect leading us to conclude the larger group is ideologically unsophisticated when in fact sub-groups within that group hold structured belief systems. Given this possibility, it seems that one could make an important case for exaniining men and women separately since their political values may differ. mvestigating the mass public as a whole fails as a research strategy because it precludes any consideration of difference in opinion structuring among different groups, and removes a priori any possibility of discovering such differences among sub-groups of the population. According to Sniderman, Brody and Tetlock, People make up their minds in different ways; conversely, the insistence that people make up their rninds about political choices more or less in the same way has reinforced the impression that ordinary citizens are ill-equipped for democratic citizenship (1991: 8). It seems plausible to suggest that the political belief systems of women and men, and the manner in which they go about deciding their individual public opinions, may not be similar given the possibility that gender structures political thinking. If the priorities and values women and men ascribe to various political goals differ, then this alone could -72-result in a different patterns of political thinking. But the justification for exanrining the belief systems of women and men separately also stems from feminist theory's assertions that women and men hold different fundamental values. If, as Gilligan suggests, and the gender gap evidence reinforces, women and men hold different values to be fundamental, then these crowning postures would shape their opinions into different packages, or systems, based on the explicit connections that individuals would make between these fundamental values and less specific issues. The 'ethic of care' and the relevance of egalitarianism over individualism among women might mean that the set of opinions they hold would be more tightly constrained around questions of the community, and of the place of society in protecting those individuals who cannot protect themselves, including but not limited to its children. The role of government in the economy, as responsible for the well-being of its citizens, might take on greater relevance than the place and position of the individual in the capitalist system. Indeed, one might anticipate that the free enterprise system would not likely occupy as relevant a position in the system of women's beliefs. Moreover, feminist belief, given the importance it ascribes to sex in the determination of women's opinions, should play a central position in women's belief systems. Alternatively, we might anticipate that men would more closely connect their attitudes on the basis of the value of rights and individual freedoms. Thus questions of capitalism, and the free enterprise system, would dominate in their belief systems. Questions of civil liberties and taxation would also play a greater role in the determination of their beliefs than would be the case for women. As 'crowning postures' -73-and 'heuristic tools', economic questions and questions of right might hold greater sway in determining men's than women's opinions. In addition, feminist belief should play a less than central position in their belief systems given that men can feel an affinity to feminist causes but they cannot identify with the group itself. Investigating the belief systems of women and men separately would allow for an investigation into this possibility. F R A M E W O R K F O R PROCEEDING Based on the preceding discussion, the examination of the gender gap, the structure of opinions and the belief systems of women and men will begin by combining the large number of attitudinal variables included in the 1988 Canadian National Election Study into a smaller number of more manageable 'packages' of attitudes. This step in the investigation will be outlined in detail in Chapter 3. The next step is to account for the gender gap in opinion in Canada on the range of issues included in the study, followed by an investigation into the possible sources of these gaps. Table 2.1 outlines those variables that will be examined as sources of the gender gap and as detemiinants of opinions more broadly. (1) (2) -74-Table 2.1: Potential Detenriinants of Political Opinions Women Men Core Values Religion Religion Feminism Feminism Self-interest Income Income Variables Public sector Public sector Employment Employment Employment Status Employment Status Education Education Age Age Social Group Income Income Memberships Education Education Union Membership Union Membership Age Age Voluntary Voluntary Organizations Organizations Region4 Region Language . Language Gender Role Marital Status Marital Status Socialization Homemaker ***** Children Children Education Education Employment Status Employment Status Feminism Feminism Feminism Education Education Marital Status Marital Status Employment Status Employment Status Note: Several variables appear in more than one category of determinants. These variables will only appear once in the statistical analysis. 4 Region and language will only appear in the analysis on the social structure of opinions. It is not held to help in explaining the gender gap in opinions, but are important components of the social differentiation in women's and men's opinions. -75-The statistical equations that appear above model the statistical procedure employed in this investigation to measure and account for the sources of the gender gap in opinions. In every case the regression model employed is ordinary least squares. Equation (1) is simply the regression of each opinion index on sex in order to account for the gender gap on that issue. The gender gap in opinion is thus b\. Equation (2) introduces into the regression equation each of those variables which have been set out in this chapter as modifiers of gender gaps in opinions. As such, the coefficient on SEX in Equation (2), R\, represents the gender gap on opinion index Y controlling for the impacts of each of those variables. By means of statistical control, this coefficient on sex represents that portion of the gap that remains unaccounted for. Thus a comparison of the coefficients on sex found in Equations (1) and (2) allows for a determination of the impact of the variables included in Equation (2) on the gender gap in opinion. The investigation will then begin an inquiry into the social stnictoiring of opinion for women and men. The relationship modelled in Equation (2) assumes that the relationship between opinions and each of the deterrninants of opinions is similar by gender. But if interaction effects exist between each of the independent variables introduced in Equation (2) (X^X^ and sex, then failing to account for them will inflate the standard errors of the coefficients in Equation (2). While interaction terms could be introduced into Equation (2) in order to allow for the possibility of gender-specific relationships, the resulting equation would be tremendously unwieldy. Thus for ease of interpretation, the relationship as modelled in Equation (2) is run separately for women and men. Each of those variables introduced as potential deterrninants of the gender gap -76-in opinions in Equation (2) will be regressed on the opinion indexes created from the election study but every equation will be run separately on the women's and men's samples. Regional dummy variables will be added to these independent variables in order to determine the degree to which regional differences in opinions are common for women and men. (3a) YrV^tnkymkXik^im (3b) 7 / = a r + S A X * + e i f Thus Equations (3a) and (3b) differ from Equation (2) in the absence of a dummy variable for sex. Equation (3 a) is the regression set for men and Equation (3b) is the regression set for women. Separate regressions allows for a direct comparison of each variable's impact (i.e. coefficient) on opinion by gender. The last step is an examination of each gender's belief system in order to test the degree to which varying crowriing postures dominate in the determination of attitudes. This investigation will employ factor analysis, rather than correlation coefficients alone, in order to more accurately model this element of political flunking. The factor analysis statistical procedure will be outlined in Chapter 7. -77-Chapter 3; Creation of the Opinion Indexes In this chapter I will provide some background information on the 1988 Canadian National Election Study (CNES)1, the manner in which it will be employed in this analysis, and the manner in which the attitudinal variables were collapsed into opinion indexes. This investigation is based on 1988 CNES and the large number of attitudinal variables it contains. CNES respondents were interviewed in three waves: by telephone in the campaign period study, by telephone in the post-election study, and at the end of the study by a self-aohninistered mailback questionnaire. For this investigation questions were drawn from each of these three waves of the study in order to increase the number of variables available to create opinion indexes. Thus of the original 3,609 respondents contacted throughout the study, only the responses of those who completed interviews/questionnaires in all three parts of the study are utilized here in order to minimize the number of cases lost due to missing responses. Minimizing the number of lost cases is important since the listwise deletion of missing responses in factor analysis can drastically reduce the number of cases available for study. The remaming respondents, 2,115 Canadians, form the core of this investigation. 'The 1988 Canadian National Election Study was directed by Richard Johnston, Andre Blais, Henry E. Brady, and Jean Crete. The Institute for Social Research at York University carried out the fieldwork and the Social Sciences and Humanities Research Council of Canada funded the project. The original conductors of the data and the institutions involved bear no responsibility for the analyses and interpretations presented here. -78-Men make up 51.2 per cent of the sample (1075) and women the remaining 48.8 per cent (1024).2 Creation of Opinion Indexes The 1988 CNES contains over 100 questions tapping into a broad range of opinions which makes it a useful instrument for assessing not only public opinion on various issues but also belief systems. The investigation at hand combines a focus on the structure among individual opinions, that is belief systems, with a focus on the structure of individual opinions. It will also evaluate the existence and sources of gender gaps among a diverse set of issues! As such, the creation of a set of indexes, created from the questions in the survey, allows for the analysis and inclusion of as many of the questions available in the survey as possible. Any attempt to evaluate the opinions of women and men on each of the attitudinal questions included in the 1988 CNES would result in a rather lengthy and tedious investigation. The creation of indexes makes the task somewhat less onerous. Nevertheless some information is lost by summing across survey questions. But the loss of information is more than offset by the increased possibility of discovering patterns in opinions across a smaller set of indexes, rather than across a much larger number of individual opinions. The creation of indexes makes the task more manageable and potentially provides for a more successful investigative exercise. Apart 2 Responses to the variable employed in this investigation for detennining the sex of the respondent, ZRSEX, found in the mailback questionnaire portion of the study, were missing in 16 of the 2115 cases. Thus the responses of2099 women and men were available for use in this investigation. -79-from reasons of parsimony, however, the collection of opinions on a large number of issues and questions also allows for a more proper modelling of the belief systems of Canadian women and men than could a smaller set of opinions for two reasons. First, the concept of a belief system adopted in this paper is that outlined by Campbell et al. (1960) subsequently elaborated upon by Converse (1964). Belief systems are centrally based, widely applicable, elaborate structures of attitudes. The ability to conclude with confidence that one has properly modelled a group's belief system is directly dependent on the set of opinions included in the analysis. The greater the scope of issues and topics covered by the opinions, the greater the likelihood that the belief system will be properly modelled. Attempting to describe or measure belief systems on the basis of a small number of measured attitudes boldly ignores the definition of a belief system. In this case more is better. Second, in the measurement of attitudes it is accepted that cumulative measures of single attitudes are more likely to reduce error in measurement than are single ones (Abramson, 1983). The 1988 CNES, with its large number of attitudinal variables, asked a number of related attitudinal questions which allow for the creation of summative indexes. These indexes, because they are created from a number of related questions, are likely to provide a more accurate picture of opinion than are single attitudinal measures. Although some information is lost in the summation of individual opinions into a single index, the random fluctuation that will normally appear in responses to individual opinions is less problematic as a number of related opinions are added together. That is, an index has a greater likelihood of representing the 'true' individual opinion on that -80-issue since some of the noise that occurs within individual responses is eliminated by summing across related items. The goal in the exercise of creating indexes is to create as parsimonious a set of indexes as possible balanced with the need to maintain singularity of subject within each of them. While individual attitudes regarding particular policies or issues are important in and of themselves, this exercise is directed at evaluating beliefs at a more general level. For example, opinion on U.S. investment in Canada is included within the greater category of beliefs regarding continentalism. While some information-is lost in this aggregation, it does allow for greater manageability of the large number of opinion questions included in the survey. Thus these 'shorthand' packages are created not only as a device for discussing the myriad of opinions within the survey, but also because it is logically sound to assume that individuals will base similar opinions, i.e. those witliin the same broad categories, on similar principles. Moreover the exercise of assessing the structure of opinion should not rely on a limited number of opinions. The greater the breadth of issues examined, the greater is the level of confidence that one can place on the results obtained. In responding to attitudinal questions respondents at times answer that they "don't know" and have "no opinion". For the purposes of this study, such responses were recoded as middle neutral categories in every attitudinal variable where they appeared. This step was taken in order to immunize the number of missing cases in each sub-sample. As noted above, missing cases can pose a serious problem because of listwise deletion of such cases in the statistical procedures. Objections can be raised on the inclusion of such -81-responses as middle categories. There is evidence that women tend to have a higher proportion of "don't know" and "no opinion" responses and to include them as middle categories will artificially inflate the number of women providing valid responses to the survey questions. The decision to proceed with this step was nevertheless taken on the basis of three considerations. First, there is some evidence that the trend of women's greater numbers of "don't know" responses is diminishing as new generations of politically active and sophisticated women enter the electorate. Second, Shapiro and Mahajan (1986) reported that the results of their investigation of trends in gender differences in policy preferences were unaffected by the inclusion or exclusion of "don't know" responses in their analysis. Finally, and as mentioned previously, listwise deletion of missing values can become a problem when using factor analysis to investigate the existence of belief systems. Thus and in light of the first two considerations, ensuring that the sample size not drop below an acceptable level, "don't know" and "no opinion" responses were coded as middle categories. The method employed for breaking the issues into distinctive categories proceeded in two stages. First, the issue questions were grouped into a number of general policy dimensions based on the logical connections between them. The primary dividing criteria for grouping was the general policy area or ideology to which the question referred. These included such categories as foreign policy, economic policy, capitalism, social services, civil liberties, and labour issues. As an example, an "economic" package of opinions would include those related to the economic functioning of the country, ranging from the privatization of Crown corporations to subsidies for certain sectors in the -82-economy. The next step in the creation of indexes was to assess how tightly the questions fit together in each of the categories. Thus an attempt was made to determine that the assumed connections between the variables did indeed exist. To do this, each opinion question was statistically tested for its level of constraint to the general policy index. Constraint, as it is used here and throughout this thesis, means 'the success we would have in predicting, given initial knowledge that an individual holds a specified attitude, that he (sic) holds certain further ideas and attitudes' (Converse, 1964:207). As an example, a reasonable level of constraint within an economic policy category allows us to assume that an individual supportive of the privatization of Air Canada is not likely to be a strong supporter of government's involvement in business ventures generally. Stated differently, if an individual expresses conservative economic principles on some issues, one should be able to predict at some minimum level that he/she will also hold conservative opinions on related issues. Two statistical tests were used to determine the level of constraint between the variables making up each of the indexes: Pearson's correlation coefficient and principal component analysis (as employed in factor analysis). The correlation coefficient was used to measure the strength of the relationships between all pairs of variables in every index created in the first step. These bivariate correlations were examined and any variable unable to meet a pre-set rninimum average inter-item correlation was subsequently dropped from that index. This ininimum average inter-item correlation was set to 0.10. Although this may seem rather low the small range of each of the individual -83-variables results in generally small bivariate correlations. The working assumption throughout the index creation stage was that respondents were bound by some minimal level of ideological predisposition and thus would respond in a predictable fashion to each of the items included in an index. While bivariate correlations tell us how closely each pair of survey questions within an index are related, they are a weak measure of how well the questions fit together as a group. Principal component analysis allows us to measure the association between each variable in the index and the respondent's assumed underlying ideological predisposition and provides us with a measure of the 'tightness' of the group of questions witiiin an assumed index. Thus as a second statistical test, principal component analysis was used to measure the extent to which an index represents a single opinion structure. When running every index through principal component analysis, two considerations came into play. First, the principal factor extracted from every index needed to account for a large percentage of the variance across the variables, indicating that the responses to the questions in the index are generally bound by a single ideological predisposition. Second, the loadings of each variable to the principal component had to meet some pre-set minimum level in order to be taken as significantly constrained to the factor. This level is set at 0.40. There are no hard and fast rules for deterniining when factor loadings should be treated as significant, but the absolute value of .40 or greater is commonly employed. It should be noted that the index creation process proceeded in a circular fashion. Several variables could have been included in two or three different indexes depending on -84-how one chose to interpret the question and the pre-set responses. In such cases, the variable was added to the index to which it appeared to have the greatest logical connection. If the variable was then subsequently dropped from the index because it failed to pass the criteria in the second stage of the process, the process was repeated, including the said variable in another appropriate index and the testing undertaken again. In several cases the strongest logical connections were not immediately obvious. When this occurred the variable was tested in every category that seemed logically appropriate and the tightest fit determined the index in which it was included. A number of variables failed to meet the minimum standards for inclusion in an index; in such cases, the variables were removed from the indexes. Some of these variables are important nonetheless since they tap into an important issue that is not part of a greater index. As a result, in some cases these will be evaluated singly. It can be argued that some of the indexes logically belong together while others should be broken up into smaller more tightly constrained indexes. Despite the use of 'sophisticated' methodological techniques in the creation of these indexes, the end result is as much a matter of personal decision as it is one of statistical requirement. The general rule of thumb employed here was to keep indexes as simple in attitude measured as possible; that is, unless there was a clear undeniable logical reason for keeping variables together in an index when the principal component analysis suggested otherwise, they were separated. Given that factor analysis will combine those separate indexes that should be combined in the greater belief structure, separation of such variables at an earlier stage seemed the least difficult route to chose. The alternative, -85-keeping variables together that should have been separated, could have resulted in some indexes correlating strongly with more than one underlying factor which would have made interpretation of the factor solution more difficult. Given the focus of this thesis, as a last step and in order to provide an accurate picture of the opinions of both Canadian women and men, it was deemed important that the indexes be equally applicable to both genders. The working assumption in this analysis is that the political cultures of the two sexes differ, and neglecting to examine the application of the indexes to each gender separately would negate that assumption. In order to ensure the indexes accurately reflected the beliefs of both genders, the sample was split by gender and the entire procedure for creating the indexes begun anew. The results of this test, using the same statistical criteria outlined above, revealed several differences in the composition of indexes for the two samples (a total of six questions did not load significantly in both the women and men subsamples). Because of this, the indexes were scaled down to include only those variables that entered significantly into the indexes for both genders. This step allows for direct comparison of attitudes on the indexes between gender. Table 3.1 provides the results of the procedure for creating the opinion indexes. It includes the list of variables that comprise each of the indexes, as well as the results of the principal component stage of the analysis. The table reveals that every variable included in an index loads significantly on the principal component and that many of the principal components account for over fifty per cent of the variance in the component variables. Those indexes which do not account for a large share of the variance across -86-the variables are often those made up of a larger number of component variables which in and of itself makes it more difficult for a single factor to account for a large share of the variance. Thus the results of the index creation procedure appear to be robust. Those questions that were unable to meet the niinimum criteria for inclusion in an index, and that will subsequently be evaluated as stand alone questions, are not included in this list but will appear in the examination of the gender gap in Chapter 4. The next step was to create the indexes by summing the values of the responses to each of the variables included in the indexes. The one exception in this recoding procedure is the feminism variable. The creation of this variable differed slightly from that of the remaining indexes and a complete description of the creation and coding of this variable will appear in Chapter 5. All indexes are coded such that increasing values correspond to increasing support of the indexes label and were rescaled to fall within a zero to one range. This step was undertaken for ease of comparison and to ensure that in statistical procedures differences due to differences in total variation alone were minimized. Table 3.2 provides the summative statistics for the indexes for the whole sample. As a last test in the creation of the indexes, the correlation matrix for the complete set of indexes was examined to assess the degree to which any redundancy appears in them (results not provided given the large number of bivariate correlations involved). Only 6 of the 190 correlations are moderately strong (greater then .30) and they are all among pairs where some degree of association is expected: between capitalism and union support (.31); capitalism and continentalism (.41 - partly driven by the Free Trade -87-Table 3.1: Variable Composition of Opinion Indexes The complete wording of questions can be found in the CNES codebook. Included are the factor loadings of each variable on the principal component, as well as the share of the variance accounted for by the principal component and the eigenvalue of the principal component: Principal Component Factor Loadings RELIGIOSITY INDEX A/D that the bible is the actual word of God .83 A/D that religion is an important part of my life .83 Eigenvalue =1.38 Percent of Variance explained= 68.9 ANTI-FEMINISM INDEX How do you feel about feminist groups? .81 Do you think the government should do more/less for women? .81 Eigenvalue =1.30 Percent of Variance explained = 64.8 ANTI-UNION INDEX How much power do you think trade unions should have? Paying union dues should be mandatory The use of strikes is often necessary/never justified Eigenvalue =1.61 Percent of Variance explained= 53.7 .76 .74 .70 EXTRA BILLING INDEX To cut the deficit, allow doctors to bill patients directly .93 To cut deficits, allow hospitals to bill patients directly .93 Eigenvalue =1.74 Percent of Variance explained= 86.8 -88-Table 3.1 Continued: Variable Composition of Opinion Indexes Principal Component Factor Loadings DEFENCE POLICY INDEX Canada should stay in/get out of NATO (campaign) .58 Canada should stay in/get out of NATO (post-election) .60 Support/oppose the buying of nuclear submarines (campaign) .77 Support/oppose the buying of nuclear submarines (post-election) .77 To cut the deficit, government should reduce the defence budget .61 Eigenvalue = 2.25 Percent of Variance explained= 45.0 CYNICISM INDEX Quite a few/hardly any of the people ruiming government are crooked People in government waste alot/not very much the money we pay in taxes Can just about always/only some of the time trust people in Ottawa to do what is right Almost all/few people in government seem to know what they are doing Eigenvalue = 2.03 Percent of Variance explained= 50.7 CONTINENT ALISM INDEX Canada should have closer ties/distance itself from the U.S. .78 Support/oppose the Free Trade Agreement .84 Free Trade Index - measures support for the policy on a 7 point scale. .84 Canada needs more U.S. investment .57 No amount of money would get me to move to the U.S. .52 Eigenvalue = 2.62 Percent of Variance explained= 52.5 LOW TAXATION INDEX To cut the deficit, increase taxes on consumption .81 To cut me deficit, increase personal income tax rate .81 Eigenvalue =1.30 Percent of Variance explained= 65.2 .72 .71 .73 .69 -89-Table 3.1 Continued: Variable Composition of Opinion Indexes Principal Component Factor Loadings ANTI-SOCIAL SPENDING INDEX To cut the deficit, reduce spending on universities .58 To cut the deficit, make it harder to get unemployment insurance .54 To cut the deficit, reduce welfare payments .68 To cut the deficit, reduce spending on arts and cultural activities .67 Eigenvalue =1.54 Percent of Variance explained= 38.5 ANTI-DUALISM INDEX How much do you think should be done to promote French? Favour/oppose extension of services to language minorities (campaign) Favour/oppose extension of services to language minorities (post election) Right to French education outside Quebec Support/Oppose Quebec Sign Law Agree/Disagree Canada has two founding peoples, British and French Eigenvalue = 2.41 Percent of Variance explained^ 40.2 ANTI-STATE INVOLVEMENT INDEX To cut the deficit, sell CNR to private investors .68 To cut the deficit, sell Petro-Canada to private investors .69 Poorer regions in Canada normal market result/demand government job incentives .49 Farmers and fishermen should find new work/get government help .64 To cut the deficit, reduce subsidies to farmers .63 To cut the deficit, reduce CBC government grants .47 Eigenvalue = 2.20 Percent of Variance explained= 36.7 .79 .72 .70 .59 .49 .44 -90-Table 3.1 Continued: Variable Composition of Opinion Indexes CAPITALISM INDEX Principal Component Factor Loadings Important industry decisions should be left to management alone .53 Unskilled workers receive wages that are about right for level of skill required .55 A person's wages should depend on the importance of the job .49 Working people earn what they deserve 64 A/D some form of socialism would be better than the system we have now .53 Profit system teaches people value of hard work and success .50 Getting ahead a matter of ability and hard work .43 Competition leads to better performance and a desire for excellence .47 Most business executives do important work and deserve high salaries .53 When business allowed to make as much money as they can everyone profits .59 Eigenvalue = 2.80 Percent of Variance explained= 28.0 ANTI-SEPARATE SCHOOL FUNDING INDEX A/D public money should not be used to support religious schools .81 A/D where religious schools have gotten public funds they should continue to receive them .83 A/D public money for schools should be made available to any religious group .69 Eigenvalue =1.83 Percent of Variance explained= 61:0 ANTI-CIVIL LIBERTIES INDEX Equal rights for homosexuals damage Canadian moral standards .49 Society sometimes has to forbid some things from being published .54 Persons in positions of responsibility are entitled to be treated with special respect .49 Too much emphasis placed on individual freedom .60 Too little emphasis placed on respect for institutions .55 Eigenvalue = 1.44 Percent of Variance explained= 28.9 -91-Table 3.1 Continued: Variable Composition of Opinion Indexes Principal Component Factor Loadings ANTI-IMMIGRATION INDEX Should Canada admit more/fewer immigrants? .85 A/D that new immigrants make Canada a better place to live .85 Eigenvalue = 1.43 Percent of Variance explained= 71.5 BRITISH TIES INDEX A/D that Canada will always have a special relationship with Britain .83 It is important to keep the monarchy as a link to our past/no place in Canada today .83 Eigenvalue =1.38 Percent of Variance explained= 68.9 TRADITIONAL ROLES INDEX More women should stay home/ have careers Women make less money because hers is supplemental income/result of discriniination Eigenvalue =1.23 Percent of Variance explained= 61.4 .78 .78 PRO-LIFE INDEX Abortion should never be peirrutteaVpermitted after need established/ .89 is a matter of the woman's personal choice (campaign) Abortion should never be permitted/permitted after need established/ .90 is a matter of the woman's personal choice (post election) If R.C. hospital only one in region, it should perform abortions .70 Eigenvalue = 2.11 Percent of Variance explained= 70.2 -92-Table 3.1 Continued: Variable Composition of Opinion Indexes Principal Component Factor Loadings LOW ENVIRONMENTAL PRIORITY INDEX Land and natural resources should be turned over to private owners A/D that protecting the environment is more important than creating jobs Eigenvalue =1.11 Percent of Variance explained ENERGY INVESTMENT INDEX To prevent energy shortages, encourage development of oilfields and dams .79 Big energy projects are vital strategic investments .79 Eigenvalue =1.26 Percent of Variance explained= 62.8 RISK AVERSION INDEX People who try to reform things are busybodies who do more harm than good .81 Trying to make sweeping reforms is much too risky .81 Eigenvalue =1.32 Percent of Variance explained= 66.0 .75 .75 = 55.6 -93-component of the index); capitalism and state involvement (.33); state involvement and continentalism (.37); state involvement and social spending (.35); and abortion and religiosity (.41). Some of these correlations may be argued to be of such strength that an argument could be made for collapsing some of the indexes together. The decision not to collapse the indexes was made in order to keep each index as unique in focus as possible. To have collapsed indexes might have meant a loss of information that was important to the investigation of gender gaps. For example, although continentalism and state involvement are moderately correlated, the indexes do appear to be measuring distinct issues. Moreover, the size of the gender gap in continentalism relative to that in state involvement would not have been apparent had the indexes been collapsed (see Chapter 4). Some discussion of the results in Table 3.2 is in order. In general, Canadians do not consistently reveal conservative (nor liberal) opinions across the range of opinions tapped by these indexes. Indeed they reveal a mix of opinions. Canadians are conservative in their lack of support of unions and in their support of the capitalist system and relations with the United States. Similarly, they are 'hawkish' in their support of defence measures and support mamtaining ties with the British Commonwealth. Thus Canadians appear to have endorsed the main tenets of classical liberalism, referred to here as conservatism. And this conservatism extends to the religious sphere of their lives. But this conservatism is not found on all issues. Canadians are liberal in their endorsement of government involvement in the economy and in their support of the welfare state. They are also supportive of the dualistic nature of Canada and are more -94-Table 3.2: Summary Statistics of Opinion Indexes for the Full Sample Index Mean Standard Deviation Religiosity .55 .37 Anti-Feminism .39 .21 Defence .59 .26 Continentalism .52 .31 Anti-Social Spending .42 .20 Anti-Dualism .40 .26 Anti-State Involvement .45 .21 Low Taxation .80 .21 Capitalism .61 .21 Anti-Civil Liberties .46 .22 Traditional Roles .28 .27 Pro-Life .33 .29 Anti-Union .58 .25 Extra Billing .22 .27 Anti-Separate School Funding .60 .34 Anti-Immigration .62 .37 British Ties .66 .35 Low Environmental Priority .26 .28 Energy Investment .66 .31 Cynicism .71 .24 Risk Aversion .37 .32 -95-pro-choice than pro-life. They are also egalitarian in their support of women's issues and in promoting individual freedoms more generally. And despite a generally supportive view of attempting reforms in society, the cynicism directed at politicians and governments makes it unlikely that they have much faith in these institutions being able to bring about such reform. Thus Canadians have an ideological mix of opinions that at first glance may appear somewhat disjointed. -96-Chapter 4; Is there a Gender Gap in Canadian Opinion? The evaluation of the gender gap in the 1988 Canadian National Election Study is the first step in an investigation of gender and political behaviour and this chapter provides such an evaluation. Increasing evidence suggesting that gender, in combination with other social forces, is an important factor in political behaviour in many countries has served to increase the attention directed towards the notion of the gender gap in the media and in academic circles. Its discovery, both in voting and opinion data, precipitated a number of investigations into political behaviour and gender. Many of these investigations have attempted to account for gender differences in attitudes, but few have evaluated women's attitudes and belief systems in as comprehensive a fashion as is the goal of this investigation (Conover, 1988a). Investigating the existence of gender gaps in opinion in the 1988 CNES is an important starting point, for their existence alone would suggest a need for reassessing belief regarding gender and political thiriking. Moreover it stands as an important continuing step in the comparative evaluation of gender gaps across countries arid over time. The Gender Gap: Past Evidence ^ The notion of a gender gap first appeared in the literature examining American elections in the early 1980s. Reagan's bid for the Presidency revealed that women were more supportive of Democratic candidates than men and this gave rise to greater interest -97-in the "woman's vote". In 1980 women were 4 percent less likely to support Reagan as presidential candidate than men, and by 1984 that figure had grown to 9 percent (Kenski, 1988:44). Kathleen Frankovic attributed this gap to women's unwillingness to be more aggressive in foreign policy (1982:446). The discovery and discussion surrounding this voting margin was significant for it signalled a recognition of the potential women held as a powerful voting bloc. This significance was underscored by the discovery of voting gender gaps in other countries. In Canadian elections between 1974 and 1988 women's support for the Liberal party ranged from 3.4 percentage points to 9.1 percentage points higher than that of men's, and in 1984 British women abandoned the Conservative party, despite having supported it in greater numbers than men since the 1960s (Wearing and Wearing, 1988:344; Norris, 1986:120). This evidence suggested that the gender gap was not simply an American phenomenon but rather a general tendency in voting behaviour across developed countries. The discovery of the voting gender gap was followed by research noting that voting differences are matched by gender gaps in political attitudes. Since 1980, election studies and survey research have documented such gaps in a number of countries including the United States, Canada and Britain. Conventional wisdom on the gender gap in political attitudes holds that women are more likely than men to oppose the use of nuclear energy and military force, give more support to "compassion" issues, and reveal more conservative tendencies on a number of "traditional" issues. The most longstanding difference in the attitudes of men and women occurs on the -98-issue of defence policy. Women's opinions reveal a consistent and strong trend away from men's on the issues of defence and force. Robert Shapiro and Harpreet Mahajan report that in U.S. surveys conducted between 1952 and 1983 the average gender percentage point difference on questions dealing with the issue of force was 8.1, with men selecting the more violent options (1986:49). The surveys examined in their analysis included the General Social Survey, Gallup and Harris data, and the American National Election Studies. Others have provided supportive evidence of this difference of opinion among the U.S. electorate (Abzug, 1984; Baxter and Lansing, 1983). And in a related issue, Clark and Clark found that in 1988 women were 10.2 percentage points less likely to support the death penalty (1993:38). Canadian evidence is more limited, but is similarly supportive of a gender gap on pacifism issues. Examining Gallup survey data gathered in 1983 and 1984, John Terry reports the existence of even larger differences on this issue than those recorded in the United States: gender differences from 12 to 18 percentage points existed on questions dealing with the issues of allowing cruise missile testing in Canada and the chances of nuclear war (1984:12). Similarly, Janine Brodie cites a CBC News 1984 Federal election survey in which a 13 percentage point difference existed in support of allowing cruise missile testing in Canada (1988:180). The gender differences on questions dealing with these issues are consistent across surveys, and are statistically reliable: their size usually exceeds the margin of error for the samples on which they are based. A further commonly reported division in the attitudes of women and men appears on what are commonly referred to as "compassion issues". As with the defence gender -99-gap, men consistently reveal opinions that are less liberal than women's on compassion issues: women tend to endorse a strong role for government on issues concerned with welfare policies, aid to the poor, the unemployed and others in need. But these differences are often not as large as those reported for the issues of force and defence. In U.S. research, Shapiro and Mahajan report that survey data from 1952 to 1983 showed the average gender percentage point difference on 'compassion' issues to be 3.3, while it was 5.8 on questions dealing with support for regulation and protection (1986:50). Further confirmation of this tendency is provided by Sandra Baxter and Marjorie Lansing who report that Hero found women "several percentage points more favourably disposed towards such programs as aid to dependent children, unemployment assistance, old age benefits, Medicare, anti-poverty programs and 'relief generally" (1983:60). Similar results have been reported for other surveys (Poole and Zeigler, 1985; Deitch, 1988). Comparable Canadian evidence supports the existence of a compassion gender gap. Based on Gallup data, Terry reports that women are more likely to support expanding social welfare programs by a 6 percentage point difference and are less likely to say that the future development of Canada depends on private enterprise by 15 percentage points (1984:11). Using 1979 Quality of Life survey data, Kathryn Kopinak found that Canadian women 'ideologically favour a redistribution of power in the workplace as well as a redistribution of valued goods and services by government' (1987:29). She reports that a statistically significant difference of means between women and men on an attitude scale measuring support for government welfare efforts existed in 1979. This attitudinal difference is further supported by Gallup data reported by Joanna -100-Everitt (1994). As revealed in her analysis, by the late 1980s and early 1990s women and men were showing between 6 and 9 percentage point gaps in opinion on the issues of daycare support, Family Allowance, Pensions and a Minimum Annual Income level. Consistent differences recorded in various surveys from both the United States and Canada support a conclusion that a gap on compassion issues exists, even if smaller in size than differences on defence issues. There also exists evidence that women hold more conservative opinions than men on questions touching upon moral issues. Women score higher on religious fundamentalism scales (Poole and Zeigler, 1985; Conover, 1988); are more likely to endorse school prayer (Clark and Clark, 1993); support in greater numbers the censorship of TV programming and pornographic material (Clark and Clark, 1993; Terry, 1984); and give greater support to a rise in the legal drinking age (Terry, 1984). The evidence of the existence of gaps on these 'traditional' issues is less conclusive however as results are usually smaller than those recorded in the defence and compassion issues and contradictory evidence does exist. As examples, Elizabeth Adell Cook and Clyde Wilcox (1991) found no significant difference between women and men in support for school prayer, and Cynthia Deitch in her evaluation of General Social Survey data from 1973 to 1984 found only one significant difference in support for school prayer (1988:205). Finally, Shapiro and Mahajan record an average percentage point difference of only 2.3 for such survey questions across the examined data sets, a difference which they interpret as "small" relative to other recorded differences in the data (1986:52). While there is evidence supporting the existence of gender gaps in opinion on -101-these three issues, there has been little reported research on the existence of gender gaps on other issues. The research extant is limited, and sometimes contradictory, making it difficult to substantiate the existence of any patterns. Nevertheless, those issues which have been examined in previous studies shall be presented in order to lay the groundwork for an examination of any differences in the 1988 CNES. On what are sometimes referred to as women's issues, specifically the questions of abortion and child care, evidence on the existence of a gender gap is mixed. First, on the issue of abortion, Keith Poole and L. Harmon Zeigler report that in National Election Studies conducted in 1972, 1976 and 1980, the gap in the most permissive abortion response category ranges from 3 to 4.7 percentage points with women moving from less (in 1972 and 1976) to greater support for permissive abortion policies (1985:83). And from the 1988 American NES, Clark and Clark report a small percentage point gender difference (3.9 points) in support of a liberal abortion policy, with men holding the more liberal opinions (1993:38). But in her evaluation of General Social Survey data from 1973 to 1984, Deitch reports no significance differences on attitudes toward abortion (1988:205). With the gains made by women in increasing the attention given to the abortion issue, and feminist arguments regarding the right of a women to control her own body, one might expect to see greater support for choice among women than has been the case. Yet the gains made by pro-lifers in generating support for their stance, often closely linked to religious arguments and groups, are as likely to have an impact on the attitudes of many women, given reportedly greater levels of religious conviction. Religious -102-arguments may be competing with rights and gender arguments among women, and if so, could serve to reduce the degree to which a simple gender gap is apparent. The gap may exist yet be invisible due to the confounding impact of religion on the relationship. Removing the impact of religion on abortion opinion may liberate the gap on this issue. Second, on the issue of child care, there is insufficient reported evidence to discern any patterns. The only available evidence suggests a gap exists on the issues. Clark and Clark report that in 1988, American women were more likely by a gender percentage point difference of 9.7 to dunk spending on child care should be increased (1993:39). The only evidence for Canada comes from Terry who reports that in June of 1983 women were more likely than men to diink spending should be directed at child care centres rather than distributed as allowances (1984:13). Evidence is also mixed on the existence of a difference of opinion on the issue of civil rights for minorities. Although they concede that exceptions exist, Baxter and Lansing conclude that "over the years, there is sufficient difference to state that women have been more supportive of efforts to achieve racial equality than have men" (1983:59). In direct contradiction however, Poole and Zeigler's research led them to conclude that "Men and women do not differ greatly on civil rights issues" (1985:148). The expectations for opinions on this issue in Canada are not clear, but here again, it may be the case that religiosity muddies the relationship between attitudes on civil liberties and gender. On environmental issues, Abzug documents a 1981 CBS/NY Times poll in which 48 percent of women and only 41 percent of men favoured continuing environmental -103-improvement programs regardless of the cost. (1984:125). Poole and Zeigler report that in 1980 38.4 percent of men favoured relaxing environmental regulations, while only 29.6 percent of women agreed to the statement (1985:70). Based on this limited evidence, it can tentatively be said that women are somewhat more concerned with protecting the environment than are men. Finally, one study found that women were slightly more cynical than men regarding politicians and politics in general. Clark and Clark note that in 1988, American women were 6.5 percentage points more likely than men to think that politicians are corrupt (1993:39). The gender gap evidence is compelling, at least for attitudes on compassion and defence issues, for two main reasons: first, the difference in attitudes on these questions has appeared over time, and second, the difference is consistent despite survey effects, particularly question wording. Alternative measures of attitudes employed to survey different populations of men and women are providing consistent and hence reliable conclusions. It is important to extend the investigation to include new issues, new surveys, and new populations to assess the confidence that can be placed in concluding that a gender gap exists. The 1988 CNES allows for an investigation into whether such differences in attitudes persist, and moreover, allows for the investigation to address issues and policies not previously assessed in Canadian research. The review of the gender gap evidence is perplexing on one score: a gender gap in opinion is less likely to be consistently found on issues that are directly tied to questions of gender. That is, where one might expect to see a gender gap because the issues are -104-fundamental to women and their freedoms, for example on the abortion issue, one is less likely to find a gap than if one looks at a question which is less immediately connected to gender alone, for example on the issue of defence. As suggested above, the competing values of religious belief (or the degree to which religion is an important and relevant factor in one's life) and feminism, may explain part of this puzzle. It may be that among women for whom religion is of tremendous importance (and this is likely to be the case more often among women than among men), gender concerns are secondary in importance. That is, among religious women, questions surrounding women as a group, and their equality, freedoms, and societal benefits, may be superseded by religious ones. This degree to which this may be the case will be examined in Chapter 5. It should be pointed out at this point in the investigation that when a gap in opinions exists and persists it is still the case that a significant number of Canadian men and women are in agreement on the policy and/or issue. Thus differences are often in intensity rather than direction. This fact should not be overlooked or lost in the focus on difference. That the means of responses to survey questions are statistically and substantially different by gender does not indicate that women and men are fundamentally at odds. The deviation around the mean for both genders should not be forgotten, and often represents a substantial overlap in opinion.1 1 I thank Carol Bacchi for pointing out to me the synedochic fallacy: 'the tendency to refer to a group or collectivity in terms of its central tendency, as if it had no internal variation' (Bacchi, 1986:63). The point is well taken but measures of central tendency are nevertheless an important statistical tool for describing group attributes. -105-The Gender Gap in 1988 Table 4.1 presents the results of difference of means tests for the twenty indexes created in the manner outlined in Chapter 3. The entries in the first two columns of the table are the means on that index for men and women respectively. The next two columns report the standard deviations around the means on that index, again respectively for men and women. The final column reports the difference of means, its statistical strength and its direction. Given that every index ranges from 0 to 1, the difference of mean can be expressed as a percentage of the range. Moreover, given that all indexes were coded so that increasing values correspond to increased conservatism, a positive gap means that women are more liberal on that issue. Alternatively, a negative gap means that men are more liberal on the issue. Religious belief and feniinisiri are the two first entries in the table given their hypothesized role as core values. The remaining indexes appear in the table in descending order of the value of the gap on that issue, i.e. from the largest gap in opinion to the smallest. Table 4.2 provides the percentage point differences in opinion on the number of stand-alone questions included in the study. Eleven of the twenty indexes reveal statistically significant differences across the means for men and women, and the responses to five of twenty-seven survey questions reveal percentage point differences between gender categories that are greater than ten. It is not immediately clear from these results whether women and men live in different political worlds, but it does suggest that on some political issues they hold differing opinions. The important task of this chapter is to identify on which issues these differences appear and to deterrnine their consistency with differences noted in the past. -106-Core Values: Religiosity and Feminism It has been suggested that part of the gender gap story lies in a competition of values between religious and feniinist beliefs. The evidence from Table 4.1 provides some support for this explanation. Substantive and significant differences are recorded on both the feminist and religiosity indexes. On these two core values, women and men are at odds although the difference is greater on religious belief. On the religiosity index, women reveal a strong and statistically significant difference from men: the percentage point difference between the genders is equal to 7 percent of the index's range, the third largest gap recorded in the indexes. Women score higher on an index made up of questions asking respondents whether they take the Bible to be the word of God and whether they agree that religion is an important part of their lives. This gap corresponds with differences that previously have been recorded in the U.S. (Poole and Zeigler,1985; Clark and Clark, 1993). It also reinforces Bernard's (1981) suggestion that religion has played a fundamental role in women's culture and continues to be more important for women than men . The relevance of religious belief for political opinions in Canada has also been reported (Johnston, 1986). The feminism index reveals a moderate but nonetheless statistically significant difference of means, 2 percent of the range. The feminism index is comprised of two survey questions; the first measures feelings about feminist groups and the second measures belief concerning whether the government should do more for women. Previous studies have reported the absence of significant differences on various -107-Table 4.1: Gender Differences in Opinion Indexes ( Index Mean Standard Deviation Difference of Means cf cf Religiosity .52 .59 .38 .35 -.07** Anti-Feminism .40 .38 .21 .20 .02* Continentalism .59 .44 .30 .30 .15** British Ties .61 .70 .36 .34 _ 09** Capitalism .64 .58 .22 .20 .06** Anti-State Involvement .48 .42 .22 .19 .06** Defence .61 .56 .27 .25 .05** Energy Investment .68 .64 .32 .30 .04** Anti- Sep. School Funding .62 .57 .34 .34 .04** Traditional Roles .29 .26 .29 .26 .03** Anti-Immigration .61 .64 .38 .36 -.03 Pro-Life .32 .35 .28 •30 -.03* Extra Billing .23 .21 .28 .26 .02 Risk Aversion .38 .36 .32 .31 .02 Cynicism .70 .72 .24 .23 -.02 Anti-Social Spending .43 .41 .20 .20 .02 Low Environmental Priority .26 .27 .29 .28 -.01 Anti-Dualism .41 .40 .27 .26 .00 Anti-Civil Liberties .46 .46 .23 .21 .00 Low Taxation .80 .80 .22 .20 .00 Anti-Union .58 .58 .26 .25 .00 Note: d"=men and ¥=women. A positive gap means that men are more conservative on the issue; a negative gap means that women are more conservative. ** p<01; * p<05; -108-Table 4.2: Percent Agreement with Individual Issues bv Gender Men Women Percent Difference Stand Alone Questions A. State Involvement in Economy 1. Privatize Air Canada 65.0 45.6 19.4 2. Energy prices should be left to the market 37.9 25.2 12.7 3. Government should control inflation 35.2 27.0 8.2 4. To cut deficit, reduce regional development subsidies to industries 32.8 26.7 6.1 5. To cut deficit, allow C B C more advertising 60.9 52.6 8.3 6. Gov't regulation of business does more harm than good 26.6 18.9 7.7 B. Workers and Management 1. If private enterprise abolished, few people would do their best 58.9 47.8 11.1 2. Workers and management share the same interests in the long run 69.2 68.2 1.0 3. If people fail, they weren't given a good enough chance 30.7 24.7 6.0 4. People don't work hard because their job is dull, unpleasant, or unimportant 40.4 36.9 3.5 5. Corporations don't pay their fair share 77.5 79.2 -1.7 C. Social Policy 1. If parents want child care, they should pay for it themselves 30.3 25.9 4.4 2. Adults should be prevented from buying pornographic material because it degrades women 20.1 37.4 -17.3 3. Owners of ethnic restaurants should have to hire anyone qualified 73.7 78.0 -4.3 4. Capital punishment is never justified 27.1 33.2 -6.1 5. Should try to encourage immigration from countries most like us 45.6 30.1 15.5 D. Morality/Religion 1. Teach children that all men created equal, but we all know that some are better than others 50.3 40.8 9.5 . 2. It is the duty of the strong to protect the weak 78.8 74.8 4.0 3. Canada is a country of many religions and faiths 76.4 82.1 -5.7 4. Institutions should refrain our selfish instincts 7.0 4.8 2.2 E. Other Issues and Policies 1. Support the Meech Lake Accord 43.8 29.8 14.0 2. Native peoples should have self-government as long as it is democratic in principle 67.1 65.4 1.7 3. Best for native people to be assimilated 51.7 43.8 7.9 4. English Canadians who move to Quebec should have a right to school children in English where numbers warrant 66.5 68.4 -1.9 5. To cut the deficit, make Family Allowance payments to low income families only 69.8 63.9 5.9 6. Must ensure an independent Canada even if it means a lower standard of living 51.6 51.9 -0.3 7. If provincial law conflicts with the Charter, the Courts should have the final say 48.6 43.3 5.3 Note: A positive gender gap means that men are more likely to hold a particular attitude; a negative one shows that women were more likely than men to have the attitude in question. -109-feminism measures. It has been suggested that the negligible difference in opinion on feminism is partly the result of men's increasing responsiveness to the messages being sent out by the women's movement and as a result, their increased willingness to agree with those messages (Deitch, 1988; Clark and Clark, 1993). The discovery of a significant and substantive difference of opinion in the 1988 CNES, however, downplays the universality of this "coming together" of opinion and instead reinforces the division of opinion between the genders that exists on this issue. Force and Defence The 1988 data provides evidence that a statistically significant gap on the issues of force and defence exists in Canada. Women score lower than men on an index comprised of questions on support for NATO, the purchase of nuclear submarines, and arguments for mamtaining the defence budget level. The difference of means for the defence index is five percent of the range, and although there is no objective measure for deterrmning the substantive significance of a difference of this magnitude, we can state that it falls at the middle range of the statistically significant differences found in this sample of indexes. Nevertheless, as has often been found in the past, in 1988 Canadian women were more pacifist, dovish if you will, than men. It is important to point out however, that on average both women and men are generally supportive of defence policies, but women tend to be somewhat weaker in their support as a group. -110-Compassion As noted above, gender gaps have traditionally appeared over "compassion issues". Past evidence indicates that women tend to have more sympathy for the downtrodden, and that this sympathy translates into a greater acceptance of a dominant government role in the economy and society. I choose to include the capitalism, state involvement, and social spending indexes as compassion issues. Many of the differences discovered in these indexes and several of the stand alone questions asked in the 1988 CNES signify the existence of this compassion gender gap. Confirming the results reported by Terry (1984), women are less likely to endorse capitalism and its tenets. The difference of means on this issue is 6 percent of the range for the index, moderate over the range of differences recorded. The capitalism index is a combination of questions related to the free enterprise system and economic individualism. Hence the index touches on the need for state involvement in general, and the notion that hard work and competition are the proper tools for bringing about the 'best' and the 'most' in individuals. It is a compassion issue in the sense that it addresses belief in economic individualism, and by connection, the ability to succeed in society without the help of the government. As seen in Table 4.1, both genders overwhelmingly support and endorse capitalism's basic tenets, but men are more unified in their support than women. In short, women are somewhat less likely to agree that Adam Smith's invisible hand is sufficient and necessary for bringing about a fair and equitable economic system. Recalling for a moment the discussion in Chapter 2, one can suggest that such differences stem from -111-either of two sources: self-interest in that women are less likely to benefit equally from capitalism and as such are less supportive of the system overall; or a value difference in that women bring with them to the political arena a set of values that differs from men and as such, emphasize notions of responsibility and caring that are to a degree incompatible with the values embodied in capitalism (most notably individualism). But regardless of its source, the difference itself means that women are generally less supportive of the system that dominates in the economic sphere and infiltrates the political and private spheres. Capitalism is intimately connected with liberalism as a governing principle, and as such bears scrutiny if groups in society are less cornmitted to it as a guiding principle. A glance at the stand alone questions provides less support for an ovemhelming gender difference in support for capitalism. Few substantive differences of opinion appear across the set of stand alone survey questions on capitalist questions that were not included in the capitalism index (see Section B of Table 4.2). One substantive difference of opinion is recorded, a difference of 11.1 percentage points on the degree to which competition motivates individuals, but the remaining differences for these questions range from 6.0 to 1.7, and as such they are not as robust as the other recorded differences. Moreover the differences do not always appear in the direction anticipated given previous results. The largest difference, and in the direction anticipated by previous results, appears over the question as to whether the private enterprise system provides the motivation for individuals to do their best. Women are not as likely to see private enterprise as the -112-motivating factor for good work habits: only 47.8 percent of them agree with the statement, while 58.9 percent of men agreed with it. The remaining differences, despite their sometimes unanticipated direction, are not substantive and suggest little division on these questions. The fact remains, however, that the bulk of the evidence supports the conclusion that women and men have a difference of opinion on the question of support for capitalism. As a second component of the compassion issue, the state involvement index reveals an intergender difference of means that is as robust as that found in the capitalism index, and unlike opinion on capitalism and the free enterprise system, many of the stand alone questions in this area also exhibit strong differences between the genders. The index itself is a combination of questions regarding the privatization of Petro-Canada and CNR; economic redistribution and subsidies for the less viable regions and segments of the Canadian economy; and public financial support for the CBC. It comprises a fundamental aspect of the 'compassion' question in that it queries respondents about their support for redistributive government policies for those in need. Table 4.1 shows that women endorse a strong role for government in the economy as indicated by their higher mean score on the index (the statistically significant difference of mean accounts for 6 percent of the index's range). These results are similar to those reported by both Terry (1984) and Kopinak (1987) for Canadian data. Support for a division of opinion on this issue also appears in the stand alone questions associated with this policy area (Section A of Table 4.2). Gender percentage point differences on the six state involvement in the economy questions are all greater -113-than 6, and two are greater than 10. In all cases, women endorse a greater role for the state in the economy than men whether it be on specific questions such as the financial support of the CBC (8 percentage point difference), the privatization of Air Canada (19) or on the reduction of regional subsidies tp industries (6). More generally, women are less supportive of allowing the market to detennine energy prices (13) which is anticipated by and associated with their lesser support of capitalism and its tenets. Similarly women are less likely to agree with the statement 'Government regulation of business does more harm than good'. Thus women as a group are more supportive of state involvement in the economy, and this greater support is not confined to programs or spending designed to aid women as a group. The overall conclusion of an opinion gap on state involvement in the economy is supported in 1988. Contrary to expectations however, no substantive difference appears in support of social welfare policies, an area that has traditionally been labelled a gap issue. The social spending index is composed of questions suggesting reductions in spending on universities, welfare payments, arts and cultural activities, and making it harder to get U.I. payments, all in an effort to reduce the deficit. There is a small percentage point difference by gender on the spending index (roughly equal to 2 percent of the range) but there is little reason to conclude a gender difference in opinion on this issue since the difference is neither substantive nor statistically significant. In the stand alone questions, women were also less likely to agree with the statement 'To cut the deficit, make Family Allowance payments only to low-income families', but only by 5.9 percentage points, a modest difference at best. -114-The lack of difference in the social spending index is intriguing since differences on social welfare questions have been recorded in the past, although the bulk of the evidence comes from the United States. It is not immediately apparent why minimal differences exist on this policy issue. The difference may be due to the interpretation of what constitutes 'social welfare' survey questions by researchers. It may simply be that some of the reported 'social welfare gaps' employed survey questions that were dissimilar to those included in the social spending index used here. For example, Everitt's (1994) conclusion of differences in support for social welfare programs in Canada is based on Gallup survey questions directed at child care, pensions, guaranteed annual incomes, and family allowance payments. There is no correspondence between this definition of social welfare and the social spending index employed in this analysis. The only correspondence is to the stand alone family allowance question, not included in the social spending index, but which did report a modest difference in the CNES data set. If previous researchers selected questions other than those selected here as indicators of opinion on social welfare, then we should look to the questions that might have been labelled 'social welfare' questions by others in order to determine if this selection of questions would have resulted in a conclusion of difference. For instance, the subsidy questions that are part of the state involvement index specifically refer to aid to economically disadvantaged groups, questions which are often considered social welfare questions. Had they been part of this social spending index, then the differences reported in them would support the existence of the gap in the social spending area as defined by others. -115-It remains unclear, however, why the questions tapping into support for unemployment insurance and welfare payments that are part of the social spending index did not reveal greater differences than those reported. Such issues have shown sharp gender contrasts in past studies. It will be interesting to see whether there has been a weakening of the gap in these areas, or if it is simply an anomaly of this survey. Despite the absence of difference on these few questions it is still true that Canadian women were more liberal in opinion than men on compassion issues in 1988. Traditional Issues Gaps have also been recorded over traditional issues, a third area which includes questions of a religious nature and questions of morality. Some evidence has reported that on these questions, contrary to attitudes in the compassion and defence areas, women tend to reveal more conservatism than men. The evidence in 1988 is mixed, and a conclusion of differences here depends heavily on one's choices of questions to include in this category. On the civil liberties index there is no reportable difference in the opinions of women and men. On an index made up of questions of regarding support for individual rights and freedoms, homosexual rights, publication restrictions, and respect for those in positions of authority, the opinions of both genders correspond. But substantive differences do appear in the stand alone questions. Women are much more likely to agree that the sale of pornography should be restricted because it degrades women (by a 17 percentage point difference) and are less likely to agree with the statement 'We have to -116- . teach children that all men (sic) are created equal, but almost everyone knows that some are better than others' (by a 10 percentage point difference). Only a moderate difference exists with women being more likely to agree with the statement 'Canada is a country of many religions and faiths' (by 6 percentage points). The only other Canadian evidence on attitudes of this type comes from Terry (1984) who reports that women were 14 percent more likely to agree with TV censorship. Thus substantive differences in attitudes on civil liberties appear in one survey question on restrictions to the sale of pornography, a question whose wording makes specific reference to the degradation of women by the material which in and of itself likely increases the salience of the question for women respondents more than it does for men, and on a separate question on the inherent equality of "men". These differences provide little reason for concluding that there is a fundamental difference of opinion on the issue of civil liberties. Evidence of attitudinal difference on questions of civil liberty in the U.S. has also been less than conclusive (Baxter and Lansing, 1983; Poole and Zeigler, 1985). Other Issues The 1988 CNES includes a large number of survey questions that extend beyond these three issue areas and which allow for an assessment of the degree to which gender gaps exist in other areas. One attitudinal difference which stands out is that found in the continentalism index. This statistically significant difference, which is equal to 15 percent of the range of the index, is the largest difference recorded across the set of -117-indexes, and is fifty percent greater than the second largest index gap. Apart from the size of the difference alone, the gender gap on this issue is interesting for the continentalism index includes individual questions regarding free trade with the United States. This issue, the Free Trade Agreement, dominated the 1988 election campaign and the consciousness of Canadian voters (see Johnston et al., 1992). Given the increased salience of this issue over the other questions that make up the continentalism index, it stands as an important exercise to examine the degree to which the question on free trade is driving this gender gap in opinion. Table 4.3 presents the breakdown of the continentalism index, and reports the differences found in support for the survey questions that comprise it. The most staking elements in this table are first, the size of the recorded gaps in opinion, and second, their consistency across each of the questions in the index. The gap between women and men on the question of free trade was substantial in 1988. A gap of 19 percentage points existed in both survey questions on the issue. The salience of the issue was due in part to the campaign levied both for and against it and to the fact that Canadian women's groups were active in voicing their opposition to the agreement (Bashevkin, 1989a). Critics of the Agreement argued that it was a direct attack on the interventionist state and warned that if signed, the Agreement could threaten Canada's social welfare programs. In light of these arguments, and the previous evidence of women's greater support for government intervention in the economy and their lesser support for capitalism and unfettered competition, it is understandable that women rejected the agreement in greater numbers than did men. In her analysis of opinion on free trade in the 1988 CNES, -118-Table 4.3; Breakdown of Public Opinion on Continentalism Issues by Gender Percent Agreeing Men Women Percent Gap 1. Think Canada should have closer ties to U.S. 2. Canada needs more U.S. investment 3. No amount of money could get me to move to the U:S. 4. Support for the Free Trade Agreement (campaign wave) 5. Support for the Free Trade Agreement (post-election) 50.7 38.8 47.5 48.1 64.0 29.7 25.9 29.5 29.1 44.9 21.0 12.9 18.0 19.0 19.1 Note: The post-election free trade question is a combination of three forced-response questions combined into a nine-point scale. Question 3 reports on the percentage of respondents disagreeing with the statement. Elizabeth . Gidengil (1995) concluded that 'Women's opinion [on free trade] was shaped by their commitment to the welfare state rather than their assessments of the health of the national economy and they were much more likely to men to accept the anti-free trade argument that Canada's social programs were at risk' (403). What is clear from this table is that the gender gap in continentalism was not driven singly by opinion on free trade; substantive differences appear in each of the index's component questions. Besides differences on the trade question, women were also less likely to think ties should be strengthened between the two countries or that Canada needed more U.S. investment. Canadian women were also less willing to move to the U.S. Interestingly, if the tradeoff to Canada's independence is a lower standard of living, the difference of opinion disappears (see Section E of Table 4.2). This supports the conclusion that the gap was based less on differences in nationalism than it was on questions of economic self-interest. It is true that the gaps on the individual questions in -119-the continentalism index may have resulted from a tendency to associate all U.S. questions with the free trade issue. Future surveys will reveal the degree to which the salience of the free trade question in 1988 drove opinion on associations with the United States overall. The remaining substantive differences in opinion appear across a diverse set of issues. As seen in Table 4.1, a gender opinion difference is recorded on the question of mamtaining ties to Britain and the monarchy. A substantive difference of mean (9 percent of the index's range) is revealed between women and men on this issue. Women score higher on an index made up of questions regarding Canada's special relationship with Britain and the monarchy as an important link to our past, revealing that they place higher importance in such ties than do Canadian men. This is one of the few issues on which women reveal a greater conservatism in opinion than men. A statistically significant difference of means also appears in the energy investment index, although the size of the difference is small. Women tend to score lower on this index, a combination of questions regarding the need for strategic planning in order to ensure continued energy supplies in the future. Women's lower score means they are somewhat less likely to endorse intervention in securing a source of energy, and lean more towards the idea of cutting down on use than do men to ensure a continued energy supply. There is also a intergender difference in support for the public financing of religious schools: women are somewhat more supportive of the public disbursement of funds to support religious schools (difference of means is 4 percent of the range). This -120'-difference may tie in with their increased willingness to endorse government action in order to ensure equality, or with the greater importance of religion in their lives, and hence to the support for religious education. On the issue of capital punishment, one of the stand alone questions reported in Table 4.2, there is a small difference in the opinions of women and men: a third of all female respondents agreed that capital punishment is never justified while only twenty-seven percent of male respondents agreed, a difference of six percentage points. This is only a moderate difference, but it can be linked to women's more dovish stance on questions of defence and the use of force. Responses to the stand alone question on immigration also reveal a gender difference of opinion. Women are less likely, by a fifteen percentage point difference, to agree with the statement that Canada should encourage immigration from countries most like us (see Table 4.2). While close to half the male respondents agreed with this statement, a little less than a third of the female respondents did the same. This result corresponds with that found in the immigration index. Canadian women are less likely to believe that immigrants make Canada a better place and are less likely to agree that Canada should admit more immigrants. A significant gender difference of opinion is also found on the traditional roles index. Canadian women are more supportive than Canadian men of greater freedom of role choice for women and are less likely to believe that women's incomes are supplementary to those of men. These two measures are about gender-role equity: allowing women free choice between home and careers. The difference in opinions is to -121-be expected given that women are the direct beneficiaries of greater support for equality. Similarly, on the question of abortion there exists some evidence of a fundamental difference of opinion between women and men, although the direction of the difference is not that anticipated. The pro-life index reveals a substantive and statistically significant difference of opinion with men somewhat more likely to report pro-choice opinion on the issue. Women appear less likely to choose pro-choice responses than are men when queried on their beliefs regarding abortion, by a difference of 3 percent of the index's range. One would anticipate that women would be more likely to adopt a pro-choice stance for a number of reasons. First of all, religious men may be less likely than religious women to find the right to control one's body argument compelling. Men have never directly faced the challenge of deciding whether to terminate a pregnancy, and as such the pro-choice arguments might be argued to have less immediacy for them. It is also true that women who might not have faced the abortion decision themselves are nevertheless more likely than men to know someone close to them that has faced the decision, a friend or sister perhaps, and so pro-choice arguments might take on greater relevance for them in that light. Therefore one might expect that men would be less likely to endorse the pro-choice stance. Women are however more religiously committed than men, and thus for many women, opinion on abortion may be more difficult to render. For religious women, the competing values of morality and women's individual rights requires a difficult choice about value priority. As shown in Table 4.4, the relationship between the pro-life index -122-and both the religiosity and traditional roles indexes are weaker in the women's sample of respondents than in the men's sample of respondents. This should not be taken to mean that women are less "logical" in the connections they make between these issues. Rather, the evidence suggests that these issues are competing ones for women and as such are less easily resolved by women. Given the less straightforward association between the issues for women, finding weak connections between them is more likely, as shown in the 1988 data. Table 4.4: Association Between Abortion and Other Indexes by Gender Issue Index Pro-life index Men Women Religiosity 0.4286** 0.3841** Traditional Roles 0.2395** 0.1841** Sample size N 103? N=971 Note: Entries are Pearson correlation coefficients. ** p<.01; gender differences in the correlation coefficients are riot statistically significant. The last of the issues that can be labelled a women's issue in the sense that it is of special significance to women is the topic of child care. But on this issue there is little evidence of a gap in opinion between women and men. Women are somewhat less likely to believe that parents wishing for child care should pay for it themselves, but this difference is not substantive (Section C of Table 4.2). Terry (1984) provides evidence -123-that opinion on child care funding differs by gender but that opinion is dependent among women on their employment status: working women, more so than working men, are likely to endorse government support of child care facilities. The impact of third factors on differences of opinion will be addressed in the following chapter. In several other issue areas women and men reveal similar opinions. On the question of official bilingualism and Canadian duality, there is general agreement: Canadians are relatively supportive of extending and continuing French language services within the country. However, support for the Meech Lake Accord was not similarly divided (Section E of Table 4.2). While 44 percent of men supported the Accord, only 30 percent of women were willing to endorse it. This difference may have to do with the large numbers of women adopting the neutral categoiy on this question: 46 percent of women answered neither or undecided to this survey question, compared with only 27 percent of men (numbers not reported here). It may also have to do with the National Action Committee's vocal opposition to the Accord. It is true however that opinion on the Meech Lake accord was more evenly divided across women than it was among men. A further finding is that women do not lend greater support to unions than do men, contrary to the finding reported by Kopinak based on the 1979 Quality of Life Survey (1987). According to our anti-union index, women do not favour a redistribution of power in the workplace any more than do men. Moreover, both are somewhat less than supportive of union power in that both means fall above the midpoint for this index. While Kopinak suggested the finding was related to women's greater support for a redistribution of valued goods and services by government, such a conclusion is not -124-supported by our results. The two issues do not appear to be connected. On the issue of native peoples in Canada, the results are mixed (Section E of Table 4.2). A substantial majority (sixty percent) of both Canadian women and men endorse the principle of native self-government 'as long as their system of government conforms with the principles of Canadian democracy'. On the other hand, women are less likely to agree with the statement 'in the long run, it would be best for native people to be completely assimilated into Canadian society' than are men. While half of the male respondents agreed with this statement, only forty-four percent of the women did the same. A Gender Gap in Opinion? Based on the differences exhibited in opinion, it seems clear that a gender gap in opinions existed in 1988. This gap in opinion did not extend across all issues; neither were the differences so large as to constitute connasting approaches to political issues and social relations. This should not be interpreted as an indication the differences were neither relevant nor substantial. The three issue areas which previously revealed differences in opinion do so here as well. The defence gender gap, perhaps the most longstanding, is evident in this survey and reinforces women's more dovish stance on issues of war and peace. The existence of a compassion gender gap is also supported by the evidence presented here. The capitalism and state involvement indexes, and their related stand alone questions, show significant differences by gender. In every case women are likely to endorse a more -125-liberal stance than men. They are more egalitarian, at least on economic issues, and less committed to the free enterprise system and its workings. Significant gaps were also recorded in several of the issue areas that were investigated here for the first time. The largest difference of all those recorded is found in the continentalism index which included questions on free trade. Women were less likely to support free trade and other associations with the U.S. The connections between free trade and the capitalism index are obvious, but the differences extended beyond mere free market questions to more fundamental ones in feelings towards the United States This can be contrasted to women's greater support for continuing ties with Britain and the monarchy. Note however that ties to Britain in no way constitute a threat to individual economic security, unlike ties with America. This more liberal outlook is not true of all the recorded gender opinion differences: women scored significantly higher than men on the religiosity index and were more likely to endorse restricting the sale of pornographic material, results that reinforce women's previously recorded greater conservatism on these issues. Men were also more pro-choice than women. The direction of difference, whether of greater conservatism or liberalism, depends on the nature of the issue investigated. On some issues in which gender gaps were previously recorded, there was little evidence that gaps existed in 1988. Women and men think similarly about child care and no fundamental differences appeared in the civil liberties and environment questions, although women are somewhat more supportive of reducing demand for energy supplies rather than investigating new sources. Finally, women and men are equally cynical about -126-Canadian politicians and politics, and equally supportive of attempts at reform. What are we to make of these findings? Most obvious is the conclusion that women and men have differing public opinions. Many of these recorded differences are similar to those found in previous studies and so suggest that the gap is not a fleeting phenomenon. This provides some basis for concluding that the factors that shape public opinion may vary by gender. Perhaps less obvious, but equally important, is the nature of these gaps in opinions. The gaps that are recorded appeal" in areas that are fundamental elements of political ideology: economic individualism, the free enterprise system, equality of opportunity, defence policy, and associations with other nations. If we concede that these are fundamental political issues, then one can tentatively state that the political ideologies of women and men differ. This appears to be a liberal political gap. And on the two fundamental values proposed to be of particular importance to the detennination of women's opinions gaps are also apparent. On both religiosity and feminism, women and men reveal a difference in thinking. This may account for the lack of gender gaps on more proximate issues of particular concern to women. -127-Chapter 5: Explaining Gender Gaps in Opinion Gender gaps exist in a number of the values and opinions tapped by the indexes created from responses in the 1988 CNES. Many of the gaps appear on questions of a ,. fundamental nature and as such provide evidence of the existence of different political cultures. This chapter will attempt to uncover the sources of these gaps. A number of competing theories regarding the source of the gender gap in opinions were provided in Chapter 2: self-interest; gender role socialization; feminism; and social group memberships. This chapter will empirically evaluate the ability of each of these competing theories to account for the gap in political opinions. Before proceeding with the investigation into the source of the gender gap, however, the importance of core values, namely religiosity and feminism, must be determined. They have been introduced as both important elements of opinion and as important components of the gender gap puzzle. Thus the chapter begins with an examination of the social structure of support for both of these values, and continues with an analysis of their role in the gender gap puzzle. The final part of the chapter will proceed with the comparative assessment of the theories put forward for explaining 1 gender gaps in opinion. Value Pluralism and the Gender Gap Puzzle It has been suggested that the failure to find gender gaps on some issues may be • .-128-due to a competition of values: religiosity against feminist belief. If religious values and feminist values are competing for dominance in the shaping of opinions, then gender gaps on some issues may not appear at the zero order. Only through controlling for the effects of these values might any actual gender gaps appear. But in order to fully understand how each of these core values is shaped by gender and other social factors, they themselves must be investigated. As a first step then, the social stjucture of support for these two values is evaluated by regressing each on the social variables introduced as potential determinants of the gap in attitudes. The analysis is first undertaken for the whole sample, and then individually by gender in order to determine whether the social structure of these two core values differs between women and men. The Creation of Measures Some explanation of the variables to be employed in the analysis in this chapter and the next is in order. In every case where gender appears as a variable it is coded so that men=l and women=0. Three independent variables are included as measures of economic self-interest: the nine-point family income scale recoded into a 0 to 1 range; an employment status dummy variable coded 1 for "working now" and G otherwise; and a dummy variable for public sector employment coded 1 for employed by a government organization or a government-owned enterprise and 0 otherwise. The impact of gender role socialization is tapped by three independent variables: a dummy variable for homemaker status coded 1 for homemaker and 0 otherwise; a dummy variable for marital -129-status coded 1 for married/partnered and 0 otherwise1; and the seven-point children in the home variable recoded to a 0 to 1 range. The remaining social group memberships are included as a set of dummy variables in order to estimate then- role as sources of the gender gap in opinions: an education dummy variable with some college or university attendance coded 1 and 0 otherwise; an age in years variable recoded to fall within a G to 1 range; a union dummy variable coded 1 for respondents who are members of a union or who live with someone belonging to a labour union and 0 otherwise; and a set of dummy variables for membership in various organizations (political parties, cultural organizations, service/trade organizations and issue groups) with 1 coded for membership in the group and 0 otherwise. The examination of the social structure of opinion oh religiosity and feminism will also include a series of dummy variables for region of residence (Atlantic, Quebec, Prairies and British Columbia) allowing for a comparison of each with the excluded region, Ontario, and a Francophone dummy variable coded 1 for respondents for whom French is their mother tongue and 0 otherwise. A -The Social Structure of Religious Belief The religiosity index includes two variables that touch upon religious belief: acceptance that the Bible is the word of God and the importance of religion in one's life.2 This index ^though the coding is such that married and partnered couples cannot be distinguished this category will be discussed as though it includes only married couples. In any case, married couples are very likely to make up the largest share in this marital status category. 2 The exact wording of these variables is: 'The Bible is the actual word of God and is to be taken literally word for word"(Mainly Agree, Mainly Disagree, No Opinion) and -130-was regressed on a set of social measures in order to determine the relative gap that existed on this core value. Table 5.1 presents the results of a structural analysis of religious belief. The results from the full sample reveal that a significant gender gap remains on this core value despite controlling for a number of social factors (the gap value on this index drops from -.07 to -.05). That the gender difference on this core value is not fully explained by the social factors examined here suggests that women and men come at this subject at some more fundamental level. Either biological, early life socialization or cultural differences may account for the variance in religious belief between women and men; the evidence presented here provides a few clues as to which is the more likely explanation. Among the social determinants examined, both income and education appeal" to play some part in shaping religious belief: increased income and higher education are both associated with weaker religiosity. The causal links are not immediately discernible, although the increased tolerance associated with advanced education may -play a part. A strong positive association is also found to exist between religiosity and age suggesting that there are life cycle or generational effects at play. Canadians may become more religious as they age, or younger Canadians may be less religious than their parents by virtue of cultural shifts in beliefs. There is also some evidence that the presence of children in the home and marital status are relevant to religious belief: married respondents and those with children in the home are stronger in their religious "Religion is an important part of my life" (Mainly Agree, Mainly Disagree, No Opinion). -131-Table 5.1: The Social Structure of Religious Belief: Regression Results Independent Variables Full Sample Sex -.05** Family Income -.16** | 7* * -.12* Working -.03 -.05 -.01 Public Sector Employment .01 -.01 .02 Homemaker .01 .02 Married .07** .10** .02 Children at home 1 j * * .08 .15** Post-secondary Education _ | | ** _ | ] ** _ j j * * Age .26** francophone -.05* -.07* .04 I nion .03 .03 .0-1 Member - Political Party .03 .06 .01 Member - Cultural Org. .07** .10** .04 Member - Service. Business Club -.01 -.01 -.01 Member - "Issue" Group -.06** -. 10** -.02 Atlantic 12** 1 i ** II** Quebec -.04 -.01 -.07 Prairies. .04 .06* .01 British Columbia _ I | * * _ j i ) * * Constant .64** .62** .62** R-: .16 .17 .15 (N) 1867 994 873 Note: * p<.05; **p<.01;cf=men,?=women. Entries are unstandardized regression coefficients. See text for operationalization of the variables. -132-convictions. The most likely explanation for this finding is that the more religious are more likely to marry and raise children than are non-religious individuals. It seems less likely that marriage and the raising of children would lead to greater religiosity. Regional variation also exists in religious belief: the two outermost provinces provide opposite poles of belief. The Atlantic provinces show the highest levels of religious conviction while British Columbia provides the secular anchor. Finally, religious belief also varies with memberships in social organizations. While members of cultural organizations are somewhat more religious, members of issue groups are less religious than non-members. Regressing religiosity on these variables separately by gender reveals some interesting differences in the social patterning of opinion. Women and men exhibit many common patterns: age, region, family income and post-secondary education each divide women and men along the same lines. Thus part of women's greater religiosity may be due to their lower average incomes, to fewer women attending post-secondaiy institutions and to women's longer life span. Four factors enter significantly only for men: marital status, language, and membership in both cultural and issue groups. Married or partnered men are closer in opinion to women overall, that is they are more religious than non-married men. A number of possibilities exist for explaining this result. A first is that once married, women are able to exert a religious influence on their husbands which is evident in their responses to questions on this issue. An alternative explanation is that women seek out those men that are more religious as partners. Thus the question is not one of changing -133-one's husband's views, but rather of selecting a husband whose religious views most closely resemble one's own. On the other hand, it may be that men who are more religious are more inclined to marry and thus it is not the act of marrying itself that accounts for their differing views but their selection into the category. Marital status is not a predictor of women's religiosity but the number of children in the home is associated with their religious attitudes.3 Among women, increasing numbers of children in the home is linked to increased religiosity. Although it may be that increased numbers of children in the home bring with them a tendency for women to become more religiously committed, the alternative seems more likely: women who are more religious (and perhaps less supportive of women having careers) have more children. B - The Social Structure of Feminist Belief Quantifying feminist consciousness with the variables available in the 1988 CNES presented a challenge. The survey does not include a subjective measure of feminist identification which has been argued to be a more valid measure of support for feminism.4 Respondents are not directly asked if they consider themselves to be 'feminists'. The only alternative is to assess feminist consciousness through an objective 3 The children in the home variable just failed to meet statistical significance at the .05 level in the men's sample. 4 Conover (1988a) discusses the problems associated with objective measures of feminist identity and employs a subjective, more direct measure of feminist identity in her analysis. The variables in the CNES do not provide this option. -134-measure. To properly quantify feminist consciousness, the objective indicator should tap into the presence of both identification with feminists as a group and a belief in the need for action to better the group's fortunes. Two variables presented acceptable criteria for characterizing respondents as a 'feminist': a feeling thermometer ranging from 0 to 100 measuring how "positively" the respondent feels towards feminists as a group5; and a second variable which asks respondents whether or not they believe that the government should do more for women.6 As noted in Chapter 3, the procedure employed in creating the feminism measure differed slightly from that employed for the other indexes. The two variables were not simply added together to form the anti-feminism index given that the range of one variable far exceeded that of the other (100 compared to 4). As a means of overcoming this difference in range, responses to the thermometer scale were weighted by 1/100 and responses to the second variable were weighted by 1/4 in order to avoid losing variance in these items. After summing the two items, the index was then rescaled to fall within the 0 to 1 range, a step taken with all the indexes in order to allow for comparability across the measures. It should be noted at this point that there is little in this measure that differentiates 5 While the thermometer variable is 'acceptable' it is not ideal. Positive feelings for feminist groups are likely to be closely related to identification with feminist groups, but the two are not one and the same. 6 The exact wording of these variables is: 'How do you feel about feminist groups?' with response recorded on a thermometer scale mruiing from 0 'very negative feeling' to 100 'very positive feeling'; and 'Do you think the government should do more, do less, or do about the same as now for women?.' -135-between feminist belief in women and in men. Many arguments have been made that feminism in men is limited to an ideological position rather than one of 'consciousness' (Fulenwider, 1980; Klein, 1984; Conover, 1988a). The lack of variables in the 1988 CNES restricts the ability to construct a measure which differentiates between male and female consciousness. The result is that the feminism measure employed in this investigation may be weak in its ability to differentiate between women's and men's feminism and in some instances may be acting as a partial proxy for the holding of a liberal ideology more generally. The inability to construct a useful measure of ideology from the variables in the study removes the possibility of including such a measure in the analysis that would serve to control for the influence of ideology on opinions. Table 5.2 provides the results of the structural analysis of feminist belief. As shown in the results from the regression on the full sample, the gender gap in feminist belief remains even after controlling for a number of social factors, although it is no longer statistically significant (the gap value on this index stays at .02). In fact, feminist belief is significantly structured along very few social dimensions. The exception is family income: support for feminism decreases with economic security. Given that women have less financial security than men, controlling for income should reduce the size of the gap. That it does not suggests that income is less influential in the determination of feminist belief for women than it is for men. A moderate relationship also exists between feminist belief and membership in an 'issue' organization. There is unfortunately no way of determining the type of issue group that is at play here. The only other association of significance is that feminist belief is stronger in Quebec and among -136-Table 5.2: The Social Structure of Support for Feminism: Regression Results Independent Variables Full Sample Sex .02 —— family Income .10** .13** .06* Workiim .02* .04* .00 Public Sector Employment .01 .02 -.00 1 lomemaker .04 .01 Married .00 -.01 .03 Children at home -.01 .05 -.08 Posl-secondarv Education -.01 .00 -.02 Age -.01 -.07 .04 liancophone -.04* -.02 -.05* Union -.02 -.01 -.02 Member - Political Pai1\ -.01 -.01 .00 Member - Cultural Org. -.03 -.03 -.02 Member - Service. Business Club -.01 -.01 -.01 Member - "Issue" Group -.04** -.04* -.05* Atlantic -.02 -.04 .00 Quebec -.05** -.08** -.03 Prairies .00 -.02 .03 British Columbia .01 .01 .01 Constant .38** 37** R" .07 .10 .08 (N) 1801 960 85.1 Note: * p<05; **p<01; • ?=-women; crimen. Entries are unstandardized regression coefficients. -137-Francophones. Not surprisingly these patterns are found to differ by gender. Although anti-feminism increases with economic security among women in the sample, the strength of this association is much weaker than that discovered among the men. This likely reflects differences in the importance of economic self-interest calculations on this issue for women and men: men are much more willing to endorse greater government support of women as a group when such support does not threaten their own economic self-interest. Men at higher family income levels are more likely to feel this threat than men at lower income levels. Although support for feminism is also likely to vary with income among women, the relationship is much weaker and employment status does not differentiate the strength of women's feminism. Aside from economic calculations, Francophone women and Quebec men are found to be more supportive of feminism reflecting the different status of feminist groups within that province. Their greater profile in the province reflects itself in higher levels of feminist support. Moreover members of issues groups are likely to more be feminist than non-members. This finding holds for both women and men. C - The Gender Gap Puzzle While feminism has been advanced as part of the reason behind women's greater liberalism on many issues, its influence is likely to be mediated on some issues by women's greater religious belief. An attempt to model the relationship between gender, core values and opinions appears in Figure 5.1. The positive impact of gender on both -138-religiosity and feminism reflects women's greater conservatism and greater liberalism on these two respective values (see Table 4.1). But each of these core values influences more proximate beliefs in opposite ways. Feminism, in its support of egalitarianism and denunciation of the use of force, is a source of liberal belief. Religiosity, in its support of social conservatism, exerts a conservative influence on opinions. The result of these two competing influences is a push towards a middle ground: exarnining means on opinions indexes makes it appear that women fall at the middle when they are actually in two opposing camps. This competition of values is only likely to occur on issues with some moral component however; only on such issues will religious beliefs come into play. Figure 1 Gender Opinion The empirical investigation of this model proceeds by first regressing opinion on sex alone, and then on sex and both core values. Any change in the gap on opinion, as indicated by a comparison of the two coefficients on sex, would indicate the combined influence of both core values on that opinion. The expectation is that on those issues without some moral component including feminism as a determinant of opinion will serve to reduce the gap in opinions given women's greater support of feminism. On the other -139-hand, for those issues with some moral component the expectation is that the competition of values will result in little or no change on the size of the gender gap. Table 5.3 provides the results of this analysis. For ease of presentation, dependent variables appear in the left hand column, and independent variables appear across the top row. The gender gap in support of an issue is that coefficient found in the first SEX column. The gender gap controlling for both religiosity and feminism is that coefficient found in the SEX* column. As in Table 4.1, positive gaps indicate that men are more conservative on an issue and negative gaps indicate that women are more conservative in opinion. Both religiosity and feminism have been scored so that higher values correspond to greater conservatism on the issue. Hence, a positive coefficient on religiosity indicates a positive relationship with the index: as religious belief becomes stronger, conservatism on that issue also increases. A positive coefficient on the anti-feminism index also indicates a positive relationship with the issue but the interpretation of the relationship differs: greater support of feminism is associated with weaker conservatism on the issue. The results in Table 5.3 provide support for the model outlined in Figure 5.1. Religiosity exerts a conservative force on a number of the indexes: separate school funding; traditional roles; abortion; civil liberties; and British Ties. Thus apart from its association with opinion on British Ties, which presumably reflects traditional values, religiosity enters significantly on opinions that have some moral dimension. Feminism also enters as an important predictor of opinion: it enters significantly into sixteen of the nineteen regressions. Its influence is felt across a broader range of issues than religiosity, -140-Table 5.3: Influence of Religious Belief and Femiriism on the Gender Gap: Regression Results Index Sex Sex' Rel. Fund. Femi-nism R : N Continentalism .15** 14** -.04* .23** .09 1758 British Ties -.09** -.08** .12** .09* .03 1927 Capitalism .06** .06** -.02 .24** .08 1783 Gov't Involvement .06** .05** -.03** .24** .08 1890 Defence .05** .05** .04* .21** .04 1920 Energy .04** .05** .08** .03 .01 1897 School Funding .04** .02 -.23** .10** .07 1937 Women's Issues .03** .04** .15** ->3** .07 1930 Immigration -.03 -.03 .16** .13 1926 Abortion -.03* -.00 .32** .15** .17 1917 Extra Billinii .02 .02* -.04* II** .01 1932 Reform .02 .02 00** .19** .03 1933 Cvnicism -.02 -.02 -.02 .02 .00 1931 Social Spending .02 .02 .06** .27** .09 1913 Environmental Priority -.01 -.01 .08** -.02 .01 1921 Dualism .00 .00 .01 .33** .07 1903 Civil Liberties .00 .02* .15** .13** .08 1896 Taxation .00 .00 .02 .06* .00 1939 Union Support -.00 .00 .04* .31** .06 1926 Note: ** p<01; * p<05. Entries are unstandardized regression coefficients. -141-and it is strongest on economic issues and on issues of particular concern to women. Its influence is also in the anticipated direction, weaker support of feminism is associated with more conservative opinion on issues. Our interest extends beyond each variable's independent influence, however, into their influence on the gap in attitudes. A comparison of the two coefficients on sex reveals that the two values appear to be competing for influence on five issues. On two such issues, abortion and school funding, the gender gap is substantially reduced. This is to be anticipated on the separate school funding issue given the direct connection between religiosity and public funding of separate schools. The stronger one's religious beliefs, the more likely one is to support the public funding of separate schools. But the gender gap at the zero order appears to have been due to women's greater religiosity. Although feminism exerts some liberal pressure on the issue, religiosity dominates in its influence on support for this issue. Women are more likely to endorse the public funding of separate schools because they are more religious than men. The disappearance of the gender gap on abortion opinion suggests that part of women's greater conservatism on the issue stems from their greater religiosity. The degree of competition between core values on this one issue sets it apart from all others and underscores the unparalleled struggle of values that women encounter on this issue. Controlling for religiosity and feminism reduces the zero order conservative gap on abortion to naught. Women become as liberal on the issue as men. But religiosity seems to exert stronger pressure on this issue than does support of feminism: the coefficient on religiosity is almost twice as large as that on feminism. On this one issue, the -142-competition of values seems clear. One issue area rivals abortion in the competition between the two core values: traditional roles. Unlike the case of abortion, however, feminism exerts the greater influence on opinion on gender equity. Moreover, contiolling for both serves to increase the liberal gap that appears on the issue. In spite of women's greater feminism, controlling for women's greater religiosity serves to increase the size of the gap found on the issue. A similar result is found in the case of opinion on civil liberties. On this issue, the removal of religiosity and feminism as competing values 'unlocks' the gender gap. Women's greater support of feminism may well reduce the gap on this issue, given the egalitarianism that accompanies feminist belief, but controlling for the influence of religious values removes the conservative pressure such a value imposes on civil liberties belief and on this issue produces a gap where none existed at the zero order. Because religiosity is associated with a belief in the need to protect the moral order, it is associated with greater conservatism on civil liberties issues. Thus its removal from the relationship between gender and opinion on the issue reveals women's greater support of egalitarianism on civil liberties issues even while controlling for their greater support of feminism. The one remaining issue in which a competition of values appears is unmigration. Both feminism and religiosity appear to exert similarly strong pressures on opinion on this issue. However controlling for both these core values does not change the size of the gap on the issue. It remains small in absolute size (relative to the other gaps) and -143-statistically insignificant. The picture that emerges from Table 5.3 is that on some issues, those with a particular moral component, religiosity and feminism compete for dominance. On some of these issues, religiosity holds greater sway; on others, feminism is more influential. That women are more religious than men, and also score higher on feminism, means that feminism must exert a stronger pressure on those issues with a moral component to overcome this conservative pressure. The model in figure 1 appears to have been supported. Explaining the Gender Gap in Opinions Religiosity and feminism provide some help in understanding the difference in the opinions women and men hold on some issues. But in order to fully account for gaps in opinions several other possible explanations must be investigated. Before begirming with that investigation, however, some discussion of the competing explanations for the gap is in order. Economic Self-interest Women's greater liberalism on a number of issues has been argued to stem from their weaker financial positions. Economic self-interest is often assumed to be a determinant of political opinions: rational individuals will form opinions based on their own economic interests. The concept is normally operationalized with income and employment variables, as objective indicators of an individual's economic interest. -144-Before employing them in a similar manner it is imperative that they be examined to assess more accurately their meaning and their impact on each gender. It should not be assumed that income and employment 'mean' the same things for men and women, indeed much has been written about gender differences across them (Armstrong and Armstrong, 1993; Duffy and Pupo, 1992; Statistics Canada, 1990). Women have not, for instance, historically constituted a share of the Canadian labour force equal to that of men. But over the last forty years, women, married women in particular, have increased their relative share of the paid labour force. In 1951, only 10 percent of married women worked outside the home but by 1991 this number had risen to 60 percent (Armstrong and Armstrong, 1993: 11)-. And by 1991 women accounted for 45 percent of the total Canadian labour force (Armstrong and Armstrong, 1993: 15). It is now the case that the majority of married, single, separated, and divorced women work for pay at some point during the year. A majority of widows, on the other hand, remain mil-time homemakers. The increase in labour force participation has not been accompanied by equality in income. Despite increases in participation, a wage gap remains between working Canadian women and men, and women are found in disproportionate numbers among the lowest paying jobs in the economy. In 1988 full-time male workers earned an average salary of $33,558 while similarly employed women earned only $21,918 or 65 percent of the male average salary (Duffy and Pupo, 1992: 36). Part of the reason for this difference is occupational ghettoisation: women are concentrated in a limited number of occupations, and within these occupations, they make up the greatest share of employees. -145-In 1991 more than four out of five women's jobs were concentrated in the service industries: trade; finance, insurance, and real estate; community, business, and personal services; and public achninistration and defence (Armstrong and Armstrong, 1993:24). Their lower average salary also owes to the fact that within these broad service categories women's jobs are disproportionately concentrated in the lowest paying occupations, jobs characterized by poor benefits and oppoitunities for job promotion and security (Armstrong and Armstrong, 1993:24). Employed women are quite likely to be found working as secretaries, salespersons, bookkeepers, tellers, cashiers, nurses, food and beverage workers, office clerks, primary school teachers, receptionists and child care workers, and making up the greatest share of employees in these occupations. They also earn less on average at these occupations than men in the same jobs (Armstrong and Armstrong, 1993:41). According to Armstrong and Armstrong, "there are not only men's jobs and women's jobs but also men's wages and women's wages" (1993:45). It is also true that women are more likely to be employed in government positions. Thus when comparing full-time employment for Canadian men and women, neither income security, benefits, job status, job promotion nor the job wage itself are equivalent. As a result of such employment and income patterns, women are more likely to ^ live below the poverty level than are men, a fact commonly referred to as the 'feminization of poverty' (Dooley, 1994). One of the main contributors to this phenomenon is the increasing divorce rate, and the concomitant increase in single-parent (often female-headed) families. Another distinction in the work patterns of women and men is the greater number -146-of women working part-time. At present about one in seven Canadian workers is a part-timer, and of these workers, over seventy percent are women (Duffy and Pupo, 1992:41). Table 5.4 reveals that while men take on part-time work in the early stages of their careers in order to juggle work with their studies, women take on part-time work at every stage in their careers. The typical woman's career pattern fluctuates between part-time and full-time work, and not does exhibit a simple movement from part-time to full-time (Duffy and Pupo, 1992). Women's traditional work pattern is full-time employment until marriage and/or motherhood results in a departure from the workforce. The departure can be for Table 5.4: Break Down of Part-Time Labour force by Gender and Age, based on 1986 Census data Age Category Men Women 15 to 24 years 19.2% 23.0% 25 to 44 years 4.1% 3 1 3% 45+ years 5.1% 17.4% Total Share of Part-Time Workforce Source: Duffy and Pupo. 1992:42 28.4% 71 7% either months or years, and is often followed by re-entry into the work force on a part-time rather than a full-time basis due to the added workload brought on by the presence -147-of young children in the home. Once the children are grown, full-time work resumes and continues through to retirement. And in retirement, unlike the pattern in men, a share of women return to work part-time in order to supplement limited pensions or savings. While this is the traditional career pattern for many women, it should riot be assumed that part-time work is always the preferred type of employment for all women. Rather "many women's decisions about part-time work are not premised on notions of individual choice and action but rather reflect family responsibilities, the structure of the economy, the availability of child care and family services, and the nature of job opportunities'' (Duffy and Pupo, 1992:77). While some women, even those with young children at home, would prefer full-time work a number of factors limits their ability to do so, most notably the availability of child care, the availability of jobs outside of the service industry which is dominated by part-time work, the sharing of responsibilities at home by husbands and partners to lessen the 'double day' workload, and societal expectations regarding the responsibilities of mothers to young children. As much as times have changed, polls still show that a majority of Canadians believe mothers should stay home with young children if at all possible rather than look for work outside the home (Duffy and Pupo, 1992:34). Such expectations structure the decisions women make regarding work, as much as the other factors listed. Although men have slowly started sharing household and child care duties with their spouses and partners, the choice for them in no way encompasses the same structural' difficulties as it does for women. Thus what may at first glance appeal' to be a simple employment decision is more often a decision reached in light of limited alternatives. -148-The 1988 CNES limits the degree to which differences in employment and income can be examined as determinants of political values and opinions. In the first instance, the available measure of income is family income. Personal income might be a preferred measure of self-interest. Susan Carroll (1988) has argued that women are more likely to develop independent political opinions if they have both financial and psychological independence from men. The opinions of wives are likely to minor those of their husbands because their evaluations will be based on similar circumstances, those of the husband. The individuation of a woman's opinions away from those of her husband and the male norm depends in part on her ability to develop a sense of financial independence. A woman's sense of financial independence can be gained through financial security, and is most likely achieved with an independent and substantial means of income. On the other hand, her sense of psychological independence varies with marital status, and married or partnered women are considered to be the least psychologically independent from men. For Carroll, the minimum requirement for the expression of 'true' women's opinions is, however, personal income. While a high family income implies financial security, it does not equate to economic independence for women in those families. If Carroll is correct in her assessment of the importance of economic independence in allowing women to derive their own political opinions, then family income will help little in detennining the nature of that impact. For the purposes of this study, psychological independence (as measured by marital status) is the only element of Carroll's equation that can be included in the analysis. -149-The preceding discussion is not meant to suggest that the family income measure is without value. Despite its limitations, the measure provides a means of assessing the impact of self-interest on the formation of public opinions. The theoretical importance of self-interest rests on assessments of the relative impact of public policies or public actions on individual circumstances, i.e. what impact will this policy have on my economic fortunes? This cognitive assessment is likely to rest at least partially on the economic impact to the family, since the impact of many policies depend on the family income (taxes, tax credits). Personal income, while an important measure of economic independence, does not accurately reflect the overall economic position of individuals within families. For instance, in terms of self-interest, how comparable are the situations of a woman and partner each earning $25,000 for a reported family income of $50,000, and a homemaker, with no reportable income, married to a man earning $50,000? The psychological impact of the differences in personal incomes between the two women is likely to be substantial, but in terms of economic self-interest arguments, the two families are economically similar and so the family income measure and not the personal income measure is the preferred measure. Although the homemaker has no personal income, her economic situation is not comparable to a single person with no income, since she enjoys a certain amount of financial security (so long as she believes the marriage is secure). Another restriction encountered in the 1988 CNES is an inability to differentiate between part-time and full-time work status. The data set only records whether or not an individual is employed. The only possible comparison is thus between the employed and the non-employed, a comparison which restricts the ability to evaluate some of the -150-elements that differentiate women's work from men's. While working may provide some degree of independence for women, in keeping with Carroll's analysis, such independence is limited if the job is part-time and low-paying. A woman's sense of independence is likely to increase as the job approaches a full-time career. But the given measure does not allow for a more sophisticated distinction in employment status. Table 5.5: Familv Income and Employment Status by Gender Men \\ omen Family Income** (%) (%) Under $30,000 v-1.9 46 0 $30,000-$59.000 41.4 30.6 Over $60,000 23.6 14.3 Employment Status** Working Now 74.2 19.2 Laid OfT/Unemployed 5.6 Disabled/Student 5.7 Retired 13.S 16.4 Homemaker 0.7 21 4 Employed in Government Sector** 22.5 77.5 68 7 Note: Entries are percentage in each category. Columns may not add to 100 due to rounding. ** p<01. Bearing these discussions in mind, Table 5.5 outlines differences in the family income and employment status variables of women and men surveyed in the 1988 CNES. Substantially more women than men are in families with incomes under $30,000, and fewer women than men enjoy family incomes greater than $60,000. Such figures suggest that on the whole men are more likely to enjoy the security provided by a satisfactory level of income, and their political opinions are likely to reflect this difference in economic security. -151-In terms of employment status, differences are also apparent. While almost three quarters of men in the survey are currently employed, less than half of the women surveyed are drawing salaries. Part of this difference is due to the fact that over 20 percent of the women surveyed considered themselves to be 'homemakers'. This category of employment is of little relevance in the male sample as less than one percent of men selected this category for themselves. Also of note is the slightly higher number of women than men in the retired category. This is no doubt due in part to their longer life span. The figures also reveal that women were more likely to be employed in either a government organization or a government-owned organization in 1988. It has been suggested that gender gaps in opinion result from women's greater dependence on the welfare state and its programs, and from the greater number of women employed in sectors dependent upon continued welfare funding (Erie and Rein, 1988). If women are more supportive of government involvement in the economy, it may reflect their own self-interest; government involvement in the economy is more likely to guarantee security of employment for them. The employment status measure provides an important tool for evaluating the predictive strength of employment categories on opinions. Nevertheless, it provides information based solely on the respondent's employment status at one point in time, i.e. at the time of the interview. As previously mentioned, not all women who select the homemaker employment status category in the survey will remain at home throughout their careers, and not all of the working women in the survey have always been in the -152-workforce. This movement in and out of the workforce is not captured in the employment status variable but nevertheless must be kept in mind when assessing the results. Table 5.6: Gender Percentage Point Differences in Family Income Level Categories by Marital Status Famih Income Category Marital Status Married Divorced' Separated Widowed Never Married Ml Under $30,000 -5.4 -21.5 -.37 -12.6 -11.1 S30.000-S59.000 -.09 5.5 6.2 1.0 1.8 Over S60.000 6.4 16.0 -2.5 1 1.7 9.3 Note: Entries are the percentage of all men in category minus the percentage of women in same category. Negative entries indicate the share of women in the category is greater than the share of men. Tables 5.6 and 5.7 elaborate on differences in the income levels of women and men. While it has been shown that more women than men live with family incomes below $30,000, the differences in the shares of women and men in this income category vary across marital status and with the presence of children in the home. As shown in Table 5.6, divorced and separated women fare worst financially: 20 percentage points more divorced and separated women than men of similar marital status have family -153-Table 5.7: Gender Percentage Point Differences in Family Income by Presence of Children in the Home Family Income Children in the Home No Children in the 1 lome finder $30,000 -13.9 -9 $30.000-$59.000 0.4 3.6 Over $60,000 8.6 10.2 Note: For further explanation see Table 5.6. incomes under $30,000. A full 64 percent of divorced or separated women are in this income category while only 43 percent of divorced or separated men are in the same income category (figures not shown). The family income difference between women and men is also large among never married individuals: never married women fall into the under $30,000 income category by 13 percentage points more than do never married men. And as shown in Table 5.7, greater numbers of women with children •in the home (almost 14 percentage points) live in the lowest family income category than do men with children in the home. These differences in employment and income are likely to affect women and men and be translated into differing values and beliefs. Many women live in economically difficult situations, and this burden is often accompanied by sole responsibility for raising the children of failed marriages. The fact that more women than men live in poverty seems a logical reason for their greater support for welfare state policies and government involvement in the economy. As such, differences in political opinions may be the result of economic self-interest: women as a group are more economically vulnerable, and this translates into greater support for welfare state programs. But this vulnerability is more -154- . pronounced among certain groups of women: divorced and separated women, and women with children in the home. Gender Role Socialization As a potential source of the gender gap, the socialization of women into the homemaker role may account for beliefs that are more liberal than those of men. A gender role ideology that suggests the proper role for women is that of mother and homemaker and that the proper place for women is in the home may translate into a belief among such women of the need for caring and compassion, and for assuming a level of social responsibility not matched among men. It is suggested that this traditional gender role ideology, due to the socialization of women into the role, should be included as an important detenninant of women's political opinions since the adoption of that role plays a part in the determination of individual values, and hence, political opinions. In order to properly assess whether the adoption of this traditional gender role has any influence on political opinions, women who have adopted the homemaker status and may therefore endorse the traditional definition of the role must be compared to women who have not chosen this role. The previous discussion of employment status highlighted the number of women who designate their employment status as 'homemaker' (over twenty percent - see Table 5.5). This share is not matched among the male respondents, and there is no equivalent employment category for them. Thus work differs both as a socializing agent and as a financial status determinant by gender, as differences in their categories of employment -155-status makes clear, and thus care should be taken to ensure that there is a complete understanding of how this difference may translate into varying political values and opinions. The traditional definition of the homemaker is a married woman dependent on the family's principal wage earner (normally the husband) for economic support. But there are several competing arguments behind a woman's subjective selection into the homemaker category. Homemaker status can and often does reflect more than simply an occupational status category. The possible interpretations of the relevance of the homemaker category are several. The category may represent nothing more than a simple occupational status category: its selection may coincide with a woman's temporary departure from the workplace due to the presence of children in the home. As discussed above, a high proportion of women, more so than men, engage in considerable movement in and out of the labour force during their lives (Armstrong and Armstrong, 1994; Rinehart, 1992). As a determinant of opinions, such a temporary departure is not likely to have a large impact on women's attitudes. Instead these women's opinions are likely to reflect their earlier connections with the paid labour force and will not be greatly influenced by the short time spent as homemaker. Alternatively, for some women the selection of the homemaker label may be chosen as an alternative for 'laid off, or 'unemployed'. Although the survey allows for the selection of unemployed/laid off as an employment status category, some women may choose instead to characterize themselves as 'homemakers'. The societal status accorded -156-a homemaker would be higher than that accorded the unemployed and laid off categories, and thus the choice is understandable. The ability to select the homemaker category is an option open principally to women: Canadian society has not yet fully adopted the idea of male homemakers. In the case of unemployed/laid off women who select the homemaker category, their political opinions are likely to minor those of women also temporarily out of the workforce caring for children in the home. The assumption here is that such women would choose to work if given the oppoitunity, and their opinions would reflect this reality. Finally some women may select homemaker status precisely because they identify with the traditional societal definition of the role. As such the selection of the status may identify women who hold the set of values traditionally associated with a gender role ideology that suggests the proper role for women is that of mother and caregiver and that the proper place for women is in the home. This is the psychological motivation for the selection of the category of interest for this investigation, as it serves as a proxy for women who believe that women belong and should remain within the domestic sphere, and that as homemakers, their main concern should be the care of the home and their families. This traditional gender role ideology should be included as an important determinant of women's political opinions since its adoption may play a principal role in the determination of deeply held values, and hence, political opinions. The competing motives behind its selection are difficult to disentangle empirically as a result of the simple categorization of employment status employed in the 1988 CNES. There is insufficient information for deteirnining the motives behind the selection -157- .. ' of the homemaker category. Hence caution must be used in employing the category as a proxy for adult gender role socialization. Homemakers make up 21.7 percent of all women respondents in the 1988 CNES. Comparing homemakers to all other women on a number of socio-demographic variables reveals some significant differences between the two groups (see Table 5.8). Homemakers are older than other women, are less likely to have attended college or university, are more likely to often attend religious services, are more likely to have children living at home (and in greater numbers) and are almost all married or partnered. This offers some cause for confidence that the women tapped by the measure are likely to adopted the traditional definition of the role since most of the couples in this category are married rather than partnered.7 It also points to the fact that most homemakers are living with a man. The smaller proportion of homemakers in the 18 to 30 age category and the larger share of homemakers in the 51 to 60 age category are likely due to the impacts of the feminist movement and a changing society on younger women. Young women who came of age during the second wave of Canadian feminism are less likely to adopt the more traditional definition of "a woman's place". On the other hand, the high proportion of homemakers found in the 31 to 40 age category is likely due to the fact that many women 7 It should be noted that homemakers are not likely to appear in high numbers in any other marital status category than married/partnered given the financial restrictions that accompany the adoption of homemaker status. -158-Table 5 8: Selected sncio-demo&ranhic differences between Homemakers and other women All 1 lomemakers Other Women Age** 18-30 13.8 26.9 31-40 33.5 24.5 41-50 13.3 16.8 51-60 24.8 10.6 Over 60 14.7 21.3 Education** No college/l niversity 60.7 51.0 Some college/l 'niversity 30.3 49.0 Religious Attendance** Less than once per month 36.1 46.1 Few times per month 19.8 20.9 Once a week or more 43.1 32.4 Marital Status** Married.'Partnered 90.4 56.1 Divorced/separated 2.8 11.4 Widowed 10.8 Never Married 21.6 Children in the home** None 42.9 60.1 One 18.4 17.1 Two 22.1 15.8 Three or more 16.6 6.9 ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ (202) (731) Note: **p-,0l . in this age categoiy are the principal care givers to young children. The causal links are not immediately apparent in the differences in religious attendance and education and homemakers: lower educational attainment and more -159-regular religious attendance could be the result o/having adopted the traditional role ideology or they may be partly responsible for its adoption. In any event, homemakers appear to be me women who account in part for the group's greater religiosity. The adoption of a traditional gender role ideology is associated with two other characteristics: marital status and the presence of children in the home. Both are linked to the traditional gender role that defines women by their private sphere responsibility and should thus also tap into its likely adoption among women. Similar qualifications to those made in the use of the homemaker proxy are in order here. Not all women with children in the home will have adopted a 'privatized' outlook nor will all married women adhere to the traditionally defined role of the homemaker. The validity of the marriage proxy is questionable given the collapsing of the categories of 'married' and 'living with partner' in the 1988 CNES. Such muddling of proxies is urifortunate and likely attenuates our measure of the impact of gender role socialization on opinions. The higher numbers of children in the home may coincide with the values inherent in the traditional definition of the homemaker role: without career or full-time employment responsibilities, such women are more likely to have open to them the possibility of large families. Marriage is also likely to provide them with the financial security necessary to make this choice economically feasible. The task is to determine the degree to which gender roles --particularly that of homemakers — are linked to women's values and political opinions. As an initial step in this process, difference of means tests between homemakers and other women were onducted on the twenty issue indexes. Table 5.9 presents the results of these tests. -160-Table 5.9: Differences in Opinion Indexes Between Homemakers and Other Women Index Mean Difference of Means* All Other Women Homemakers Religiosity .57 .66 -.09** Anti-Feminism .38 .41 -.03 Defence .56 .56 .00 Capitalism .58 .58 .00 Anti-State Involvement .42 .42 .01 Anti-Social Spending .41 .42 .01 Anti-Civil Liberties .45 .48 -.02 Continentalism .44 .46 -.02 Anti-Union .58 .59 -.01 Cynicism .72 .73 -.02 Anti-Dualism .40 .41 -.01 F.xtra Billinii 21 .19 .02 F.ovv Taxation .80 .79 .01 Anti-Sep. School Funding .58 .54 .04 Ami-Immigration .63 .67 -.04 British Ties .71 .68 .03 Traditional Roles .24 .35 -.12** Pro-l.ifo .32 .14 -.12** Low Environmental Priorit\ .27 .27 .00 Enemy Investment .65 .61 .05 Risk Aversion .36 .37 .00 Note: * p<05; ** p<01. -161-The results suggest that the homemaker gender role is linked to some political opinions: statistically significant differences of means were recorded in three of twenty indexes. Although the differences recorded between homemakers and other women are not as significant as those recorded between women and men, women's gender roles merit attention in the structuring of opinion on particular issues. The most notable of these differences appear in the religiosity, traditional roles and abortion indexes which record differences of means greater than nine percentage points. Homemakers exhibit opinions that are significantly more conservative than other women on these three issues: less pro-choice, more religious, and less supportive of extending women effective choice in their lives. That homemakers lend less support to equality for women increases the confidence that can be placed in the adoption of homemaker status as a proxy for women who have taken on the traditional gender role ideology. That homemakers are more religious and weaker in the support of reproductive freedom lends preliminary support to the hypothesis that homemakers employ private sphere values, adopted as a result of their particular gender role, in the determination of their political opinions. Their opinions on these two issues could stem from a belief in the responsibility for acting as guardians of morality and protectors of children. Thus some gender gaps in opinion may be explained by the particular set of values held by homemakers. Feminist women and homemakers are not merely opposite ends of an ideological or gender role spectrum. Although the two variables are correlated, the strength of the correlation is rather weak (1-.O6) and statistically insignificant. Thus while there may be -162-some reason for assuming a priori that a feminist as defined here is not likely to be a homemaker, this is not borne out by the data. Moreover, although homemakers score three points higher on the feminism index (and thus are less feminist), this difference is not as large as that recorded on other issues. This reinforces the necessity of understanding that for some women homemaker status is simply a temporary departure from the workplace. It also means that homemaker status must be questioned as a valid proxy for gender role socialization. And as made clear by Rinehart (1992), gender consciousness need hot be liberal in orientation, but can be directed towards gaining benefits for women that are conservative in nature. As an example, the Canadian group R . E . A . L . women exemplifies group consciousness among women directed towards ensuring the importance of traditional women's roles in society. Given the preponderance of Canadian feminist groups with more liberal aims however, and supported by the results that reveal homemakers adopt more conservative stances on many issues, there is something to be found in the homemaker proxy. But it is incorrect to assume that homemakers cannot be gender conscious or feminist; the share of women falling into this category is low but it is not zero. It is important then to include both measures in the analysis, since the impact of one is not directly associated with the impact of the other. Other Demographic Sources of the Gender. Gap ,. The impact of social groups upon the political opinions of women and men have also been suggested as possible factors behind the existence of gender gaps. Women and men '• -163-are likely to vary in both group membership and intensity of involvement. Some memberships will be gained through a measure of choice (e.g. union membership) while inclusion in other groups will be based on observable characteristics (e.g. age cohort). The relevance of social group memberships to political opinions is twofold: the groups women and men join will affect the values they hold and the salience of various issues for them (and so groups become an important source of opinion difference as membership in them varies by gender); and as members of the same social group, women and men are likely to receive different values and messages from that group, linked specifically to their gender. For instance, post-secondary education is linked to the adoption of more liberal political values and political opinions, but among women and men that have attended post-secondary institutions the impact of that attendance need not be similar. Thus even as members of the same social groups, the impact of that membership on political thinking is likely to vary. This possibility is not often entertained in investigations of gender and public opinion. It is nevertheless relevant to assess the potential for differential impacts by social groups. It would be misleading to assume that only women's private roles or their explicit rejection of that role alone account for their political opinions. Like men, women are social beings beyond their roles as homemakers and mothers. Women join social organizations, attend universities, and are members of a particular social class. Such pressures are not likely to be mmimized as a result of their private sphere roles. As a preliminary step in the investigation of social group memberships as sources of the gap in opinions, gender differences in membership rates in those social groups -164-Table 5.10: Shares in Various Socio-demographic Groups by Gender Men Women Marital Status* Married/With Partner 70.7 63.7 Dhorced/Separated 8.1 10.6 Widowed 2.2 8.8 Never Married 19.0 16.9 Children in the Home None 55.2 56.3 18.1 17.4 : i | ! i | |§ | l | | : | 17.8 17.2 Three or more 8.9 9.0 Age Cohort* 18-30 23.8 24.1 32.0 26.1 41-50 16.1 16.0 51-60 12.6 13.7 Over 60 15.5 19.8 I 'nion Member 34.4 36.0 Non-l nion Member 65.6 64.0 No College I iiivcrsity 53.4 55.0 College I iiivcrsity 46.6 45.0 Parly Member 9.5 8.8 Non Parly Member 90.5 91.2 Member Ethnic Ciroup 10.2 8.7 Non-Member 89.8 91.3 Member Service Business Club* 28.5 22.2 Non-Member 71.5 77.8 Member Issue Group 13.3 11.5 Non-Member 86.7 88.5 Note: Entries are percentage in each categoiy. Figures may not add to 100 due to rounding. * p .05. -165-which could have a liberalizing effect on values and opinions were computed. Table 5.10 presents the results of this step in the analysis. Gender differences on this front are not ovemhelming, but do highlight some of the distinctive choices made by women and men as well as differences in their group memberships more generally. Marital status is found to vary by gender: women are more likely than men to be widowed (by 6.6 percentage points) and men are more likely to be married (by 7 percentage points). This is undoubtedly linked to women's longer life span. Small differences also appear in the age distribution of the respondents: more men are in the 31 to 40 year categoiy, while more women are in the Over 60 age group. Women are also somewhat less likely to have attended a college or university and are more likely to have been a member of a union (or to have someone in the household that was a union member). Greater differences are found in the rates at which women and men belong to various social organizations. Women are less likely to be members of each of the groups measured by these items. They are less likely to be a member of a political party, an ethnic organization, a service or business organization and an issue group. The small differences in the proportion with which women and men make up of the categories of these social groups dismisses the possibility that nominal membership differences alone account for the gender gap in opinions. The only significant gender differences in these group memberships are found in marital status, age group and service club membership. Thus if these social group memberships are relevant to the gender gap in opinions it is more likely to be due to relative differences in the impact of these -166-memberships on political values and opinions for each gender, and each gender's reaction to their memberships in these social aggregates. Comparing the Theoretical Models A number of factors were introduced as potential determinants of the gender gap in opinion: self-interest; gender role socialization; feminism; and social group memberships. In order to test the competing theories, a series of regression analyses were run on those indexes in which gaps in opinion were recorded after controlling for both religiosity and feminism. A first regression was run including only a dummy variable for sex coded 1 for men and 0 for women in order to measure the size of the gap in opinion without controlling for any factors. A second regression was run controlling for the complete set of variables discussed as potential determinants of the gender gap. Any changes to the gap as a result of these controls can be detemiined by comparing the first and second regressions' coefficients on the sex dummy variable. The results of the analysis are found in Tables 5.11 through 5.13. Defence Gap As shown in Table 5.11, the gender gap in opinion on the defence issue is not changed by the introduction of the various control variables. This suggests that the models play little part in accounting for women's consistently weaker hawkishness. Nevertheless, a number of the factors are related to opinion on this issue. Self-interest, as measured by family income, is strongly related to the defence index: support for defence measures -167-Table 5.11: Explaining Gender Gaps in Opinion: Regression Results Independent Variables Defence Capitalism Anti-State Involvement Sex . 0 5 * * .06** . 0 6 * * Sex* . 0 4 * * 0 4 * * . 0 3 * * Religiosity . 0 4 * -.01 - . 0 3 * Anti-feminism 1 9 * * . 2 0 * * . 2 2 * * family Income . 0 5 * . 2 0 * * 0 9 * * Work inn .01 .00 .01 Public Sector Employment .00 -.01 - . 0 4 * * Homemaker - 0 1 -.01 - .04 Married -.01 - . 0 0 . 0 2 * Children al home - . 02 - .03 . 0 7 * Post - secondaiA Education -.01 .01 .01 Age .0.1 . 1 6 * * 1 3 * * L'nion - .02 - . 0 6 * * -.01 Member - Political Partv - . 02 .01 - .02 Member - Cultural Orn. - . 0 5 * -.01 - .02 Member - Service. Bus. Club . 0 5 * * . 0 3 * * .01 Member - "Issue" Group - .03 -.01 -.01 Constant . 4 5 * * . 3 9 * * . 2 6 * * R 2 .06 .20 .13 ( N ) 1755 1635 1724 Note: * p<05; **p<01. Entries are unstandardizedregression coefficients. -168-rises with family income. It is likely that low income families characterize the tradeoff . between defence spending and social spending as a zero-sum game: money spent on defence is money lost to low income families. But since controlling for women's greater financial insecurity does not serve to reduce the gap suggests that income is not as strongly associated with defence opinion for women as it is for men. Some evidence is also found that women's lesser involvement in social organizations, namely service and business groups, may also be relevant. These groups' members are found to be more supportive of defence measures; and women are, of course, significantly less likely to join these groups. Members of cultural organizations, on the other hand, are likely to be less supportive of such policies by an equal measure. Women, however, are as likely to join these groups as men. And both women and men are much more likely to be a member of a service organization than they are a member of an ethnic or cultural organization. Feminism exerts an impressive liberalizing impact on defence opinion. Feminism, possibly in its heightened support of democratic and liberal principles which include a rejection of power and force as a means of resolving disputes, decreases support for defence measures (Conover and Sapiro, 1993). However, controlling for women's stronger feminist belief does not significantly reduce the gap on this issue. Compassion Issues Gap Two indexes within the group of compassion indexes revealed gaps in opinion: the capitalism index and the state involvement index. Table 5.11 provides the regression -169-results for the two indexes. On the capitalism and state involvement indexes a significant reduction in the gender gap, by one-third and one-half respectively, results when controlling for the hypothesized deterrninants of the gender gap. Part of this reduction in the gap comes from differences in women's and men's economic self-interest calculations. Self-interest enters significantly into the regression equations, but mainly through family income. Income is associated with increasing conservatism in each of the two issues, although the relationship is much stronger for opinion on capitalism. On this index, a unit change in family income is associated with a twenty percent increase in support for capitalism. The only other self-interest variable to play a role in determining opinion on these issues is public sector employment: individuals dependent in some manner on government for their income are more supportive of government involvement in the economy overall than other individuals. This is a clear- and direct manifestation of economic self-interest at work in the determination of issues. Women's weaker economic status plays some part in reducing the gap on this issue. Feminism is also strongly associated with opinion on compassion issues and this is likely due to a strong adherence to liberal principles. In each of the two indexes feminists are more liberal in opinion than non-feminists and significantly so. Thus controlling for women's greater feminist belief may play some part in the recorded reduction in the gender gap. Age also reveals a strong and significant impact on each of the two compassion issues. In each case increased age is associated with increasing conservatism. There is little to suggest whether these are life-cycle or generation effects. As part of the life cycle explanation, increasing age is often associated with increasing -170-conservatism in opinion. It is not clear why this might be the case for economic issues: one might anticipate that a weaker financial status might bring with it a liberalization of opinion on compassion questions, particularly on the involvement of government in the economy and in redistiibutive arena. As part of generation effects, the coming of age of different generations of adults during particular social periods is likely to result in opinions that by virtue of generation alone are likely to deviate from those of individuals coining of age before or after them. In this case, the greater liberalization of younger respondents on the compassion issues may be due to generation effects: coming of age during after the 1960s these younger generations are likely to hold opinions that are on average and by virtue of their birth alone more liberal than their parents. In any event, women's longer life span should increase the size of the gap on this issue, unless older women are more likely to believe themselves economically vulnerable to cuts in government programs than men of the same age. Any remaining effects are weak and/or inconsistent across the two compassion . issue indexes. Of particular interest is the weak support for the impact of gender role socialization variables on the gender gap in compassion issues. Found only on the state involvement index, marital status and children in the home enter significantly into the equation, the latter recording a stronger association than the former. Homemakers, on the other hand, are no different in their opinions from all other respondents. The direction of the association between marriage and children and opinion on compassion issues is contrary to that predicted by an ethic of care: married respondents and those with children in the home are more conservative in their opinions on government involvement in the -171-economy. This provides some support for the conservatism associated with the traditional institution which may be relevant in both women's and men's opinions. Thus the higher proportion of married men may account in part for the liberal gap on this issue. Although the strengths of the associations are moderate, union membership and membership in a service association exert competing pressures on support for capitalism. Union members are less supportive of capitalism than non-union members while members of service and business/trade associations are more likely to support capitalism. The associations are to be expected given the raison d'etre of each group. Although no significant differences were recorded between women's and men's association with unions, men were found to be members of service clubs in significantly greater numbers than women. The existence of the gap may also stem in part from women's lesser involvement in such organizations. Continentalism Gap The continentalism gap was the largest gender gap found among the indexes created from the 1988 CNES. Table 5.12 provides the results of this analysis. The factors introduced as potential determinants of the difference in opinion do not go very far in explaining this difference of opinion: the gap drops by a mere 3 points to 12 percentage points. Evidence is found of competing sources for the drop in the gender gap on this issue. Unequalled in its strength as a determinant of opinion and of the gap on this issue is feminism: the difference in support for those strongly opposed to ferninism and those -172- . Table 5.12: Explaining Gender Gaps in Opinion: Regression Results Independent Variables Extra Billimi Continentalism British Tics. Sex .02 .15** -.09** Sex' .00 .12** -.08** Religiosity -.04* -.03 .12** Anti-Feminism .08** 19** .08 Family Income 14** .15** -.00 Working -.01 .02 -.02 Public Sector Employment -.03 -.04** -.01 1 lomemaker -.03 .00 -.08* Married -.01 .01 -.03 Children at home .02 -.02 .01 Post-secondary Education .01 .00 .00 Age 22** 1 j ** .13** I iiion -.01 -.04** -.03 Member - Political Partv .01 .02 -.03 Member - Cultural Org. .01 -.06* -.15** Member - Sen ice. Bus. Club .04** .05** .00 Member - "Issue" Group .03 -.01 .04 Constant .08* .30** .63** R: .07 .13 .06 (N) 1762 1612 1757 Note: * p<05; **p<01. Entries are unstandardized regression coefficients. -173-most supportive of feniinism is a full 23 points. Feminist belief exerts an unparalleled negative influence on support for continentalism. Thus if a particular "women's voice" exists, evidence from the continentalism gap suggests that feminist consciousness provides the key to unlocking this voice rather than the adoption of an adult gender role associated with a particular pattern of beliefs and opinions. Self-interest also plays a large part in the determination of the gender gap as low income levels are associated with weaker support for trade and economic ties with the United States. This is due in large part to economic self-interest calculations. Public sector employment also plays a part in deterrnining opinion on this issue: government employees are less supportive of continentalism and thus women's greater dependence on the government as employer is likely to contribute to the gender gap in opinion in this issue. And the existence of a union member in the family is also shown to account for some of the liberalization on this issue. This can be interpreted as self-interest at work in the determination of opinion: part of the rhetoric of the free trade discussion included the ability of unions to continue to play a strong part in the Canada economy in the face of American pressure to restrict a number of longstanding Canadian economic and welfare practices. In sum, women's weaker economic positions appear to account in part for their greater liberalism on this issue. Evidence is also found of generational differences in support of continentalism: older Canadians are more likely to agree with increased ties with the United States than are younger Canadians. This may stem from either generational shifts in support of such ties or from the fact that for younger Canadians financial security is more dependent on -174-the economy than older, retired, Canadians. In any event, the association between age and support of continentalism may explain the weak drop in the gap: more women are found in the highest age category than men and a greater share of women are retirees. These differences would offset women's greater liberalism on this issue. Service and business club members continue their pattern of greater conservatism by being slightly more pro-American. Women's lesser involvement in such groups may explain their greater liberalism on this issue. Members of cultural organizations are, on the other hand, less likely to endorse continentalism than other respondents. Women and men join such groups equally, however, and so this finding is likely to matter little for the gender gap in attitudes on continentalism. Extra-Billing Gap As shown in Table 5.12, the gender gap on the extra billing issue is reduced to naught when controlling for the set of variables included as potential determinants of the gap. This reduction most likely comes as a result of controlling for three factors: age, family income and anti-feminism. Generational differences in support of extra-billing are significant: the gap from young to old in support of the policy is 22 points. Younger Canadians it seems are more concerned with the costs of such a policy: whether this is due to individual self-interest as a result of families and children, or because of a different culture associated with younger Canadians is not clear. Controlling for age, because women live longer than men, should serve to increase the gap on this issue. That the gap is reduced suggests that older woman may feel more economic vulnerability than older -175-men: extra billing may present a greater threat to their financial status. If so, controlling for age may not serve to increase the gap on this issue. Family income also plays some part in reducing the gender gap on this issue. The economic security afforded by a higher income is associated with support of a policy allowing doctors to pass off some costs to patients. Thus women's lower average income may translate into weaker support on this issue. But feminism is also part of the explanation. Feminist women and men are less likely to support such a measure. A stronger belief in feminism is associated with greater liberalism on this issue. Part of the gap may then be explained by women's greater support of feminism overall. Religiosity is also associated with opinion on extra billing but the association is a negative one: greater religiosity corresponds with weaker support for extra billing. Although the strength of the association is small, women's greater religiosity can not be discounted for its part in generating the gap on this issue. British Ties The gender gap in support for continued British ties is somewhat reduced by the introduction of controls on the relationship between gender and opinion as shown in Table 5.12. Religiosity and age are both significantly and positively associated with support for maintaining such ties. Age differences are more likely due to generational than life cycle effects: earlier generations of Canadians are likely to have memories of the close ties between Canada and Britain during the war, and on that score would seem more likely than younger generations to argue for maintaining those ties. Differences in age -176-between women and men may then explain part of women's greater conservatism on this issue. The same is true of religiosity: women's stronger religious beliefs may explain part of their desire for mamtaining ties with Britain. Members of cultural and ethnic organizations are also less likely to endorse such ties and by 15 points on the index. This is a significant difference but its connection to the gap is limited at this point in that women and men are members of such groups in roughly the same numbers, mterestingly, homemakers are significantly less supportive of mamtaming British ties than all other respondents. It is not immediately clear why homemakers would hold distinct opinions on this issue. What is clear is that a conservative gap remains on this issue despite the fact that homemakers are more liberal than all other women on this issue. Traditional Roles As shown in Table 5.13, controlling for all factors on the traditional roles index serves to double the size of the gap rather than diminish it. Much of the increase in the gap likely comes from controlling for religiosity. Women are more religious and hence controlling for it increases the size of the gap recorded between women and men. Part of the increase in the gap may also come from controlling for age. The age opinion gap, a full 32 points on the women's index range between an eighteen year-old and a ninety year-old respondent, far surpasses any gender differences in opinion recorded on the indexes. On this issue age effects are likely to be generational; the impact of the women's movement is likely to be strongest on younger scores of Canadians who have come of age since the beginning of the second wave of the Canadian feminist movement. Similar results have -177- • Table 5.13: Explaining Gender Gaps in Opinion: Regression Results Independent Variables Traditional Roles Anti-Civil Liberties Energy Investment Sex .03** .00 .04** Sex* .06** .02* .04* Religiosity .08** .13** .06** Anti-Feminism .26** .12** .01 Family Income -.04 .04* .04 Workiim -.03* .01 .00 Public Sector Employment -.03* .02 .01 Homemaker .04 .01 -.04 Married -.02 .00 -.04 Children at home .07 .02 .05 Post-secondary Education -.03* .01 -.04* Age .32** .23** .15** Union .02 -.02 .02 Member - Political Party .01 .01 -.03 Member - Cultural Org. .01 .02 -.05* Member - Service. Ejus. Club -.02 -.01 .06** Member - "Issue" Group .01 -.01 -.05* Constant .04 .22** .57** R : .16 .13 .04 (N) 1759 1731 1731 Note: * p<.05; **p<01. Entries are unstandardized regression coefficients. -178-been found in previous studies (Abzug, 1984; Baxter and Lansing, 1983). And such generational changes in opinion on this issue are likely to be greater among women than men. Feminism is similarly associated with a more liberal position on the issue. Since feminism is in part grounded in issues of equity for women this association is also understandable. Despite controlling for women's greater support of feminism, the gap nevertheless is increased. Interestingly homemaker status is not significantly associated with the traditional roles index.8 This might reflect the weakness of the variable as a proxy for gender role socialization. Both working and public sector employment are associated with more liberal attitudes regarding women's income and the benefits of women having careers. The self-interest associated with this issue position is clear of working women; the connection of self-interest to the adoption of similar attitudes among men is less straightforward. Civil Liberties Similar to opinion on women's issues, controlling for various factors serves to significantly increase the size of the gender gap recorded on opinion regarding civil liberties. The gap on this issue increases to 2 points as shown in Table 5.13. The similarities with women's issues continue when considering the factors that likely account for this difference. Both age and religiosity are significantly and positively 8 Although a moderate coefficient exists between homemaker status and the tiaditional roles index, the coefficient fails to meet the 5 percent level of statistical significance. -179-associated with increasing conservatism on this issue. Controlling for women's greater : religious conservatism and their longer life span unlocks the gender gap: women it appears are more egalitarian on this issue. Feminism is also associated with opinion on civil liberties questions. Feminists are less likely to support restrictions on fundamental freedoms. And the gap is increased despite controlling for women's greater feminism. Energy Investment The final gender gap, on the energy investment index, remains intact despite the introduction of controls. Neither self-interest nor feminism are significantly associated with opinion on the issue. Instead social group memberships, particularly age, are strongly and positively associated with opinion on this issue. Generational differences in support of energy projects are also apparent: older Canadians are more likely to give their support to such projects. Thus controlling for age should serve to increase the gap on this issue, that it does not suggests that age's influence on this issue is gender dependent. Religiosity is also associated with conservative attitudes on this issue; controlling for women's greater religiosity should also translate into a large gender gap. Like age, then, its influence depends on gender. Organizational memberships are also found to differentiate opinion on this issue. Service club members are more conservative on this issue as they were on other issues. Part of women's greater liberalism may come from their weaker association with such groups. On the other hand, both cultural and issue group members are more liberal on the energy issue: they are more likely to support reducing energy consumption than -180-government efforts to increase supply. If the salience of such group memberships differs between women and men, this may account for part of the liberal gap on this issue. Conclusion In sum, mixed support is uncovered for the feminist, self-interest, gender role socialization and social group membership hypotheses. When simultaneously introduced as potential determinants of opinion on a number of issues, the hypotheses reduce gender gaps in opinion only slightly. ; Economic self-interest cannot be dismissed as a determinant of opinion and seems to play some part in producing measured gaps in opinion. It is a significant predictor of opinion in seven of the nine indexes in which gaps were recorded. And its impact is largely restricted to the family income measure. Increased family income brings with it an increased conservatism on political issue positions. Thus some of the difference in the opinions of women and men stems from the fact that women are less likely than men to be found at higher income levels. Feminism's role in the creation of gender gaps is similarly supported by the evidence presented, In seven of the nine examined indexes, feminism was found to have a significant liberal effect on attitudes. Women's more liberal opinions thus stem in part , from their stronger support of feminism. Despite the weaknesses in the measure employed in this investigation, the social structure of support for feminism reveals that women and men come at this belief in different ways.: Apart from some differentiation by income, there is little social differentiation in women's support of feminism. Accounting -181-for this value in women requires an investigation which is beyond the parameters of this study. For men, however, feminism is dependent on their financial status, their involvement in the workforce, and living in Quebec. It is not, then, as personal a belief for men as it is for women. But feminism's influence on the gender gap is often mitigated by the strength of women's religious belief. On the defence, civil liberties and traditional roles indexes, religious belief exerts a strong conservative force on opinions that mitigate feminism's ability to account for gaps. Women's greater religiosity accounts in part for the failure to significantly reduce the size of gender gaps in opinions. Social group memberships, particularly the respondent's age and membership in the some social organizations, contribute to opinion on a number of issues. On the extra billing, civil liberties, and traditional roles indexes in particular, the importance of age in differentiating opinion is unmatched by any of the remaining measures. In every case age is associated with increasing conservatism on issues. That women constitute a larger share of the oldest generations means that controlling for age should serve to increase the gap that appears on issues. And in the case of opinion on civil liberties and traditional roles issues this difference is most likely generational. The failure of the private sphere factors, marital status, children in the home and homemaker status, to significantly account for gender gaps suggest that the political relevance of women's private sphere roles has been exaggerated. The only instance in which the homemaker status significantly entered a regression was in the case of British Ties. And in this instance, homemakers were more liberal in attitudes than other women - 1 8 2 -thus increasing rather than decreasing the gap on the issue. And when children in the home and marital status appeared as relevant factors to opinion, on the state involvement index, their influence was conservative. At least at this stage in the investigation, their importance must not be exaggerated as factors relevant to the gender gap in opinion. A much larger gap in opinions would exist were it not for two groups of women: older women and women with children in the home. Both these groups of women were more religious than all other women which explains in part the failure to reduce the size of the recorded gaps. Mothering is of little relevance to women's political opinions except through its association with religious belief. It is not clear whether this association is likely to dimmish over time. If, on the other hand, the young women coming of age in a generation that is much less religious remain less religious as they age, then gender gaps on many issues may increase over time. The failure to explain gender gaps in opinion suggests that biological differences or early life socialization account for part of the differences recorded in women's and men's opinions. It also suggests that differences in political culture exist: there is a evident difference in women's and men's political thinking. These gaps are, admittedly, less than overwhehning in size. Many of the social factors investigated here reveal larger gaps in opinion than those recorded between women and men. But their small size is easily explained: the dominant political culture is very likely to bring women and men together in their thinking. The political world is very much a man's world. That small differences in women's and men's opinions exists nonetheless is impressive, and suggests a women's political culture might very well exist. -183-Chapter 6: The Social Structure of Opinions Compared This chapter will reproduce the analysis undertaken in Chapter 5 but do so separately for women and men. Moreover it will extend the analysis to the full set of indexes included in this investigation rather than limit it to those indexes on which gender gaps appeared. In other words, the purpose is to comparatively assess the sources of opinion on the full range of issues included in the indexes; the emphasis is on understanding the roots of women and men's political thinking, rather than the roots of the gender gap. While several factors were investigated in Chapter 5 to determine their contribution to gender gaps in opinion the analysis must be extended and modified. The manner in which the sources of the gender gap were investigated, by subrmtting men and women in the sample jointly to regression analyses, allowed for the comparative assessment of the relative predictive abilities of each of those factors and thus their likely contribution to those gaps. But submitting the full sample to such analyses prevents each factor's predictive strength from varying by gender. That is, the computed regression coefficients are averages of women's and men's independent regression coefficients. Thus if income is a strong predictor of opinion among men but not among women submitting them jointly to regression analysis is likely to lead to the conclusion that income is only a moderate predictor of opinion. Hence the need for separate analyses. In order to uncover the social structure of each gender's opinions, a series of -184-regressions were run with each of the opinion indexes serving as dependent variables. In addition to the core values, self-interest, gender role, feminist and social group membership variables included in the gender gap analysis, region and language are also included here in order to compare their influence on opinion by gender. Two sets of analyses were undertaken: one each for women and men. Like a number of the other variables included in the analysis, regionalism has a lengthy history in investigations of political behaviour in Canada (e.g. Elkins and Simeon, 1980). While it may not contribute to the gender gap in opinion, it nevertheless is likely to enter as an important source of opinion. It may have a differential impact on opinion , by gender. As an important element of political culture, regional variation in political opinions has been interpreted as evidence of variations in culture in Canada (Elkins and Simeon, 1980). The dominant political culture, that is the male political culture, varies by region of residence even while controlling for various other social determinants of opinion. However it is suggested here that this dominant culture only partly relates to women's political opinions. Women do not engage in traditional politics to the same degree as men and thus this cultural variation may not be found in their opinions. It is also possible that regional political questions, very much the subject of traditional politics, may simply have less relevance for women given their weaker financial positions. The politics of federalism may matter less, and thus their opinions may vary less across provinces than do men's. Moreover, regional variation in the strength of feminism may translate into gaps that are apparent in some provinces, but not in others. Thus region is included in this investigation of the social structure of opinion. -185-The possibility that social group memberships like region can, and often do, vary substantially by gender is often overlooked. The basic assumption employed in much analysis of political behaviour is that there exists "no reason to expect that the larger world view produced by higher education is very different for women or men, nor are there many gender differences in their occupational and social experiences" (Baxter and Lansing, 1983:44). This assumption, as will be shown, is often unsound. The investigation begins by splitting the sample by gender, and then regressing each of the factors on the indexes to allow for a comparison of their relative impacts. The result is thirty-eight separate regression equations Given this large number of regressions and the difficulty in attempting to uncover patterns across them, the number of variables included in the analysis was kept to a minimum. There are, for instance, no interaction variables for the social indicators and many of the independent variables are dummy variables which, by definition, suppress any potential differences in impact across the categories of other independent variables. While the model employed necessarily simplifies the social structure of opinion in Canada, the alternative is to unnecessarily complicate the investigation by including such a large number of variables that uncovering any patterns in the results is unlikely. Regional differences in opinion are tapped by a series of four dummy variables (Atlantic, Quebec, Prairies and British Columbia) that allow for a comparison of opinion in each region with that in the excluded province, Ontario. The remaining variables are operationalized in a manner similar to that employed in the investigation of gender gaps undertaken in Chapter 5. -186-Comparing the Social Determinants of Opinion; Results The regressions produced in the analysis are robust1 and provide evidence of a difference in women's and men's thinking on political questions. Although differences in the coefficients for women and men are small, this is not inconsistent with the differences exhibited in their mean levels of support for various issues (Green, 1992). Nevertheless many of the variables found to be significant determinants of opinion confirm conclusions reached on the basis of previous research; namely that region, family income, education, age, and language are all relevant social determinants of Canadian opinion. This conclusion holds true in many cases for both Canadian men and women. But these results do not preclude one from MgUighting that over several issues the group basis of attitudes is not equivalent for the two groups. Put differently, depending on the issues addressed, one is likely to discover differences in group support by gender. Moreover, the results ovemhehningly support the need for including both feminist belief and gender in discussions of the sources of belief for Canadians. We will now turn to an examination of these results in detail. Defence Table 6.1 presents the results of the regression procedure undertaken for both the female and male subsamples on the defence index. An examination of the structure of opinion 1 The relatively low R2 values obtained for a number of the regressions are not unusual in structural analyses of opinions. The regression attempts less to "explain" the psychological process of opinion formation than it does to provide a framework within which such opinions are formed. -187-Table 6.1: Determinants of opinion: Defence Independent Variables Women Men Religiosity .01 .06* Anti-Feminism .13** .18** Family Income 1 [ ** .00 -.02 .05* Public Sector Employment -.01 .01 Homemaker -.03 — Married -.01 -.01 Children at home .03 -.07 Post-secondai\ Education .00 -.03 Age .16** -.07 Francophone -.02 -.07** Union .01 -.01 Member - Political Party -.02 -.03 Member - Cultural Org. - 09** -.01 Member - Service. Bus. Club .06** .05* Member - Issue Group -.02 -.02 Atlantic .01 -.03 Quebec -.09** -.04 Prairies -.01 -.04 British Columbia -.06* -.11** Constant .47** .58** R' .10 .08 (N) 820 934 Note: * p<05; **p<01. Entries are unstandardized regression coefficients. -188-reveals key differences in the underlying determinants of opinion on this issue. Several gender differences among the social deterrninants of opinion included in this investigation appear which were not apparent in the gap analysis. In that analysis, family income, feminism and language appeared as significant predictors of opinion. Upon further inspection, however, only feminism acts as a common determinant of defence orientation. Feminist men and women are less supportive of defence spending than non-feminists. This suggests that thinking about women in group terms and supporting increased government support for their status directly corresponds to decreased support for government involvement in the area of defence. This may involve a tradeoff: finite government resources may limit the number of causes adopted, it may also stem from the traditional pacifist stance normally and historically adopted by women's groups, particularly in the first wave of Canadian feminism. The more likely explanation is that the feminism measure adopted in this analysis is simply tapping into a general liberal ideology, particularly among men. The dominance of age and family income as determinants of women's defence orientations is unmatched in the men's sample. That significant differences exist across levels of these two variables suggests that a gender gap on the issue may stem from younger generations of women, perhaps socialized in a manner unique from young men and from older generations of women. It may also partly stem from women whose financial status is such that they stand to lose directly from government expenditures on defence budgets. Even at the highest income levels, women are less supportive of defence spending than men. -189-For men only, on the other hand, employment status is engaged in defence sentiment although it did not appear as a significant predictor of opinion in the gap analysis undertaken in Chapter 5. Working men are more pro-defence than men who are not working. The relevance of employment status to defence opinion is perhaps linked to the fact that many of the men not gainfully employed are retirees. Among men, employment status may serve as a proxy for financial security: every other category of employment status is characterized by some measure of financial insecurity. The same is not true of women. Homemakers, while not earning income for their labour, are likely to enjoy some measure of financial security (at least in terms of family income) that mirrors that of working women. Religious belief also enters significantly into the men's regression on defence sentiment. Strong religious belief is associated with somewhat stronger support of defence expenditures. That no similar association is found among the women's sample suggests this may also play some part in producing the gender gap found on this issue. There are some similarities in the association of social interaction in various organizations with defence sentiment for men and women. Women and men who are members of service and business/trade clubs are more likely to support defence spending. That women account for a smaller proportion of the membership in such groups suggests that this accounts for some of the gap on this issue. Most interesting is the finding that, for women, membership in a cultural association is associated with a significantly lower level of support for defence measures, by almost 10 points. The absence of a shnilar trend for men indicates that this may be connected with the gap in attitudes as well. -190-Although regional variation in defence attitudes is generally limited for both women and men, some regional distinctions do appear in each group. Neither group displays a strong or consistent regional pattern of support. A typical man or woman in every province will be generally hawkish and supportive of defence programs and policies. However two exceptions should be noted: a typical woman in Quebec or British Columbia is likely to be less hawkish than women elsewhere in Canada; and men in British Columbia stand apart from all other men in their weaker support of defence spending. Francophone men show a similar pattern of support to women in Quebec. While the strength of feminism generally in the province of Quebec may account for the difference in defence opinion among women in that province (and perhaps among Francophone men), it is less clear why men in British Columbia would display a similar distinction in attitudes. The particular political culture in that province translates into weaker support of defence measures among both genders, although the relationship is somewhat weaker among women. Compassion Issues A second set of issues which often reveal gender differences in opinion are 'compassion issues'. Table 6.2 presents the regression results for the capitalism, state involvement and social spending indexes. A number of common determinants of opinion are found among the indexes. Among both women and men, family income, feminism and age each play a part in structuring attitudes on the profit system and the role of government in the economy as was the case in the gender gap analysis. In most cases, higher income enters -191-Table 6.2: Deterrnmants of Opinion: Compassion Issue Anti-State Anti-Independent Capitalism Involvement Social Spending Variables Women Men Women Men Women Men Religiosity -.03 .00 -.03 -.02 .03 .05** Anti-Feminism .22** ,21** 1 7 * * .25** .28** 23** Family Income .21** .20** .07* 09** .05 Working .00 .00 .01 .01 .01 .00 Public Sector .01 -.03 .05** -.03 -.03* .01 Homemaker .00 .-- -.04* -.01 — Married -.02 •02 .04** .01 .01 .01 Children at Home -.04 -.02 :o5 .10* .05 :00 Post-Secondary .02 .00 .01 .02 -.01 -.03* Age • •/•'.- 1 7 * * .13** .10** 1 7 * * .13** . .06 Francophone .03 -.02 -.01 -.05* .01 -.02 Union -.07** . 07** .00 -.01 .01 -.02 Political PartyMember -.02 .03 .02 -.06* -.01 .06** Cultural Org. Member -.05 .02 -.02 -.02 - -.03 -.01 Service ClubMember .03 .04* .01 ,01 .03* ,03 Issue Group Member .01 -.04 -.02 .00 -.03 -.04* Atlantic .02 - -.05* .03 .00 -.03 . -.08** Quebec -.01 .00 .02 .02 -.03 -.05* Prairies -.02 .00 .02 .02 -.04* -.02 British Columbia -.04 -.01 .01 .03 _ 07** -.08** Constant 39** .43** .28** 27** .22** .31** R2 .18 .20 .10 .15 .16 .16 (N) 757 878 817 917 816 . 930 Note:* p<05;** p<01. Entries are unstandardized regression coefficients. as a significant predictor of support for capitalism and a laissez-faire approach to government involvement in the economy. The only exception is the failure of family income to enter significantly into the men's regression on social spending. Age is also a significant predictor of opinion: increased age is associated with opinions that are increasingly pro-capitalist (i.e. more conservative) for both genders. Here also, however, men's opinions on social spending are not significantly associated with age. These results are to be expected: financial security underpins a greater acceptance of competition and the advantages of capitalism, while a tendency for individuals to become increasingly conservative with age means that older survey respondents support capitalism more than younger respondents. An equally plausible explanation for the relationship between age and opinion on capitalism is that younger generations of Canadians are generally less willing to endorse free-market principles than older ones as a result of generational shifts in culture. Finally, feminism in both cases moves opinion to the left towards more support for compassion issues: the strongest feminists are more supportive, sometimes by more than twenty-five points, of compassion issues than the most anti-feminist. The association between feminist consciousness and opinion on these issues is not surprising. In striving to ensure equality of opportunity in society, feminism brings with it a weaker endorsement of unfettered economic practices. Ferriinism, in its general willingness to use the state as a means of securing equality for women, can be expected to bring with it a willingness to endorse an expansionist state involved in programs with goals other than gender equity. -193-Considering each of the three regression pahs individually provides a better sense of where the genders differ in the structure of opinion on these issues. Opinion on capitalism, perhaps the more abstract of the three indexes, reveals little in the way of differentiation in its sources between women and men. Even union membership, which fails to enter significantly on the two other issues in this group, exerts a similar pressure on women's and men's opinions regarding capitalism. The few distinctions between the structure of opinion on capitalism between the genders include the weaker support for capitalism among men in the Atlantic provinces and the stronger support for capitalism among men who hold memberships in service and business organizations. Women's opinions on capitalism are also more likely to vary with age than men's. That men's involvement in service and business clubs is associated with greater support of capitalism supports the idea that women's lesser involvement in social organizations precludes their opinions from coming closer to those of men. This was the only social group for which significant differences in women's and men's rates of membership were found. The added difference in the influence of such memberships make this factor seem of particular importance to gaps in opinion. Greater variation appears in the roots of thinking on government involvement in the economy and social spending. With respect to social spending, family income and age are significant predictors of women's, but not men's, opinions. The generational transmission of cultural shifts, or life cycle effects, and financial self-interest are significantly associated with opinions regarding government's social spending. No similar patterns are found among the men's opinions. The offsetting influences exerted -194-by age and income among women may account for the small gap recorded on this issue (i.e. women are older but they also have lower average incomes). No similar patterns exist among men's opinions. For men, religious belief enters into the equation: support for social spending varies inversely with the strength of religious conviction. This factor is offset by the fact that post-secondary education is associated with greater liberalism on this issue for both women and men. However, the strength of the association is much stronger among women. In every case, the coefficient on the variables for women is larger in value than the coefficient produced for men indicating a stronger relationship between the variables among women. The common influence of age, family income and feminism for both women and men exists in the state involvement index. Important to note is the stronger influence of feminism on the opinions of men. Feminist men, all else being equal, are as supportive of government intervention in the economy as feminist women. Important as well is the weaker association between opinion on the issue and age for women. Support for government involvement in the economy is greater among older women than it is among older men, ceteris paribus. Among the remaining variables, marital status proves significant in differentiating women's opinions although its impact is small. Marriage appears to move women's opinions closer to those of men on this issue. No comparable effect appears among the men. Marriage may well serve to restrict women from speaking in their 'own voice'. Somewhat strikingly, the presence of children in the home shapes men's opinion on this issue but not women's, increasing numbers of children in the home increases the -195-conservatism that men reveal on the issue of government involvement in the economy. At the very least this finding underscores the often forgotten potential for men's private lives to be linked to their public attitudes. mterestingly, homemakers are more liberal on this issue, net of all other effects, than other women. Although self-selection remains a possible explanation, on this issue gender role socialization may be a factor in accounting for the gender gap. But the combined effects of marital status, public sector employment and family income in structuring women's opinion on this issue suggests that economic self-ihterest and financial security, dependent in large part for many women on the combined elements of both family income and marital status, play a fundamental role in determining opinion on government involvement in the economy, which in and of itself can have a direct impact on a woman's financial stability. A far greater degree of regional differentiation is found among men's opinions on the anti-social spending index. On this issue, men are found in two regional camps: Ontario and the Prairies versus the remaining provinces, with the former provinces supporting reductions in social spending to cut the deficit by almost 10 points on the index. The placement of the Atlantic provinces and Quebec in the camp opposing reduction suggests that this is at least partially a case of have against have not provinces; British Columbia, Manitoba and Saskatchewan are exceptions to this pattern. Regional variation among women's opinions on social spending is more limited. For women, the more relevant regional comparison is between British Columbia and, surprisingly, the Prairies with the remaining provinces. Women in these two regions are -196-distinct from other Canadian women in their strength of support for social spending. Particularly intriguing is the finding of greater support for such programs among women in the Prairies. But the greater support for social spending among women in British Columbia, similar to the difference found among men's opinions, suggests that on this issue the doniinant provincial culture accounts for the opinions of both genders. The remaining differences in regional variation in opinion on this issue do, however, provide some evidence for considering that the context within which questions of social spending policies get played out, and hence opinions formed, is gender dependent. While women's involvement in service groups serves to bring their opinions closer in line with those of men, men's involvement in political parties serves to increase the gap. In short, women's and men's commumty contacts do not reproduce similar attitudinal patterns. The claim is limited, however, in that regional differentiation on the capitalism and government involvement indexes is not as sharp as that found on the social spending index. It is also found that Francophone men are more supportive of government involvement in the economy than all other men with no similar distinction apparent among women. The fact that men's, but not women's, involvement in service clubs and in political parties is significantly associated with stronger support for capitalism and stronger support for government involvement in the economy, respectively, reinforces the relevance of social interaction. In some instances it serves to converge opinion on issues; on others it moves women and men further apart in their political thinking. -197-Civil Liberties Table 6.3 presents the social structure of opinion on the anti-civil liberties index. Gender differentiation in the structuring of opinions on this issue is not overwhelming. Among both women and men, support for protection of civil liberties diminishes with age, increases with feminist belief, and decreases with religious belief. Moreover, the strength of each of these associations varies little with gender. The most visible difference between women and men is found in the degree to which the structure of male support for the defence of civil liberties varies regionally. Ontario stands alone among the provinces in the men's sample for its lower level of support for the protection of individual liberties and freedoms. No such regional variation exists in women's opinions on this issue. The only regionally connected variable which enters significantly into the women's equation is language. Francophone women support individual rights more strongly than all other women. The only evidence that opinion on this issue is grounded in economic evaluations comes from the significant association of family income with civil liberties attitudes among men: this is more likely to stem from class differences in support of the issue, than it is from self-interest calculations. For women, however, support for the issue varies little with family income level. Continentalism Although some similarity is found in the source of opinion on this issue, several differences are also to be found in a gender comparison of the structure of opinion on -198-Table 6.3: Determinants of Opimom^Civil Liberties Independent Variables Anti-Civil Liberties Women Men Religiosity Anti-Feminism 11 II J4*=i 09*" family Income 02 06* Working -.00 01 Public Sector Employment 01 03 Homemaker! 01 Married -02 02 Children at home 06 00 Post-secondary Education 03 00 Age 21 22* Francophone -OS 00 Union -02 -02 Member - Political Party -02 04 Member - Cultural Org. 03 01 Member - Service. Bus. Club -.01 -00 Member - Issue Group 02 00 Atlantic -01 - 0(>** Quebec -03 -07: Prauies 00 - 04* British Columbia 02 - 06* Constant 26* 27 R 2 15 16 820 921 Note: * p<05; **p<.01. Entries are unstandardized regression coefficients. -199-continentalism. As revealed in Table 6.4, similaiities in the social structure of opinion on this issue exist in the association of both family income and feminism with opinion on continentalism. Feminism exerts liberal pressure on support for ties with the United States for both women and men. Financial security, on the other hand, leads to the adoption of a more conservative stance; that is, support for increased ties with the United States. Among the remaining sources of opinion, age exerts significant pressure on women's opinion: the large gap between young and older cohorts of women (.17 points) is not matched in the corresponding male analysis. The clear liriking of the issue to women as a group and their fortunes appears to have been most successful in mobilizing opinion against the Agreement among feminists, and was a message of particular relevance to younger women whose economic fortunes were more mtimately tied to the Agreement itself. Feminism has been argued to be of particular relevance to young women regardless of their willingness to adopt feminism explicitly. Due perhaps in pail to this differential impact, age does not appear as a significant predictor of opinion on this issue among men. Evidence that self-interest arguments drive women's opinion on continentalism is found in the strong relationship between opinions and family income which is positively related with opinion on continentalism. This relationship is much stronger than that found among the men's sample. Women may take stronger cues from assessments of economic self-interest on issues that are directly linked to their economic fortunes in the political arena. This is likely to at least partly explain the gap on this issue. The free -200-Table 6.4: Determinants of Opinion: Continentalism Independent Variables Women Men Religiosity .02 -.06* Anti-Feminism .23** .24** Family Income .22** .12** Working .02 .04 Public Sector Employment -.03 -.05* 1 lomemaker .01 — Married .01 .02 Children at home .02 -.10 Post-secoiidarv Education .00 -.01 Age ]7** .07 Francophone .00 .03 I 'nion -.04 -.06** Member - Political Part\ .05 -.00 Member - Cultural Org. -.10** -.02 Member - Service. Bus. Club .05 .04 Member - Issue Group -.01 -.01 Atlantic .02 .05 Quebec 14** .18** Prairies .04 .08** British Columbia .01 .07 Constant 17** 37** R; .12 .12 (N) 747 865 Note: * p< 05; **p<01. Entries are unstandardized regression coefficients. -201-trade component of the continentalism index dominated the 1988 election campaign, and the issue was directly linked to women's economic fortunes by women's groups in the country (Bashevkin, 1989a). Thus the negative association made between economic fortune and free trade for women in particular was both direct and highly visible and reinforced by the myriad of other arguments made against the Agreement (dominated to an extent by questions of cultural independence). But the limited degree to which self-interest alone is able to account for gender differences in opinion on this issue is made clear by the comparison of women's and men's opinion at different family income levels. Women at the highest family income level show a level of support for continentalism equal to that found among men at the lowest income level. That women are much more likely to be found among the lower family income levels generally underscores the importance of self-interest arguments for women on this issue (see Table 4.2). Economic self-interest stands alone as a factor of greater relevance for women than men. But women's economic self-interest calculations appear unaffected by public sector employment. In fact, only among men is a negative association found between public sector employment and support for continentalism. Although employment in the government sector may be linked to women's attitudes on free trade with the United States, it does not appear to have much impact on the broader question of general relations with their southern neighbour. And neither does the presence of a union member in the home appear to play much part in the determination of women's opinion; only among men is union status an important factor in opinion on continentalism. But some association is found between women's memberships in cultural -202-organizations and continentalist sentiment: women in these groups are likely to be much less supportive of close ties with their American neighbour. This distinction is not matched among men. The most visible distinction in the social structuring of men's opinion on continentalism is that it occurs on a decidedly regional basis. Men in Quebec form an unusual partnership with men in the Prairies in their support of continentalism. In contrast, the importance of region in the women's sample begins and ends with the Quebec/ROC distinction. Apart from Quebec, the provincial context within which questions of continentalism play themselves out has a greater impact on men than women. In Quebec, the dominance of the Free Trade issue in the media and elsewhere resulted in a envhonmental impact equal for women and men. In the remaining provinces, however, the issue played itself out in a provincial context that was of limited consequence for women. Only among the men does social interaction in the dominant political culture provide a means of reproducing opinion on this issue. For women outside Quebec, it appears the issue was an economic one, and little else. The only remaining distinction in men's opinions on this issue is the negative association of religious belief with support of continentalism. It is not immediately clear how these two factors are connected, but the absence of such an association among women makes the finding particularly intriguing. Traditional Roles and Abortion Opinion on women's traditional roles and abortion presents further validation of the -203-importance of assessing the social structure of opinion separately by gender. In this case it appears that gender plays an fundamental role in shaping the degree to which self-interest and group memberships impact on opinion. This does not mean, however, that the gender gap in opinion is large (recalling Table 4.1 it falls in the middle of the recorded gender gaps in opinion). The lesson from the evidence in Table 6.5 is that sharp differentiation in the roots of opinion need not translate into sharp differences of opinion. What stands out in a comparison of the roots of women's and men's thinking about women's traditional roles is the limited commonality. Religious belief, feminism and age reveal similar pressures on opinion on women's issues for the two gender groups. Support for equality of opportunity for women as a group is weak among older age cohorts for both genders and, as to be expected, feminists are more supportive of such equality than non-feminists. Religious belief, associated with support of family values, is likely to bring with it weaker support of gender equity. The association is only slightly weaker among women. The similarity between women and men ends here. Among women, the structure of opinion on women's roles is, after religious belief, age and feminist consciousness, dominated by three factors: homemaker status, post-secondary education and public sector employment. Homemakers are more conservative in their support of women's equality. If one expected gender role socialization to matter at all for political opinions, then this is the issue on which any distinction should appear. And it does. Gender role socialization, or the self-selection of women into the homemaker category, plays a part in deteimining opinion on the gender roles for women but it is of limited relevance in the determination of attitudes on other issues. The only -204-Table 6.5: Determinants of Opinion: Women's Roles and Abortion Independent Variables Traditional Roles Anti-Abortion Women Men Women Men Religiosity .08** .10** .27** .31** Anti-Feminism .27** ~) j ** .13** Family Income -.05 -.04 -.07 -.04 Working -.01 -.06** -.04 .01 Public Sector Employment -.04* -.02 .02 -.02 Homemaker .06* .05 Married -.01 -.04 .05* -.02 Children at home .08 .10 .08 .16** Post-secondarv Education -.04** -.02 .04 -.02 Age .30** .36** -.01 -.06 Francophone -.04 .02 .08* .04 Union .03 .01 -03 .00 Member - Political Party -.05 .05 .01 .02 Member - Cultural Org. .03 -.02 .02 .04 Member - Service. Bus. Club .01 -.04* -.04 -.02 Member - Issue Group -.01 .03 .02 .03 Atlantic -.01 -.06* .05 .02 Quebec .03 -.03 -.06 -.03 Prairies .01 .00 .02 .02 British Columbia .02 -.02 - 03 .04 Constant .05 .11** .08 .12** R2 .19 .16 .20 .21 (N) 820 939 823 936 Note: * p<05; **p<.01. Entries are unstandardized regression coefficients. -205-other issue on which it enters significantly into the equation is for government involvement in the economy. On that issue homemakers were more liberal in their attitudes. On this issue they are more conservative. Attendance at a post-secondary institution also liberalizes women's thinking about then status in a manner unmatched among men. It seems that an education, among other things, provides women with tools that allow them to endorse greater equality for women. Among men however, this liberalization comes in part through employment status. Recalling Table 5.2, retired males make up the largest comparison group with working men; the dummy variable in this instance may be tapping into further generational divisions on the issue. Greater support for gender equity is also found among men who are members of service and business groups. This is in fact the only time that such membership is associated with more liberal attitudes on an issue. The greater sensitivity of men in such groups may be due to the possession of greater knowledge of the problems women encounter in the workplace. But men in Atlantic Canada stand apart for then-greater support for gender equity: why this should be the case is not certain. This certainly does not seem typical of the more 'traditional' culture often attributed to this region of the country. The structure of opinion on abortion is less differentiated by gender than opinion on women's roles. For both men and women, increasing religious commitment strongly corresponds with weaker support for choice on the abortion issue. And feminists among both gender groups are much stronger in then support of increased access to abortion than those who are most anti-feminist. Feminism does however exert a much stronger -206-influence on women's opinions on this issue than it does for men. Such a result lends support to the idea that the feminism measure is tapping into something different in women than in men. However, this 'difference' is only apparent on certain issues. The presence of children in the home also exerts conservative pressure on attitudes towards abortion. Fatherhood may trigger an awareness that extends the issue beyond one of the rights of women to the rights of the unborn child as well.2 Among women, conservative pressure is also found in language group membership: Francophone women are less supportive of increased abortion access than are non-Francophones. The language variable may be tapping into Catholic belief (given its correlation with language) as a source of conservative opinion on the question of abortion access. But Francophone women are also more likely to be feminists so it would be incorrect to assume that Francophone women are any less supportive of abortion availability than all other women. Canadian Dualism Table 6.6 shows that opinion on the longstanding question of the relationship between Canada's two linguistic groups is marked by a pattern of support structured along a number of dimensions: region, education, age, language and feminism are each linked to opinion on this issue. This attests to both the salience and divisiveness of the issue. As a 'contested' issue central to much of the political debate of the 1980s, it is understandable 2 The coefficient on the children in the home variable in the women's sample is strong but fails to reach statistical significance at the 5 per cent level. -207-that this issue reveals such a high degree of social stratification at the time of the 1988 election (Johnston, 1986). It is also true that social stratification on this issue is remarkably similar between women and men. Quebec men and women stand apart from the other provinces in their overall support of dualist policies. This regional support is strongly reinforced by the supportive stance adopted by Francophone women and men. The distinction of showing the lowest degree of support for dualism goes to men in the Prairies. Their level of support (0.45 on the index) ranks significantly below the lowest regional level of support among the women (0.38 in British Columbia). Put differently, the lowest level of regional support for dualism among the women lies roughly mid-way between the highest and lowest levels of support for the men. Women and men in Ontario are the moderates on this issue, with both genders in the Atlantic provinces revealing only slightly greater support than in Ontario. Feminism, no doubt because of its strength in Quebec and its acceptance of diversity, also increases support for dualism. It is important to note that the strength of the feminist coefficient among men is much larger than it is among women. Put somewhat differently, knowledge of a man's feminist position provides greater insight into his opinion on support for bilingualism in the country than would the same knowledge about a women. If women's adoption of feminism is rooted in existence as a woman and therefore more personal than men's, for whom it stands as one of a set of related ideological positions about social issues, then one would expect to find that feminism is more closely associated with opinion on bilingualism policies among men -208-Table 6.6: Determinants of Opinion: Dualism Independent Variables Anti-Dualism Women Men Reliuiositv .03 -.01 Anti-Feminism .15** .28** Family Income .03 -.02 Working .02 -.03 Public Sector Employment -.02 .00 Homemaker .00 Married .05* .01 Children at home -.02 -.02 Post-secondar\ Education -.03 -.05** Age .15** .12** Francophone -. 14** -. 19** Union .02 .03 Member - Political Party .00 .01 Member - Cultural Org. -.03 .00 Member - Service, Bus. Club -.00 .02 Member - Issue Group -.04 -.02 Atlantic -.04 -.02 Quebec -.10** -.08** Prairies .09** j j ** British Columbia 10** .06* Constant .28** .34** K- .31 .38 (N) 818 923 Note: * p<.05; **p<01. Entries are unstandardized regression coefficients. -209-than it is among women. For women, support for bilingualism would be less intimately connected with feminism in that its direct connection with the issue is more limited. Age cohort similarly separates various support groups: the age coefficients in the women's and men's samples reveal an increasing conservatism among older age cohorts. That support for institutionalized bilingualism decreases among older Canadians may suggest that for older ROC Canadians resistance to the 1969 Official Languages Act may not have died, and the continuous play of the question on the political agenda would ensure it little chance of doing so. Younger Canadians, having witnessed little of the issue's potential for extremism may as a result make them less inclined to dismiss official bilingualism as easily. It is also true that younger ROC Canadians are likely to have had greater exposure to the French language than older Canadians (e.g; French immersion programs) and perhaps less negative experiences as a result (e.g. refused employment due to language requirements). Such factors would account for the age gaps in support. The only factors that do not have consistent effects across the sexes are marital status and post-secondary education. Married women are less likely to support equal status for English and French in Canada than are non-married women. Alternatively, marriage is of little help in predicting men's opinions on this issue. As noted earlier, marriage also served to bring women's attitudes closer to those of men's in the state involvement index. These findings provide limited support for Carroll's assertion that marriage restricts women from developing independent political orientations: on the questions of Canada's duality and support for government intervention in the economy, -210-married"women are more conservative in outlook than never married, divorced, and widowed women! What is less clear is why marital status is relevant among women on just these two issues. The results further show that men with some post-secondary education experience are more supportive of the two founding nations version of Canadian history than other men, due perhaps to the increased tolerance associated, with a higher education. No similar association is found within the women's sample. British Ties On the symbolic issue of continued ties with the monarchy, the enduring dichotomy between Quebec and the ROC is clearly evident. As shown in Table 6.7, the opposition of Quebecois to the traditional and historical association with the Crown is manifested on two fronts: Quebec residence and Francophone status strongly predict opposition to continued British ties. This demarcation is found among both women and men, although it is somewhat stronger among the women. Membership in a cultural association is also likely to be associated with weaker support for the Monarchy: women and men in such associations are equally less supportive of continuing our historic ties to Britain. In contrast, those who are more religious are more likely to support these ties. Among women, a number of additional social indicators are associated with opinions on the issue. Only among women does age enter as a significant predictor of attitudes on this issue: older women are much more supportive of ties to the Crown than younger women. It is not immediately clear why men do not exhibit the same.pattern by age. Post-secondary education also tends to have a modest conservative effect on -211-Table 6.1: Detenninants of Opinion: British Ties and Anti-Separate School Funding Independent Variables British Ties Anti-School Funding Women Men Women Men Religiosity .15** .12** -.20** -.25** Anti-Feminism -.03 -.01 14** -.08 Family Income -.00 -.04 .04 .01 Working -.01 -.03 -.02 .03 Public Sector Employment -.08** .03 -.03 .03 Homemaker -.05 -.06 — Married -.05* -.01 .02 .02 Children at home -.03 .09 -.04 -.09 Post-secondarv Education .07** -.04 -.03 .01 Age .10** .09 .10 .13* Francophone -.18** -.10* -.16** -.13** Union .05* -.04 .02 .00 Member - Political Party -.09* .01 .03 -.02 Member - Cultural Org. _ 17** -.15** -.02 -.04 Member - Service, Bus. Club .04 -.03 .03 .00 Member - Issue Group .01 .07* .04 -.03 Atlantic -.06 -.03 .01 .01 Quebec _ 19** -.14** .01 -.03 Prairies -.07* .00 .05 .04 British Columbia -.01 -.01 .07 .03 Constant .72** (,9** .62** 73** R2 .21 .12 .13 .14 (N) 820 937 826 941 Note: * p<05; **p<01. Entries are unstandardized regression coefficients. -212-women's support for a continued relationship with the Crown. However these effects are not found among men. For women, any liberal push comes from employment in the public sector. While education is likely tp introduce individuals to alternatives to continued ties with Britain, it is not clear why employment in the public sector would have the same impact. Neither is it clear why these factors impact on one gender's opinions and not the other's. Public Policy: Anti-Separate School Funding The separate school funding index stands together with abortion in the degree to which attitudes on this question are directly associated with religious belief, as shown in Table 6.7. Besides religious belief, women and men share only one common determinant of attitudes on this issues. For both men and women greater support for public funding of separate schools is strongly and directly associated with language: Francophones are significantly more in favour of public funding for separate schools. Only among men is age an important predictor of opinion. Increasing age is associated with less support for school funding, supporting evidence that opinions become increasingly conservative with age or that there is some generational change in support for the public funding of separate schools. However, only among women is feminism associated with attitudes on this issue, feminist women are much more likely to support public spending for separate schools. Feminist men, however, are no different from all other men in their support for such policies. For men, support for such policies is linked to their religious beliefs and the language they speak; among women, in addition to these two sources of opinion, -213-feniinist belief exerts an influence on such policies. It appears that among women feminists in particular an acceptance of diversity extends more fully to public support of that diversity than it does among feminist men. The egalitarianism associated with feminist belief is more generally adopted by feminist women than it is by feminist men. That the issue involves the education of children may be the source of this difference: if feminism allows women to speak in then particular voice, a voice that includes women's natural nurturance and compassion, then the evidence on this issue suggests that feminist men will by virtue of then gender alone never speak in a similar voice. Feminism for men will designate a different set of priorities that it will for feminist women. Public Policy: Low Environmental Priority and Energy Investment Table 6.8 reveals that the sources of concern for the envhonment are very different for men and women. For women, post-secondary education is likely to increase one's belief in the priority that should be given to envhonmental concerns. Alternatively, religious belief is likely to dimmish the priority given to the envhonment. For men greater concern comes with employment in the public sector. That no similar association is found among women is intriguing and possibly stems from differences in the types of occupations women and men are likely to hold in the public sector. In a similar patterning, Francophone women and Quebec men are likely to adopt more conservative attitudes on the issue, although the differentiation is slightly greater among women. And unlike the pattern for many other issues, women in Atlantic Canada are likely to be different from the other provinces in the weaker priority they give to the envhonment although men -214-Table 6.8: Determinants of Opinion: Low Envirorimental Priority and Energy Investment Independent Variables Low Environmental Priority Energy Investment Women Men Women Men Religiosity .08** .03 .06* .04 Anti-Feminism -.03 .08 -.02 .13** Family Income .03 -.04 .07 .05 -.02 -.02 .01 -.01 Public Sector Employment -.02 -.06** .02 -.00 Homemaker -.02 -.04 Married -.00 -.01 -.03 -.04 Children at home -.08 -.02 -.01 .09 Post-seeondarv Education -.08** -.03 -.06* -.05* Age -.06 -.02 .10 .21** Francophone j | ** .05 .09* .07* Union -.02 .02 .00 .01 Member - Political Part} -.05 .04 -.05 -.02 Member - Cultural Org. -.06 -.03 -.04 -.05 Member - Service. Bus. Club -.01 .00 .07** .04 Member - Issue Group -.05 .00 -.01 _ Q9** Atlantic .06* .05 .00 .04 Quebec .06 .08* .03 14** Prairies .04 .01 09** .08** British Columbia .00 -.06 -.03 .05 Constant .28** .24** .54** 49** R" .12 .07 .07 .10 (N) 817 037 806 925 Note: * p<.05; **p<01. Entries are unstandardized regression coefficients. -215-in Atlantic Canada are not similarly distinguishable. Greater differentiation also appears among the sources of concern over energy shortages between women and men. Three common determinants are found on this issue. Post-secondary education reduces support for encouraging investments in energy investments while Francophones are more willing to support the development of such projects. The only other significant common finding is that both women and men in the Prairies are more conservative on this issue. In addition, however, men in Quebec are significantly more likely to endorse public investment programs in energy supplies than men in any of the other provinces. This regional differentiation reflects each particular region's status as either an energy producer or an energy consumer and the particular interest calculations likely to accompany that status. Among the differences in sources of opinion on this issue is the finding that age predicts opinion among men only: older men are significantly more likely to endorse such programs. Moreover, only among men is feminism associated with less support for such programs. Neither age nor feminist consciousness are significant predictors of opinion on energy investments among women. The differential impact of feminism in this case again point to the possibility that the measure is tapping into something fundamentally different across the two groups. On this issue, which appears to have little connection to feminist principles, feminist men are nevertheless found to be more liberal than non-feminist men. For many men in the sample, the feminist variable may very well be tapping into an ideological predisposition, rather than one of feminist consciousness. Memberships in some social organizations reveal significant associations with -216-opinion on this issue, although the patterns differ by gender. Knowing that a woman is a member of a service organization tells us something else: the woman is likely to be more supportive of the funding of big energy projects. On the other hand, knowing that man is a member of an issue group means that he is likely to be somewhat less supportive of such projects. In both the environmental priority and energy cases, women's opinions are likely to vary with religious belief. This is true in neither case for men. As a core value, religious belief is associated with less support for dealing with environmental concerns and with greater support for the development of new oilfields and dams, but only among women. It is not immediately clear why this would be the case. Public Policy: Low Taxation and Extra Billing As shown in Table 6.9 opinions on both taxation and extra billing by hospitals and doctors in Canada reveal limited social stratification. Nevertheless gender differences are to be found in the sources of such opinions. The presence of a union member in the home is moderately associated with approval of tax increases among both sexes, but this is the only common deteiminant. Among women, employment and membership with a political party are significantly associated with support for tax increases as a means of cutting the deficit. Feminism, on the other hand, is associated with a levelling off of support for tax increases but feminism is not associated with tax opinion among men. Among men, opinions on taxation have their roots in age: older men are significantly less likely to endorse tax increases as deficit-reducing measures. The lack of a similar, -217-Table 6.9: Determinants of Opinion: Low Taxation and Extra Billing Independent Variables Low Taxation lixtra Billing Women Men Women Men Religiosity .02 .03 -.04 -.03 Anti-l:eminisrn .09* .03 .11* .09 Family Income -.02 -.02 .18** .10** Working .04* .01 .01 -.01 Public Sector Employment -.02 .02 -.02 -.03 I lomemakev .00 _ _ -.02 — .02 .01 -.02 -.02 Children at home -.07 .04 .04 .02 Post-secondary Education -.00 -.01 .01 .03 Age -.03 -.12** .24** .01 .03 .01 -.03 Union .04* .04* .00 -.03 Member - Political Party .07** -.10** -.02 .04 Member - Cultural Org. .00 -.04 -.01 .02 Member - Service, Bus. Club -.00 .01 .02 .05* Member - Issue. Group -.02 -.02 .02 .03 Atlantic -.02 .00 -.01 -.03 Quebec .02 .01 .07 .04 Prairies -.01 -.02 .01 .00 British Columbia -.05* -.04 .04 .00 Constant .74** .80** .01 .11* R* .06 .09 .08 .07 (N) 827 940 824 938 Note: * p<.05; **p<01. Entries are unstandardized regression coefficients. -218-relationship among women is intriguing. But most notable is the finding that membership in a political party is negatively associated with opinion on taxation, the opposite of the finding among women. This is the only time that memberships in a social organization produces diametrically opposed results: women in political parties endorse tax increases as a means of deficit reduction while men in political parties endorse, presumably, spending cuts. The significance of this finding is clear: women and men in political parties have very different priorities. Opinions on extra-billing similarly reveal limited social stratification. For both sexes, much of the differentiation in opinion is to be found across income levels and age groups. Higher income and age levels are associated with weakened opposition to extra billing by doctors and hospitals! For men, membership in a service or business organization is associated with greater support for extra billing. Among women, feminism enters as a strong predictor of opinion: feminists are stronger in their opposition to such policies than non-feminists. This may reflect an understanding among feminist women that is not matched among feminist men; namely that extra billing would exact an unequal burden on women. The same is also likely true of feminism's relevance for women's opinions on taxation: taxation as a means of deficit reduction would not exact as heavy a toll on women as would cuts to government programs. Public Policy: Immigration Inrmigration is an issue which arouses deep sentiment and triggers beliefs on a number of fronts: compassion for those less fortunate than ourselves, questions concerning the -219-ability of the economy to sustain an increasing labour force, the ability of the welfare state to bear the burden caused by inunigrants with few language and other skills, and justification for restricting entry to refugees and those with substantial financial assets. Canadians in 1988 were slightly less than positive about increasiiig immigration numbers, and no significant gender gap appeared on thisi issue. Nevertheless, Table 6.10 reveals that the construction of opinions on this issue varies significantly by gender. Among men, weaker opposition to increased immigration is associated with post-secondary experience, increased family income, feminist belief and increased age. The latter is likely a function of the perception among young men of the greater degree of job competition introduced into the labour market by increasing numbers of inunigrants; with greater job security Or retirement in hand, such concerns are of lesser importance for older male cohorts. And the effects of the three former variables are in the anticipated direction: higher levels of family income bring increased financial security and less concern for economic self-interest; feminist belief increases one's acceptance of plurality and diversity; and higher education is often associated with a liberalization of attitudes; Each serves to reduce opposition to increasing the number of immigrants allowed into the country. Religious belief is associated with weaker support for increased immigration numbers while membership in a cultural association is closely associated with support of immigration. For women, this liberal push is revealed among members of issue groups. Among the women in the sample, the impacts of post-secondary education, religious belief and feminist belief mirror that of Canadian men, but without the Table 6.10: Determinants of Opinion: Anti-lnmugration and Anti-Union Independent Variables Anti-Immiiiralion Anti-Union Women Men Women Men Religiosity .12** .11** .04 .08** Anti-Feminism .18** .14* .28** 29** Family Income -.05 -.12* 11 ** .04 Working .02 .04 .00 .02 Public Sector Employment -.02 -.06 -.08** -.08** Homemaker .01 _ _ -.00 Married .01 -.01 .00 -.03 Children at home -.05 -.05 .01 .06 Post-secondary Education -.10** -.14** .03 .02 Age -.09 -.24** .06 -.05 Francophone .08 .04 -.04 -.02 Union .01 .00 -.10** -.18** Member - Political Party -.08 -.08 .00 .00 Member - Cultural Org. -.07 -.10* .05 -.02 Member - Service. Bus. Club .03 -.01 .04* .02 Member - Issue Group -.10** -.05 -.01 -.01 Vtlanlic .06 .03 -.02 .00 Quebec .01 -.05 .08* .04 Prairies .05 .06 -.02 .05* British Columbia .16** .02 -.04 -.02 Constant .54** 71 ** .43** 47** K2 .10 .13 .17 .22 (\) 825 938 824 935 Note: * p<.05; **p<01. Entries are unstandardized regression coefficients. -221-comparable effects found among age groups or income levels. Thus opposition to immigration is less rooted in financial concerns for women than for men. Instead regional differences exert competing pressures on opinion. Women in British Columbia are less likely to support increased immigration than women in the remaining provinces. It is not immediately clear why this issue in particular evokes a regional dimension in women's attitudes that is not matched in men. In sum, for men immigration policy is a matter of economic self-interest; for women, it is more a matter of social group cues and feminist influence. Unions and Labour Groups Women and men are remarkably similar in their levels of support for labour and unions, but the sources of such attitudes differ. As shown in Table 6.10, women display greater variance in their level of union support across the factors included in the analysis. The common deterrninants of opinion on labour and unions are the presence of a union member in the home, feminist consciousness and public sector employment. Each of these factors produces a strong liberalizing force on opinions, although the presence of a union member in the home acts as a stronger predictor of opinion for men. This may be due to the fact that more men in the sample are likely to have directly participated in union activities; for women, it is more likely the union influence came through the union membership of someone else in the home. Another statistically significant result in the men's sample is the weaker support among Prairie men for unions compared to opinion in the other provinces. Men who are more religious are also less likely to support union -222-activity. Among the women, however, a number of additional deteirninants appear. The importance of self-interest in shaping opinion is supported: support for unions lessens with increased levels of family income. The correspondence can be assumed to stem from either the visible gains connected directly to union activity (e.g., in wages and/or benefits) or higher union membership numbers among lower income families. But it is the lack of a similar correspondence between income and union support among men that is intriguing. This suggests that economic self-interest, at least in this case, may be of greater importance for women. It may also point to the fact that union membership is more widespread across income levels for Canadian men which would account for the negligible impact of income on labour support. Union support is also predicted by residence for women: women in Quebec are less supportive of union activity. Among men, the prairie provinces are distinctive in their lesser support for unions; among women, it is Quebec that holds this distinction. Finally, women's membership in a service or business association is likely to bring their opinions closer to those of men: women in such organizations are less supportive of union activity. Political Cynicism and Risk Aversion As shown in Table 6.11, there is little evidence in either the women's or men's sample to suggest that cynicism is structured by socio-demographic considerations. A common determinant for both women and men is the presence of a union member in the home: a -223-slightly elevated level of cynicism is found among union members or among individuals with a union member in the household. Among men, cynicism also has financial and political roots: higher income leads men to be less cynical about politicians and government in general and membership in a political party is associated with less cynicism about politicians. But among women, greater trust in government requires in part direct contact with its institutions: public sector employment decreases women's level of cynicism: This result is suggestive of the degree to which income and financial status are limited political vehicles for women. Attitudes on risk aversion show a slightly higher level of socio-economic variation and greater gender difference in the sources of opinion on this issue than is true of political cynicism. Feminist consciousness and post-secondary experience exert common pressures on women and men in then support of attempts at social reform: both are associated with greater support for such attempts. Among men, the desire for reform also has regional and religious roots: men in the Prairie provinces are less likely to advocate reforms than men in the remaining provinces and religious men are less reform minded than other men. Some of this regional dimension appears in women's opinions but in a different pattern. Women in Quebec and British Columbia are less willing to endorse reforms than women in the other provinces. Moreover, public sector employment is slightly associated with women's willingness to attempt reform: women employed in the public sector are slightly more willing to attempt reform than those who are not similarly employed. It is perhaps the case that women who are more reform minded choose employment in the public sector as an avenue for bringing about such reforms. The final -224-Table 6.11: Detemiinants of Opinion: Cynicism and Risk Aversion Independent Variables Cvnicism Risk Aversion Women Men Women Men Religiosity -.00 -.03 .05 .08** Anti-Feminism .01 -.02 .26** .16** FamiK Income -06 -.07* -.07 -.03 Working .01 .03 -.03 -.01 Public Sector Employment -.05** -.01 -.05* -.00 Homemaker .03 „ -.06 Married .00 .01 .03 .00 Children at home .04 .01 -.08 -.OA Post-secondarv Education .00 .00 -.05* - 07** Age .03 .01 -.07 -.01 Francophone -.05 -.04 -.02 .00 Union .05* .04* -.01 .02 Member - Political Party .04 -.05* -.02 -.03 Member - Cultural Org. .01 -.01 .01 - | | * * Member - Service. Bus. Club -.01 -.01 -.03 .01 Member - Issue Group .02 -.01 -.04 -.01 Atlantic -.02 .02 .03 .02 Quebec -.02 -.04 .09* -.04 Prairies .02 -.01 -.03 .08** British Columbia .02 .00 .08* -.01 Constant 7 | ** .73** .33** .33** R .04 .04 .09 .08 (N) 822 940 826 938 Note: * p<.05; **p<01. Entries are unstandardizedregression coefficients. -225-distinction is in the finding that men who are members of cultural organizations are far more likely to think that reform is worth attempting than men who do not belong to such organizations. Conclusion The goal of this chapter was to determine the degree of differentiation that exists in the social structure of women's and men's opinions across the range of issues tapped by the indexes created from the 1988 CNES. The evidence suggests that the differentiation is subtle, but nonetheless significant. The cues that women and men look to for help in determining their political opinions and variation in the strength by which social forces influence each gender's opinions cannot be dismissed. Gender does matter to political thinking. A complete understanding of the nature of political thinking in Canada requires the addition of gender, as a relevant determinant of opinion, to those factors traditionally considered as determinants of attitudes. In combination with other social forces, accounting for gender allows for a better understanding of political tninking in Canada. But gender alone explains only a small portion of the variance that occurs in opinions. Women and men look to many of the same factors for cues in the determination of opinion on a number of issues. On some issues, the strength of those cues is similar; on others, it varies considerably. Understanding the nature of the issue addressed increases the likelihood one has of determining the extent to which gender matters. Nevertheless, an attempt to look for simple patterns across the range of issues -226-examined would underestimate the complexity of the stiucmring of opinion. Gender role socialization appears to matter little in the shaping of women's political opinions. The most traditional of gender roles, and the most easily measured in an empirical sense, is that of the homemaker. But, all things being equal, homemakers do not differ in their opinions from other women. As a determinant of opinion, gender role socialization fails miserably. The adoption of the homemaker role was significant for thinking on only two of the issues investigated here: on opinions regarding equality for women and government involvement in the economy. That the measure should enter significantly in the women's roles index is to be expected; the index itself included a variable which forced respondents to choose between staying at home or pursuing a career. That it entered into only one of the remaining indexes is more surprising. However, these results replicate those found in other studies of the gender gap and women's opinions (e.g. Sapiro, 1983: 143-167). It is . of course possible that the failure to find homemaker status playing a role in the determination of opinions may be partially due to the weakness of the measure as a true proxy for the adoption of a traditional gender role. Some women work although they believe their proper place is in the home; others stay home temporarily in order to care for young children only to reenter the workforce at a later date. Such factors are likely to reduce the likelihood that homemakers are found to hold opinions which differ significantly from other women. The lack of relevance of women's private roles to their public attitudes is also suggested by the poor showings of both the marital status and the children in the home -227-variables. Indeed, these variables appear as often as predictors of opinion among men. Marriage, when it enters significantly as a determinant of opinion, appears to make individuals more conservative in their opinions. This is true of men in the case of the core value of religious fundamentalism. And marriage acts as a conservative force on women's opinions on abortion, dualism and government involvement in the economy. The only exception to this rule is found in the case of opinion on British ties: married women are likely to hold more liberal views on this issue than all other women. It would appear that married individuals bring with them to their partnerships a set of opinions distinct from other individuals, or that the partnership coincides with a particular milieu that exerts conservative pressure on individuals within it. A similar result is found with the presence of children in the home. In each of the three cases where it appears as a significant predictor of opinion, an increasing number of children in the home corresponds with greater conservatism. This is true of women for the core values of religious belief and feminism, and of men in the pro-life and state involvement in the economy indexes. Thus it seems that on some issues the private live's of individuals shape their beliefs. More importantly this is true of both women and men. Three issues appear particularly sensitive to private factors like marriage and children; namely, religiosity, abortion and government involvement in the economy. In each case the two private social forces are linked with more conservative opinions. The links between the first two issues and private factors are relatively clear; with the issue of state involvement in the economy, however, the association is less easily discerned. Mothering, -228-at least with the measure employed here, matters little for women's political opinions. In contrast, fatherhood, although relatively weak compared to other factors of relevance in the determination of their opinions, has a greater impact in the shaping of these opinions than does motherhood. It should not be forgotten that mothering mattered for women in the deterrnination of the two core values examined in Chapter 5: feminism and religiosity. The causal links are likely from feminism and religiosity to mothering: women who are religious and concomitantly less feminist are more likely to become mothers. ' On the other hand, the measure of feminism employed in this investigation does explain much of the variation in women's and men's opinions: It fails to enter significantly in just over one quarter of the thirty-eight regressions run in this analysis. Moreover, whenever feminism significantly entered into a regression, it usually did so at a level far exceeding that of the other determinants of opinion. And more often than not, its impact differs with gender. Recalling the discussion in Chapter 5, there are a number of weaknesses in the feminism measure employed here. The most relevant is its inability to distinguish between a feminism for men which may be ideologically driven and a women's feminism that is rooted in gender consciousness. This weakness is borne out by the results derived in this chapter. Feminism enters more strongly for men on issues that are not directly connected to feminist concerns. For example, feniinism enters as a significant determinant only for men's attitudes on the energy investment index and it enters more strongly than is true for women in the anti-dualism index. On concerns that can be -229-considered of greater interest to feminists, the measure has a stronger impact On women's opinions than men's: it enters more strongly on the pro-life index, a decidedly feminist issue; it enters significantly only among women on the anti-separate school funding index perhaps due to their more limited means of providing for such schooling personally or to a belief in egalitarianism that extends to children's schooling; and it enters significantly only among women on the low taxation and extra billing indexes perhaps due to an awareness of the burden such policies would force many women to bear. The need for the inclusion of variables in surveys better suited for the derivation of feminism measures which are both more reliable and differentiate between feminism as part of a general ideological position and feminism that is grounded in women's everyday existence might bring us closer to understanding the nature of each gender's political thinking. For the moment at least, feminism is a relevant determinant of •opinions that must be.included among those traditionally considered to shape opinions in Canada. Also significant are the differences discovered in the predictive abilities of both self-interest and social group memberships in explaining political thinking. Self-interest appears as a relevant determinant of opinion in fifteen of the twenty indexes and social group memberships appear to matter in every index investigated. These results add to those generated by previous research in public opinion in Canada emphasizing the importance of such factors. Does the relative importance of self-interest,in shaping opinions vary for women and men? The answer to this question depends on the issue under investigation. On most -230-issues, family income acts in a similar manner in structuring the opinions of women and men: in all cases, increased income brings with it a reduction in concern for those most dependent on the welfare state. A self-interest argument suggests that the lessened concern for one's own economic fortunes that accompanies higher income levels corresponds with the endorsement of economic individualism and competition, and subsequently lesser support for government programs designed to improve the lot of those with less of the competitive spirit and desire to ensure self-sufficiency. And this is the case on all the economic issues investigated here. For cynicism and attitudes on immigration policy, however, the reverse is true: higher income brings with it a liberalization of attitudes on these two issues. This pattern likely emerges as a result of differences in socio-economic class. On five issues, family income appears to have a differential impact on opinion: on the defence, anti-social spending; continentalism, and anti-union indexes the variable enters significantly in only the women's regressions, or at the very least at a stronger level than for men; on the anti-immigration, anti-civil liberties, energy investment and cynicism indexes, the variable enters significantly onto only the men's regressions. Given the greater number of women at lower family income levels it seems their more liberal opinions stem in part from lesser financial means. It also means that on key gap issues, defence, continentalism and social spending, women's opinions are more dependent on their personal economic circumstances than is true of men. Self-interest was also tapped by two other variables: employment status and public -231-sector employment. Either alone or together, these two measures were significant in twelve of the twenty indexes. And these measures show a significant degree of differentiation by gender as well: in none of the regressions does either of the two measures significantly enter for both women and men. And what stands out is the relevance of public sector employment to women's opinions: on seven issues it acts as a liberal force on opinions. Unanticipated, this result begs for further analysis to uncover the relationship beneath these results. Finally, and in conjunction with the results obtained through previous research, social group memberships are important components of opinion for both women and men in Canada. The age, language, region, and education measures enter significantly into many of the indexes included in the investigation. The presence of a union member in the home has relevance for fewer issues, but is nevertheless an important component of opinion structure. Few surprises are found among these results. Nevertheless the results show that it is often the case that the construction of opinions for women and men varies. First, the generation gap in opinion appears more often among the opinions of women than men and may play a part in four of the recorded gender gaps in opinion. On the British ties, defence, social spending and continentalism issues, younger women are significantly more liberal in then opinions than older women. No similar generational division exists among men on these issues. This difference reflects the change in political outlook that separates generations of women. Such a gap in opinion may diminish over time if the opinions of coming generations of women closely -232-match those of today's youngest generation of women. If so, then gender gaps on these issues may increase over time. In addition, regional differences in opinion are largely a male phenomenon. More often than not the regional variations which structures the opinions of Canadian men are not significant for women. Regionalism, an important differentiating factor in much political research in Canada, has a greater impact in men's attitudes than in women's. The social groups within local communities that allow for the transmission of the local political culture may be less relevant for women than for men: the difference lies either in the social and political groups that they choose to join, or in the salience of those groups to their male and female members. The relevance of this finding extends beyond public opinion, and undoubtedly demands greater attention. Qualitative research on local communities may prove beneficial in this regard. Some hints are provided by the social group membership variables included in this analysis. More often than not, membership in a political party is likely to influence men's opinion more than women's. The only time that membership in a party mattered more for women was in serving to bring their attitudes on British Ties closer to those of men. On the issue of taxation, however, party membership revealed an interesting distinction in support of taxation as a means of overcoming deficits: women in political parties support it more than other women, while men in parties support it less than other men. Service organizations are unique in their ability to bring women's opinions more closely in line with men's. This is understandable given that these organizations, like -233-political parties, are dominated by men. Involvement in them serves to bring women opinion's, and hence political culture, closer to that of men's. Issue groups, on the other hand, although they enter significantly into regressions less often than service groups, seem to move men's opinions closer to women. Men who are involved in these organizations are more liberal in their support of social spending and energy production versus conservation, and are more likely to endorse British Ties. mterestingly, on two key gap issues, defence and continentalism, women who are members of cultural or ethnic organizations are further away from men in their attitudes. For men, such organizations serve to move their opinions on immigration and reform away from those of women. This may suggest that such groups are more likely to be men-only and women-only groups, if not as a matter of policy, then at least as a matter of course. This is of course highly speculative. But further investigation might provide answers that reveal the relevance of social interaction for women's and men's political opinions. -234-Chapter 7: The Belief Systems of Women and Men The ideological underpinnings of opinion, that is how individual attitudes relate to each other in a hierarchical fashion and are constrained by core beliefs, provide an additional test of the existence of distinctive gender cultures. This is the direction taken in this chapter. With the use of factor analysis1, the core beliefs which mdeipin women's and men's opinions will be isolated in an effort to assess their similarity. The working assumption is that women's and men's opinions are at some minimum level driven by ideological beliefs, but that the patterning of opinion around these core beliefs, or heuristic tools, varies by gender. More specifically, the degree of constraint that exists in women's and men's belief systems may differ, as well as the patterns that appear among their opinions. Exploratory factor analysis provides a means of determining the validity of this assumption. The first part of this chapter reviews the debate surrounding the existence of belief systems in order to situate this analysis within it. The second part elaborates on the method of analysis, factor analysis, employed here to measure belief systems, and the third part discusses the results obtained in the analysis. The final part will assess whether the working assumption stands up under scratiny. Gender and Belief Systems Belief systems are the interconnections between various opinions and the degree to 1 For a detailed explanation of the factor analysis procedure see Harman (1967). For a less mathematically daunting look at factor analysis see Kline (1994). -• . -235-which those interconnections are hierarchical. They are systems in the sense that core beliefs, found at the highest level in the Merarchy/determme or shape opinions on issues found at a lower level in the hierarchy: Ideology is often assumed to underpin opinions on a range of issues. An ideology is a "coherent arrangement of attitudes towards groups of issues that is characterized by constraint: the positions taken on individual issues by the ideological voter must 'hang together' in some logical way" (Harrop and Miller, 1987: 115). In this sense, knowing whether an individual's core ideological belief lies on the left or right allows you to predict their opinions on a range of more specific issues. The relevance of ideology to the shaping of attitudes has not heretofore been directly evaluated in the investigation: the impact of feminism as a potential proxy for ideological orientation has been introduced but the statistical procedures employed so far have not permitted tests of relative importance. Including it in this stage of the analysis will allow for the determination of its relevance to the opinions of men and women. Belief systems are not, however, in and of themselves ideologies^  Belief systems are more general than ideologies: their elements may be connected psychologically rather than logically. Thus the failure to detect logical constraints need not be interpreted as an indication that individuals have little or no connections among the opmions they hold. Those connections may exist but at first glance may not appear to have any logical rationale behind them. They can nonetheless constitute belief systems. A complete understanding of political minking requires a move away from the social deterrninants of opinions towards the underpirrnings of opinions. The debate on -236-the degree of constraint that exists in the belief systems of the mass public is ongoing. In his seminal article, Converse (1964) argued that ideological constraint was limited in American public opinion. Analysing American election data, Converse concluded that, for the most part, Americans were devoid of any sophisticated system for determimng their opinions on various public issues. As he wrote, A realistic picture of political belief systems in the mass public, then, is not one that omits issues and policy demands completely nor one that presumes widespread ideological coherence; it is rather one that captures with some fidelity the fragmentation, narrowness, and diversity of these demands (1964: 247). His conclusion, given its ramifications for democracy and the fact that it contradicted standard belief, was met with scepticism and much subsequent research has attempted to overturn it. Converse's conclusion was based on his findings of low correlations across a range of connected issues in surveys of the general public, and to low correlations across repeated measures in panel data. He concluded that while a minority of the population adopted real unchanging attitudes, the greatest share held 'non-attitudes' that fluctuated over time. The measured randomness in response was due to "people with no real attitudes on the matter in question, but who for some reason felt obliged to try a response to the item despite our generous and repeated invitation to disavow any opinion where none was felt" (1970: 175). It was incorrect, he concluded, to impute belief systems to the mass public, and more than this, evidence suggested that a majority of Americans thought little about political issues and as a result held no real opinions on them. -237-In response to Converse's minimalist conclusion, subsequent research attempted to refute the notion that the mass public was ideologically unsophisticated. The first attempt, by Nie et al. (1976), pointed out an increase in correlations between issues in American public opinion in the period since Converse first reported his findings, and concluded from this that ideological consistency was dependent on the stimulus received from the political arena. If the mass public was as ideologically unsophisticated as Converse, claimed it was due to the lack of both political cues and politically salient issues at the time of his investigation. The 1960s, a particularly politically salient period, became the instrument for gelling together the opinions held by Americans. Unfortunately, Nie et al. based their conclusions on evidence generated from measures that had changed over the two periods of study, and the increased correlations were an artifact of these changes. As pleasing as their conclusions were, they lacked empirical support. Others have looked to methodological reasons for the low levels of constraint found among the mass public. Achen (1975), in particular, has shown that much of the randomness surrounding attitude response is due to methodological considerations, rather than ideological ones. The vague language used in survey questionnaires is directly responsible for much of fluctuation seen in measured opinions. Having "cleaned" Converse's data of its methodological noise, Achen was able to show that the public was more attitudinally consistent that Converse's conclusions would have us believe. Yet others have noted that the assumptions required to remove this methodological noise are as arbitrary as those made by Converse (Neuman, 1986). -238-•N Three separate and distinct paths have recently been taken by work in this area and suggest different avenues for dismissing the minimalist claim. The first takes as a given the fact that abiding and consistent opinions are less than abundant in the general public, and thus are unlikely to be measured by survey mstruments. This should not be taken to mean, however, that the mass public's belief systems are predominantly made up of non-attitudes. The point is, suggests Zaller, that public opinion fluctuates over time with the salience of particular issues, dictated to some extent by media coverage, question phrasing and placement on the questionnaire, as well as personal experiences (1992: 266). Thus there is little reason to assume that individuals will give the same response to a survey question from one time period to the next. Opinions do and will vary over time. Neither should variation in opinions be interpreted as a lack of political sophistication or reasonableness in the public. There is a general range within which individuals will respond to survey questions, the range determined by then selective consideration of the mix of messages presented to them daily. The more informed an individual is politically, the less likely such messages are to have a substantial effect on his or her political opinions. Similarly, personal political predispositions, or political values, will predispose individuals to accepting certain messages and rejecting others and thus determine the range of responses given to survey questions. Although this variation is normally limited to a particular range, attitude change can occur "not by producing a sudden conversion experience but by producing gradual changes in the balance of considerations that are present in people's minds and available for answering survey questions" (Zaller, 1992: 266). -239-A second branch of recent work suggests the minimalist claim may rest on the methods employed to study ideological structure, rather than on the true underlying structure of opinion. Most investigations of attitudes have relied on measures of bi-variate correlation which consistently underestimate the strength of response consistency across items. Factor analysis, as will be briefly explained below, attempts to overcome the limitations in the use of bivariate correlations for estimating the constraint which exists across opinions. A third branch of work attempting to reverse the rninhnalist conclusion accepts that many individuals lack political sophistication and that many pay little attention to politics. These researchers claim, however, that reasonable political thinking can take place in spite of these two factors. Individuals develop judgemental shortcuts, or heuristic tools, that allow them to manage politically (Sniderman et al., 1991). A further hypothesis is that the minimalist conclusion may rest on the assumption of causal homogeneity: the calculus employed by individuals in the determination of their opinions is assumed not to vary across the mass public. Sniderman et al. question the merit in such an assumption. There is little reason to assume a priori that all individuals employ the same processes to reach their attitudes and opinions. Such an assumption could mask important elements of belief systems within these separate groups and hide important differences that exist in the structuring of their opinions, in effect leading us to conclude the larger group is ideologically unsophisticated when in fact sub-groups within that group hold structured belief systems. Given this possibility, it seems that one could make an important case for examining men and women separately since their political outlooks - 2 4 0 -may differ. It must be emphasized that the assumption of a difference in women's and men's belief systems is not meant to imply that women and men look to radically different ideological sources for opinions. Canadian men and women are after all engaged in the same political system with its concomitant political debates and ideological strains. The salience of particular political issues is unlikely to vary dramatically between the two groups. The difference, it is suggested, might lie in the details of those belief systems, that is in the strengths of association of opinions to core beliefs. One could speculate on the possible differences in the belief systems of women and men. Retxirrring to Gilligan (1982), if the ethic of care and emphasis on egalitarianism over individualism that is argued to guide women in their moral decision-making extends to their political opinion s^etting, then one could speculate that their belief systems would be constrained around an ideology that includes community responsibility for individuals. The ability to separate economic beliefs and community concerns might be less true of women than men: women who believe that government has a role to play in the economy might also believe the role should extend to the responsibility for governments in ensuring that all are treated equally in economic and other areas. The same set of opinions might be less tightly constrained for men if they are able to simply disconnect the two sets of issues. Men's core economic beliefs may play less of a role in deterrnining their opinions regarding social government intervention. The more interesting speculation lies in the positioning of feminist belief among other opinions. As revealed in Chapter 6, women are slightly more supportive of -241-feminism than men. Moreover, men's support of feminism is more grounded in social context than is true of women: women reveal much less social differentiation in feminist belief. If this is extended to belief systems, it may prove to be the case that the role of feminist belief is more central in women's belief systems than men's given the fact that it is less likely to be shaped by social context. If women's support of ferninism stems from gender consciousness then it may relate to a greater number of.issues for them. Measuring Belief Systems Factor analysis provides a statistical method for exarnining those core beliefs underlying political opinions and the degree to, which these core beliefs constrain opinion. Thus the method allows for the determination of the degree to which the opinions of women and men are 'structured', for an investigation of how women and men connect political and other issues, and hence for a comparison of this particular element of their political opinions. Although a full explanation of the technique is out of place, a brief discussion of the method is in order. The usual technique employed for assessing attitude constraint is to compute the mean level of bivariate correlation across a number of response items. But as Johnston explains (1988: 58-59), these bivariate correlations are less than adequate measures of ideological structure for two main reasons. Bivariate analyses are hot applicable to hierarchical structures because they generalize from pans of associations to the larger system. While pairs of associations give us some information on the strength of association in a set of variables, this information will only provide a partial and somewhat - 2 4 2 -distorted picture of the larger system. Second, bivariate correlations of items related to a third unmeasured underlying factor will consistently underestimate the strength of that relationship. If the true relationship between the two items is to an underlying factor, the correlation between the two items will underestimate the level of constraint that exists in the system. For example if item A and item B are each correlated to Factor 1 at the same level, for example r=0.4, then using bivariate correlations as a measure of constraint would yield a bivariate correlation of only r=0.16 (i.e.rab=ral-rbl=0.40-0.40). The bivariate correlation will underestimate the true underlying constraint between the two items. Thus only in estimating the hierarchical structure of opinion is one able to get an accurate measure of the strength of ideological constraint in the mass public. The usual statistical method employed for measuring such systems is factor analysis. Factor analysis allows for the simplification of a complex set of data by condensing the information into a smaller set of factors. Factors are constructed from the correlation matrix underlying the data using an iterative process; those factors which explain the greatest share of the variance among the data are said to underlie them, and in this sense a hierarchical relationship between factors and items is derived. The goal in factor analysis is parsimony: that solution which most simply explains the correlation matrix is best. Hence in this case factor analysis can be used to answer the question: what constructs or factors might account for the correlations between the indexes? Undertaking the analysis separately for women and men allows for the comparison of these underlying constructs, or factors, between them. The analysis consists of submitting the indexes employed throughout this -243-investigation to the factor analysis procedure.2 Both the feminism and religiosity measures are included given the arguments concerning their central, albeit differing, importance for the determination of men's and women's opinions. The evidence presented in the chapters on the gender gap and the social structure of opinions supports the relevance of both core values to the determination of opinions, and thus provides justification for their inclusion in the analysis of belief systems. Including both in the factor analysis investigation tests their overall relevance in belief systems and allows for assessments of differences in their relevance in that system by gender. The first step in the analysis consists of computing the correlation matrix for each sample in order to determine the appropriateness of the factor model and to identify variables that do not appear to be related to the other variables in the model. Computing the correlation matrix for the indexes employed in this study revealed that one variable, or index in this instance, poorly correlated with the other indexes. Given the lack of correlation between the low taxation index and the remaining variables, it is unlikely that the index shares common factors with them. As a result the low taxation index was dropped from the factor analysis procedure. With listwise deletion of missing data in the factor analysis, the sample size drops from the original 2099 to 1406. The final sample consists of 749 men and 657 women. The second step in the analysis, factor extraction, determines the number of factors to be employed in the rotation stage of the analysis. It is possible to compute as many 2 The estimation program employed is SPSS for MS WINDOWS release 6.1. -244-factors as their are variables in the analysis; each successive factor would account for a portion of the residual variance among the indexes until the total variance is accounted for. However, this would not comply with the goal of the analysis: parsimony. Thus in this stage of the analysis a decision must be made regarding the smallest number of factors to extract in order to adequately represent the data: A number of methods are available for deterrrmring the number of factors to extract. In this investigation, both principal component and maximum likelihood methods of extraction were employed.3 The results of these tests showed that five factors could adequately represent the women's index data and that six factors would suffice for the men's. Maximum likelihood was employed for the extraction stage of the analysis for the remainder of the investigation. The final step in the analysis, rotation, attempts to transform the factors in an effort to make them more interpretable. The goal of rotation is to get each index to load significantly on only one factor and thus to achieve a more simple structure. There are a variety of algorithms used for rotation, but the two principal choices are between orthogonal and oblique rotation. Orthogonal rotation results in factors that are uncorrelated; oblique rotation, on the other hand, allows for correlation among the factors. In this investigation oblique rotation (OBLIMIN) was employed for the rotation stage of the analysis since it is unlikely that the extracted factors, or core beliefs, will not covary to some degree. OBLIMIN allows for this possibility and hence is likely to more 3 The eigenvalues, scree plots and goodness-of-fit tests were examined as criteria of the number of factors needed to adequately represent the data. Each confirmed the selection of 5 factors for women and 6 factors for men. -245-accurately model the belief systems of Canadian women and men. Results The results of the factor analysis procedure appear in Tables 7.1 through 7.3. Table 7.1 presents the results of the extraction stage of the analysis for both women and men. Tables 7.2 and 7.3 provides the factor pattern results, the sum of squares values4, variance explained by each factor, and the factor correlation matrix derived in the rotation stage of the analysis. Table 7.2 presents the results of the women's sample and Table 7.3 presents the results of the men's. The results suggests that some structure exists in the political thinking of Canadians and that any variance in the belief systems of women and men is limited. At the extraction stage of the analysis, five factors emerge among the women accounting for 47.2 per cent of the total variance among the indexes. Among the men, six factors emerge accounting for 53.4 per cent of the total variance. Given the sheer diversity of opinions included in the analysis, these are robust results for both gender groups in that common factors 'explain' half of the variance found among the indexes. The indexes that load on the first common factor extracted at this stage provide further evidence that some measure of ideology permeates the opinions of Canadian 4 Sum of squares are equivalent to eigenvalues in the unrotated solution in the sense that this number divided by the number of variables in the factor analysis is equal to the proportion of variance accounted for by the factor. In oblique rotation, however, they represent the direct contribution of each factor to reduction in variance removed of any contribution due to correlation between factors. -246- . -Table 7.1: Factor Analysis Results: Women's and Men's Extraction Stage Results^  Women Men First Common First Common Factor Communalities Factor Communalities Defence .39 .44 .34 .26 Continentalism .54 .47 .54 .48 Anti-State Involvement .53 .42 .61 .47 Anti-Social Spending .56 .42 .56 .52 Anti-Feminism .47 .27 .46 .39 Anti-Union .45 •21 .47 .25 Capitalism .58 .45 .66 .58 Extra Billing .33 .20 .39 .23 Anti-Dualism .25 ' .28 .39 .39 Religiosity .09 .35 -.11 .56 Pro-Life .11 .36 -.10 .41 Traditional Roles .25 .23 .16 .25 Anti-School Funding .15 .37 .18 .31 Anti-Civil Liberties .31 .29 .18 .10 Anti-Immigration .21 .32 .16 .32 British Ties .17 .18 .06 .16 Low Enviro. Priority -.01 .29 .11 .12 Energy Investment .17 .16 .13 .10 Cynicism -.04 .21 -.14 .41 Risk Aversion .14 .17 .18 .22 Initial Statistics % of %of Eigenvalues Variance Eigenvalues Variance Factor 1 3.01 15.0 3.09 15.4 Factor 2 2.00 9.8 2.09 10.4 Factor 3 1.76 8.8 1.90 9.5 Factor 4 1.40 7.0 1.36 6.8 Factor 5 1.32 6.6 1.23 6.2 Factor 6 — — 1.01 5.1 Notes:a Extraction method is Maximum Likelihood. -247-women and men. In each case the first common factor accounts for 15 per cent of the variance found among the indexes. Although much of the variance remains unaccounted for, a common ideological underpirining cannot be discounted. The index that loads most highly for both genders is the capitalism index: its importance and relevance to the ideological debate lends support to the idea that Canadians do rely at some level on ideology as a means of making sense of political questions. Moreover many of the remaining indexes that load significantly on this common factor are issues which have driven much of the ideological debate in Canada: the role of the government in the economy, the social support of groups and organizations, relations with the United States, the position of unions, defence policy, and the health care debate over extra billing. Of particular interest is the fact that the feminism variable loads significantly on this first common factor for both women and men. Although this was anticipated given its significance in many of the indexes in the regression stage of the analysis, it does suggest that feminism has played itself out as an ideological debate linked with other political debates. One index loads on this first factor among women but not men: the anti-civil liberties index which might be interpreted as evidence that community concerns are more fundamental to women than men given its failure to load on the first factor in that sample. Upon rotation, however, attitudes on civil liberties fell within a grouping of indexes that was nevertheless identical between women and men. The anti-dualism index loads significantly on the common factor for men but not for women. For men, opinion on language policy is very much tied up with other -248-ideological considerations, particularly economic ones. For women, however, attitudes on language policy are not connected with core ideological beliefs, and form a separate set of considerations from economic ones. This difference did not sort itself out upon rotation. The connection of opinion on Canada's dual nature to other opinions is quite different for women and men. The rotation stage of the analysis provides the most relevant information in that it attempts to simplify the relationship uncovered between the variables. Simplification in factor analysis refers to a solution in which variables load significantly on just one factor and in which each factor has a number of variables that load on it but for which the remaining variables have close to zero loadings. In order to interpret the factors, each index is included in that factor upon which it has the highest factor loading but only if that factor loading reaches the minimum acceptable level of .30. Tables 7.2 and 7.3 present the results of the rotation stage and for ease of interpretation the loadings also appear in descending order of strength by factor. The women's factors are labelled 'Economic Liberalism', 'Secularism', 'Social Tolerance', 'Environmentalism', and 'Internationalism'. In the women's sample every index meets the cut off point and the only variable to load significantly on more than one factor is the continentalism index. This suggests that women were likely to see the question of ties with the United States as both an issue of economics and nationalism. The absence of further multiple loadings suggests that a relatively simple structure exists among their opinions. 'Simple' in this context applies as a criterion in the factor analytical sense and not as a more general evaluative assessment. Simple implies a lower -249-Table 7.2: Factor Analysis Results: Women's Pattern Matrix (Oblimin Rotation) Economic Social Inter-Liberalism Secularism Tolerance Environment Nationalism Anti-State Involvement .63 -.13 .161 .05 -.05 Anti-Social Spending .54 .17 • 19 .03 -.04 Capitalism .52 .03 -.25 -.13 .23 Continentalism .48 -.05 -.22 .22 .31 Extra Billing .46 -.05 -.11 .06 -:o8 Anti-Fenrinism .39 16 .15 -.14 +.04 Anti-Union .37 •07 .08 -.07 .11 Pro-Life .01 .61 -.05 .01 -.11 Religiosity -.13 .52 .13 .15 .09 Anti-Civil Liberties .11 .46 -.08 -.09 .15 Anti-School Funding .17 -.44 .29 -.26 .06 Traditional Roles .13 .40 .11 -.15 -.03 Anti-Immigration -.01 .07 .47 •20 .23 Cynicism -.03 -.07 .44 -.04 -.14 Anti-Dualism .12 .02 .39 -.27 •08 Risk Aversion .05 .05 .38 15 .03 Low Environmental Priority .04 :03 .17 .52 -.07 Defence .13 .01 -.05 -.06 .58 British Ties -.10 .05 .12 -.17 .37 Energy Investment .03 -.04 -.03 .24 .32 Sum of Squares (Variance Expl lined) 2.34(11.7%) 1.31(6.5%) 1.07(5.4%) .68 (3.4%) .61(3.0%) Factor Intercorrelations Economic Liberalism 1.000 Secularism .073 1.000 Tolerance .095 .107 1.000 Environment -.068 .016 -.083 1.000 Internationalism .246 .159 .037 -.011 1.000 Table 7.3: Factor Analysis Results: Men's Pattern Matrix (Oblimin Rotation) Inter- Identity • Economic Liberalism Secularism Immigration Nationalism Environment Politics Extra Billing .49 -.03 -.04 -.02 -.00 -.02 Capitalism .49 -.00 -.14 .43 -.12 .16 Anti-Social Spending .48 .08 .41 -.12 -.04 .14 Anti-State Involvement .45 -.12 .04 -.05 .22 .29 Continentalism .39 -.12 -.10 .30 .32 .10 Anti-Union .25 .10 .02 .20 .10 .22 Religiosity -.03 .68 .24 .05 .03 -.19 Pro-Life -.08 .63 -.10 -.06 .04 .08 Anti-Separate School Funding .15 -.41 .28 -.12 -.18 .02 Anti-Civil Liberties .20 .38 .06 .11 -.31 .01 Traditional Roles .08 .33 .06 -.20 -.05 .26 Anti-Immigration -.01 -.02 .57 .10 .08 -.06 Risk Aversion -.06 .07 .38 .01 .04 .18 Cynicism -.14 -.03 .29 -.52 .12 • 13 Defence .06 -.06 .17 .44 .06 .09 British Ties -.14 .04 .17 .32 -.15 .03 Anti-Feminism .00 .00 -.06 .09 -.00 .63 Anti-Dualism .08 -.10 .31 -.07 -.18 .39 Energy Investment ,00 .02 -.01 .27 .14 .02 Low Environmental Priority .07 .08 .16 .01 .29 -.04 Sum of Squares (Variance Explained) 2.49 (12.4%) 1.49 (7.5%) 1.30 (6.5%) .63 (3.2%) .50 (2.5%) 33(17%) Factor Intercorrelations Economic Liberalism 1.000 Secularism -.055 1.000 Immigration .059 .082 1.000 Internationalism .159 .074 -.009 1.000 Environment .085 -.012 -.022 .069 1.000 Identity Politics .391 .071 .256 .044 -.028 1.000 level of constraint among the factors in the belief system. Factor inter-correlations suggest this might be the case: the only significant bivariate correlation is that between the Economic Liberalism and Internationalism factors reflecting in part the significant loading of the continentalism index on the Internationalism factor. The men's factors are labelled 'Economic Liberalism', 'Secularism', 'Immigration', 'Internationalism', 'Envhonmentalism', and 'Identity Polities'. Three indexes fail to meet the minimally significant loading level on any factor: anti-union, energy investment and low environmental priority. The anti-union index just fails to meet the minimum criteria for significance in three of the factors: Economic Liberalism, Internationalism and Identity Politics. The greater the support a man gives to unions, the more willing he is to allow government involvement in the economy (i.e. more economically liberal); the less supportive he is of international ties and defence spending; and the more supportive he will be of political claims made on the basis of identity. Although the labour index fails to load significantly on a factor, its connection to several factors indicates that the issue connects to a number of fundamental beliefs, rather than to just one alone. That it loaded significantly on the first common factor suggests its overall relevance, rather than irrelevance, in men's belief systems. Unlike union support which failed to load significantly on any index, a number of indexes loaded significantly on more than one factor in the men's sample providing evidence that the structure of men's belief systems is more constrained than women's. First, the capitalism index loads on both the Economic Liberalism and Internationalism factors. Men who support capitalism are also likely to be less cynical about politicians : -252-and more supportive of British ties and defence spending. Second, the anti-social spending index loads on both the Economic Liberalism and Immigration factors. Weak support for social spending, on programs like welfare, unemployment and education is likely to coincide with weak support for increased immigration numbers. Third, the anti-dualism index loads on both the Immigration and Identity Politics factors. Support for official bilingualism is likely to coincide with support for increased immigration numbers. Fourth, the anti-civil liberties index loads on both the Secularism and Environment factors. Support for the protection from government interference on civil liberties issues is connected with weak religious conviction and heightened environmental concern. Finally, and similar to women, men respond to the issue of continentalism in a number of ways: it is at the same time an economic issue, an international issue, and also a post-materialism issue in that it loads onto the environment factor (reflecting perhaps the cultural side of the debate). The greater complexity in men's belief systems is suggested by the factor correlation matrix as well: moderate correlations are found between both the , Economic Liberalism and Inrmigration factors and the Identity Politics factor. The conclusion that women's belief systems are less constrained than men's is not borne out by a comparison of the mean inter-factor correlations. The higher the mean inter-factor correlations, the tighter the constraint that exists among the factors. The mean inter-factor correlations for the women's matrix is xr=0.0895 and xr=0.0944 for the men. There is little reason to conclude on this basis that women's belief systems are less constrained than men's. The only suggestion that this might be the case is the finding of multiple loadings for several indexes in the men's pattern matrix. As such, there is less -253- . than ovemhelming evidence of differences in constraint. The patterns that emerge upon rotation of the factors reveal that much of women's and men's political thinking occurs along similar lines. Although only one of the factors is identical in composition between them, many of the remaining factors differ only slightly in their composite indexes. Moreover, many of the grand political debates that have taken place in Canada are reflected in these patterns: women and men in Canada have taken part in these debates at the level of their opinions. There is in fact little reason to believe a priori that women's and men's belief systems would show much deviation in that both are responding to the dominant political world. If, as Zaller (1992) suggests, opinions are likely to fluctuate with the salience accorded individual issues, and because the forum within which such issues are debated is similar for men and women, then there is little reason to assume substantial deviation in the connections women and men make between issues. The factor which accounts for the greatest share of the variance among the indexes for both women and men is the Economic Liberalism index. The Economic Liberalism factor essentially taps into the debate regarding the market system and the place of the individual in that system. Within the framework developed by Christian and Campbell, the Economic Liberalism factor reflects the business liberalism strain of Canadian ideological debate.5 For both men and women, opinions on the role of business in the marketplace, competition and private enterprise go along with opinions about the See Christian and Campbell (1983). -254-privatization of crown corporations, the size and strength of trade unions6 and extra-billing by doctors and hospitals. There are, nevertheless, slight differences in the relative loadings of the component indexes between women and men. For women, state involvement in the economy and anti-social spending sentiment drive opinion on this set of issues. For men, there is little differentiation among the strongest four loadings on the group of opinions. The Economic Liberalism factor also includes opinion on continentalism and free trade with the United States. It appears that for most Canadians the heightened debate that occurred during the 1988 election campaign concerning the future of the Canadian state as an independent economic and cultural nation did not take place in a vacuum. The issue appears to have been placed within a larger context of government regulation of the market place and support for capitalism itself. A rejection of the basic tenets of business liberalism corresponds with a rejection of increasing economic and other ties with the US, and with an acceptance of economic nationalism. While the Economic Liberalism factor reflects a general pattern among Canadian men and women, on one of its component issues the two groups deviate. Among women the feminism measure loads significantly on the Economic Liberalism factor. The anti-feminism index does not load on this factor among men. Thus for women feminist belief is rooted centrally among their most fundamental economic considerations: to support government efforts for women goes hand in hand with government support in other areas. 6 Although the labour index failed to load on the Economic liberalism index in the men's analysis, its highest loading was on this factor. -255-On the other hand women who reject government efforts towards women are likely to be less supportive of government's involvement in the economy generally. For men, however, central beliefs regarding government support in the economy do not extend to feminist concerns; instead they form a separate, less.central, set of concerns, as discussed below. And the connection between feminism and continentalism (including free trade with the US) is clear: to support feminism is to reject continentalism. The centrality of feminist belief to women's opinions and not to men's is made clear by these results. Thus the core ideological belief, Economic Liberalism, explains the largest share of the pattern in both women's and men's opinions. Only among women, however, does this pattern include feminist belief. A second factor, Secularism, is driven largely by religious belief and includes policy issues that have a direct connection to those beliefs. This factor is the only one that is common to both women and men. That abortion opinion loads on this factor among both groups suggests its connection to both religious belief and attitudes on civil liberties. The public funding of private schools issue also loads on the Secularism factor rather than the factor addressing the degree of state economic intervention. This issue, which has a long history in Canada, is fundamentally regarded as a religious question. To be religiously fundamental corresponds with support for the public funding.of separate schools regardless of one's beliefs concerning the appropriateness of government intervention in the economy. The anti-civil liberties index also loads significantly on this factor. The inclusion of this index, which ties into such issues as homosexual rights and , • . '-256- • publication bans, suggests that moral imperatives rather than rights issues are most important for assessing opinion on individual liberties. Stronger religious fundamentalism corresponds to greater support for state restrictions on civil liberties, as a means of protecting social values and morals. Accepting the need for achieving equality for women is also linked to secularism: acceptance of women's equality, as tapped by our traditional roles index, requires a low level of religious belief. This mirrors the women's movement argument against organized religion because of its unequal treatment of women and the restrictions it imposes on women. The third common factor, labelled Internationalism, taps into those issues that reflect Canada's ties to the world and other nations within it. Canada's position in NATO and defence spending as components of the defence index are linked together with the question of Canada's ties to Britain. The factor suggests an understanding of international political questions: to argue for maintaining ties with Britain implies an opinion that rests not only with the historical importance of those ties, but also with the international consequences of severing those ties. Why the energy investment index loads significantly on this index is unclear. But for men, concern regarding defence spending and British ties is also connected to levels of cynicism: in order to endorse such policies men must believe that governments and politicians are trustworthy, honest, and capable. Among both groups, the low environmental priority index does not 'fit in' easily with the other issues included in the analysis. For women the environment index is the only variable to load significantly on the Environment factor. It is unwise to label factors -257-with so few components, and less wise to take them at face value, but at the very least its exclusion does suggest that opinion on environment issues is managed independently of other issues. This makes intuitive sense: environmentalconcern has existed at one level outside the debates on government involvement in the economy. However, the Environment factor is less straightforward for men. The low environmental priority index itself just fails to meet the criterion of significance on the factor. Two additional indexes, continentalism and anti-civil liberties load significantly on the factor but their strongest loadings are on other factors. It is not clear what to make of this pattern. In the remaining factors minor differences in women's and men's belief systems appear. For instance, both women and men link support for an open inunigration policy with a willingness to attempt reform in society. This combination completes the fourth Immigration factor among the men's sample. For women, however, in addition to these two indexes the factor includes the dualism and cynicism indexes. That women's opinion on French-English relations in Canada combines with the anti-immigration, cynicism and risk aversion indexes is quite telling. It suggests that support for bilingualism stems in part from an acceptance of diversity and pluralism generally as well as faith and trust in government. Social reformers tend to agree with government social policies designed to overcome perceived social concerns: immigration policies should be relaxed to allow for greater numbers of immigrants, and official bilingualism is an acceptable method of reconciling Canada's two solitudes. Social conservatives, on the other hand, are less likely to endorse that government adopt such social roles. Labelling*this factor 'Social Tolerance' is an explicit -258-attempt to highlight this fact.7 For men, however, language policy appears more clearly to be an issue of identity, or group, politics. The anti-dualism index loads with the feminism variable on the sixth 'Identity politics' factor. For men, the language debate appears to have been understood in terms of groups demands on the state. By the same token, feminist sentiment was also largely one of identity politics: to support feminism meant that one was also likely to support policies designed to protect the French language in Canada: It appears that the question of bilingualism and Quebec's place in Canada is perceived to be one of support: 'subsidizing' the French language is accepted alongside of support for another group, namely women. In contrast, for women feniinist sentiment was very much tied up with core beliefs on capitalism and government's role in the economy generally: it was not seen as an issue of group demands but rather as a constituent element of economic liberalism. To further investigate the nature of constraint among women's and men's belief systems, the factor analysis procedure was undertaken using controls for political sophistication. Converse (1964) suggests that the degree of constraint among opinions increases with political sophistication. Education is employed here as a proxy for political sophistication: those respondents having some post-secondary education are considered to be more politically sophisticated than those without. Table 7.4 and 7.5 71 struggled with the labelling of this factor. It did not seem to be an immigration factor, per se, given its additional loadings. The use of the 'tolerance' label is not meant to imply that women have a greater or lesser level of tolerance than men; it merely taps into the relevance of social reformism to this factor. -259-Table 7.4: Factor Analysis Results: Women with College/University Experience Pattern Matrix (Qblimin Rotation) Economic Liberalism Secularism Group Support Internationalism Immigration Energy Anti-State Involvement .66 -.09 .29 .10 .04 .01 Capitalism .61 -.00 -.14 -.22 .06 -.11 Continentalism .59 .02 -.16 -.09 .03 .15 Anti-Social Spending .49 .09 .01 -.04 -.25 -.03 Extra Billing .47 .02 -.06 .08 .10 .11 Anti-Feminism .38 .07 .08 -.05 -.23 -.25 Anti-Union .33 .04 .16 -.03 -.18 -.18 Religiosity -.02 .70 .14 -.01 .04 .11 Pro-Life .00. .63 -.08 : .10 -.01 -.13 Anti-Civil Liberties .10 .40 -.03 -.24 .04 -.01 Cynicism -.08 .03 .56 .15 -.08 .04 Anti-Dualism .02 .04 .50 -.27 .02 -.06 Anti-Separate School Funding .13 -.33 .41 -.09 .08 -.11 Defence • 16 -.05 -.12 -.56 -.11 .00 British Ties -.07 06 .06 -.47 -.00 .06 Risk Aversion -.11 -.07 .00 -.06 -.69 -.07 Anti-Immigration .05 .05 .23 -.18 -.39 .16 Energy Investment .12 .06 .02 -.22 ' -.09 .50 Low Environmental Priority .05 .09 -.06 .16 -.26 .19 Traditional Roles .02 .21 .04 -.11 -.10 -.29 Sum of Squares : 2.77 1.36 1.23 .61 .57 .48 Variance Explained 13.9% 6.8% 6.1% 3.1% 2.9% 2.4% Factor Intercorrelations Economic Liberalism < 1.000 Secularism ! ,111 1.000 Group Support [ .052 -.097 1.000 Internationalism ; -.266 -.165 -.134 1 1.000 Immigration -.166 -.297 -.202 : .138 1.000 Energy .023 .039 -.175 ; .108 -.000 1.00 i i Table 7.5: Factor Analysis Results: Men With College/University Experience Pattern Matrix (Oblimin Rotation) Government Involvement Secularism Capitalism Tolerance Social Concerns Civil Liberties Anti-State Involvement .72 -.09 .10 .00 .14 .00 Anti-Feminism .55 .09 .00 .07 -.12 -.10 Continentalism .50 -.07 -.20 -.00 .32 .28 Anti-Dualism .48 -.05 .16 •15 -.39 -.11 Anti-Union .48 .16 -.14 .09 .07 .08 Extra Billing .36 -.08 -.14 -.08 .15 -.14 Pro-Life .14 .77 .14 -.12 -.06 .08 Religiosity -.23 .65 -.05 .21 .04 -.13 Anti-Separate School Funding .13 -.39 .16 .16 -.12 -.05 Traditional Roles .18 .30 .08 .04 -.01 -.25 Cynicism .07 .03 .64 .13 .00 -.00 Capitalism .49 -.10 -.59 -.05 -.03 -.11 Anti-Immigration . -.10 -.09 .06 .68 .07 .02 Anti-Social Spending .36 -.05 .01 .44 .19 -.42 Risk Aversion .08 .05 .04 .34 -.07 -.06 Defence .21 .03 -.27 .28 •01 .20 Low Environmental Priority .09 .03 -.02 .18 .35 .04 British Ties -.03 .02 -.22 ' 19 -.32 .16 Anti-Civil Liberties .00 .25 -.24 16 -.08 -.26 Energy Investment .05 .08 -.21 .06 .04 .12 Sum of Squares 3.08 1.55 1.34 .70 .53 .34 Variance Explained 15.4% 7.8% 6.7% 3.5% 2.7% 1.7%. Factor Intercorrelations Government Involvement 1.000 Secularism -.014 1.000 Capitalism -.184 -.173 1.000 Tolerance .294 .132 -.094 1.000 Social Concerns .130 -.033 -.122 -.102 1.000 Civil Liberties -.185 -.125 -.234 -.086 .092 1.000 provide the results of the rotation stage of this analysis for women and men respectively.8 A number of interesting results emerge from this analysis. First, the mean inter-factor correlation rises for both women and men: among women with some college/university experience the mean correlation is xr=0.132 and among men with similar educational experience it is xr=0.133. This increase suggests two things. First, women and men with greater political sophistication connect their opinions to a greater degree than do those without. Thus tighter constraint is associated with greater sophistication for both women and men. Second, six factors emerge from the analysis of the opinions of women with some post-secondary education. The pattern that emerged for all women included only five factors. Despite this change, however, the first factor to emerge in this second analysis comprises the same indexes that were included in the first factor for the full sample of women. Changes are only to be found in the indexes that compose the remaining factors. This suggests that as women's political sophistication increases they make a greater number of distinctions among issues and thus connect them into more distinct groupings. Third, and of greatest interest, among men with some college and/or university experience the first factor that emerges upon rotation includes the anti-feminism index. This was not the case when all men in the survey were submitted to the factor analysis procedure. Thus if feminism has ideological undeipinnings among men, this connection is only likely to be made among men who are 8 Although the analysis was also undertaken for women and men with no college and/or university experience, low correlations of some of the indexes to the remaining indexes in the male sample made for an unstable pattern matrix, and so the results of this test are not included here. -262-? • . ; more politically sophisticated, or at the very least with the higher degree of tolerance associated with higher education. Based on his findings, Converse (1964) suggested that the alignment of group membership and political behaviour is sharpest among the most politically sophisticated and fades as the level of political sophistication decreases. These findings suggest, however, that for women the relevance of membership with women as a group (women qua women) does not depend on political sophistication. For women feminism forms a core element of belief irrespective of their levels ofpolitical sophistication. If feminism is part of women's ideologies, then, this is an ideology that stems from something other than political sophistication and tolerance. An argument can be made that it is an ideology that stems from the personal, rather than the political. Women's belief systems are in essence extensions of their experiences as women. Conclusion The investigation in this chapter attempted to address two main questions. The first concerned the degree of ideological structure that exists in the attitudes of Canadian women and men. The second focused on the degree to which these structures, if substantial, diverged between these same two groups. In the case of the first question, one can conclude with a degree of confidence that there does exist an underlying pattern of belief in those attitudes included in the factor -263-analysis procedure.9 The measured factor patterns account for nearly 50 percent of the variance across all the indexes submitted for analysis in both the men's and women's samples. And in the rotated factor matrix a number of substantive factors emerge that correspond with a number of basic Canadian political issues and concerns: the proper role for government in the economy and as a protector of moral standards; religious beliefs as they connect with various political issues; multiculturalism, at least at the level of immigration policy; international concerns on the importance of relations with other countries; environmental concerns; and the identity basis of politics. It would appear that Canadian women and men do have belief systems driven in part by ideological concerns. The second question concerned distinctions that might appear in comparisons of women's and men's belief systems. The evidence presented here suggests that there is little divergence between the belief systems of Canadian women and men. Differences are apparent but these differences are in the detail rather than the substance of women's and men's belief systems. The basic core beliefs guiding women and men overlap to a great extent. The divergence exists in the connections of lower order opinions to these core beliefs. In the first instance there is little to suggest that there is a difference in the level of constraint exhibited in their belief systems: women and men appear to connect their opinions more or less in the same manner. There is little evidence that Gilligan's claim of a different voice extends to women's and men's belief systems. A greater ethic of care 9 One must concede that the factor analysis procedure is dependent on the mix of variables entered into the analysis. -264-and the heightened relevance of community concerns did not appear among patterns in their opinions. It may well be that the dominance of the male political culture restricts the degree to which such differences might appear in belief systems in the first instance. The connections individuals make between issues are likely to be guided bythe dominant political discourse. This discourse reflects the dominant political culture, the male culture, and as such very little of a women's political culture is likely to appear among those connections. The most significant difference that exists in women's and men's belief systems is the disparate positioning of the feminism measure among the factors. This provides substantive support for concluding that the measure taps into different considerations for women and men. At most, the belief systems' evidence suggests that feminism is a fundamental concern for women: it is an element of general liberal belief. For men, however, feminism exists as an element of general liberal belief only among the politically sophisticated. In the absence of control for level of political sophistication, feminism for men is reduced to one of a number of demands that groups make on governments on the basis of their identity. As an element of their belief systems, the issue is both less fundamental and less personal. Thus thinking about women's feminism as something other than an ideology because it is rooted in the personal is incorrect. Feminism is part of women's ideology, rooted in their gender consciousness, and connected to core beliefs. For men, on the other hand, feminism as ideology is dependent on social context: it only becomes part of a fundamental group of beliefs with political sophistication. -265-Chapter 8: Gender Difference as Cultural Difference That women and men are different would come as no surprise to students of biology. That women and men hold different views on political issues might not astonish the proverbial person on the street. However, such differences have not been accepted by a large number of political scientists who argue that gender differences are small and that such differences are essentially artifacts of differences in social or economic status. This study rejects that position and has demonstrated differences between women and men both in their political attitudes and in the manner in which their attitudes are constructed. Moreover, it provides strong evidence that gender differences are not eliminated by considerations of socio-economic status. Using suryey evidence gathered at the time of the 1988 federal election, this study has attempted to take some steps towards investigating the existence of dual political cultures in Canada, a women's political culture and a men's political culture. This chapter will review the findings gathered throughout this study in an effort to elaborate on these elements of women's political culture. The evidence suggests that the patterning of women's and men's opinions and distinctions in the sources of those opinions point to distinctive political cultures. Three separate investigative paths were followed in the attempt to discover the existence of distinctive political cultures. In the first part of this chapter the findings of each investigative path are summarized and linked to the distinctive culture thesis. The -266-second part reflects on the potential sources of this women's political culture. In the final part, suggestions for future research in women's political opinion studies are made, based in part on the limitations encountered in this study, and on the need for gathering longitudinal evidence on the existence of this culture. A Women's Political Culture Political culture is deemed worthy of study because it is assumed to have a direct impact on the means and ends of politics (Stewart, 1994). And political values and opinions constitute a key element of political culture. According to Inglehart, "culture is a system of attitudes, values and knowledge that is widely shared within a society and transmitted from generation to generation." (1990:18). Values constitute preferences and priorities. They are held for their own sake, and provide a means forjudging human activity. They are yardsticks by which activity and choices are measured. Opinions, on the other hand, are expressions of attitudes and beliefs. They are also a reflection of values. The particular set of opinions and values held by women constitute a women's culture. Women are at the same time more religious and more feminist than men. This competition of values is important for the development of this culture, to be addressed at greater length below, but it is not the sole component of a particular women's culture. This culture is reflected in attitudinal differences which persist despite controls for socio-demographic differences and the liberalizing impact of feminism. This cultural difference is also found in the influence of various social and economic factors that shape attitudes -267-generally, even on issues that do not reveal gender differences in support, and it is found in the importance of gender in the shaping of women's belief systems. In an effort to explain this cultural difference one may look to biological or physiological differences. In terms of a women's political culture based on their distinct values and opinions, Bashevkin (1993) suggests just such a possibility based on women's and men's differences in opinion on political interest, efficacy, party support and selected political issues. Her account, however, was brief in its investigation of women's attitudes. This study builds on this earlier work and continues gathering evidence on the existence of a women's political culture. The opinions of women and men surveyed in the 1988 CNES suggest an emphasis that differentiates women from men. Consistent with a number of earlier studies on gender and opinions, women are found to be less 'hawkish' in their support of defence spending and membership in international defence organizations. This difference in belief is perhaps the most longstanding and the most persistent of those found between women's and men's opinions. Evidence is also provided that women exhibit opinions that conform to a broad-gauged humanitarianism, or 'agape ethos' (Bernard, 1981: Chapter 21). Their weaker conservatism on a number of issues suggests a greater concern for the well-being of individuals than is generally true of men: women are less supportive of the capitalist system in general; they endorse a greater degree of government involvement in the economy; they are less supportive of increasing economic ties with the United States; and once the influence of religious belief is removed, women are more egalitarian in then -268-support for the protection of civil liberties. This gender gap in opinions is comparable to many of the gaps on these issues found at different times and in different countries. It suggests that Gilligan's insight that moral decision-making in women constitutes an ethic of care may have validity with respect to their political opinions as well. According to Gilligan, The morality of rights is predicated on equality and centred on the understanding of fairness, while the ethic of responsibility relies on the concept of equity, the recognition of differences in need. While the ethics of rights is a manifestation of equal respect, balancing the claims of other and self, the ethic of responsibility rests on an understanding that gives rise to compassion and care. (1982: 164-165) This ethic of responsibility, rooted in women's psychological development, seems to pervade women's political opinions at a certain level, despite the overall paucity of women's participation in formal politics and women's relative inability to see their values embodied in that political arena. The differences in opinion extant, regardless of their size, provide evidence of this particular focus or ethic. On the remaining, often less generally measured indexes, the evidence of gender gaps in opinion varies. No significant differences are found in defence of official bilingualism. And stereotypes notwithstanding, women are not more reformist than men. Neither are women found to be any more cynical of politicians and governments. Women are, however, more likely to support equality for women, both in their career choices and in government support of women as a group. Intriguingly, in one of the strongest attitudinal differences recorded, men are found to be less likely to support efforts to maintain ties with Britain. Women it seems are -269-more likely to be Monarchist than men. These internationalist ties do not, however, extend to a support of increased immigration nor to support of continentalism. Women are no more likely to endorse policies whose aim is to increase the number of immigrants coming into Canada and are less likely, by a fairly wider margin, to endorse increased ties with the United States. The pattern that exists among these various issues suggests that women are generally more liberal than men. What might appear to be the existence of gaps on more broadly worded questions (e.g. government involvement versus extra billing) is in many cases simply an artifact of the composition and labelling of indexes. The only index composed of questions that are very general in their wording is the capitalism index. The remaining indexes contain precisely worded questions. Thus it seems unlikely that greater precision in the wording of survey questions, for instance in the likely beneficiaries of a particular policy change or program, is responsible for the absence of gaps on some issues. On the separate question of differences in core values, religious belief and feminism, gaps are also found to exist between women and men. Women are more conservative in their religious belief; in other words, women are more religiously committed. This greater religious conservatism has been reported elsewhere (Conover 1988; Poole and Zeigler, 1985). The central role of religion and the Church in female culture includes its use as an arena for commumty participation, an avenue for community 'good works' including visiting the sick and caring for the poor, and, given the role of the family in teaching faith to children, women's role as mother insures in part their commitment to religion and religious faith (Bernard, 1981: 496-497). Thus religion ' . • -270- . . . plays a greater role in many women's daily lives, and in turn, a greater role in the determination of their values and hence political opinions. As an example, the initial impression that men were significantly more liberal in their support of the availability of abortion services disappeared when the impact of religious belief was removed. Women are also more likely to support feminism than are men, although by a smaller margin than that found in religious belief. The limited difference in support of feminism may stem partially from the limitations of the measure employed in this study. The difference in both the social structure of support for feminism and its position relative to other opinions in belief systems suggests that despite such limitations, some of the difference in this core value is picked up. There is little that is new in this finding. The suggestion that feminism in men and women is different has been made by a number of researchers (Klein, 1984; Fulenwider, 1980). The debate has centred on the distinct roots of women's and men's feminism. Klein argues that women's feminism is rooted in their gender consciousness and self-interest but that feminist belief for men is part of a more general ideological framework. The evidence presented in this study contradicts this hypothesis. It suggests, in support of Fulenwider's findings, that feminism is centrally rooted among women's attitudes but that it is less dependent on self-interest and social status than is the case for men. It appears to be rooted in women's existence as women. Feminism forms a central element of many women's political ideology. For men, support of feminism is dependent on self-interest and social status; it is very much shaped by then social circumstances. Feminism at times forms an element of men's ideology but generally only for men who are more politically sophisticated. Feminism -271-becomes part of men's world view only when self-interest becomes a secondary concern. What then are the most relevant findings in the gender gap stage of the analysis? First, the existence of gender gaps on issues that have consistently been found to be gap issues suggests that the phenomena of gender gaps is not fleeting. Their persistence discounts the possibility that changes in women's situational experiences, e.g. increasing educational and income levels and workplace activity, which might be hypothesized to bring their attitudes closer to those of men, have brought about significant changes in their attitudes. Thus, like the results found by Bashevkin (1993), women's political integration seems unlikely to significantly reduce the size and significance of gender gaps. Second, the inability to explain away the liberal gender gap through social and economic factors provides support for the existence of a distinct women's political culture. The limited size of these difference should not be taken as grounds for discounting the importance of that distinction since the fact that women and men share many of the same experiences alone is likely to ensure a correspondence in their opinions. The fact that any difference exists at all suggests that unique experiences or a particular viewpoint contribute something distinctive to each gender's opinions. The ability to conclude with some measure of confidence that divergent cultures exist between women and men rests partially on the ability to discount other potential sources of attitudinal difference. That a number of competing hypotheses could not fully account for gender gaps in opinions increases the likelihood that gender is a fundamental consideration for opinions and for their structure more generally. -272- . • • Part of the gap story is to be found in the socio-economic differences that exist between women and men. Women's lower incomes and their lower levels of education partially explain their greater support of government involvement in the economy. Obviously, it is a rational response to their economic situations. Another part of the gender gap story lies in the differential support for and influence of feminism on women's opinions. Generational differences in opinion also form part of the story: many gaps across the youngest and oldest women in the survey were larger than those found across similar age differences in men. These generational differences across women help explain their unique political thinking. The finding of distinctions in the social structure of women's and men's opinions lends further support to the existence of gendered cultures. The results suggest that there is no one general answer to the question of what fundamental difference drives women's and men's opinions in such a way as to result in gaps on some opinions and a similarity of opinion on others. Nevertheless the evaluation was not undertaken for naught. Some preliminary conclusions can be drawn from the findings. First, it seems fairly certain that women's adult gender roles are not the primary factor in the formation of their opinions. Being a homemaker appears to have very little influence on women's opinions. The only case in which homemaker's opinions were found to differ from those of other women was in the case of attitudes on feminism and women's equality. On these issues homemakers are more conservative in their attitudes than other women. The relevance of gender role to these issues seems clear. Yet net of all other effects, these differences between homemakers and other women are small. -273-That homemakers are significantly more religious than all other women links gender role socialization and religious belief in the determination of women's attitudes and may partially account for the absence of gaps on some issues. Religious belief is in fact a key element in the shaping of women's attitudes that has been overlooked in previous research on women's opinions. While early research on differences in men's and women's political attitudes suggested that women's greater social isolation accounts in part for their more 'parochial outlooks', the evidence here suggests that despite their isolation, homemakers are not different from other women in many of their opinions (see Bashevkin, 1993:41-42). Religion does, however, account in part for the unique structuring of women's attitudes. Women are more religious than men due only partially to their lower average incomes, their lower educational attainment and age differences between women and men: after controlling for these factors a significant gap in religious belief remains. This 'traditionalism' is not then simply an artifact of women's weaker socio-economic status; it rests in part on nothing less than gender. This stronger religious belief is connected with a number of factors among women; in particular, there exists a positive association between religious commitment and the number of children in the home and homemakers are significantly more religiously committed than other women. Thus absent of the impact of religion, neither homemaker status nor children in the home have significant connections to women's opinions. Religious belief appears to shape women's political opinions to a greater extent than gender role socialization or the care of children. Second, this traditionalism is very much offset by the importance of feminism in the -274-deterrmnation of many of women's opinions. As a liberal source of opMon, feniinism goes directly against the traditionalism embodied in the stereotype of the homemaker: women should be free to choose their roles. And feminism accounts in part for the greater liberalism exhibited in many more of women's opinions. The evidence suggested that the relevance of feminism in the determination of opinion depended in part on the connection between the issue and women's concerns as a group; as the links between women's fortunes and the particular issue at hand increased, so too did the relevance of feminism to women's opinions. The findings in this study leave little doubt as to the necessity of further investigation of feminism's role in the shaping of women's opinions. A difference in culture is also suggested by the evidence gathered on a number of other factors responsible for shaping women's and men's opinions. Self-interest seems to have a differential impact on opinion on some issues; on economic issues, self-interest calculations appear to matter more for women, while on non-economic issues income seems more relevant as a determinant of men's opinions. In addition, the generational gap that exists among women on many issues is often greater than that found among the men; this hints, as will be discussed further on, at the possibility of a culture in transition. Also intriguing is the difference found in regional variation in opinion between women and men. Regional variation in opinion was usually greater among men than among women. This undoubtedly reflects women's and men's patterns of membership in social organizations; these organizations, particularly political organizations, are the vehicles that allow for regional political cultures to thrive. Differences in women's memberships in such groups might explain the absence of regional patterns of opinion -275-among them. Although the social groups examined in this study hint at the differences that membership in such organizations bring to opinions, the lack of specificity in the group variables included in the CNES permitted only a preliminary investigation of this topic. Further research in this area is required. At the very least, social group memberships provide a vehicle for a women's culture to flourish outside of the traditional political arena, and as such both reflect and feed this difference. The argument that women's political attitudes and behaviour reflect a particular culture has been entertained by others, although it does not form a cornerstone of belief regarding women's political behaviour. Vickers (1988) provides an account of women's distinctive political culture based on women's accounts and views of their involvement in political life. According to Vickers, "A culture—a set of ways of thinking about the world—means nothing, however, outside the material circumstances of its birth and development and the social arrangements that keep it alive. [...].A people's culture reflects and guides the daily way of life." (1988: 25). Women's participation in political life, according to Vickers, takes place within dual spheres and with double vision. Formal politics, dominated by men, is based on conflict and the acquisition of power. That arena is dominated by men in part because they seek to 'get ahead', to share in that power. Many more women, however, choose to engage in local informal politics, as many, stated, 'because they wished to get things done.' Moreover, many women choose this less formal political arena because it is community-oriented and located close-to-home. The difference in emphasis underscores a difference of values. It also means that -276-women's involvement in a separate arena reinforces the holding of a particular set of values; there is a reciprocal relationship between values and external environment. To understand women's political thinking more fully, Vicker's emphasis needs to be expanded in order to get a better sense of how, where and why women engage in community activities. Correspondingly, women's involvement in the dominant political world needs greater investigation in order to determine the contribution of women's values and culture to that world. The development of social arrangements that allow a culture to thrive are as relevant as the study of the particular elements of that culture. The decision by many women to engage in political activity in a different arena and with a different set of guiding principles supports the notion of the existence of a particular ethic, or value. Women's distinctive political priorities may explain in part women's participation in organizations outside of dominant politics. This is not a difference in women's and men's concern for reform. Women and men appear no different on this front. But in addition to concern for 'getting things done' rather than 'getting ahead', women's political culture emphasizes in part those particular tasks which must be accomplished. Sources of a Women's Political Culture Some help in understanding women's political culture might come from some speculation on the potential sources of this particular culture. While biological determinism cannot be dismissed as a potential source of women's and men's distinctive cultures, it is difficult to dismiss the influence of socialization processes particular to each -277-gender. To cite Inglehart at length: The political culture approach is distinctive in arguing that (1) people's responses to their situations are shaped by subjective orientations, which vary cross-culturally and within subcultures; and (2) these variations in subjective orientations reflect differences in one's socialization experience, with early learning conditioning later learning, making the former more difficult to undo. (1990: 19). Thus a women's political culture is likely to be shaped by childhood learning experiences, which are then conditioned and/or reinforced by learning later in life. It is commonly accepted that the various agents of cultural learning, namely the family, schools, the media, and social and peer groups, provide a different set of cues to women and men. In each of these social institutions, the feminine is depicted as distinct from the masculine. Thus part of the understanding of difference is to be found in the manner in which society depict as proper masculine and feminine behaviour. Wherever one looks there are clear signs of what being a women means, at least as it is culturally defined. Schools provide subtle and not so subtle cues to girls and boys on then particular gender roles. Gilligan (1982) provides evidence that even as children, girls and boys play different games and respond differently to conflict within those games. The learning must then begin at a very young age. And a peek at employment patterns provides a clear indication of the distinctions made in societal definitions of women's work and men's work. The cultural definition of motherhood articulates a clear vision of the feminine role: neither dominant nor aggressive but emotional and sympathetic. This depiction of the feminine, I would argue, shapes every girl's and woman's life and in such a manner -278-that the actual bearing of children may make little difference to the adoption of this cultural definition. Women are in constant touch with society's understanding of the key role played by mothers. The three elements of the myth of motherhood, according to Oakley, are: first, children need mothers; second, mothers need children; and third, motherhood is the sole true means of self-realization (1974:186). Some women consider such myths to be inMbiting for women and thus in need of challenge; others argue that these are celebratory statements that elevate women's distinctive ethic of care. Such normative assessments are beyond the scope of this analysis. Whether these expectations should be celebrated or dismissed, women, today, exist in a society that provides clear signals on the importance of their role as potential mother. This may explain in part the distinctiveness in women's opinions that within this study is considered characteristic Of a particular political culture. Chapter 6 suggested that the presence of children in the home had little effect on women's opinions. One could suggest that this downplays the relevance of the myths of motherhood to women's opinion. The failure of the measure to tap into differences in women's opinions may stem from the measure itself: women who had children in the home during some earlier period would not have been tapped by the measure itself because at the time of the study there would have been no children in the home. This would have been true of many of the older respondents in the study. Thus the measure does not provide a valid distinction between women who have cared for children, and those who have not. Differences in thinking between these two groups would therefore not appear. -279-Alternatively, and part of the argument made here, maternal thinking need not depend on having actually cared for children. The present-day myths and culture that support a particular vision of woman as mother are strong enough for all women, not just mothers, to feel their effects. According to psychologist Shari Thurer, "A sentimentalized image of the perfect mother casts a long, guilt-inducing shadow over real mother's lives" (1994: xii). This shadow, I would argue, enters most women's lives. A Culture in Transition If women's political culture exists, it is a nascent culture and one in transition. The culture reflected in the attitudes of women shows a distinct pattern in belief: young women are generally more liberal than are older ones. According to Inglehart, "central parts of culture may be transformed, but they are much more apt to change through intergenerational population replacement than by the conversion of already socialized adults" (1990: 19). Women's greater egalitarianism, their weaker support of defence measures, their weaker support of continentalism, and their weaker endorsement of capitalism appear most strongly among younger women. Even on issues that revealed little gender difference, intergenerational differences in attitudes are still to be found among women, and often these differences are greater than those found among men. The impact of feminism on the attitudes of younger cohorts of women, quite apart from whether these women consider themselves to be feminist, might explain this shift in women's attitudes. In endorsing greater freedom for women, and in providing women with the confidence to speak in 'a different voice', the feminist movement has influenced -280-the attitudes of many women. But feminism does not wholly explain this shift; its failure to account fully for gender gaps in attitudes makes this clear. Part of this cultural shift among women is also to be explained by the intergenerational differences found in women's religious comnhtment. Younger women are less religious than older women. Although this difference is also apparent among men, the shift in beliefs among women allows for a greater liberalization in many of their attitudes. As such, the shift should result in larger gender gaps over time. The movement away from religion, combined with the salience of gender brought on in part through feminism, may very well result in the discovery of larger gaps in the future. Future Research Directions The ability to conclude with a degree of certainty that a women's culture exists is weakened by the fact that the data evaluated in this study are only from one point in time. Political cultures are assumed to be enduring and transmitted from generation to generation. The 1988 CNES does not provide a means of determining women's opinions prior to 1988, or a means of determining how such a culture is changing over time. The 'synchronic approach', using data collected at one point in time, ignores how opinions change with the passage of time. An alternative, and certainly one to be suggested for the study of a women's culture, is the 'diachronic or historical approach', employing surveys taken at different points in•-time or other kinds of data (Bell, 1992: 12). Surveys can provide little help in determining the existence of a women's political culture prior to the 1960s; surveys have only since then become part of the tools of -281-political science. In order to investigate the existence of differences prior to the 1960s, researchers have to look for visible signs of a women's culture in language, symbols, myths, and discourse. Much of this work is currently underway, and some already exists. Liriking it to women's political culture, and accounting for changes in that culture is a step yet to be taken. But the visibility of culture in women's symbols, language, myths and discourse should not be the exclusive purview of historical research; these elements of culture should be very much part of current investigations as well. Such investigations, combined with surveys designed specifically to address women's political opinions, would bring us closer to understanding the existence of a women's political culture. Qualitative research would bring us a step closer to deterrnining women's political priorities; such research would also help in deterrnining the questions which need to be asked on political surveys. Investigations of women's participation in organizations not traditionally considered to be engaged in politics should also be undertaken to assess their part in the shaping of women's particular value frameworks. And it seems clear that more work needs to be done on women and their patterns of integration in local political communities. This study has shown that women's engagement in some organizations, namely service and business groups, can bring women's opinions closer to those of men. But their involvement in other organizations, ethnic and issue groups for example, can have the very opposite effect on their opinions perhaps by providing vehicles for the legitimation of their particular priorities. Future research must investigate these patterns of participation and ascertain their role in the development of women's political culture. -282-Conclusion Women's political beliefs can only be characterized as parochial in the limited sense that changes in social and economic factors will bring those beliefs somewhat closer to those of men. Such changes will not, however, make gender irrelevant to political thinking. A women's political culture is distinctive in its political priorities, exists by virtue of gender alone, and continues to exist perhaps as a result of women's lesser participation in the dominant political arena. It seems certain that women's involvement in non-traditional political arenas allows it to flourish. The evidence is accumulating that women's political behaviour is not properly portrayed by traditional political science. This investigation has documented that the uniqueness of women's political thinking can be added to the many accounts of political behaviour. All that remains is to have this distinction recognized in the wider political science community. -283-Bibliography Abramson, Paul (1983). Political Attitudes in America. San Francisco: W.H. Freeman and Co. Abzug, Bella with Mim Kelber (1984). The Gender Gap. Boston, MA: Houghton Mifflin. Achen, Christopher (1975). "Mass Political Attitudes and the Survey Response." American Political Science Review 69:1218-31. Allen, Sr. Prudence (1987). "Sex Unity, Polarity, or Complimentarity" in Women and Men: Interdisciplinary Readings on Gender, edited by Greta Hofmann Nemiroff. Montreal: Fitzhenry and Whiteside. Armstrong, Pat and Hugh Armstrong (1994). The Double Ghetto: Canadian Women and Their Segregated Work. Toronto: McClelland and Stewart. Arscott, Jane and Linda Trimble, editors. (1997). In the Presence of Women: Representation in the Canadian Governments. Toronto: Harcourt, Brace. Bacchi, Carol (1986). "Women and Peace Through the Polls." Politics: Journal of the Australasian Political Studies Association 21:62-67. Bashevkin, Sylvia (1985). Toeing the Lines: Women and Party Politics in English Canada. Toronto: University of Toronto Press. • (1989a). "Free Trade and Canadian Feminism: the Case of the National Action Committee on the Status of Women." Canadian Public Policy 15:363-75. . (1989b). "Political Parties and the Representation of Women." In Canadian Parties in Transition, edited by Alain G. Gagnon and A. Brian Tanguay. Scarborough, ON: Nelson Canada. . (1991). "Women's Participation in Political Parties." In Women in Canadian Politics, edited by Kathy Megyery. Lortie Commission on Electoral Reform and Party Financing, vol.6. Toronto: Dundurn Press. -284-__ . (1993). Toeing the Lines: Women and Party Politics in English Canada. 2nd edition. Toronto: Oxford. Baxter, Sandra and Marjorie Lansing (1983). Women and Politics: The Visible Majority. Ann Arbor, MI: The University of Michigan Press. Bell, David V.J. (1992). The Roots of Disunity, revised edition. Toronto: Oxford University Press. Bernard, Jessie (1981). The Female World. New York: The Free Press. Black, Jerome H. and Nancy E. McGlen (1979). "Male-Female Political Involvement Differentials in Canada." Canadian Journal of Political Science 12:471-97. Black, Naomi (1988). "The Canadian Women's Movement: the Second Wave." In Changing Patterns: Women in Canada, edited by Sandra Burt, Lorraine Code, and Lindsay Dorney. Toronto: McClelland and Stewart. Blishen, Bernard, William Carroll, and Catherine Moore (1987). "The 1981 Socioeconomic Index For Occupations in Canada." Canadian Review of Sociology and Anthropology 24:465-85. Brodie, Janine (1985). Women and Politics in Canada. Toronto: McGraw-Hill Ryerson. . (1988). "The Gender Factor and National Leadership Conventions in Canada". In Party Democracy in Canada, edited by George Perlin. Scarborough, ON: Prentice Hall Canada. Brodie, Janine, with the assistance of Celia Chandler (1991). "Women and the Electoral Process in Canada." In Women in Canadian Politics, edited by Kathy Megyery. Lottie Commission on Electoral Reform and Party Financing, vol.6. Toronto: Dundurn Press. Campbell, A., P.E. Converse, W. Miller, and D. Stokes (1960). The American Voter. New York: John Wiley. Campbell, Kim (1996). Time and Chance: The Political Memoirs of Canada's First Prime Minister. Toronto: Doubleday Canada. Carroll, Susan J. (1988). "Women's Autonomy and the Gender Gap: 1980 and 1982." In The Politics of the Gender Gap, edited by Carol M. Mueller. Beverly Hills, CA: Sage Publications. Chodorow, Nancy (1978). The Reproduction of Mothering. Berkely, CA: University of -285-California Press. Christian, W. and C. Campbell (1983). Political Parties and Ideologies in Canada. Toronto: McGraw-Hill Ryerson. Citrin, Jack and Donald Philip Green (1990). "The Self-interest Motive in American Public Opinion." In Public Opinion. Vol. 3 of Research in Micropolitics. edited by Samuel Long. Greenwich, CT: Jai Press. Clark, Cal and Janet Clark (1993). "The Gender Gap 1988: Compassion, Pacifism, and Indirect Feminism." In Women in Politics: Outsiders or Insiders?, edited by Lois Lovelace Duke. Englewood Cliffs, NJ: Prentice-Hall. Clarke, Harold D., Jane Jenson, Laurence LeDuc and Jon H. Pammett (1984). Absent Mandate: The Politics of Discontent in Canada. Toronto: Gage Publications. . (1991). Absent Mandate: Interpreting Change in Canadian Elections. 2nd edition. Toronto: Gage Publications. Clarke, Harold D., Laurence LeDuc, Jane Jenson and Jon H. Pammett (1980). Political Choice in Canada. Abridged edition. Toronto: McGraw-Hill Ryerson. Conover, Pamela Johnston (1984). "The Influence of Group Identifications on Political Perception and Evaluation." Journal of Politics 46:760-85. . (1988a). "Feminists and the Gender Gap." Journal of Politics 50:985-1010. _. (1988b). "The Role of Social Groups in Political Thinking." British Journal of Political Science 18:51-76. Conover, Pamela Johnston and Virginia Sapiro (1993). "Gender, Feminist Consciousness, and War." American Journal of Political Science 37:1079-1099. Converse, Philip E. (1964). "The Nature of Belief Systems in Mass Publics." In Ideology and Discontent, edited by D.E. Apter. New York: Wiley. Converse, Philip E. (1970). "Attitudes and Non-Attitudes: Continuation of a Dialogue." In The Quantitative Analysis of Social Problems, edited-by Edward R. Tufte. Don Mills, ON: Addison-Wesley. Cook, Elizabeth Adell (1989). "Measuring Feminist Consciousness." Women & Politics 9:71-88. -286-Cook, Elizabeth Adell and Clyde Wilcox (1991). "Feminism and the Gender Gap-A Second Look." Journal of Politics 53:1111-1122. Copps, Sheila (1986). Nobody's Baby. Toronto: Deneau. Cox, Kevin R. (1969). "The Spatial Stracturing of Information Flow and Partisan Attitudes." In Quantitative Ecological Analysis in the Social Sciences, edited by Mattei Dogan and Stein Rokkan. Cambridge, MA: M.I.T. Press. Deitch, Cynthia (1988). "Sex Differences in Support of Government Spending." In The Politics of the Gender Gap, edited by Carol M. Mueller. Beverly Hills, CA: Sage Publications. Dooley, Martin D. (1994). "Women, Children and Poverty in Canada." Canadian Public Policy 20:430-443.