Sex and Gender
Sex and gender as sources of heterogeneity in political attitudes and behaviours
Kristi Marie-Diana Winters
A thesis submitted for the degree of Doctor of Philosophy
Department of Government
University Essex
Spring 2009
Copyright Kristi Winters, 2009.
Contents
Contents 2
Acknowledgements 4
Abstract 7
Chapter 1: Analysing the gender gap 18
Why examine political preferences in terms of sex and gender? 21
The history of the gender gap and American gender gap theories 28
How ‘gender gaps’ are produced 29
Women’s growing autonomy 32
Feminist consciousness 34
Childhood and adulthood socialisation and gender 38
Social learning theory and gender role socialisation 39
Social role theory 41
Tying it all together 42
The British gender gap and issue preferences 45
The British gender gap and vote choice 47
Conclusion 56
Chapter 2: Measuring Gender 59
Results of the Personal Attributes Questionnaire 65
Selecting the measures 67
Adopting neutral language 70
Conclusion 72
Chapter 3: The Internet survey and gender measures 74
Designing the survey 77
Testing and selecting the measures 79
Gender measures: Interval or categorical variables? 85
Conclusion 96
Chapter 4: Sex, gender and political behaviour 98
Turnout 100
Conclusion of intention to vote analysis 110
Vote Choice 112
Models of electoral choice 115
Conclusion 125
Chapter 5: Sex, Gender and Political Attitudes 128
Left-right self-placement 131
Equality and kindness 139
Partisanship 143
Conclusion 149
Chapter 6: A very British gender gap? Testing gender gap theories with British data 152
Conclusion 169
Implications of the research 174
Future research 179
Bibliography 182
Appendix A – Twenty-four item Personal Attributes Questionnaire 192
Appendix B – Eighteen item Personal Attributes Questionnaire 193
Appendix C – Pretest Survey Questions Version A 194
Appendix D – Pretest Survey Questions Version B 200
Appendix E – Pilot Survey Questions Version A 204
Appendix F – Pilot Survey Questions Version B 218
Appendix G – Main Survey Questions 232
Acknowledgements
Although it is my name on the front of this thesis, I could not have completed this work without the support and guidance of many individuals. I would like to acknowledge the important ways in which others have made this research and thesis possible. First, thank you to my mom and dad, Diane Wojcik and David Wasieleski, and to their spouses, Bob Wojcik and Lynn Wasieleski. I am sure the idea of moving my entire life to another country was one that was both surprising and in its own way difficult for them, but they have given me nothing but encouragement and support – including of the financial kind – which was vital to my success.
I would also like to thank those individuals who gave me the opportunities to further and expand my career. To the members of the British Election Study team, Harold Clarke, David Sanders, Marianne Stewart and Paul Whiteley, thank you for the amazing opportunity to work with you on the 2005 British Election Study. To Justin Fisher (Brunel), Daniel Monk (Birkbeck, School of Law), Ed Bacon (Birkbeck, Department of Politics) Joni Lovenduski (Birkbeck, Department of Politics) and Meg Russell (ULC), I am deeply grateful for the teaching opportunities you gave me. The teaching I did during the course of my doctoral work improved my knowledge of many subject areas tremendously. I am deeply grateful to the British Academy for funding the 2005 focus group research work which was the basis of my first jointly authored publication with Rosie Campbell, and to the ESRC for funding the Internet survey investigating sex and gender as sources of heterogeniety in political attitudes and behaviours, the data from which is the basis of all the statistical analysis which follows.
A special note of thanks goes to Rosie Campbell who played a key role throughout my PhD research. I was first introduced to Rosie and her work through David Sanders in 2004 and it was, to quote a famous film, ‘the beginning of a beautiful friendship.’ Rosie and I have collaborated on two research projects and several publications since that first meeting. More importantly we have gotten to know each other on a personal level and my life has been enriched by it. I would also like to thank John Bartle for his advice and friendship over the years, and Elizabeth Evans and Rainbow Murray, both of whom assisted me in my research methods courses at Birkbeck and who now have exciting academic careers of their own.
There were many individuals who were important for their friendship as well as their amazing intellects: Deborah Savage, Peter Bloom, Eleni Vezirigiannidou, Nick Allen, Sam Dallyn, Ben Clemens, Alexia Katsanidou, David Hugh-Jones, Jane Kim, Paul Stoneman, and my acoustic guitar partner, Laurence Horton. Thank you to Michael Landewe for asking the two questions which inspired my pursuit of a Masters in the first place: ‘what do you want to do, and where do you want to live?’ They were questions which literally changed the course of my life.
I am deeply indebted for the many contributions my partner, Matthew Bennett, has made to my life personally and professionally. Intellectually he has contributed to this work with his proofreading skills; any mistakes which remain in the text are due to my negligence and not his. He has listened patiently when I’ve talked about my research and tried to understand the opaque statistical language I use, and he has shared with me an interest in the roles of sex and gender in society and the thorny questions which surround them. He has challenged me conceptually and intellectually, encouraged me to work when I wanted to procrastinate, and to relax when I was stressed due to my teaching commitments or finishing the thesis. Thank you, Matthew, for everything.
And finally my supervisor, mentor, employer and guitar lessons instructor, David Sanders. In 2003 I decided to attend Essex rather than the other universities at which I had been accepted because, after reading an article he had co-authored with Pippa Norris, I knew immediately that he was working on the topics which also interested me. It was the right choice and one which I am very glad I made. David supervised my MA dissertation and my doctoral thesis (with Paul Whiteley) and over the years I have been privileged to get to know both him and his wonderful family. He has been a role model to me in how to be a rigorous social scientist, an inspiring teacher, a well-balanced person and a fulfilled human being. I dedicate this work to him.
Abstract
Using data from an 2007 Internet survey (N = 2890) designed to investigate gender and sex as sources of heterogeneity in political preferences, this thesis evaluates sex and gender as separate measures for analysing British political attitudes and behaviours. Statistical findings are reviewed that suggests sex and gender (as measured by questions from the Personal Attribute Questionnaire) may be considered distinct concepts which should be conceptualised and operationalised separately, as they may offer unique explanation in empirical political analysis. In investigating the role of sex and gender in a particular form of political behaviour, intention to vote, I conclude that there is partial evidence to support the proposal that the higher an individual’s sense of their own agency – regardless of their biological sex – the more likely they are to indicate they would vote. I present findings which indicate that a high sense of one’s own agency is associated with right-leaning political attitudes, partisanship and intention to vote, whereas a high sense of communion is associated with left-leaning political attitudes. In the final empirical chapter I conclude that aspects of American gender gap theories, including the socialisation hypothesis, can contribute to a more precise understanding of British political preferences. This thesis recommends that the gendered psychological disposition of individuals is an area which should be subject to further consideration, investigation and research in political science.
Introduction
‘Everything is vague to a degree you do not realize till you have tried to make it precise.’ – Bertrand Russell
The subject category under which my research falls has widened since I first began this thesis. What started out as question about political behaviour, specifically the explanatory roles of sex and gender as sources of heterogeneity in political attitudes and behaviours, has widened beyond political behaviour to include issues of construct validity and political psychology[1]. By the end of this thesis I provide an answer to the fundamental question which provoked my research: are sex and gender distinct concepts that require separate measures in social science research? The short answer is yes, they do. However, the short answer is insufficient in light of the other results and implications which my analyses suggest. As is often the case with scientific research, I end this investigation with more questions than I started out with. Over the next several chapters I will review the findings that answer the central question as well as those that inspire new questions. Above all else it is my hope that the impact my research has on those who read it will be to plant the seed of a question in their minds, and to encourage them, whenever they read a study or hear a presentation about gender, to ask: ‘what does s/he mean by ‘gender’?’
The driving force behind this thesis is the application of the concept of gender to a political phenomenon, in this case political behaviour. Gender is a more complicated, often more nebulous term than sex. It is a concept that can be used as a euphemism for ‘males and females’, that is to say, biological sex; it can also be used to refer to the social construction of men and women’s identities or as a description of men and women in relation to each other (gender roles) (Manzu and Goetz, 2008). As I note in the first chapter, while biological sex is a stable concept (in that males remain males and females remain females over the course of their lives, for the most part), gender as a descriptive category for human behaviour or as a norm for human behaviour has been transformed over time. Gender ‘norms’ for a woman in 2008, wearing trousers and sporting a short haircut whilst studying at the Master’s level for a degree in biology, would have been shocking in 1908.
When we think of the women’s movement, some may mistakenly think this movement began with the fight for female suffrage. However, the real picture was very different. In fact, the suffrage movement was part and parcel of a wider women’s movement that was questioning traditional ‘gender norms’ as to women’s role in society[2]. Women’s attempts to redefine gender norms began long before any organised movement to grant women the vote. To examine the case of Britain in particular, women as early as the 15th century were asking why feminine perspectives of religious experience and females as individuals were not more respected within Christianity. ‘Just because I am a woman, must I therefore believe I must not tell you about the goodness of God?’ asked one early protofeminist, Julian of Norwich. In a tone that might still strike some religious conservatives as radical even today, she went on to write about Jesus through a female lens, through the female experience of giving birth and motherhood:
‘…our Saviour is our true mother in who we are eternally born and by whom we shall always be enclosed…We are redeemed by the motherhood of mercy and grace…to the nature of motherhood belong tender love, wisdom and knowledge, and it is good, for although the birth of our body is only low, humble and modest compared with the birth of our soul, yet it is he who does it in the beings by women it was done,’ (Spearing 2002, p. 201).
Such protofeminists were advancing sentiments that were later made more explicitly political by their descendants: females, their experiences and feminine perspectives have value. By the latter half of the 19th century these beliefs had coalesced into political movements with the intention of improving women’s social, economic and legal status (Bryson, 2003). However the first organised movements to advance the status of women were not focused on the vote but often to educate girls. In 1694 Mary Astell’s A Serious Proposal to the Ladies, for the Advancement of Their True and Greatest Interest was among the first, following on from Descartes, to assert that men and women are equally capable of reason. Astell further argued that women’s appearance as frivolous and incapable of reason was a product of their upbringing and not a natural disability, which demonstrated the need for women’s education. By the middle of the eighteenth century a few affluent British women formed an informal group called the Blue Stocking Society. Focused on social issues and education, the group supported (though never formally worked for) increased education for girls in order that they might be better wives and mothers. Neither Astell nor the bluestockings could be considered ‘feminist’ – Astell herself was a Tory and a supporter of the monarchy and the bluestockings have been regarded as anti-feminist (Bryson 2003) – yet they still challenged the norms as to women’s intellectual nature and abilities and, as a consequence, how women were seen in society.
By the nineteenth century feminists criticised prevailing notions of the ‘ideal woman’. Marion Reid, in A Plea for Women (1988 [1843]), criticised the idea that womanly behaviour was defined by attending to her husband, her children and her household, and argued that this ideal of self-renunciation, living only to serve, was in fact a form of self-extinction. The importance of education continued as a theme among first-wave feminist women and men who promoted the education of girls and called for legal reformation to aid disadvantaged legal status married women[3]. In 1825 William Thompson published Appeal of One Half of the Human Race, Women, against the Pretensions of the Other Half, Men, to Restrain them in Political and thence in Civil and Domestic Slavery. In his book Thompson (with Anna Wheeler) attacked the laws that made married women little more than property: at that time period women lost all legal rights over their earnings or property upon marriage, the legal owners of their earning and property were their husbands, women were denied rights to their children in the case of divorce and could lead lives treated as little more than an ‘upper servant’ by their husbands. John Stuart Mill’s better known On the Subjection of Women (1869) argued not only that women’s subordination was wrong, but that it was also a hindrance to human improvement. In the second half of the 19th century several women formed a group known as ‘the Ladies of Langham Place’, which initiated campaigns around women’s need for better education, increased possibilities for employment and improving married women’s position in law (Walters 2005). It was from these groups, what we would now call ‘grassroots’ movements, that the seeds of the suffrage movement were planted and in 1918 British women over the age of 30 were given the vote (Ibid). Although the initial legal battle was eventually won, movements to further increase women’s status in law to act as an equal citizen to men, to increase women’s educational and employment opportunities to allow them to earn an independent living continue in the developed and developing worlds even today.
Women’s formal entrance into the political process in Britain and in the United States began less than 100 years ago[4]. Although there was speculation that extending the franchise to women would result in the emergence of a female voting bloc, the female vote was initially split along similar lines to men. Instead of introducing a radicalising force into domestic politics, women as a group tended to vote centre-right. The Conservatives in the United Kingdom benefited slightly from women’s enfranchisement (Tingsten 1937, p. 42-45; Durant 1949; Ross 1955), as did Republicans in the United States (Andersen 1996, p. 65). Subsequent to the realisation that women’s partisan behaviour was not radically different from men’s, interest in women’s political behaviour faded. It was not until after the emergence of second-wave feminism (the American Presidential election in 1980) that women’s political behaviour, distinct from men’s, again became an area of interest[5]. Thereafter, much of the research into men and women’s political attitudes and behaviour focused upon the discovered ‘gender gap’ in the 1980 American Presidential election, and this notion of ‘gender gaps’ has framed much of the subsequent research (inter alia Frankovic 1982; Gilens 1988; Shaprio and Mahajan 1988; Hurwitz and Smithey 1998; Mattei and Mattei 1998; Golebiowska 1999). This binary framework of male versus female was the initial springboard for my own investigation: in what ways are men and women political attitudes and behaviours different, and in what ways are they similar?
Early on in my doctoral research it occurred to me that to organise a cross-tabulation by gender category, or more accurately into ‘males’ and ‘females’, was not really measuring the associations of ‘gender’ with political attitudes, but of the respondent’s biological sex. This, of course, was not an original thought but this conceptual distinction has been widely overlooked and under-investigated within political science[6]. It is this distinction, and in particular the conceptual operationalisations of sex and gender, which is at the heart of my research. A bit of disclosure seems appropriate at this point: while I am undoubtedly both a beneficiary of and have been inspired by the work of feminists, this thesis is an empirical analysis of the application of a particular set of survey measures which attempt to capture the concept of ‘gender’ to several different political behaviour models. Several times within this work I reference feminist theory, drawing upon and testing feminist accounts of ‘gender gaps’. However, I would like to take this opportunity to clarify that this thesis is not a normative feminist endeavour, but one which is concerned with the concept validity of the term ‘gender’, the measurement validity of sex as a proxy for gender, and how applying a gendered lens to political behaviour increases (or fails to increase) our understanding of human behaviour. As noted by the Enlightenment philosopher David Hume (2003) one cannot derive ‘ought’ from ‘is’ (p. 334), and I do not attempt to do so. The main questions that I attempt to answer with this research are ‘is’ (not ‘ought’) questions: ‘is sex a valid proxy for the concept of gender’ and ‘are sex and gender systematically associated with certain political attitudes and behaviours’? That is not to say that there are not normative implications that others may wish to draw from reading the conclusions of my findings. However, the time to assert normative positions supported by evidence comes after accurate information has been arrived at through scientifically rigorous empirical analysis.
This thesis uses data from an original Internet survey of political attitudes and behaviours that incorporate measures of sex as measured as the biological category of male or female and gender as measured by the Personal Attributes Questionnaire, incorporated from the field of psychology. To answer the questions I posed above (is sex a valid proxy for the concept of gender and are sex and gender systematically associated with certain political attitudes and behaviours), this thesis has been organised into six chapters and a conclusion. Chapter one provides the context and basis for the thesis by providing the historical and theoretical frameworks for the analysis of sex and gender in political science. In addition I specify the terminology used throughout the rest of the thesis and provide a justification for studying British men and women’s political attitudes and behaviours when it would appear there is so little difference between them. I review the American gender gap theories, which I organise into three general headings 1) increased women’s autonomy (or rational choice); 2) socialisation accounts; and 3) feminist accounts and review the theories behind how gender gaps emerge: through differences in attitudes or differences in issue salience. I also review the previous findings of investigations into a British ‘gender gap’ and the British gender gap theory, the gender generation gap and Campbell’s findings on the importance of interaction effects. In addition this chapter will review theories from sociology as to when and how gender socialisation occurs, reviewing both childhood and adulthood socialisation theory accounts.
Chapter two builds on the distinction between sex and gender and reviews the development of measures of gender from the field of psychology. With two possible methods of measuring gender available to me, I review both survey instruments and justify my selection of the Personal Attributes Questionnaire (PAQ) as the measures for gender in my study. I explain how I selected a subset of PAQ questions for inclusion into an exploratory Internet survey dataset and why I have changed the term ‘masculine’ to ‘agency’ and ‘feminine’ to ‘communion’ in my description of the gender measures. These gender neutral concepts replace the more normatively ‘loaded’ terms of masculine and feminine and are used to interpret the results of the statistical analyses in the final three chapters. The implications of this change in terminology are further expanded upon in the concluding chapter.
In chapter three I review the study design for the Internet survey that was completed in January of 2007 by 2890 participants to analyse the role of sex and gender in political attitudes and behaviours. In addition to reviewing the study design and the reliability of the measures, I also conduct ordinary least squares analysis on the three gender measures to better understand which demographic factors are relevant to an individual’s gender perspective. The regression analyses of chapter three demonstrate that sex is only weakly related to gender suggesting that sex and gender provide separate information which requires the use of separate measures in order to be more accurately operationalised. I also review the rationale for using the measure of gender as interval, rather than categorical, independent variables in statistical analysis.
Having reviewed the concepts, terms, the survey instruments and the study design, chapter four is the first of three chapters devoted to the statistical analysis of sex and gender. Chapter four tests the explanatory power of gender when included in models of intention to vote and vote choice. Informed by the work of the British Election Study team’s investigation into intention to vote, I replicate (as much as possible) several models from Political Choice in Britain (Clarke et al 2004) and evaluate where and when sex and gender contribute to explaining an individual’s intention to vote as well as their vote choice. The results of these analyses demonstrate that in certain models an individual with a high sense of their own agency is more like to express an intention to vote to vote and support the Conservative party, while in some models a high sense of communion is associated with support for the Labour party. In addition, the sex variable adds little or no information to the explanation of either intention to vote or vote choice.
Chapter five is an exploratory investigation into political attitudes. Using a model composed of socio-demographic variables and the measures of gender, I conduct ordinary least squares and logistic regression on several political measures including left-right self placement, scales of libertarian/authoritarian attitudes and socialist/laissez-faire attitudes and political partisanship. The findings of this chapter demonstrate that agency has a relationship with right-leaning political attitudes while communion is related with holding left-leaning political views. In addition the sex variable is rarely significant in the analysis, more evidence that sex and gender require separate operationalisation.
In chapter six I attempt, for the first time in political science, to evaluate the competing gender gap theories (as outlined in chapter one) with British data. Drawing on the theories as outlined in chapter one, I operationalise the American and British gender gap theories with measures from the Internet dataset and the measures of agency, communion and emotional vulnerability from psychology to assess which theories provide the most robust explanation of vote choice for men and women. Perhaps counter-intuitively, the gender gap measures employed originally to provide an explanation of women’s preference for left-leaning parties provide a higher amount of the variance explained for men’s vote choice than for women’s. Agency is shown to be related to both men and women’s right-leaning political vote choice and communion is associated with men’s left-leaning political preferences.
The concluding chapter presents an overview of the findings of my research, examines some of the conceptual implications of using measures of agency, communion and emotional vulnerability and points to areas for future research. I conclude that sex and gender measures should not be conflated in social science analysis. My findings demonstrate that measures of agency, when statistically significant, are consistently associated with right-leaning political preferences; communion measures, when statistically significant, are associated with left-leaning political preferences, while the sex variable is rarely significant and shows no systematic patterns as are found with gender measures. Although previous researchers have concluded there is no ideological ‘gender gap’ in Britain (Norris 1986), the statistical findings indicate there is an ideological ‘gender’ gap in Britain – provided that one is using measures of gender to measure gender, and not measures of sex, when conducting the analysis. In the final chapter I also reflect upon the measures of gender themselves and question to what extent are the Personal Attributes Questionnaire actually measure an individual’s ‘gender’. Are these measures of agency and communion capturing the socially constructed norms for men and women, or are agency and communion more appropriately thought of as aspects of the human experience? Finally, I review areas of future research and analysis which my findings suggest.
The core themes that were part of the women’s movement from its earliest known expression -- the idea of women’s perspectives as valuable and of equal consideration, the importance of education, of economic independence and the empowerment which comes from employment -- can be understood as an attempt by women to challenge norms as to what was appropriate behaviour for ‘women’ and ‘men’ by proposing an alternative framework for women’s behaviour which included a high level of autonomy and agency for women through increasing their education, and their legal and economic autonomy (Walters 2005; Bryson 2003). Feminist have argued that the aspects of ‘humanity’ (or ‘human nature’) normally idealised in the feminine, such as empathy, nurturing and connectedness with others are not specific to one sex. This has challenged wider social norms such that previously sex-stereotyped attributes have been re-conceptualised. The process of disambiguating the term ‘gender’ from the concepts of sex, sex roles and social norms underpins my investigation and informs the conclusions I present at the end. As I mentioned above, including separate measures for sex and gender provides us with separate pieces of information to explain political attitudes and behaviours.
My aim is an attempt to ‘gender’ quantitative political analysis. Gender scholars have identified analytic gaps in existing social science concepts and made suggestions as to how to better incorporate gender into those concepts, sometimes developing new gender-specific concepts in the process (Mazur and Goertz 2008). By incorporating measures of gender from psychology and including them with the variable of sex, my research will address a gap in the existing social science measures and may provide a way to better incorporate the role of ‘gender’ into our empirical analyses.
Chapter 1: Analysing the gender gap
The aim of this thesis is simple: to determine whether separate measures for sex and gender provide unique explanations for an individual’s political attitudes and behaviours. However, testing this simple question is not as easy as framing it. To set the stage for the terms, concepts and theoretical frameworks that are employed throughout this thesis, a substantial number of ideas must be introduced, explained and then conceptually linked. The purpose of this chapter is to lay that theoretical groundwork. It opens by reviewing the terms ‘sex’ and ‘gender’ to explain their different uses in popular culture and establish how they will be used in this investigation. Next, I explain why investigating sex and gender as explanatory variables require investigative research in political science. In the third section I briefly review the history of sex differences in political attitudes and behaviours in the United States (the political context from which most ‘gender gap’ theories have been developed), and survey the various competing theories currently employed in American political science research. In addition, as this thesis differentiates between biological sex and the socially constructed notion of ‘gender’, I provide a brief overview of sociological theories of gender identity. Following on from that, I introduce two concepts that are fundamental to this research (agency and communion) and make the case that these two psychological measures underpin much of the American gender gap theory. Finally, given that this is a study of British men and women, I review the history of sex as an explanatory variable in British politics by examining the results of political science research to date.
As mentioned above, before beginning an investigation into the existence and extent of sex and gender differences in men and women’s political behaviour and preferences, it is necessary to provide definitions for terms that will be used throughout this thesis. In the media and in popular culture there is a tendency to conflate the terms sex and gender. However, my analysis will use the terms to represent distinct concepts. The term sex will refer to the biological category of either male or female; the term gender will be used in this chapter in reference to socially-constructed notions of masculinity and femininity (Helgeson 2002, p.44).
Although this thesis is not focused on investigating of the role of biological sex as a material causal source of political difference, the biological measure of ‘sex’ should not be conceived of as having a simplistic, uniform effect on all respondents. There may exist a range of biological differences within and between the sexes which might contribute to variation in political attitudes and behaviours. Biological sources of political attitudes and behaviours can be found in the work of Alford et al (2005), who found genetics play a role in political attitudes and that variation in the genetic make-up of individual men and women may contribute to their political orientation. Fowler et al (2008) found that there is evidence of an explanatory role of genetics in turnout behaviour, and it is possible that such genetic explanations have a role to play in understanding the causal relations of biological sex differences and similarities in men and women’s political behaviour. Although I included as a traditional demographic variable of sex as the simple conceptual binary of ‘male’ and ‘female’ the range of potential biological variation which underlies this simply binary should be kept in mind.
The important of distinguishing between ‘sex differences’ and ‘gender differences’ was highlighted by a recent publication. Campbell and Winters (2008) demonstrated that British men and women are interested in different types of political issues. Our findings showed that men express a higher level of interest in partisan politics and foreign policy than women, while women express higher levels of interest in news items about education or the NHS than men. This is an example of a sex difference in political interest. In addition, our statistical analysis indicated that individuals with a high sense of their own agency, whether male or female, were more likely to be interested in partisan politics and foreign policy, and that men and women who had a high score on measures of connection with others were more likely to be interested in domestic political issues. As the measures of agency and communion are derived from measures of gender used in psychology, those distinctions could be called ‘gender differences’. In addition, Campbell and Winters’ results point to the effects of the traditional sexual division of labour on interest in politics and possibly the mutable effects of lifecycle factors on the sex variable which should be considered. Men’s interest in politics was consistent across the age cohorts. However, there was a curvilinear relationship in women’s interest in politics. Women in their teens and early twenties report similar levels of interest in politics as men in that age group, which is followed by a steep decline for women’s self-rated political interest during their child-bearing and child-rearing years. After this decline, women in their fifties and sixties report similar levels of interest in politics as their male counterparts at levels comparable to those of women and men in their teens and early twenties (Campbell and Winters 2008). Within this earlier investigation, then, there are three types of sex/gender difference: 1) sex differences in issue preferences; 2) differing levels of interest in politics by sex-related life-cycle; and 3) gender differences in the types of political issues in which agentic or communal individuals express interest. This finding indicates that sex and gender are relevant in accounting for the observed patterns in British political behaviour.
When beginning an analysis of what is often referred to as ‘the gender gap’, a linguistic distinction between a ‘sex gap’ and a ‘gender gap’ is an important one to articulate. Although sex remains a stable biological category over the course of one’s life, the social construction of gender and societies’ views of appropriate gender roles or gendered characteristics have altered over time. Shifts in the norms of appropriate gender roles (especially since the rise of second-wave feminism in the West) are hypothesised to have an impact on people's political preferences and behaviours (Carroll 1988; Norris 1999; Inglehart and Norris 2003). Therefore ‘gender gaps’ in political attitudes and behaviours must be considered in light of wider social changes. This thesis will do so by conceptually and operationally separating out concepts which are associated with gender from the stable biological category of sex. Previous studies of men and women’s political attitudes and behaviours attempted to account for differences in political outcomes by referring to the impact of childhood and adulthood socialisation, feminism, rational choice theory, generational differences or the division of household labour on men and women; however these studies often only used the biological measure of ‘sex’ to differentiate between individuals. This thesis will report on the results of a study that incorporated separate measures of sex and gender into a political science survey. In so doing, I aim to provide a clearer understanding of the distinct explanatory power of sex and gender as individual variables on men’s and women’s political attitudes and behaviours in Britain.
Why examine political preferences in terms of sex and gender?
In the course of my investigations into what is called ‘gendered political behaviour’ (a phrase which covers both sex-based and gender-based political analysis) and in this doctoral investigation in particular, I initially struggled to explain my own interest in this topic[7]. Why investigate political attitudes and behaviours in light of sex and gender? Why bother expanding upon sex, a demographic variable which does not offer much in terms of explanatory power when used in model testing? Or to sum up these concerns more bluntly: assuming there are some small differences between men and women on political measures, so what?
One reason to study the impact of an individual’s sex and gender on political preferences is that sex, unlike other demographic variables, is a simple binary variable that is (very nearly) universally applicable[8]. If the simple fact of being born a male or a female has an impact upon how an individual perceives the world and political events or situations, then the biological category sex may contribute a stable amount of the variance explained in the analysis of political decision-making across nations and political system. If a man or woman is influenced by a socially-constructed notion of appropriate ‘masculine’ or ‘feminine’ perspectives, then the social category gender, or the psychological characteristics associated with a particular gender, may contribute a portion of variance explained in the analysis of political decision-making across political systems and cultures.
It could be argued that there is no compelling reason to study sex and gender differences in Britain because there are not the large systematic ‘gender gaps’ in vote choice as in the United States. However, recent empirical research reveals evidence that points to the influence of sex and gender in political decision making. In the absence of extensive information about a candidate, British men and women draw upon gendered frameworks for evaluating political leaders (Johns and Shepard 2007; Winters and Campbell 2007).
Recent research has also found the existence of sex differences in political issue preferences (Campbell 2004; Campbell 2006a; Campbell and Winters 2006b) and, as mentioned above, that sex and gender differences may influence an individual’s interest in politics (Campbell and Winters 2008). For social scientists interested in explaining an individual’s political views and decisions, it is useful to know the extent to which distinct sex and gender patterns can be found within political attitudes and behaviours. Regardless of whether or not sex or gender differences in political preferences always add up to sex or gender gaps affecting political outcomes, it is important to have a clear understanding of if, when, and how different sex and gender frameworks impact the way in which men and women think about politics. It adds to our understanding of political calculations if we can determine if, when and how sex and/or gender frameworks may result in either identical or different political outcomes.
Kimura (1999) addressed the value of investigating sex differences in light of the vast similarities men and women share. In her book Sex and Cognition she writes:
'If we want to develop an accurate account of how people's problem-solving behaviours originate we cannot, a priori, wilfully exclude any potential source of variation across individuals. It is not only unjustified scientifically to take such a biased view of how behaviours are determined, it is contrary to common sense. The business of science is to find out how the world really works, not how it ought to work according to some wishful schema or another. Scientific explanations change as more information comes in, but at any one point in time a scientific analysis attempts to encompass all relevant facts,' (p. 2).
A note of caution may be warranted at this point. Regardless of whether one is reviewing biological or psycho-social variables, it is important to avoid making essentialist claims about men and women[9]. It is unreasonable to make a claim such as ' all women' or ' all men' possess certain characteristics, values or outlooks, or to expect such a measure to be found. Each individual, male or female, will have a combination of factors influencing their political decisions. Carol Tavris (1999) contrasted two conflicting perspectives of gender differences: essentialism and social constructionism. The essentialist perspective as defined by Bohan (1997) regards gender-related traits, behaviours and attitudes as being internal, persistent and consistent across time and situations. In the view of social constructionists, however, there is no essence ‘of masculinity and femininity, for these concepts and labels are endlessly changing,’ (p. 5) Tavris writes:
‘Unlike essentialist theories, which regard gender as the independent variable (“If I know whether you are a man or a women, can I predict how aggressive you are?”), constructionist approaches regard gender as the dependent variable. Constructionists want to know what factors predict how we come to define ourselves, label ourselves, and behave as a man, a woman, or something else,’ (p. 6).
Tavris (1999) then lists four points that highlight the intellectual pitfalls of essentialism’s false imposition of gendered limitations. First she notes that essentialism confuses snapshots with blueprints. Studies of gender provide us with a snapshot of individuals within a certain time frame, but trends can change over time. One example is that sex differences in verbal and mathematical skills have decreased (Halpern 1989); but by generalising a snapshot into a blueprint some may infer that a difference reported in a given study is an essential and inherent disparity between the sexes and could then wrongly conclude that there are more profound differences between men and women than there may be in reality or over time. In her second point, Tavris notes that essentialism falsely universalises a trait or attitude. This is something which Tavris uses as a critique specifically against biological research. She gives the example of a study, based on only 14 human brains, which reported a sex difference in the size and shape of the neurological connector between the two hemispheres of the brain, the corpus callosum (de Lacoste-Utamsing and Holloway 1982). These initial findings and speculation as to their implications received a lot of attention when the study was first published. However, attempts to replicate the findings in 22 other studies conducted since 1982 found no difference at all (Tavris 1999). Tavris thus shows the danger of universalising findings based upon small sample sizes.
As her third point, Tavris (1999) argues that essentialism fosters stereotypic thinking. An example of the error this type the popular stereotype that women talk more than men. However, recent research found that women do not consistently talk more than men (Mehl et al 2007) rather it depends upon the context in which men and women are speaking.[10] According to Marianne LaFrance:
‘…the research is consistently showing either no sex differences in the amount that men and women talk, or if there is a difference, then it depends on the context. For example, in a professional context, men actually outspeak women by a long shot,’ (Mundell 2007).
Researchers who avoid essentialist stereotypes will be more alert to other potential influences for an observed behaviour. As mentioned above, when analysing how much men and women talk the contexts in which the conversation take place is important (in a profession or casual setting, for instance) when analysing how men and women communicate. Tavris’ critique on the use of stereotyping men and women’s gendered attitudes and behaviours is especially relevant to the ways in which other studies have attempted to use ‘sex’ to capture stereotypical gender roles. Tavris writes:
‘The essentialist stereotype that man are unemotional or cold and women are emotional and warm also causes us to overlook the situations in which men and women are permitted to be expressive and those in which both must suppress feelings. It causes us to overlook the way culture, social grade, occupation, age and situation affect whether and how people will express their feelings – and to whom,’ (p. 11).
Finally, Tavris raises the critique that essentialist theories conflate sex with circumstances, noting that many apparent gender differences vanish when the analyst takes into consideration context, power relations, etc. Refuting the dichotomous assumption that women make moral decisions based on compassion and caring and men make decisions based on the abstract notion of justice (adopted in much of the American ‘gender gap’ literature; (see Gilligan (1982)) Tavris notes that this sort of stereotypical thinking fails to take into account the fact that men and women employ both compassion and justice when making moral decisions, dependent upon the situation they are in and/or reasoning about (Cohn 1991; Clopton and Sorell 1993; Wark and Krebs 1996).
In light of Tavris’ observation that previously observed sex differences do not constitute pronouncements on what ‘all’ men or women do or think, we can say that recent British research into sex differences provides insights into, but not universal assertions about, men and women’s attitudes and behaviours. In many analyses of political outcomes the demographic variable ‘sex’ in and of itself may not produce significant gaps in terms of vote choice. However, British analysis shows that the sex variable interacts with other demographics and that these interactions result in patterns of distinct political preferences among subgroups of men and women (Norris 1993; Campbell 2004; Campbell 2006). An interaction between age cohort and sex described as the ‘gender generation gap’ is a documented phenomenon in which women born since the Second World War have become increasingly left-leaning in terms of political ideological space (Norris 1993). Recent research indicates there are ‘gender-generation gaps’ in issue preferences. For example, Campbell (2004) analysed 2001 British Election Study data and reported that women were more likely to prioritise education (especially younger women) or health care (especially older women) while men were generally more likely to cite the economy as most important, with men between the ages of 55 and 59 expressing the most concern about the European Union. Such results indicate that the sex variable has interaction effects on other demographic measures and that this should be taken into account when trying to account for people's political decisions and preferences.
Cognizant of Tavris’ criticisms, I avoid the first critique of essentialism by limiting the inferences of my findings to the time period in which my survey data was obtained. The inclusion of gender measures into a political science survey certainly provides a ‘snapshot’ of how men and women responded to questions in January of 2007, but several more ‘snapshot’ surveys must be taken before we can discuss sex or gender trends in political attitudes. I also try to avoid the trap of essentialism by not generalising beyond my own exploratory dataset. The analyses I present in chapters four and five provide a definitive answer to the question ‘do separate measures for sex and gender provide separate explanations for an individual’s political attitudes and behaviours’, but these findings must be replicated with further nationally representative probability samples before they can be generalised to the wider British population. To avoid stereotypes, I will discuss sex and gender measures as having associations with specific political behaviours or attitudes instead of asserting claims as to ‘what women or men think’. As to Tavris’ observation of the need to take into account the conflation of sex with circumstances, I attempt to control for the effects of other influences on political attitudes and behaviours by including control variables so as not to misestimate the role of sex and gender as explanatory variables. There are many factors other than sex or gender (e.g. partisan identification, education, parenthood, attention to politics, or other events) which may exert much stronger influences on individual men and women’s immediate political preferences and/or decisions. However, the goal of this thesis is not to find the explanation with the biggest adjusted r2, but rather to test and incorporate new measures in order to more accurately assess and understand the role of sex and gender in British political attitudes and behaviours. In conclusion, there are several compelling empirical reasons to accurately assess what role – if any – sex and/or gender play in political calculations.
The next sections review the history of ‘gender gap’ analysis in the United States, where the ‘gender gap’ was first observed, and in the United Kingdom. A brief history of the ‘gender gap’ is provided with an explanation of how ‘gender gaps’ can emerge, as well as a summary of the theories that dominate the American study of the ‘gender gap’. However, as American ‘gender gap’ theories are not the primary influence focus of my investigation, only an overview of the United States literature is provided. Instead, more attention will be paid to documenting ‘gender gap’ findings in the United Kingdom in order to provide the reader with the relevant historical and theoretical developments in the British context. A later chapter will test the American ‘gender gap’ theories with British data in order to determine if American ‘gender gap’ theory can inform British ‘gender gap’ study, thus familiarity with the American theories is necessary.
The history of the gender gap and American gender gap theories
Early studies of American and British political behaviour did not view sex or gender as important factors. When sex differences were mentioned, women were portrayed as lacking interest and/or as having little sense of personal political efficacy (Berelson et al 1954, p.25; Campbell et al 1960 p. 489-92). If women did participate it was assumed they were likely to behave in a manner similar to their husbands (Campbell et al 1960 p. 485-86, 492-493), to ‘personalize’ politics while men focused on more substantive issues (Greenstein 1965, p.108), or as Lane (1959) described, it was taken for granted that women’s focus was upon ‘persons and peripheral “reform” issues’ (p. 216). In Britain, Butler and Stokes (1969) barely mention sex in their book Political Change in Britain. There are no direct references to women in the tables or index. In their analysis they assigned women to their husband’s occupational class even if those women were in paid work themselves. Overlooking sex as a variable of interest in political behaviour continued into the 1980s, even as traditional assumptions about the strength of class as an explanatory variable for partisan identification were being undermined (Rose and McAllister 1986). Franklin’s (1985) Decline in Class Voting in Britain also makes no reference to sex or to women, and assigned a women’s class location according to her husband’s occupation. This oversight began to be addressed by the mid-1980s (especially thanks to the work of Norris inter alia Norris 1985, 1986, 1988 discussed in greater detail below). Another example is found in Dunleavy and Husbands (1985) who incorporate notions of sex differences into their analysis of voting. The inclusion of the sex variable showed there was considerable variation in the voting patterns of women in the manual class as compared with non-manual class (p. 128).
In the United States the discovery of a sex difference in vote choice came about in the wake of the Presidential election of Ronald Reagan in 1980. According to Bonk (1988), the New York Times released the results of Presidential vote distribution in the form of a cross-tabulation by sex. The results showed that 54 percent of men supported Reagan compared with 46 percent of women voted for Reagan (Cheney et al 1996, p.2). National Organization for Women (NOW) President Eleanor Smeal brought attention to the 8 percentage point gap between men and women voting for the Republican candidate. This was the first time a sex-based gap was noticed in American voting behaviour and the term used to describe this sex-based gap, the so-called “gender gap,” was coined in 1981 (Bonk, 1988). Following this discovery of a ‘gender gap’ in vote choice, political scientists began to analyse data and speculate as to possible causes of and explanations for the sex gap in partisan identification and vote choice. Subsequent investigation by political scientists revealed that the ‘gender gap’ in vote choice had actually started as early as 1968 (Kenski 1988). Subsequent to its initial discovery in American voting behaviour, sex-based differences have been found in issue positions and policy preferences (Frankovic 1982; Shaprio and Mahajan 1988; Hurwitz and Smithey 1998; Golebiowska 1999), presidential approval ratings (Gilens 1988; Mattei and Mattei 1998;) and economic perceptions (Welch and Hibbing 1992; Clarke et al, 2006). Further, there has been a statistically significant ‘gender gap’ in every Presidential election since 1980 (Center for American Women and Politics 2005).
How ‘gender gaps’ are produced
Before reviewing the theoretical accounts of the American ‘gender gap’, it is useful to understand the ways in which a ‘gender gap’ in political issues preferences or vote choice can be produced. Kaufman and Petrocik (1999) defined an attitude difference in political issues. This attitude difference is dependent upon two factors: 1) that men and women have different attitudes on a particular issue and 2) both men and women give the issue approximately equal importance in their political considerations. A simplistic example of this would be attitudes toward military action: in the Cold War era public opinion polls showed women suspected that Ronald Reagan might lead the nation into war (Frankovic 1982) [11]. If men and women had 1) statistically significant different opinions on Reagan and 2) if both sexes gave equal weight to and voted based on that perception of Reagan, then an attitudinal ‘gender gap’ in voting would result. In contrast the salience hypothesis states that gender differences arise not only from differences in attitude, but also from differences in the weight or importance the issue has for the individual (ibid). An illustration would be the salience of economic issues, which have been found to be of greater salience to men than to women (Welch and Hibbing 1992; Box-Steffensmeir et al 1997; Chaney et al 1998) when making political decisions. Both men and women may wish to see taxes decreased, but it is possible that the importance given to economic policies may weigh more heavily as a factor with men than women when deciding for whom to vote.
Knowing how ‘gender gaps’ can be produced, I review both the four main theoretical accounts of the ‘gender gap’ used in the United States and the major empirical studies of the vote choices, partisan identifications, and issue preferences of men and women. The causal sources of differences in men and women’s political attitudes and behaviours in the United States have been the subject of much speculation. The American studies reviewed below all use secondary analysis (often with American National Election Study data) and tend to end with similar conclusions: the ‘gender gap’ is the result of a complex set of interactions between social, economic and sociological factors. Howell and Day (2000), for example, concluded that:
‘The source of the gender gap differs from issue to issue, at times produced by an explanatory variable having a liberalizing effect on women, at other times produced by a variable having a conservative effect on men – all in addition to the contribution of simple economic, value and social role differences between males and females. Thus, there are many paths to the gender gap on political issues that lie behind the well-publicized gender gap in partisanship and voting behaviour,’ (p. 871).
Most studies of the ‘gender gap’ in the United States tend to organise ‘gender gap’ theory into the following categories: 1a) the growing legal autonomy of women related to marital patterns and divorce rates; and/or 1b) women’s increased participation in the workforce, especially public sector employment and dependency to the welfare state; 2) the role of feminism and specifically second-wave feminism; and 3) the effects of gendered socialisation of men and women (Fridkin and Kennedy 2007). As points 1a and 1b are both related to an increase in women’s autonomy (social – including legal – and economic autonomy have had a concomitant increase due to interactions of one upon the other), the theoretical review below conflates the last two points into one theoretical framework of ‘Women’s growing autonomy’. It should also be noted that in contrast to the theories offered above, other authors have provided evidence that it was men’s (and not women’s) changing political attitudes and behaviours which led to the ‘gender gap’ in vote choice and partisan identification (Wirls 1986; Norrander 1997; Norrander 1999). This was said to be due to men’s shift of political preference from the Democratic to the Republican party combined with men’s increasing identification as politically ‘Independent’ (although in elections these independent-leaning men tend to vote Republican) whilst women were less like to self-identify as ‘Independent’ instead maintaining even weak Democratic identification. Although the review below focuses upon women as the key reason for the emergence of a partisan ‘gender gap’ in the United States, these studies demonstrate that the interaction of the political behaviours of both sexes contributed to its emergence.
Women’s growing autonomy
The explanations of the ‘gender gap’ which are grouped under the title ‘women’s growing autonomy’ examine the interaction of socio-economic status and sex as variables. In certain situations it is posited that differences in men and women’s political attitudes come from (working) women’s disadvantaged socio-economic status (Erie and Rein 1988). Blydenburgh and Sigel (1983) neatly summarised this perspective when they wrote: ‘women’s political behaviour can be best understood if we think of women as a disadvantaged or vulnerable minority, a group disaffected by its status of dependence’ (p.1). It is hypothesised that women’s recognition of their vulnerability led (or leads) them to support left-wing parties supportive of funding governmental safety net programmes. Analyses of the ‘gender gap’ as explained by economic status observe that women earn less than men and that women constitute a disproportional number of recipients of social welfare. These facts were used by Erie and Rein (1988) who hypothesised that women’s liberal attitudes toward social welfare was a result of women’s rational self-interest. This self-interest, according to Erie and Rein, translated politically into votes for Democratic candidates. They suggested that, because women are more dependent upon the welfare state, they support the party that supports continuing those welfare programmes upon which they rely. The notion of women as economically vulnerable coupled with women’s dependence upon the welfare state continues to play a role in explanations of the ‘gender gap’ (Piven 1985; Erie and Rein 1988; Cheney et al 1999). Andersen (1999), for example, elaborated another form of this theory through an analysis which links American women’s vote choice to their propensity to be employed in public sector jobs (teaching and social services, for example) or closely associated with government (such as the American health care system). Such ‘gender gap’ theories assume some or all of the following:
1. Women earn low wages relative to men and their political choices are driven by their sense of economic vulnerability;
2. This sense of economic vulnerability results in women prioritising social safety net programmes due to self-interest;
3. Numerically more women than men rely upon government sponsored welfare services; and
4. Women are more likely to be employed by or dependent upon governmental bodies or agency and they therefore support the party which promises continued government funding for their public service jobs or benefits.
However, the ‘rational choice’ theory of the ‘gender gap’ (which states women vote to maximise their own economic interests in maintaining welfare programmes or employment within government) has not had much empirical support when tested competitively with the other ‘gender gap’ theoretical accounts. Contrary to such theories, several studies have found that the gender variable remained a statistically significant variable after controlling for the theorised causes of women’s economic vulnerability, such as income, education, occupation, race and age (Gilens 1988; Stoper 1989; Cook and Wilson 1991; Fite et al 1990; Wilcox 1990). As Howell and Day (2000) note, ‘class stratification does not tell the whole story of the gender gap,’ (p. 860).
Alternatively, (working) women’s improved ability to live economically independence from men has also been hypothesised as a source for the ‘gender gap’ (Carroll 1988; Edlund and Pande 2002). Lipset (1981) theorised that work exposes voters to discussion of policy and candidates (p. 217). Others have linked the disappearance of a voting ‘turnout’ gap between men and women to women’s increasing participation in education and the workforce (Andersen 1975; Welch 1977; Baxter and Lansing 1983; Andersen and Cook 1985; Clark and Clark 1986). This version of the ‘women’s growing autonomy’ theory speculates that women who work are exposed to situations that call into question traditional gender roles (Rossie 1983; Klein 1984; Gerson 1985, 1987). Carroll (1988) found that women who were the most economically independent from men, specifically those women who possessed higher levels of education, those who were unmarried (both single and divorced) and those with higher occupational status, were the most different from men in their voting behaviour and opinion of presidential approval. She writes, ‘The gender gap may best be understood as a manifestation of individual women’s increased political autonomy from individual men,’ (1988, p. 240). Manza and Brooks (1998) studied survey data from 11 elections since 1952 and concluded that women’s changing labour force participation explained the origins of the ‘gender gap’. Rosenthal (1995) concluded that it was women’s economic and psychological independence from men that, in part, explained the ‘gender gap’ in vote choice that emerged in the United States in the 1980s, while noting that this effect may be mediated by other demographic variables, such as income. Edlund and Pande (2002) looked at a specific segment of American women: middle class women. They found that employment is related to the likelihood of middle-income women’s support for the Democrats but not to that of poor or rich women. In conclusion, there is not a clear story as to how women’s increased autonomy relates to the development of a sex-gap in women’s political preferences and, more crucially, these theories, constructed to problematising women’s political behaviours, offer no insight into men’s political attitudes or behaviours.
Feminist consciousness
Attitudes toward men and women's roles have been categorized into three gender ‘ideologies’ (Hochschild, 1989). A ‘traditional’ gender ideology is one that holds that a man's sphere is work and a woman's sphere is home, with the underlying assumption that men have a greater power than women. An ‘egalitarian’ gender ideology is the view that power is distributed equally between men and women and both identify equally with the home and work spheres. Hochschild also identified a third gender ideology, ‘transitional’ ideology, which views it acceptable for women to devote energy to both work and family but should hold more responsibility for the home while men should focus more on their work. Inglehart and Norris (2003) developed a 'Gender Equality Scale' which, when measured across time, shows increasing attitudes of gender equality most notably in post-industrial nations. These findings are mirrored in the work of Twenge (1997) who found that attitudes toward men and women's work became more liberal from 1970 to 1995. In Twenge’s study, both men's and women's attitudes became more liberal during the late 1970s, an important time in second-wave feminist activities. Studies indicate that men and women are increasingly embracing more egalitarian views of how men and women can and should behave; yet women still scored higher on care measures than men, while men scored higher on measures of justice (Jaffe and Hyde; 2000) and, again consistent with accepted gender roles, men score higher on questions associated with being assertive while women score higher on measures of being nurturing (Feingold; 1994).
Feminist consciousness is the category for theories that link gender gaps in political behaviour with support for women's rights, egalitarian views toward gender roles or link women’s attitudes and behaviours to their experiences as women. According to one aspect of this theory it was women’s increasing awareness of discrimination and inequality in society, and women’s subsequent activism to remedy these problems, which is at the root of the political ‘gender gap’. Early ‘gender gap’ analysis by Frankovic (1982) drew upon the theory that feminist views, especially women’s disapproval of the way Reagan handled political issues, particularly his hawkish foreign policy, was behind the ‘gender gap’[12]. After controlling for age and education (which did not reduce the size of the difference between the sexes), Frankovic attempted to use ‘feminist’ measures to account for the difference. Controlling for the respondents positions on so-called ‘women’s issues’ such as the failed Equal Rights amendment or support for legalised abortion, these variables could not account for the difference in partisan preferences. Later studies also concluded that using variables which captured a respondent’s position on ‘women’s issues’ did not explain the ‘gender gap’ because, by and large, men and women’s views on these issues were not starkly different; attitudinal differences between men and women, which could account for the gap, were lacking (Mansbridge 1985; Cook et al 1992). Although the impact of second-wave feminism most likely plays some role in the emergence of the ‘gender gap’, this impact seems to be achieved indirectly through the changing attitudes of women and men as to what are considered appropriate gender roles for the sexes (Manza and Brooks 1998).
As noted above, the ‘feminist consciousness’ explanation can also relate to distinctly female experiences or attitudes, such as motherhood (Chodorow 1978; Diamond and Hartsock 1981; Ruddick 1989). Feminist writers have explored the possibility that parenthood and the traditional sexual division of labour could be sources of the ‘gender gap’ in political attitudes and behaviours. They suggest there is an effect on women’s political views due to women’s sex and gender role as primary caregiver of children: motherhood is seen as having an important influence on how women view politics because it constitutes a large amount of their labour. It is hypothesised that women’s greater propensity to provide primary care for their children and be present in the home while men participate in the public sphere, causes women to value caring for others and to feel a sense of connection with others. Drawing upon Freudian theory, Diamond and Hartsock (1981) claim that psychoanalytical evidence suggests women define themselves in relation to others, while men form a sense of self that is separate. Chodorow (1978) asserts that, as women are more likely to be present in the home caring for children while men are absent and active in the public sphere, women’s experiences lead them to prioritise relationships and view themselves as connected to others. Conversely, men are said to be more likely to think of themselves as autonomous individuals. Chodorow speculates that such differences will remain until men and women undertake dual parenting. In a similar vein Ruddick (1989) claims that the daily experience of mothering encourages a caring and altruistic perspective from which women think about society and politics. Men, on the other hand, do not have ‘care for others’ as the centre of their daily experience and therefore do not develop the same compassionate nature. However, Oswald (2008) demonstrated with longitudinal British data that ‘having daughters makes people more likely to vote for left-wing political parties. Having sons leads people to favor right-wing parties,’ (p. 1). Warner (1991) and Warner and Steel (1999) found changes in attitudes on gender equality are more pronounced for fathers who have daughters, and they speculate that the experience of raising a girl may undermine her father’s patriarchal attitudes. Given the contradictions between feminist theory and the initial empirical evidence there is a need for more investigations into the effects of 1) having children and 2) the sex of the child on a parent’s political attitudes.
Attempts to test feminist consciousness as a causal mechanism for the sex gap in American political preferences have been inconclusive, at best. Howell and Day (2000) write, ‘it is difficult to argue that feminism explains political and issue differences because the causal direction between the two is not yet clear,’ (p. 860). Feminist measures have rarely produced any statistically significant coefficients in American gender gap analysis (Cook et al 1992; Mansbridge 1985; Frankovich 1982). However, this may be a result of parallel shift in attitudes towards men and women’s roles across society in general, or perhaps due to improper operationalisation of the measures which need to be tested (for instance men’s theorised feelings of being ‘separate’ and women’s feelings of being ‘connected’). Evidence for the idea of parallel shifts in feminist attitudes can be found in the work of Bernadette Hayes (1997) who analysed 1992 British Election Study data and found that feminist attitudes predicted the Labour party votes of men and women equally well and that individuals with pro-feminist orientation were significantly less likely to vote for the Conservative Party than any other party. She concludes that, in the British context, it is an individual’s feminist orientation[13], and not his/her ‘gender’ (by which she means biological sex), which is the important variable.
Childhood and adulthood socialisation and gender
Socialisation accounts of the ‘gender gap’ in political attitudes and behaviour cover all social influences and types of socialisations. Childhood socialization, adult socialization and life experiences can all be considered as part of the ‘socialization’ hypothesis. The different moral frameworks of boys and girls as outlined initially by Kohlberg (1981) and later re-examined by Gilligan (1982), form part of the ‘childhood socialisation’ theory. American political science has incorporated Gilligan’s (1982) work on different moral frameworks by simplifying her findings into a simplistic dichotomous variable which predicts women’s moral frameworks will be based upon compassion, an ‘ethics of care’, whereas men’s will be based in notions of right and wrong, an ‘ethics of justice’. These childhood socialisation accounts hold that gender differences stem from ‘sex role conditioning experienced by boys and girls and [that] the influences of these differences are assumed to carry through to adulthood,’ (Manza and Brooks 1998, p.1240). Can political preferences be set as early as childhood? Recent research suggests they can. Fridkin and Kennedy (2007) studied eighth-grade American middle students (13 to 14 year olds) and found that a ‘gender gap in policy and partisanship is established early, before children reach adulthood,’ (p. 133). Based upon their results they suggest that the adult American gender gap is rooted (at least partially) in childhood socialization.
Adulthood socialisation theories hypothesise that gender differences are produced/caused by the roles which adults adopt; this is especially the case with impact of motherhood upon women (Sapiro 1983; Ruddick 1989). Adulthood socialisation theory predicts that women who have been socialised to value nurturing activities may react favourably toward parties offering social welfare policies. Both childhood and adulthood socialisation theories view differences in political preferences as coming from sex-role differentiation patterns which are built up over the course of one’s life and life experiences (Sears and Huddy 1990).
Social learning theory and gender role socialisation
Social learning theory attempts to account for how behaviour is learned. According to the theory it is learned in two ways: 1) through others’ modelled behaviour and 2) through social reinforcement. Gender roles and gendered behaviour are explained as being socially constructed and altered by exposure to new and different modelled behaviour (Helgeson 2002, p. 145). Central to social learning theory is the idea that the individual (usually a child) is a passive agent upon whom society is impacting. According to the theory, cultural forces and the influence of others determine appropriate gender behaviour. The individual responds to rewards and punishments in the environment. Observing and performing gender appropriate behaviour leads to developing one’s gender identity; the gender identity does not lead to the behaviour (Helgeson 2002, p. 148).
Social learning theory can account for how and why sex differences in cognition have changed in a brief period of time and in particular can explain changes that are too fast to be explained by biological (or evolutionary) changes in cognition (Helgeson 2002, p. 147). According to the social learning theory, sex differences are altered as social norms and cultural role-models change. The emergence of alternative models of appropriate gender behaviour influences how males and females perceive the ideal man and ideal woman, consider the ‘strong and silent’ ideal projected by John Wayne in the 1950s as compared with the 1990s macho mob boss Tony Soprano who went into therapy to discuss the emotional basis for the character’s panic attacks. How a man is expected to behave in the early part of the 21st century is quite different from the ideal of male behaviour in the 1950s; the change has been even greater for women’s social norms in both the public and private spheres (Twenge 1997). Social learning theory can be used to account for changes to girls’ and women’s economic and occupational aspirations which underpin the ‘women’s autonomy’ theories outlined above. Girls’ preferences for stereotypically female occupations have decreased over time, while their preferences for traditionally male-dominated occupations have increased (Helwig 1998). Consequently there are increasing numbers of women who have entered the workforce to occupy professions previously dominated by men (Halpern 2004, p. 147)[14]. Similar to Chodorow’s (1978) ideas on the role of dual parenting, Halpern speculates that as society shifts more toward accepting and encouraging men’s roles as active and nurturing fathers, the sex gap in empathy and nurturance may be reduced; as women become more involved in activities that require spatial analysis – such as sports – sex differences in spatial ability may be reduced (Halpern 2004).
Gender role socialisation builds upon the principles of general social learning theory to account for how individuals take on gendered roles. It posits that cultural cues in the environment encourage men to be ‘agentic’ and women to be ‘communal’ and in so doing to ‘take on’ the modelled male and female gender roles[15]. Boys are taught to control their expression of feelings and be assertive (behaviours associated with agency); girls are encouraged to express concern for others and to be more passive (behaviours associated with communion). This is encouraged either indirectly through observing the modelling of others’ behaviour or directly through rewards and punishment, for instance a parent scolding a son who cries or dissuading a daughter from acting like a ‘tomboy’. Gendered roles are reinforced through life as individuals move from childhood to adulthood, marry and become parents.
In summary, social learning theory and, more specifically gender role socialisation theory, attempt to account for changes in both sex-based differences in cognition as well as shifting cultural norms through the influence of modelling and reinforcing gender-role appropriate behaviour. Social learning theories have the advantage of being able to account for declines in those sex differences of cognitive performance that have diminished (as societies embrace gender equality) as men and women’s experiences inside and outside the home become similar (Richardson 1997). A criticism of the theory is that it views people as passive agents who are influenced by society’s values and modelled ideals, not as active agents which create social norms.
Social role theory
According to social role theory, it is the differences in men and women's social roles that create gendered behaviour (Eagly et al 2000). This theory focuses on the social conditions of a society instead of the passive reactions assumed in social learning theory. In this paradigm, the way in which labour is divided between men and women in society explains why men become agentic and women become communal. It is because men are primarily responsible for work outside the home that they adopt an agentic framework; women’s responsibility for work inside the home (even when employed) leads them to adopt a communal orientation. When women do work outside the home, their employment patterns are different from those of men: jobs that require competitiveness and benefit from some amount of aggression, such as business or law, are male-dominated while service industries which require the ability to nurture, such as nursing or teaching, are dominated by women. Social role theory does not require that men adopt an agentic outlook or women a communal one; rather, it theorises that it is because societies have differentiated men and women’s social roles (perhaps as path-dependent extensions of evolutionary roles) that gender differences exist. Wood and Eagly (2002) have advanced a modification of social role theory that takes into account the function of physiological differences between men and women as well as the social roles they occupy.
Tying it all together
At this point it might be useful to pull together what may seem to be several disparate threads and connect the topic of this thesis to American gender gap theories presented above. From the American gender gap literature I have linked and extracted two repeating concepts, implicitly contained within the theories but never clearly conceptually articulated: agency and communion. These themes will form the conceptual basis for this investigation. As the word ‘agency’ in particular has been used in different contexts to mean different things, it is necessary to clarify what the terms ‘agency’ and ‘communion’ will mean in the context of this thesis. David Bakan (1966) developed the ideas of ‘agency’ and ‘communion’ in his book The Duality of Human Existence: Isolation and Communion in Western Man, wherein he argued there are two principles of human existence: an agentic one where the focus is upon the self and separation, and a communal one wherein the focus is upon on others and connection. Agency involves self-enhancement and self-assertion, whereas communion entails group participation and cooperation with others. An agentic person exists for him or herself and forms ‘separations’; the communal person provides for the group and embeds the self in a collective (Buss 1990). Marshall (2001) defines the concepts as follows: ‘Agency is an expression of independence through self-protection, self-assertion and control of the environment. Communion is the sense of being ‘at one’ with other organisms or the context, its basis is integration, interdependence, receptivity, (p.5)’. Marshall (1984) organised the characteristics of agency and communion into categories below in Table 1.1.
Bakan (1966) also suggested that ‘agency’ was a masculine trait and ‘communion’ was a feminine one. This assertion was later somewhat supported by empirical evidence showing that, while men and women possess both agency and communion, a higher sense of agency is more common in men and a higher sense of communion is more common in women. However, it is still possible for an individual woman to have a high sense of agency and for a man to have a high sense of communion with others. Carlson (1971) classified self-reported instances of affect as agentic, communal or mixed. Affective instances reported by men were significantly more agentic than were those reported by women, supporting the notion of sex differences in modes of self-expression (Buss 1990, p.556).
Table 1.1 Characteristics of agency and communal views (reproduced from Marshall 1984 p. 64)
| |Agency |Communion |
|Main Aim |Control |Interdependence |
| |Independence |Union |
|Dominant Strategies |Assertiveness |Acceptance |
| |Control |Cooperation |
| |Change |Contact |
| | |Openness |
| | |Personal Adjustment |
|Characteristics |Achievement-oriented Change-resisting |Being |
| |Contracts |Contextually motivated Emotional tone |
| |Distance |Non-contractual cooperation; forgiveness|
| |Doing |Tolerance |
| |Egoist |Trust |
| |Formal Organisation | |
| |Physical Action | |
The three categories of American theories of the gender gap, outlined earlier, were: 1) those theories concerned with women’s growing legal and economic autonomy, the concomitant rising in women’s dependency on the welfare state and women’s increased participation in government-funded public sector; 2) those that explore the role of feminism (specifically second-wave feminism) in increasing women’s feminist consciousness: and 3) those that incorporate the effects of gendered socialisation of men and women. Each theory can be directly related to the ideas of agency and communion. I would argue that women’s sense of agency is the basis of those theories of women’s legal and economic independence. If a woman is economically dependent and reliant upon public services (what I would term a low sense of agency), gender gap theory hypothesises she will support public programmes that are designed to help people (a communal outlook) [16]. Another interpretation of women’s autonomy theory predicts that women who have gained legal and economic independence and act differently from men have done so as a result of their perception of their own agency. In feminist theory, women’s characteristic connectedness, whether it be sourced to Freudian ideas of women not separating from their mothers or to women’s experience as mothers, is grounded in what I see as an assumption that women possess or further develop a communal view of interpersonal relationships. Likewise Diamond and Hartsock’s (1981) Freudian theory of separation experienced by men can, within the paradigm I am proposing, be viewed as positing that men are generally assumed to have a more agentic, autonomous perspective on interpersonal relationships. Finally, the socially-constructed gender roles of socialisation theories can be interpreted through the concepts of agency and communion that I propose: the traditional teaching of a masculine role to males can be termed as conditioning agency, whilst the traditional teaching of a feminine role to females can be viewed as conditioning communion.
In the following chapters these concepts of agency and communion will be quantified using well-established psychological measures of masculinity and femininity. These measures were included in an Internet survey which attempted to specifically relate the separate explanatory contributions of sex and gender to men and women’s political preferences and behaviours. To set the stage for the rest of this thesis, the focus of this chapter now shifts to Britain and the analysis of gendered political preferences and behaviours in the British context.
The British gender gap and issue preferences
The final section of this chapter presents the relevant findings on sex differences in issue preferences and vote choice in the British context. As noted above, the possible explanatory power of the ‘sex’ variable in political attitudes and behaviour was overlooked from the earliest days of British political science until the 1980s. Early analysis of sex differences in political attitudes and behaviours in the British context was dominated by one researcher in particular, Pippa Norris, whose investigations into sex-differences in political preferences began in the 1980s (Norris, 1985, 1986, 1988, 1993, 1996, 1999, 2001; Norris et al 2004). More recently additional researchers have engaged in experimental and quantitative analysis into the role of sex differences (often phrased as ‘gender differences’ or ‘gendered behaviour’) in British political attitudes and behaviours (Campbell and Winters 2008; Johns and Shepard, 2007; Winters and Campbell 2007; Campbell and Winters 2006b; Campbell 2006; Campbell 2004; Hayes 1997). The works of the two main analysts of sex-differences in British political behaviours, Pippa Norris and Rosie Campbell, are reviewed below.
Pippa Norris’s 1986 article ‘Conservative Attitudes in Recent British Elections: an Emerging Gender Gap’ investigated British gender differences in political issues. Norris examined elections to determine whether or not men and women differed on salient issues (e.g. National Health Service or defence) and, if there were differences, whether those differences could be explained by consistent ideological tendencies – either left-wing or right-wing – in men’s or women’s responses. Using data from the 1979 British Election Study and a 1983 General Election Gallup/BBC survey, Norris chose a cross-section of economic, environmental, foreign policy, and social issues from recent elections. Using the 1979 data, Norris found middle-aged women were more supportive of, and men more opposed to, the government withdrawing the army from Northern Ireland. The same age group of women were more supportive of increasing funds for the National Health Service, a position Norris reports as being ‘left-wing’. Analysing ‘right-wing’ issues, Norris found men were more supportive of expanding nuclear power than women of all ages; however, women expressed a higher willingness to curb Communists within Britain. On the basis of these results, Norris concluded that there were no consistent ideological differences in men and women’s political views. When Norris examined the 1983 data, she found more similarities than differences, although young men were more likely to be critical of Labour’s proposal to give up Polaris nuclear weapons; women were more likely to support Labour’s proposed Commission to freeze (or reduce) prices. At that time Norris concluded that no ideological ‘gender gap’ in left-right policy positions was to be found in Britain.
In a later work Norris (1996) applied the ‘gender generation gap’ theory to the analysis of the left-right self-placement measures available in the Eurobarometer data. She found that in 1983 younger women positioned themselves slightly left of men, while older women positioned themselves slightly right of men, suggesting an interaction between sex and age. A decade later the ‘gender generation gap’ remained, with younger women slightly more left wing than younger men, middle-aged women slightly more right-wing than middle-aged men and older men and women being quite similar in their self-placement (Norris, 1996). Using data from the 1997 and 2001 British Election Studies, Campbell (2006) also found evidence in support of an ideological gender generation gap, similar to the findings of left-right self placement in Norris’ work (1996). Campbell found that women born following the Second World War were to the left of men in their self placement answers and that women born after 1957 were more left leaning on Heath et al’s (1994) socialist/laissez faire scale; this evidence supports the Ingelhart and Norris (2000) modern gender generation gap theory.
A more recent investigation of responses to the 2001 British Election Study’s open-ended ‘most important issue’ question revealed sex-difference patterns in men and women’s political preferences. Campbell (2004) examined the 2001 data using cross-tabulation analysis and found that women were far more likely than men to cite either education or the National Health Service as their most important issue. Campbell also found an important interaction effect between sex and age: women who were of childbearing age were more likely to cite education as their main priority and older women were most likely to rate the National Health Service as their most important issue. Economic issues were predominant for men. Campbell (2004) found that taxation was the most important issue to men aged twenty-five to thirty-five, the economy was cited by more men than women in every group over the age of twenty-five, and immigration was the top issue for men aged eighteen to twenty-five. In order to best understand sex-difference in Britain, Campbell concluded that subgroup analysis, for instance the interaction of age and sex, or parenthood and sex, are important components for the analysis of data.
The British gender gap and vote choice
In contrast to the United States, Britain’s recent elections had no equivalent gender gap in vote choice. At the aggregate level, there are some differences in men’s and women’s vote choice but not a statistically significant gap repeatedly found in the same partisan direction, as is the case in the United States. The lack of a statistically significant ‘gender gap’ in voting preferences has led in British gender scholars to focus upon subgroup differences; for example, the work of Norris (1993) which demonstrated a gender generation gap in which younger women are more likely to vote for the Labour party than younger men. The work of Campbell (2004, 2006) also builds upon this gender generation gap and adds parenthood as a variable with which to divide men and women into subgroups in order to more accurately analyse sex differences.
Norris (1993) documented the percentage difference in the men and women voting for the two major parties in British elections, Labour and Conservative (see Table 1.2). From 1945 until 1992 women were more supportive of the Conservative party, and this trend was noticed as a weak point in Labour support (Mortimore 2000)[17]. This was especially true before the mid-1970s when the gender gap in vote choice was in the double digits: 17 percent (1951 and 1955), 14 percent (1945) and 11 percent (1959 and 1970). From 1974 - 1992 the gender gap closed considerably, never again reaching double digits and exceeding 3 percentage points only twice (in October of 1974 and 1992)[18]. In her 1985 analysis Norris concludes:
Table 1.2: Vote distribution by sex (reproduced from Norris, 1993)
| |Con |Labour |Lib Dem |Gender Gap |
|Year |Men |Women |Men |Women |Men |Women | |
|1945 |35 |43 |51 |45 |11 |12 |14 |
|1950 |41 |45 |46 |43 |13 |12 |7 |
|1951 |46 |54 |51 |42 |3 |4 |17 |
|1955 |47 |55 |51 |42 |2 |3 |17 |
|1959 |45 |51 |48 |43 |7 |6 |11 |
|1964 |40 |43 |47 |47 |12 |10 |4 |
|1966 |36 |41 |54 |51 |9 |8 |8 |
|1970 |43 |48 |48 |42 |7 |8 |11 |
|1974 (F) |37 |39 |42 |40 |18 |21 |3 |
|1974 (O) |35 |37 |45 |40 |16 |20 |8 |
|1979 |45 |49 |38 |38 |15 |13 |3 |
|1983 |46 |45 |30 |28 |23 |26 |2 |
|1987 |44 |44 |31 |31 |24 |23 |1 |
|1992 |46 |48 |37 |34 |17 |18 |6 |
Note: Norris calculated the gender gap as the ‘differences between the two-party (Conservative – Labour) vote lead among men and women. Thus, Gender Gap = Women (Con % - Lab %) – Men (Con % - Lab %) (p.130). Source: Gallup Polls, 1945-59; BES 1964 – 92.
‘…initial evidence suggests that in a number of Western countries the traditional pattern of greater Conservative voting amongst women has been replaced by no significant difference between the sexes or a tendency for men to favour right-wing parties. At this stage we can only speculate about why these changes have developed and whether these gender differences in political attitudes are due to sex-role socialisation, life experiences, the socio-economic position of the sexes or inborn difference. What is clear is that the traditional assumption that women are more conservative than men is no longer valid, if it ever was, and we need to re-examine the changing relationship between sex and voting,’ (Norris 1985, p. 200).
Ipsos-MORI has calculated the sex difference in partisan support using their own data from 1974 through 2001 (see Table 1.3). The results show a waxing and waning in British women’s tendency to favour the Conservative party and men’s tendency to support Labour. In the election for which data is most recently available, the 2005 British general election, Ipsos MORI polling found ‘probably for the first time ever in a British general election, more women than men voted Labour (MORI 2007),’ (see Table 1.4)[19]. This result, while apparent in some pre-election analysis, did not appear in the post-election result analysis[20].
Table 1.3: British gender gap (reproduced from Ipsos MORI)
|Election Year |Conservative lead with women |
|1974 |+12 |
|1979 |+9 |
|1983 |+8 |
|1987 |0 |
|1992 |+6 |
|1997 |+2 |
|2001 |+1 |
Source: MORI election aggregates. 'Gender Gap' = Con % lead over Lab among women minus Con lead over Lab among men. .
Table 1.4: 2005 vote distribution by sex (reproduced from Ipsos MORI)
| | |Con |Lab |Lib Dem |Other |
|All: | |33 |36 |23 |8 |
|Gender: |Men |34 |34 |22 |10 |
| |Women |32 |38 |23 |7 |
|Gender gap | |+2 |-4 |-1 |+3 |
Note: Base "Absolutely certain to vote" (N = 10,986). Source: MORI surveys for the Observer, Financial Times, Sun, Evening Standard and Sunday Mirror, 16 May 2005. Differences are calculated by subtracting women’s percentages from men’s percentages; therefore positive numbers indicate more men than women and negative indicate more women than men.
In her 1993 article, ‘The gender-generation gap in British elections’, Norris first discussed the ‘gender gap’ as described above, indicating women were more likely to vote for the Conservatives and men for Labour, but that the differences declined over time (see Table 1.2). She cites the then-current position, quoting Rose and McAllister (1990) who wrote that gender had no influence upon voting in Britain at that time, contradicting the traditional theory that women favour Conservatives when feminist theory predicted that women ought to support Labour. Their reasoning is straightforward: on matters that are salient to voting, men and women tend to share similar political values. They asserted that on most major political issues men and women divide similarly – along lines of party or class, not gender (Rose and McAllister 1990, p 51).
Norris, however, found a trend within the sexes. Although the 1992 election result produced a six-point sex-difference with women being more likely overall to vote for the Conservatives, the ‘gender gap’ had reverse effects when the generation of the voter was taken into account (see Table 1.5). In her analysis of longitudinal data, Norris concluded that younger women were least likely to vote for the Conservative party in the elections of 1964 to 1992 (excluding the elections of 1970 – October 1974) while older women had been the most likely to vote Conservative in the same period (excluding the election of 1983). Norris concludes that, given the consistency of the trend, the pattern should not be attributed to short-term factors but needed a long-term explanation (Norris, 1993). She examines two possible sources for this: structural change and political mobilization. Linking the prominent explanation (at the time) of social grade and vote choice, Norris theorizes that women’s lives ‘have gradually been transformed by long-term shifts in women’s patterns of work and domestic responsibilities….As a result, younger women and men experience a more common lifestyle than their parent’s generation. We might expect this significant structural change to be reflected in changing political attitudes and values,’ (Norris, 1993, p. 136). Presenting an alternative account, Norris (1993) also considers the role of political mobilization by the women’s movement, drawing upon the theories produced by feminists in the United States to account for the gender gap. To test this theory, Norris uses one of the measures within the 1992 British Election Study designed to measure attitudes toward gender equality. Using multiple regression techniques, Norris analyses the explanatory power of both theories, using social grade, education, union membership, gender and the results of the gender equality scale. The results, according to Norris: ‘suggest that among younger and middle-aged groups gender continues to prove insignificant….Among older voters, however, gender continues to be an important predictor of a Conservative or Labour vote. According to these limited measures, the puzzle of why older women in Britain continue to vote Conservative cannot be reduced to either simple structural or mobilization theories,’ (1993, p. 137). In 1999 Norris again explored the gender-generation gap following the 1997 election. The analysis, not elaborated here, finds that once again the modern ‘gender gap’ is most evident in the youngest age cohort, with younger women consistently more Labour-leaning than younger men and older women more Conservative in their vote choice than older men (Norris 1999, p.156)[21]. Using logistic regression she finds that ‘the direct effect of gender on voting choice was insignificant in the 1997 election…Nonetheless the results of model 2 confirm
Table 1.5: 1992 Vote by age group (reproduced from Norris 1993)
| |Con |Labour |Lib Dem |Gender Gap |
| |Men |Women |Men |Women |Men |Women | |
|All |46 |48 |37 |34 |17 |18 |+6 |
| | | | | | | | |
|Young (>30) |51 |40 |36 |40 |13 |20 |-14 |
|Middle (30-64) |44 |49 |36 |33 |20 |18 |+8 |
|Old (65+) |45 |52 |43 |32 |13 |16 |+18 |
Norris calculated the gender gap as the ‘differences between the two-party (Conservative – Labour) vote lead among men and women. Thus, Gender Gap = Women (Con % - Lab %) – Men (Con % - Lab %) (p.130). Positive numbers indicate women as more support of the Conservatives and men of Labour, negative numbers indicate the reverse. Source: BES 1992.
an important interaction between the effects of gender and age,’ (p. 159). These differences persisted even after adding controls for structural variables and cultural attitudes[22].
The most extensive work to date on sex and British political attitudes and behaviour is Rosie Campbell’s Gender and the Vote in Britain (2006). Campbell’s main question is: ‘in what way and to what extent are women’s political preferences different to men’s as expressed in voting behaviour? (2006, p. 1)’ Building on the premise that subgroup analysis is an important component of British ‘gender gap’ analysis, Campbell found evidence to support the impact of children on women’s vote choice in the United Kingdom. Using 2001 data from the British Election Study, she found that middle or high income women with children were more likely to support Labour than others in the sample, and that those women were 1) more likely to support the Labour Party than men and 2) less likely to support the Conservative Party. This suggests that for certain income levels, mothering had a distinct impact on women’s vote choice from that of men who are also fathers. In testing the ‘gender generation gap’ theory using 2001 data, Campbell found that female respondents born after 1946 were more likely to support the Labour party than male, whereas women born before 1946 were more likely to support the Conservative party. While this seems to confirm the gender generation gap theory, Campbell also noted that in 2001 there was no gender gap in party support between men and women aged eighteen to twenty-four. Campbell suggests this lack of a gender gap may be a function of the importance of the parenthood variable interacting with the income variable, as middle income mothers who are more likely to support the Labour Party are also likely to be over the age of twenty-five. She concludes:
‘Thus there is little evidence of an aggregate-level gender gap, whereby being a woman significantly increases the chance that an individual would vote for one party or another; but instead we see a complex pattern of sub-group effects, where sex interacts with other background characteristics to produce gendered political behaviours and attitudes,’ (2006, p. 129).
However, as Campbell also notes, the British analyses presented above were all limited in that they were conducted as secondary analysis on already existing data with the variable ‘sex’ as the only measure to capture the concepts both of biological sex and the social construction of gender[23]. This limitation on the way in which concepts of gender can be operationalised limits the conclusions that can be drawn from the analysis.
To provide the reader with a preview of findings to come in this thesis, the final section of my literature review includes the most recent contributions of Campbell and Winters to the study of political behaviour in Britain. The exploratory work of Campbell and Winters (2006b) qualitatively investigated how sex interacts with other variables to produced gendered patterns of political attitudes and behaviours. In the week preceding the 2005 General Election, six focus groups were conducted by Rosie Campbell and Kristi Winters[24]. The focus groups took place in London and Colchester, Essex and the six groups were divided by sex into three groups of men and three groups of women each divided by age (under 40s, over 40s and a mixed group for each sex) in order to test for any gender generation gap effects. The sampling yielded a broadly similar group of men and women to permit comparisons by sex[25]. In order to conduct a fair test of gender differences the same questions were asked in each of the groups following a pre-designed interview schedule. We concluded that the most profound differences were not to be found in the content of what men and women discussed as regards political interest, leaders, political issues and parties, but the way in which men and women discussed them. Similar to previous findings which analysed patterns in responses given to the ‘most important issue’ question (Campbell 2004), men were more likely than women to raise taxation, Europe and the economy while women were more likely to discuss the issues of education, health care and childcare. What was noticeably different was the ways in which men and women discussed them. Women made references to other people, inserting the views of people they interacted with when discussing their position and relating their use of friends and family for political information; in short, women were more likely to speak from a communal viewpoint. In contrast, men did not bring up their relationships, their children or the needs of others; instead, they discussed the campaign in a more distant, abstracted way. Men talked about the strategic ways in which the campaigns were being conducted; the way the parties were formulating their messages; men expressed interest in the election because some wanted to see if Tony Blair would get a ‘bloody nose’ as predicted in the media; and expressed more interest in policy debates which were contentious and confrontational[26]. These are all in line with a more agentic perspective on politics.
Our most recent research, presented here as the conclusion to this literature review, is a preview of the findings presented in the later chapters of this thesis. Campbell and Winters (2008) demonstrate that sex and gender are separate concepts which contribute independent explanations to British men and women’s interest in politics. Using the measures of the agency and communion which are described more fully in the following chapters, we conducted statistical analysis on men and women’s interest in politics[27]. The ‘fact’ that, generally, women report having lower levels of interest in politics than men has been established by several empirical studies (Andersen 1975; Baxter and Lansing 1983; Burns 2001; Burns et al 2001; Campbell et al 1954; Hayes and Bean 1993; Tolleson-Rinehart, 1992). These studies assume that, when evaluating an individual’s interest in politics, what is meant by ‘politics’ has a common definition or is interpreted in similar ways by all the survey respondents. Given the insights of our focus group work, as well as other related empirical findings that men and women have different areas of interest and focus when discussing politics (Norris et al 2004, p. 44; Campbell 2004, p. 24), we questioned whether what political scientists thought of as ‘politics’ was the same as what men and women respondents conceived of as ‘politics’ when self-rating how interested they were. To unpack what might be considered political, we not only asked respondents to rate themselves with the commonly used survey question used to measure interest in politics (‘Thinking about politics in general, how much interest do you generally have in what is going on in politics?’) but also to rate their level of interest on specific issues or areas which could be considered ‘political’: The state of education in Britain; The state of the National Health Service; British foreign policy (the conflict in Iraq, Israel-Palestine, EU); Law and order and crime (including domestic security); British partisan politics (Blair and Brown, party conferences, party campaigning).
What we found with confirmatory factor analysis was that the measure of ‘interest in politics’ is not a uni-dimensional measure anchored by ‘not at all interested’ at one end and ‘very interested’ at the other. Instead, our analysis showed that there are two dimensions of political interest: one which is composed of domestic political issues including health, education and law and order, and the other which we labelled ‘general interest in politics’ which was composed of high self-reporting on the standard measure of general ‘interest in politics’, interest in foreign policy and interest in partisan politics. In light of the results of our focus groups and other empirical research, it is perhaps unsurprising that the men in our study reported higher levels of interest in ‘general’ politics and women in ‘domestic’ politics. The results of confirmatory factor analysis revealed that the measure for the self-reported general ‘interest in politics’ loads onto a single component along with foreign policy and partisan politics. Our findings suggest that it is not simply the case that women are less interested in politics than men. Instead, what respondents associate with the term ‘politics’ (namely partisan politics and foreign policy) is of less interest to women than men, and those issues which are domestic political issues (education, health policies and law and order) are of more interest to women, generally, than men. In addition to these sex-based differences in men and women’s interest in politics based upon confirmatory factor analysis, we also used ordinary least squares regression to analyse the explanatory contribution of sex and gender (using measures of agency and communion) to an individual’s interest in politics. Our analysis demonstrates that a high sense of agency had a positive effect on an individual’s interest in general politics and that a high level of communion was associated with higher interest in domestic politics, over and above the difference in interest levels accounted for by an individual’s sex (p. 67).
Conclusion
Although ‘gendered political behaviour’ in Britain is still an under-researched area since Pippa Norris’ pioneering work in the 1980s, recent investigations into areas of commonality and difference in men and women’s political behaviour indicate that there are both sex and gender differences which need be explored both qualitatively and quantitatively[28]. This chapter has reviewed the evidence for sex-differences in political attitudes and behaviours, the theories that have attempted to account for the ‘gender gap’ in the United States and the alternative theories which have been developed to understand sex differences in the British context. In the United States, sociological accounts such as different moral frameworks due to socialisation, shifts in women’s economic and legal status and the effects of feminist movements have been used to explain why women tend to favour Democratic candidates. In Britain, where systematic partisan gaps in partisan identification and vote choice have not been found, demographic accounts involving generational differences, income and parental status have been used to account for the slight, but longitudinally consistent and statistically significant differences in men and women’s political attitudes and behaviours.
In addition to reviewing the major theories and findings of gendered political behaviour in the last several decades, two concepts have been introduced, discussed and mapped onto the theoretical accounts of the gender gap in the United States: agency and communion. These concepts underpin the assumptions made by theories concerned with the nature and origins of women’s increased or decreased autonomy on their political preferences, those made by feminist accounts of gender differences, and those made by sociological theories. In the British context, ‘gender gap’ analysis has demonstrated the need for examining subgroup patterns to avoid the masking effects of looking at the data only in the aggregate. Pippa Norris’ work has demonstrated the role of generational effects and Rosie Campbell’s work demonstrates the necessity of incorporating subgroup analysis, including parenthood and income, when analysing men and women’s political attitudes and behaviours. My own work with Rosie Campbell, using the imported psychological measures of agency and communion introduced in this thesis, demonstrates the need to conceptually disambiguate the terms for and measures of ‘sex’ and ‘gender’ when analysing men and women’s political behaviours.
Chapter 2: Measuring Gender
In the previous chapter, multiple theoretical frameworks from the American and British contexts were reviewed in order to provide an insight into different ways of understanding patterns in men and women’s political attitudes and behaviours. Political scientists clarified the two ways which men and women’s views might differ, appealing either to 1) different attitudes toward the same issue or 2) the difference in the importance given to issues which informed their vote choice. Patterns in men and women’s different attitudes and issue salience might be influenced by three possible factors: the societal changes brought about by second-wave feminism; women’s increasing legal and economic autonomy (and concomitant rise in economic vulnerability); or the socialisation of men and women according to different worldviews, achieved in either childhood or adulthood socialisation processes. One of the critiques presented in the previous chapter noted the lack in political science of any measures of an individual’s gender as a separate concept from their biological category of sex. As previously mentioned, if being born a male or a female has an impact upon how an individual perceives the world and political events or situations, then the biological category sex may contribute to a stable amount of the variance explained in the analysis of political attitudes and behaviours across nations and political system. However, if being raised with specific norms of appropriate male or female behaviour has an impact upon how an individual perceives politics, the social category of gender may also contribute to the variance explained in such analyses. In order to measure the concept of gender as distinct from sex, I conducted research into measures of gender employed in the field of psychology and incorporated them into an Internet study of political attitudes and behaviours to understand if and when the differences in men and women’s political preferences can be explained by a measure of the individual’s gender as distinct from sex.
A challenge presented to those who wish to investigate ‘gender’ as a causal source of variation between and among men and women is how to operationalize an a concept based in aggregated norms with individual level measures. Or, to put it in more plain language, if we agree that ‘gender’ is the socially-constructed norms of masculinity and femininity how can we determine how ‘feminine’ a specific person is? Does it make sense to try to evaluate societal ‘gender norms’ individually? Social psychology continues to wrestle with this question (Lippa 2002). In this chapter I present one method used to try to conceptualize and measure socially-constructed gender norms on the individual level, the Personal Attributes Questionnaire. The reader may recall that in the previous chapter, the concepts of agency and communion were linked with the concepts of masculinity and femininity. In this chapter the concepts of agency and communion will be linked with the measures for masculinity and femininity, respectively. In addition, I will introduce an additional component used when measuring gender called ‘emotional vulnerability’.
To investigate the concepts of sex and gender as separate concepts direct measures of masculinity and femininity were needed. In addition, tests of the internal validity of the gender measures were necessary in order to proceed with the analysis. In 2006 Dr. Rosie Campbell of Birkbeck College, University of London received an ESRC grant[29] to conduct an Internet survey to uncover the sources of gender differences in men and women’s political behaviours. I worked with Dr. Campbell on a previous qualitative project to investigate differences and similarities in men and women’s political issues and attitudes during the 2005 British general election (Winters and Campbell 2007; Campbell and Winters 2006b). Later we discussed and agreed upon my participation on the Internet study through the creation of a special project on measurement validity, which I headed. Working on this project allowed me the opportunity to develop separate and precise measures of gender beyond the ‘sex’ variable currently used in political science’s ‘gender gap’ analysis. I turned to psychology, a field that has a long tradition of investigating the psychological components of gender. Two particular measurement instruments represent the most popular ways to measure gender: the Bem Sex Role Inventory (BSRI) (Bem 1974) and the Personal Attributes Questionnaire (PAQ) (Spence and Helmreich 1978). Both formats were created as part of the same shift in perspective on gender roles that occurred in the 1970s; I will discuss the underlying concepts common to both questionnaires before moving on to address which of them I chose to employ and the rationale for my selection.
Prior to the 1970s, psychologists generally presumed that an individual’s biological sex, their gendered sex role (masculine or feminine), and their personal attributes and characteristics were bimodal, that is to say the differences of the modal female and the modal male were distributed along a left-right spectrum with little overlap (Spence and Helmreich 1978). At one end of the spectrum were men with their masculine roles and characteristics while at the other end were clustered women, their feminine roles and characteristics. It was assumed by academics and researchers that there was a high correlation between an individual’s display of sexually appropriate characteristics and their sex roles, e.g. a rough and rugged man would not be as nurturing a parent as the feminine mother. It was also assumed that an individual’s sex, gender and gender roles were bound together, positioned either on the male/masculine side of the gender spectrum or on the female/feminine. Displays of cross-sex behaviours and attitudes were assumed in some sense to be pathological and had negative connotations (Spence and Helmreich 1978, p.11).
Those attitudes were challenged in the 1970s by investigators who began to demonstrate that men and women possess both masculine and feminine characteristics. Using methods of formal empirical research, Spence and Helmreich asked individuals to identify the socially desirable characteristics of men and of women. This collection of attributes formed the basis for a core set of attributes for each sex (Jenkins and Vroegh 1969; Bem 1974; Spence et al 1974). Derived from these collected attributes, accounts of the ‘ideal’ women tend use ‘such adjectives as emotional, sensitive, and concerned with others’ while the ‘ideal’ man is described with adjectives such as ‘competitive, active, and independent,’ (Spence and Helmreich 1978, p.17)[30]. The idealised attributes for women are similar to Bakan’s (1966) concept of communion, while the male attributes are related to Bakan’s concept of agency, as mentioned in chapter one. This is not to say that the ideal woman possesses all communion and no agency traits, or that the ideal man has agentic and no communal characteristics, but rather that an ideal individual is characterised as possessing both in differing degrees. So, for example, the ideal woman would be both minimally agentic, though not entirely non-agentic, and maximally communal. In addition, Bem (1974) made an important contribution to the investigation by including the concept of androgyny, ‘a self-concept (that) might allow an individual to freely engage in both ‘masculine’ and ‘feminine’ behaviours,’ (Bem 1974, p. 155).
The Bem Sex Role Inventory (BSRI) and the Personal Attributes Questionnaire (PAQ) both consist of a number of trait descriptions set up on a masculine bipolar scale and a bipolar feminine scale, with traits stereotypically differentiating the sexes. The BSRI contains 60 measures using a seven-point scale upon which respondents are asked to rate themselves. The PAQ short form contains 24 items with five-point self-rating scales. Masculine characteristics (denoted by Spence and Helmreich as ‘M’, a shorthand I incorporate below) are defined as those characteristics that are present in both men and women, but believed to be more desirable in males than in females (e.g. independence and competitiveness). Feminine characteristics (denoted by Spence and Helmreich as ‘F’, which I also use) contain characteristics that are considered socially desirable in men and women but are believed to be possessed in greater degree by females. Finally, Spence and Helmreich (1978) included what they deemed the ‘M-F scale’ to rate how socially desirable a characteristic is in males and in females; for instance aggressiveness is judged desirable in males, but not in females. They note:
‘…the content of the M and F scales turned out to support our theoretical preconceptions, the former containing items referring to instrumental, agentic characteristics and the latter to expressive, communal characteristics. The M-F scale, however, contained both agentic and communal characteristics, many of the latter seeming to refer to emotional vulnerability (e.g. feelings easily hurt),’ (1978, p. 19).
To determine whether gender was a bipolar measure, Spence and Helmreich (1978) examined the correlations in the M and F scales. If gender was most appropriately measured on a bipolar scale then the higher scores in the M scale should be related in a negative direction to the F scale. What they found, however, was that the correlations between the M and F scale were not only low, but positive in both men and women[31]. This low correlation between scales of masculinity and femininity was also reported by Bem (1974) using the Bem Sex Role inventory measures. The conclusions of these parallel investigations demonstrate that we cannot assume a relationship between being a woman and being feminine, nor between being a man and being masculine. Individual men and women possess both masculine and feminine characteristics in varying degrees, although the general trend is for men to score higher on masculine attributes and women to score higher on feminine ones. Further, the inclusion of the concept of androgyny allows males and females to have both a high sense of agency and a high sense of communion, while also including recognition of the ‘undifferentiated’ individual – someone who possesses neither a high sense of agency nor a high sense of communion.
Although both the BSRI and the PAQ can be used to measure gender, only the PAQ format was used for the Internet survey and for this thesis. The decision to not use the BSRI was based on several considerations. First, the Bem Sex Role Inventory contains sixty questions measuring two aspects (masculinity/agency and femininity/communion) while the original PAQ uses only twenty-four questions and measures three dimensions (M - masculinity/agency, F - femininity/communion and M-F - emotional vulnerability). Moreover the PAQ, as will be seen later in this chapter, it was possible to reduce the PAQ from twenty-four measures to eighteen. Considering the premium on question space in a large N survey, selecting from among eighteen questions which tap into three dimensions is more parsimonious than selecting between sixty questions for two dimensions. Although the PAQ and BSRI both employ numerical scales (grounded in adjectives at each extreme of the scale), the PAQ uses only five measures while the BSRI employs seven. Given that the scales are numerical there is less ambiguity in a five-point scale than a seven-point scale, making the PAQ more suitable for an Internet survey format. In addition, our attempts to replicated previous research required the use of the Personal Attributes Questionnaire format. Sochting et al (1994) used the PAQ as the method for determining the ‘gender’ of their participants. In order to try to replicate their results we use the same measures they employed in order to be able to engage in like-for-like comparison[32]. Finally, research recently published on the PAQ suggested a refinement of the scale that allowed for a reduction in the necessary measures of gender from twenty-four to eighteen (Ward et al, 2006), a summary of which is provided below. For all of the reasons given above, the PAQ was selected as the source from which new measures would be incorporated into the survey.
Results of the Personal Attributes Questionnaire
Spence and Helmreich (1978) found that the self-reported PAQ scores collected from American college students revealed significant differences between the arithmetic means of men and women on each item. Each item was scored on a scale of zero to four, with the highest number correlating to the most masculine response and zero to the most feminine response. Total scores are obtained by adding together an individual’s score on the masculine question set, the feminine question set and the M-F question set, with a possible range from zero to thirty-two. Men scored lower than women on F items and higher than women on the M and M-F items (the highest score on the M-F scales was always in the direction of masculine traits); the reverse scores occurred for women[33]. Spence and Helmreich (1978), in order to construct a method of combining scales scores, first determined the median scores on the M and F scale using the total sample scores with male and female responses combined[34]. Using cross-tabulation, Spence and Helmreich then classified individuals using a 2 by 2 table of those who scored above the median and those who scored below in each category (p. 35). Reproduced below is a replica of their scheme for classifying individuals on the M and F scores by a media split (Table 2.1). The distribution of men and women’s scores across the various categories was more complex than the simple dichotomous operationalisation of sex (male or female) as a proxy for gender would indicate. The percentage of college students falling into each of the four
Table 2.1: Schema for classifying individuals with Personal Attributes Questionnaire (reproduced from Spence and Helmreich 1978)
Masculinity
| |Above median |Below median |
|Above median |Androgynous |Feminine |
|Below median |Masculine |Undifferentiated |
Table 2.2: Percentage of College Students Falling in Each of the Four Categories Based on Median Splits on M and F (reproduced from Spence and Helmreich, 1978, p.53)
For males[35] Masculinity
| |Above median |Below median |
|Above median |Androgynous |Feminine |
| |32 |8 |
|Below median |Masculine |Undifferentiated |
| |34 |25 |
For females
Masculinity
| | Above median |Below median |
|Above median | Androgynous |Feminine |
| |27 |32 |
|Below median | Masculine |Undifferentiated |
| |14 |28 |
NN = 715 men and women (exact distribution not specified). Difference in means test produced statistical significance of p = < .01.
categories is reproduced below in Table 2.2 (reproduced from Spence and Helmreich 1978, p.53).
Helmreich, Spence and Wilhelm (1981) used a sample of high school students, college students, and parents. They found that the internal consistency reliabilities (Cronbach’s alpha, α) for the M scale ranged from α = .67 to .78 for males (high school students, college students, and parents) and from .71 to .77 for the female participants (from the same groups). For the F scale, Cronbach’s alpha ranged from α = .72 to .80 for males and from .73 to .79 for females. The M- F ranged from α = .53 to .61 and from α =.61 to .65 for males and females, respectively. According to Pallant (2005) the recommended score on the Cronbach’s α coefficient should be equal to or higher to .7 (p. 90), which was achieved for both men and women’s masculine (M) and feminine (F) measures.
This evidence, wherein both men and women can be classified as masculine, feminine, androgynous or undifferentiated, shows the seriously flawed nature of the assumption that the dichotomous biological variable ‘sex’ can be used as a proxy variable for the social construction of ‘gender’. At present political science cannot estimate the difference in political preferences and attitudes between a woman who scores high on femininity and a woman who scores high on masculine measures; neither can current political science explain the difference in preference between a man who scores high on masculine measures and a man scoring high on feminine measures. Including direct measures of gender into a large N British political science Internet survey instrument allowed for the inclusion a large number of cases. As a result, the number of cases was large enough to allow investigation into the relationship between an individual’s gender perspective and their political preferences while controlling for the explanatory power of sex, age, education and other relevant demographic variables.
Selecting the measures
Having established the measures and method for the direct measurements of gender for my investigation (the Personal Attributes Questionnaire measures), it was necessary to narrow down the measures from twenty-four to a more reasonable number. The section below reviews a publication which informed and influenced my selection of measures from the subsection of the twenty-four original Personal Attributes Questionnaire measures. The article (fortuitously released a few months before the study was designed), ‘Measurements of Agency, Communion, and Emotional Vulnerability with the Personal Attributes Questionnaire’ written by Ward, Thorn, Clements, Dixon and Sanford (2006), was used not only as a guide for which measures were to be incorporated into the survey, but also as a resource for the interpretation of the measures as interval variables and not simply as categorical variables. Ward et al also provided more gender neutral terminology for the PAQ measures, which I incorporate and will later argue are better terms to capture the concepts the Personal Attribute Questionnaire measure.
The work of Ward et al (2006) informs this thesis in two ways. The primary purpose of the article by Ward et al was to evaluate the internal consistency reliabilities of the PAQ measures through the use of confirmatory factor analysis (CFA). Using a sample of college students (N = 382), they tested the PAQ gender measures, analysis that should produce a 3 factor solution: 1) the M – masculine – scale, 2) the F – feminine – scale and 3) the M-F – masculine-feminine – scale. By examining the results Ward et al determined that three M-F items had small loadings, two of which were not significant. They then assigned those measures as ‘misspecified’ and removed them from further analysis. After removing the misspecified items and conducting further analysis, they found indications that two F scale items were misspecified. Those two measures were also eliminated, and reanalysis showed that the fit indexes for the models were improved.[36] Finally they examined two M items for potential cross-loadings and found two M items, one of which they removed and another which they retained because, although it did load on the M-F scale, the loadings on the M scale were ‘sizeable (standardized loading = .47)’ (p. 210). The final goodness of fit tests were significant and an improvement over the original scores. After they removed the six misspecified items and once again conducted analysis with another sample of college students (N = 230), they found the scores were an improvement on the original sample results. Using their new scales, they found:
‘…the resultant scales were well correlated with the original scales and differentiated men and women as well as the original scales. Eliminating one item for the M scale and two items from the F scale did not alter the internal consistency reliabilities of the two scales, and removal of three misspecified items from M-F improved reliability,’ (p. 213).
Ward et al also concluded that the M-F scales (which they relabelled ‘emotional vulnerability’ measures) were more effective in differentiating between men and women than any of the other PAQ scales. They also stated that ‘EMV could be useful for explaining sex differences in other behaviours (e.g. interests, attitudes, preferences and abnormal behaviours),’ (p.214). Up until this point in this thesis little has been said about the emotional vulnerability measures which are contained within the PAQ[37]. This is, in part, because the M scale (agency scale) maps on to characteristics associated with masculinity and the F scale (communion scale) maps on to characteristics associated with femininity. It is these two scales (M and F) which are used in combination to produce Spence and Helmreich’s (1978) categories of masculine, feminine, androgynous and undifferentiated. Adding information about an individual’s emotional vulnerability does not add additional information about an individual’s gender perspective; it only that the individual rates high or low on emotional vulnerability. Less research has been done using the EMV scale but one study did find that the EMV scale was related to sex differences in response to experimental pain (Dixon et al. 2004). Ward et al write: ‘Even when sex differences are not the primary focus, it may be important to control for EMV in studies of agency and communion (e.g., in relation to mental and physical health). Thus, further research to develop the sex differentiating aspect of EMV seems warranted,’ (p. 214). In order to contribute to research into the use of EMV scales when studying agency and communion, my statistical analysis in all future chapters will include the EMV scale as a control variable.
For the various reasons listed above, the Ward et al (2006) reduced version of the Personal Attributes Questionnaire was employed in this study of what drives gender differences in political issue preferences. The original twenty-four item scale and the reduced eighteen items scale are attached in appendices A and B.
Adopting neutral language
The second element of the work of Ward et al that informs this thesis is their change in terminology and their accompanying change in the conceptual definitions of behaviours that are ‘masculine’ or ‘feminine’ into more gender neutral terms[38]. As noted above, the ‘ideal’ woman is characterised by adjectives such as ‘emotional’, ‘sensitive’, and ‘concerned with others’ while the ‘ideal’ man is said to possess the characteristics of ‘competitiveness’, being ‘active’, and ‘independent’. Yet, as has also been noted, these idealised characteristics can exist in varying degrees within individuals regardless of sex. These traits can be labelled with terms that accurately reflect the degree to which these characteristics can be possessed by both men and women, terms that carry far fewer implications than the terms ‘masculine’ ‘feminine’ or even ‘androgynous’. The use of words like ‘masculine’ or ‘feminine’ have wider associations than those terms suggested by the PAQ gender measures, and could easily be misunderstood as referring to displays of human sexuality. Another valid reason to divorce the gender measures from sexualised labels is that, since their development in the 1970s, the M, F and M-F scales of the PAQ have not only been applied to assess ‘masculinity’ and ‘femininity’ in gender role and sex differences research (Burnett et al 1995, Celluci et al 1998; Dade and Sloan 2000; Toller et al 2004), but the same survey questions have been used to measure the concepts of ‘agency’ and ‘communion’ (Helgeson 1994; Fritz and Helgeson 1998; Fritz 2000; Abele 2003) and also as measures of ‘instrumental’ or ‘expressive’ traits (Spence 1993; Spence and Buckner 1995; Cox et al 2004)[39].
Ward et al chose to rename the factors of the revised scale to Agency (using the measures of the M scale), Communion (using the F scale feminine measures) and Emotional Vulnerability (based upon the M-F scale). In this thesis, I follow their example and from this point onward will talk about the psychological dispositions of male and female respondents in terms of their self-rating on agency, communion and emotional vulnerability, rather than describing behaviour as ‘masculine’ or ‘feminine’. There are several advantages in doing so. It should be noted that this thesis is an exploratory and descriptive empirical investigation into political attitudes and behaviours that focuses on the explanatory effects of sex and gender. Very little has been said about broader feminist theory or the normative implications of feminist critique in these areas because this research is not about feminism per se[40]. However, as this investigation deals with issues of sex and gender it is appropriate to consider relevant contributions of feminist theory and feminist critiques, and one of the many consequences of feminism has been an evaluation of the use of language in our society. It is undeniable that the use of the terms ‘masculine’ ‘feminine’ ‘androgynous’ and ‘undifferentiated’ carry normative as well as descriptive connotations. Although people may describe the ‘ideal woman’ as characterised by a commitment to caring for the well-being of others, a man who is committed to the common good and taking care of all the members in his society might not think of his values as being ‘feminine’; likewise an independent, self-assertive woman may resent the wider implications of being labelled as ‘masculine’ and its connotations with being ‘manly’ in the biological sense of the word. This thesis is an investigation into the explanatory effects of sex and gender on political attitudes and behaviours, but I suggest here that the best way of discussing the effects of ‘gender’ is to use terms for gendered perspectives which do not themselves imply sexual norms. I suggest that the description of someone as highly ‘agentic’ or ‘communal’ or even ‘emotionally vulnerable’ is more neutral than the use of the terms ‘highly masculine’ or ‘highly feminine’ Taking on board the critiques of feminism with regard to the normative implications of language, I have adopted the neutral terms Ward et al applied to their revised scales.
Conclusion
In the last fifty years Western society’s understanding of what gender is has undergone significant transformation. For most of human history, what was considered socially appropriate for an individual was dependent upon their biology and subsequent sex roles. Societies have valued independence, assertiveness and competitiveness in men, and passivity, nurturing and compassion in women. Norms for the appropriate behaviour for both men and women have been challenged and reconceived, undoubtedly as a result of the activities of and reactions to feminist critiques, combined with women’s increased access to higher education, increases in their legal and economic autonomy, and changes in the assumptions about what are ‘appropriate’ sex roles,. In psychology, researchers have moved away from an assumption of the unity of sex and gender, represented by a bimodal distribution with men and masculine traits on one end and women and feminine traits on the other. Empirical investigations have showed that men and women possessed both ‘masculine’ and ‘feminine’ traits to varying degrees and that the various aspects of gendered behaviour cannot be captured with the simple dichotomous variable of ‘sex’.
This chapter has reviewed the process by which gender measures were selected and incorporated into the Internet survey on understanding gender as a driver of political attitudes and behaviours. The justification for the selection of the Personal Attributes Questionnaire was given, as well as a review of a recent analysis which reduced the number of measures and improved the survey instrument. Links between the measures of masculinity and femininity and the ideas of an individual’s sense of agency and communion – terms that will be incorporated throughout the rest of this thesis – were also given. The next chapter will explain the construction of the Internet survey instrument and review the method used to determine which of the PAQ questions were selected. It will also evaluate the performance of the measures in an Internet survey through the use of confirmatory factor analysis, elaborate on the other questions used in the survey and enumerate the hypotheses which will be tested in later chapters.
Chapter 3: The Internet survey and gender measures
In order to conduct the necessary empirical analysis to determine whether sex and gender are separate concepts, with each contributing a separate amount of explanation to people’s political attitudes and preferences, a large number of survey respondents were needed. A multi-faceted research funding proposal to study gender as a source of political preference heterogeneity with an aim of testing possible drivers of gender (and sex) differences was accepted by the Economic and Social Research Council in 2006[41]. In order to properly study the relationships (if any) between sex, gender, and political attitudes and behaviours a sufficient number of participants for subgroup analysis, where previous British ‘gender’ differences have been found, would be necessary. As mentioned in chapter one, Pippa Norris (1993) outlined the ‘gender generation gap’ in Britain. She showed younger women were more likely to vote for Labour than younger men, and that older women were more likely to vote Conservative than older men. In addition, the work of Rosie Campbell (2004, 2006) also suggested the importance for British men and women’s political preferences of sub-group populations related to/determined by age cohort, income and parenthood. To understand wider societal patterns a minimum number of men and women across all age groups were needed; this requirement was the main impetus for the decision to use an Internet panel for the study. Although face-to-face interviews based on a random sampling selection would have been preferable, the funds needed to achieve our target of 3,000 cases for the main study was far beyond the funding available.
Thus for reasons of cost, sample size, rapid field deployment and data turn-around time, an Internet survey was used as the basis for this exploratory research. Internet surveys are increasingly being used both in business and in academic research, but the full implications of using non-probability Internet panel data as compared to other more established survey methods are not yet fully known. Malhotra and Krosnick (2007) used American data from the 2000 and 2004 American National Election study to compare the accuracy of their findings in describing populations. Their American data produced results in which the face-to-face samples were more accurate than volunteer Internet respondents. Bandilla, Bosnjak and Altdorfer (2003) administered one module of the International Social Survey Programme as a paper-and-pencil questionnaire and one as a web-based survey. The paper version was given to a representative sample of the German population and the online version to a panel representative for German Internet users. After weighting the Internet data based on socio-demographic variables, there were inconsistent findings for the two datasets. However, when a sub-sample of parallel respondents was compared - their study compared respondents with similar education levels - ‘the results demonstrate a surprisingly high level of agreement across both methods. In other words, at least as far as the opinions of the highly educated public are concerned, the ISSP survey could have been conducted using a randomly recruited online panel by means of a web survey,’ (p. 241). Sanders et al (2007) analysed British data based upon identically-worded questionnaires from both the Internet and face to face samples of the British Election Study. Their analysis showed ‘few statistically significant differences between the coefficients generated using the in-person and Internet data,’ (p. 257).
In addition to their cost effectiveness, Internet surveys have the advantages of greater flexibility in question wording and format, access to verbatim past responses – thus reducing memory bias – and the self-administered format (removing the need for an interviewer) which may reduce social desirability bias (Sanders et al, 2007). Other reasons to employ an Internet survey included the flexibility in question wording, in question ordering, in response wording and response ordering which an online survey can easily provide; therefore the flexibility an Internet study allows was an important consideration in the decision. Further, as this work was exploratory, the main concern was to analyse a large number of comparable men and women rather than making generalisations from a probability sample to the wider population. As will be seen below, the substantially reduced costs associated with an Internet study and the rapid turnaround time between field work and data analysis allowed us to test and refine our questions in ways which would have been cost prohibitive in any other format. Finally, the author acknowledges that this research is exploratory, and any findings using Internet data should in future be replicated using a face-to-face probability sample to test for any mode effects in order to make generalisable statements about the role of agency and communion outlooks in political attitudes and behaviours.
In order to successfully investigate our areas of interest, Dr. Campbell and I decided that the development and deployment of a three-stage internet survey was most appropriate for our project. The first stage consist of a ‘test’ wave, using newly created questions, existing questions imported from other existing surveys but not used in a British political science context (for example, the PAQ gender measures), and well-established political science questions which we could analyse with the newer measures. The second stage of the survey was a pilot survey in order to have a ‘dry run’ of the measures that were chosen from the testing wave, as well as to conduct some comparisons of question wording and response options. Our final stage was the full survey, with a goal of 3,000 respondents. The structure and questions used in the main survey were informed by those results analyses that were conducted on the trial measures in the initial testing and pilot waves. The next section reports on the results of the survey construction and selection of measures for the gender scales used in the rest of this investigation.
Designing the survey
A challenge for this project (and for most of those that want to include multiple-item scales into a survey) arose when deciding how to balance the need for a sufficient number of gender measures to obtain valid results with the restriction on how many questions overall respondents would be willing to answer and could answer accurately while making sufficient cognitive effort. The questions in the Personal Attribute Questionnaire were not designed for large N surveys covering several different topics. Whereas many surveys have between three and five questions to establish a scale, even the reduced PAQ developed by Ward et al (2006) had eighteen questions. For Internet respondents who were paid a nominal sum for participating, eighteen questions on gender alone were simply too many for a survey that was restricted to an overall total number of questions between eighty and ninety. It was decided that three measures for each concept (masculinity, femininity and the M-F scale, totalling nine questions overall) was a reasonable balance between sufficient numbers of gender measures to create the scales for our planned data analysis. Nine questions represented ten percent of the total number of questions asked in the survey, this allowed for a sufficient number of questions allotted to other topics that important to our investigations and traditional measures we needed to include in order to replicate certain statistical models.
As has been mentioned above, a special project on measurement validity had been included in the overall heterogeneity study[42]. In my research in preparation for this study, I noted the importance of testing questions (which the authors below term ‘pre-testing’) as well as a paucity of documentation and publication as to how questions were tested, what the results of the initial tests were, how the questions were modified as a result and the final results. As Converse and Presser (1986) wrote:
‘Pretesting a survey questionnaire is always recommended – no text in survey methods would speak against such hallowed scientific advice – but in practice it is probably often honoured in the breach or the hurry. There is never the money nor, as deadlines loom, the time, to do enough of it. There is a corollary weakness that the practice is intuitive and informal. There are not general principles of good pretesting, no systematization of practice, no consensus about expectations, and we rarely leave records for each other. How a pretest was conducted, what the investigators learned from it, how they redesigned their questionnaire on the basis of it – these matters are reported only sketchily in research reports, if at all. Not surprisingly, the power of pretests is sometimes exaggerated and their potential often unrealised,’ (p. 52).
Although the main focus of this thesis is the application of gender measures to a political science survey and the explanatory power of sex and gender, it is also my hope that this chapter in particular will contribute to the under-reported process of testing new survey measures. Several newly constructed and newly imported questions were tested in our Internet survey – some with more success than others – but as the focus of this thesis are the measures of sex and gender this section focuses upon the testing methods used to evaluate the accuracy and refinement of the available psychological measures, and how the measures were selected.
The process of survey development proceeded in the following fashion. The original proposal was based on focus group research conducted by Dr. Campbell and myself, the findings of which suggested men and women used different conceptual frameworks when discussing and evaluating politics (Campbell 2006; Campbell and Winters 2006b; Winters and Campbell 2007)[43]. Part of our investigation was an attempt better to operationalise the gendered morality framework employed to measure the sociological accounts of gender gap analysis – the ethics of justice and the ethics of compassion frameworks developed by Gilligan (1982) – into survey questions which could be used in statistical analysis[44]. These concepts, imported from qualitative formats and transformed into quantitative ones, posed the largest challenge in the survey. Our concern about devising accurate measures for the gendered morality questions led us to create the three-stage survey plan, outlined above. The first stage consisted of all the questions that were brand new, such as our morality questions, or had not been tested in a political science survey, such as the measures of gender in the Personal Attributes Questionnaire. The aim was to administer the pretest survey to a sample large enough to achieve an N of around 300 respondents. This would allow us a sufficient number of participants to look for differences, but was a small enough wave to allow us to save sufficient funds for the large Ns we would need for the two survey waves still to come.
After analysing the pretest results and choosing the best measures a second survey, the pilot survey, was administered. The pilot survey included a few experimental tests of question ordering and wording, as well as the measures that had been refined or altered based on the results of the pretest. In order to have sufficient respondents to evaluate the experiments we set a goal of 1,000 participants for the pilot survey. Finally, after conducting the analysis on our refined measures and informed by the results of our pilot survey experiments, we set a goal of 3,000 participants for the final version of the survey; the dataset from this main survey wave forms the basis of this thesis. The complete version of the questions asked in each wave of the survey can be found in the appendixes.
Testing and selecting the measures
In the testing wave of the survey, all of the questions listed by Ward et al (2006) were included in the survey questionnaire and then tested (see Appendix A). The pretest questions in survey A were administered to members of the YouGov panel between 5th November and 7th November, with a total of 361 people responding to the A version of the pretest survey[45]. As was mentioned in the previous chapter, Helmreich, Spence and Wilhelm (1981) used a sample of high school students, college students, and parents and found that their internal consistency reliabilities (Cronbach’s alpha, α) for the M scale ranged from α = .67 to .78 for males (high school students, college students, and parents) and from .71 to .77 for females (from the same groups). For the F scale, their Cronbach’s alpha ranged from α = .72 to .80 for males and from .73 to .79 for females. The M- F ranged from α = .53 to .61 and from α =.61 to .65 for males and females, respectively. According to Pallant (2005) the recommended score on the Cronbach’s α coefficient should be equal to or higher to .7 (p. 90), which was achieved for both men and women’s masculine (M) and feminine (F) measures.
To assess my data for internal consistency with the Internet responses, a scale reliability test was conducted using SPSS on the test wave data. Analysis of my agency measures produced a Cronbach’s alpha of α =.81 on 7 items, above the recommend .5 for scales of less than 10 items (Pallant 2005, p.90). Based upon the results of my analysis, removing any particular item did not change the scale reliability. According to the ‘Item-Total Statistics’, the Cronbach’s alpha for the scale would range between .80 and .77 if any one item were deleted. The measures of communion produced a Cronbach’s alpha of α =.84 on 6 items. Again, eliminating any one particular item did not improve the scale. The range of Cronbach’s alpha when removing the various questions ranged from .83 to .80. Our emotional vulnerability scale obtained a Cronbach’s alpha of .70, but in this instance removing the measure ‘indifferent to others approval’ reduced the alpha to .61. This information was incorporated into the item selection decision.
The second point of concern related to the relationship between sex and gender. We examined the distribution for each response by sex using cross-tabulation. To test the relationship we selected those questions that were statistically significant in their association with the predicted biological sex. In other words, we were careful to incorporate ‘agency’ measures that were associated with men and ‘communion’ measures that were associated with women. If the survey did not use gender measures that were the most conservative and rigorous, that is to say if they did not do a good job in predicting the respondent’s sex, the findings could be criticised as employing improperly specified measures[46]. The results of the cross-tabulation analysis, the gamma, are presented below in Table 3.1. The table presents the results of the chi square test of independence and the ordinal measure of association, the gamma. Items that produce a negative gamma (for example, not at all competitive -.31**) indicate measures that predicted more men than women at the extreme positive end of the scale; positive values of gamma (for instance not aware of other’s feelings .28*** or never cries .42***) indicate that more women answered in the extreme positive end of the scale. The questions selected for use in our Internet study -- three measures of agency, communion and emotional vulnerability -- are listed in bold in Table 3.1.
Our Internet survey was administered to 6000 members of the YouGov panel between 30th of January and 5th of February 2007. In total 2890 people responded to the survey[47]. After the data were collected it was necessary to analyse the gender measures. Analysis was run on the variables included in the main survey (the variables that are used to construct scales employed in the subsequent analysis). Scale analysis of the agency questions on the main survey variables produced a Cronbach’s alpha of α =.60 on 3 items, while the measures of communion produced a Cronbach’s alpha of α =.80 on 3 items; both were above the .5 score often recommended for scales of less than 10 items (Pallant, 2005, p.90). The emotional vulnerability measures produced α =.495 on three items, which falls just below the Pallant’s recommended .5 threshold. However, as it is common to find low Cronbach’s alpha with small scales, an alternative evaluative measure can be used: the mean inter-item correlation. Briggs and Cheek (1986) recommend an inter-item
Table 3.1: Summary statistics derived from conducting cross-tabulations between 18 items measuring gender and biological sex; test wave
|Question |Gamma |
|A- Not at all Independent |-.145* |
|A- Very passive |-.195 |
|A- Not at all competitive |-.307** |
|A- Gives up very easily |-.071 |
|A- Not at all self confident |-.340*** |
|A- Feels very inferior |-.413*** |
|A- Goes to pieces under pressure |-.127 |
|C- Not at all able to devote self completely to others |.090 |
|C- Not at all helpful to others |.162* |
|C- Not at all kind |.115* |
|C- Not at all aware of feelings of others |.280*** |
|C- Not at all understanding of others |.247* |
|C- Very cold in relations with others |.180* |
|E-Not at all excitable in a major crisis |.317*** |
|E-Indifferent to others approval |.195** |
|E-Feelings not easily hurt |.155 |
|E-Never cries |.420*** |
|E- Very little need for security |.205 |
N = 359. Source: Sex and gender as sources of heterogeneity in political attitudes and behaviours stage 1 A dataset. Chi square significance reported as ***p = < .001, **p = < .01, * p = ................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.
Related download
- member matters
- price reduced loopnet
- campaign finance report state of wisconsin gab 2
- annual report web version actors federal credit
- illinois state chartered active credit unions
- annual meeting of members access credit union
- the role of subscription based patrol and restitution in
- internship brochure edward hines jr va hospital
- labor cabinet dept of worker s claims
- sex and gender
Related searches
- women and gender equality
- women and gender issues
- women and gender articles
- gender equity and gender equality
- women and gender studies book
- women s rights and gender equality
- education and gender inequality pdf
- biology and gender roles
- gender and gender identity
- sports and gender issues
- media and gender stereotypes
- society and gender roles