close× Call Us
close×
Title Behavioural Foundations of Negative Attitudes Towards Women
Post date 02/01/2018
C1 Background and Explanation of Rationale
(this section was updated on April 13, 2018)

Why do certain individuals display negative attitudes towards women, while others do not? Why are some men unwilling to allow women to access social and economic spaces traditionally dominated by men? And what are the potential consequences of this behaviour to politics, a male-dominated arena that is largely associated with power? These are the questions the current paper seeks to address.

To answer these questions, our point of theoretical departure is the notion that individuals’ treatment of women derives from variation in levels of self-esteem. Specifically, we propose that men with low/fragile or defensive self-esteem are more likely to hold traditional views about gender roles and appraise women more negatively. We also anticipate that men with this described self-esteem profile will be particularly competitive towards women, seeking to maximize their individual fitness by reducing or removing women as competitors. As a traditionally male-dominated setting largely associated with power, men with low/fragile or defensive self-esteem are also expected to electorally reward male candidates only, especially if they hold some level of political ambition.

To test these hypotheses, we employ a survey experiment to assess the relationship between individual-level psychological characteristics and social treatment of women. First, we examine whether individuals who display low-implicit or defensive self-esteem are more likely to hold negative or traditional attitudes towards women.

We then employ a randomly assigned experimental component to investigate whether differences in negative or traditional perceptions of women translate into behaviours that are reflective of alternative strategies for social competition—whether men who hold these views seek to maximize their fitness vis-à-vis others differently, depending on whether the “other” is a man or a woman. The question here is then, why do some men deny women as capable social competitors?

Finally, we assess whether these social competitive strategies translate into individuals’ voting strategies by assessing whether men with low or fragile self-esteem are more hesitant to vote for women.
C2 What are the hypotheses to be tested?
(this section was updated on April 13, 2018)

Why do certain individuals display negative attitudes towards women, while others do not?
• H1A: In men, negative or traditional perceptions of women will be correlated with low/fragile or defensive self-esteem. In other words, we expect that men with low/fragile or defensive self-esteem will be more likely to hold negative or traditional attitudes towards women.
We also consider that this relationship might be mediated by two other individual-level factors, namely: social competitiveness and cognitive ability. As such, we also hypothesise:
• H1A1: Men with low/fragile or defensive self-esteem will be more likely to hold negative or traditional attitudes towards women, particularly if they display high levels of social competitiveness.
• H1A2: Men with low/fragile or defensive self-esteem will be more likely to hold negative or traditional attitudes towards women, particularly if they display low levels of cognitive ability.
We are also interested in the factors that lead women to hold negative or traditional perceptions of women.
• H1B: In women, negative or traditional perceptions of women will be correlated with low/fragile self-esteem, but not defensive self-esteem. In other words, we expect that women with low/fragile self-esteem will be more likely to hold negative or traditional attitudes towards women.
We also consider that this relationship might be mediated by three other individual-level factors, namely: self-efficacy, general competitiveness, and cognitive flexibility. As such, we also hypothesise:
• H1B1: Women with low/fragile self-esteem will be more likely to hold negative or traditional attitudes towards women, particularly if they display low levels of self-efficacy.
• H1B2: Women with low/fragile self-esteem will be more likely to hold negative or traditional attitudes towards women, particularly if they display low levels of general competitiveness.
• H1B3: Women with low/fragile self-esteem will be more likely to hold negative or traditional attitudes towards women, particularly if they display low levels of cognitive flexibility.
Why are some men unwilling to allow women to access social and economic spaces traditionally dominated by men?
• H2: Men with low self-esteem will be particularly competitive towards women, seeking to maximize their individual fitness by reducing or removing women as competitors. Given our research design, this means that we expect that: compared to cues about a successful man, cues about the success of a woman will negatively affect the self-esteem of men who have low-implicit or defensive self-esteem.
What are the potential consequences of this behaviour to politics, a male-dominated arena that is largely associated with power?
• H3: Men with low self-esteem will be less likely to support women occupying positions of power; as such, men with low/fragile or defensive self-esteem will be less likely to have voted for women in the past, as well as more hesitant to vote for women in the future.
C3 How will these hypotheses be tested? *
(this section was updated on April 13, 2018)

Data Collection
This study adopts a three-step research approach, in which we: 1) assess individuals’ perceptions of women; 2) estimate individuals’ levels of self-esteem and, 3) analyse whether individuals’ levels of self-esteem and its expectedly correlated perceptions of women are associated with individuals’ competitive behaviour or cognitive flexibility.

Data for all three steps will be collected through an online survey hosted by Qualtrics. For the project, participants will be recruited from Amazon Mechanical Turk (mTurk). Sample will be restricted to US residents over the age of 18. To motivate recruitment, each participant will be given a direct $0.50 payment as well as be entered into a draw for one of 20 cash prizes of$15.00 each. Prizes will be awarded to participants as Amazon Gift cards.

The survey will be composed of a combination of traditional survey elements to measure participants’ demographics and backgrounds (including their explicit views towards women, cognitive abilities, and self-reported self-esteem); and a two-step Implicit Association Test (IAT) to capture participants’ implicit levels of self-esteem and social competition strategies.

Prior to the IAT, participants complete a 29-question demographic battery, including sex, age, education, ethnicity, income, religious affiliation, religious observance, political orientation, and voting behaviour.

At this stage, participants also complete a three-question measure of cognition flexibility, a Rosenberg self-esteem scale (RSE), a self-efficacy measure by Chen et al., 2001[1], measures of dominant competitiveness and general competiveness (Newby and Klein 2014), measures of hostile and benevolent sexism (Glick and Fiske 1996), and a measure of social dominance orientation (Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. 1994).

Post-IAT, participants are invited to complete three additional batteries designed to capture ideological differences: 1) an eight-question measure of right-wing populist attitudes (modified from Schneider 2008), a Tripartite measure of conservatism (Duckitt et al. 2010), and a multi-dimensional measure of social values (Smith et al. 2013).

Variables are coded as following:

Sex is assessed as a binary response variable in which Male =0, Female =1.

Age is a continuous variable.

Political orientation is measured as a 7-point scale that ranges from -3 (very liberal) to 3 (very conservative) and in which 0 corresponds to “neutral”.

Past vote for women is a variable coded based on respondents’ answer to the question “have you ever voted for a woman for any political office?” It gives respondents the option of selecting up to six offices; as such, the variable ranges from 0 to 6, where 0 means that a respondent has never voted for a woman and 6 means that a respondent voted for a woman for all six types of offices listed.

Prospective vote for women is a variable coded based on respondents’ answer to the question “if given the opportunity, would you vote for a female candidate?” The question is only displayed to those who never voted for women in the past. The variable ranges from 0 to 4, in which 0 refers to “No, I would prefer not to,” 1=“I would vote for a woman for some offices, but not others,” 2=“Maybe,” 3=“Yes,” and 4 is assigned to those who have voted for a woman in the past.

Cognitive flexibility is measured on a continuous 4-point scale, based on the outcome of three problem solving tasks.

Explicit self-esteem is assessed in response to the question “Overall, would you say you are a person who has a high sense of self-esteem?” The question is measured as a continuous variable based on 7-point self-report Likert scale.  The scale ranges from -3 (strongly disagree) to 3 (strongly agree) and in which 0 corresponds to “neutral”.

Rosenberg Self-esteem scale. As an alternative measure of self-esteem participants also completed a Rosenberg self-esteem scale (RSES, Rosenberg 1965). The RSES captures explicit self-esteem as a continuous composite measure 10 questions on a 4-point scale.

Self-efficacy scale. As an alternative measure of self-evaluation individual participants also completed a measure of self-efficacy (Chen et al. 2001). The scale is based on a continuous composite measure using 8 questions measured on a 5-point Likert scale.

Competitiveness is measured as a continuous composite variable based on a 25-question measure, transformed into a 5-point Likert scale (Newby and Klein 2014). This measure breaks into two subgroups:
General Competiveness (12 questions)
Dominant Competiveness (13 questions)

Hostile and Benevolent Sexism is a continuous composite measure based on 22 questions, transformed into a 6-point scale. This measure breaks down into two main subcategories:
Hostile Sexism (11 questions)
Benevolent Sexism (11 questions)
Protective Paternalism (4 questions)
Complementary Gender Differentiation (3 questions)
Heterosexual Intimacy (4 questions)

Social Dominance Orientation is a continuous composite measure based on 14 questions, transformed into a 7-point Likert-scale (Pratto et al., 1994)

IAT and Experimental Component:
After answering a series of demographic questions, participants will be asked to complete two IAT tests. Following Karpinski, participants first complete an “unspecified-other” IAT in which they are asked to associate positive and negative terminology with themselves and with another unspecified individual. Next, participants are randomly assigned to complete one of two specified-other IATs. For one group of respondents, the “other” is a specified highly successful male, and for a second group of respondents, the specified “other” is a highly successful female.

Unspecified-other IAT: For this IAT, the evaluative dimension will be labelled as “pleasant” and “unpleasant” and the self-dimension will be labelled “self” and “other”. Five target words will be used for each category.

Pleasant: Smart; Capable; Success; Worthy; Proud
Unpleasant: Dumb; Useless; Failure; Inept; Ashamed

Self: I; Me; My; Myself. Mine
Other: Them; Their; They; Themselves; Its

Specified-other IAT: For this IAT, the evaluative dimension will be the same as the unspecified-other IAT. The difference is that in this IAT the “other” will be given a specific identity. Specification of the other is done in two ways: by providing respondents with a short description of the identity of one of two individuals (one female and one male) and then by employing the respective individual’s name and gender pronouns. In Treatment 1, the gender will be female, as indicated by the name “Barbara”. In Treatment 2, the gender will be male, as indicated by the name “Robert”. Assignment to one of the two specified-other IATs is done through complete random assignment.

Before the specified-other IAT, participants will be instructed as follows:

“In a few moments, you will be asked to complete the second half of the experiment. As in the first part you just completed, you will be assigning attributes to yourself and to another person. However, unlike in the first part, where the “other” was an unspecified individual, the other person you are asked to evaluate in the upcoming activity is real.

On the next page, you will read the description of this real person. To avoid potential conflicts with our results, we have withheld their full identity, but provide a short description of their background and accomplishments.”

Participants are then randomly allocated to either Treatment 1 or to Treatment 2:

Treatment 1: Female
Treatment 2: Male
Profile:
“Barbara is a real person.

She is a successful American businesswoman and entrepreneur. After completing University, she started her own business and developed it into a successful Fortune 500 company. Currently, Barbara has a net worth of over $50 Million. As a result of her success, she has been interviewed by numerous television and media outlets about her leadership in business.” Profile: “Robert is a real person. He is a successful American businessman and entrepreneur. After completing University, he started his own business and developed it into a successful Fortune 500 company. Currently, Robert has a net worth of over$50 Million.

As a result of his success, he has been interviewed by numerous television and media outlets about his leadership in business.”
IAT:
Pleasant: Smart; Capable; Success; Worthy; Proud
Unpleasant: Dumb; Useless; Failure; Inept; Ashamed

Self: I; Me; My; Myself. Mine
Other: Her; Hers; Herself; Theirself; Barbara
IAT:
Pleasant: Smart; Capable; Success; Worthy; Proud
Unpleasant: Dumb; Useless; Failure; Inept; Ashamed

Self: I; Me; My; Myself. Mine
Other: Him; His; Himself; Theirself; Robert

We employ IAT results to code the following variables:

Implicit self-esteem is a continuous variable measured as respondents’ standardized D-scores in the unspecified-other IAT. D-scores are constructed using the code provided by the IATgen package. The package uses the standard IAT procedures outlined by Greenwald et al. 2003; and Lane et al., 2007 (see Carpenter et al., 2017 for full details on the IATgen package). Data are analyzed within subjects. A standardized difference score (D-score) is calculated for each participant. D-scores have a potential range of -2 to 2. A D-score of 0 means that an individual subject was equally fast in both conditions, with no implicit preference. A positive-D-score indicates an individual is faster in the compatible block (seeing themselves as implicitly good); a negative D-score indicates an individual is faster in the incompatible block (seeing the other as implicitly good). Following standard practices within psychology, D-scores of implicit self-esteem can be interpreted as follows:
Very low implicit self-esteem (<-0.65)
Low implicit self-esteem (-0.35/-0.64)
Moderately low implicit self-esteem (-0.15/-0.34)
Moderately high implicit self-esteem (0.15/0.34)
High implicit self-esteem (0.35/0.64)
Very high implicit self-esteem (>0.65)

Defensive self-esteem refers to individuals with high explicit self-esteem, but low implicit self-esteem. The variable is constructed as a categorical variable which consists of individuals with high explicit self-esteem on the self-report measure, but low implicit self-esteem on the unspecified-other IAT. As above, following established practice for the interpretations of an IAT in psychology, we define low-implicit self-esteem based on three levels of significance (-0.15, -0.35, -0.65). Explicit self-esteem is measured on a 7-point Likert scale. Defensive self-esteem is therefore modelled based on the following:
Very high explicit
(Explicit 7, Implicit -0.15)
(Explicit 7, Implicit -0.35)
(Explicit 7, Implicit -0.65)
High explicit
(Explicit 6-7, Implicit -0.15)
(Explicit 6-7, Implicit -0.35)
(Explicit 6-7, Implicit -0.65)
Low explicit
(Explicit 5-7, Implicit -0.15)
(Explicit 5-7, Implicit -0.35)
(Explicit 5-7, Implicit -0.65)

Statistical Analyses:

H1A: Men with low/fragile or defensive self-esteem will be more likely to hold negative or traditional attitudes towards women.

H1B: Women with low/fragile self-esteem will be more likely to hold negative or traditional attitudes towards women.

We test these hypotheses through a series of standard OLS regression models with robust confidence intervals. We measure attitudes towards women as a continuous variable capturing hostile and benevolent sexism (as per the battery developed by Glick and Fiske 1996). We are interested in heterogeneous effects based on sex, so we test each model for two samples: one restricted to male respondents and another restricted to female respondents. Model specifications will include the following independent variables:

Regression Model 1: Implicit self-esteem
Model A1: Implicit self-esteem
Model A2: Implicit self-esteem and competitiveness
Model A3: Implicit self-esteem and cognitive flexibility
Model A4: Implicit self-esteem competitiveness, and cognitive flexibility

Regression Model 2: Defensive self-esteem #1
Model B1: Explicit self-esteem
Model B2: Explicit*Implicit self-esteem
Model B3: Explicit*Implicit self-esteem and competitiveness
Model B4: Explicit*Implicit self-esteem and cognitive flexibility
Model B5: Explicit*Implicit self-esteem competitiveness, and cognitive flexibility

Regression Model 3: Defensive self-esteem #2 (Explicit based on a Rosenberg Scale)
Model C1: Rosenberg Scale
Model C2: Rosenberg Scale *Implicit self-esteem
Model C3: Rosenberg Scale *Implicit self-esteem and competitiveness
Model C4: Rosenberg Scale *Implicit self-esteem and cognitive flexibility
Model C5: Rosenberg Scale *Implicit self-esteem, competitiveness, and cognitive flexibility

Regression Model 4: Self-Efficacy
Model D1: Self-Efficacy
Model D2: Self-Efficacy*Implicit self-esteem
Model D3: Self-Efficacy*Implicit self-esteem and competitiveness
Model D4: Self-Efficacy*Implicit self-esteem and cognitive flexibility
Model D5: Self-Efficacy*Implicit self-esteem competitiveness, and cognitive flexibility

We also run robustness checks with controls for age, race, political orientation, educational background, and income.

H2: Compared to cues about a successful man, cues about the success of a woman will negatively affect the self-esteem of men who have low-implicit or defensive self-esteem.

To test H2, we employ a series of one-sided T-tests. These tests assess whether individuals’ levels of self-esteem are affected depending on whether they were assigned to take the female (Treatment 1) or male (Treatment 2) versions of the specified-other IAT. Our dependent variable is thus the difference between participants’ scores in the specified-other and unspecified-other IATs (i.e., specified-other D-score minus unspecified-other D-score).

As is common practice in experimental research, we begin our analysis with an overall test of differences in means between Treatment 1 and Treatment 2 groups.

As we are predominantly interested in heterogeneous effects, we also sub-divide our sample into specific sub-groups and conduct the same T-tests:
Sex (male, female)
implicit self-esteem (men with low-implicit self-esteem, men with high implicit self-esteem) (see coding as above)
Defensive self-esteem (men with low-defensive self-esteem, men with high defensive self-esteem) (see coding as above)

Amendment: In a previous iteration of this proposal, we considered using a heterogeneous model to compare the difference in means between groups. This proposed model also contained a series of control variables. Upon further consideration, we believe that this model introduces a necessary level of complexity without sufficiently contributing to the evaluation of the core hypotheses. Consequently, we will analyse our statistical results directly through a series of T-tests. The decision to adjust our statistical approach was made prior to the review of any data collected for this study.

H3: Men with low/damaged self-esteem will be less likely to support women occupying positions of power.

We test these hypotheses using two regression models with robust confidence intervals, a general linearized regression (GLM), and an ordered logit regression. We measure respondents’ support for women occupying positions of power with two variables: past vote for women (GLM) and prospective vote for women (ordered logit). Model specifications will include the following independent variables:

Regression Model 1: Implicit self-esteem
Model A1: Sex, Implicit self-esteem
Model A2: Sex, Implicit self-esteem and competitiveness
Model A3: Sex, Implicit self-esteem and cognitive flexibility
Model A4: Sex, Implicit self-esteem competitiveness, and cognitive flexibility

Regression Model 2: Defensive self-esteem #1
Model B1: Sex, Explicit self-esteem
Model B2: Sex, Explicit*Implicit self-esteem
Model B3: Sex, Explicit*Implicit self-esteem and competitiveness
Model B4: Sex, Explicit*Implicit self-esteem and cognitive flexibility
Model B5: Sex, Explicit*Implicit self-esteem competitiveness, and cognitive flexibility

Regression Model 3: Defensive self-esteem #2 (Explicit based on a Rosenberg Scale)
Model C1: Sex, Rosenberg Scale
Model C2: Sex, Rosenberg Scale *Implicit self-esteem
Model C3: Sex, Rosenberg Scale *Implicit self-esteem and competitiveness
Model C4: Sex, Rosenberg Scale *Implicit self-esteem and cognitive flexibility
Model C5: Sex, Rosenberg Scale *Implicit self-esteem, competitiveness, and cognitive flexibility

Regression Model 4: Self-Efficacy
Model D1: Sex, Self-Efficacy
Model D2: Sex, Self-Efficacy*Implicit self-esteem
Model D3: Sex, Self-Efficacy*Implicit self-esteem and competitiveness
Model D4: Sex, Self-Efficacy*Implicit self-esteem and cognitive flexibility
Model D5: Sex, Self-Efficacy*Implicit self-esteem competitiveness, and cognitive flexibility

We also run robustness checks with controls for age, race, political orientation, educational background, and income.

[1] Self-efficacy is defined as “beliefs in one’s capabilities to mobilize the motivation, cognitive
resources, and courses of action needed to meet given situational demands” (Wood & Bandura, 1989, p. 408, in Chen et al., 2001, p 62).

C4 Country United States
C5 Scale (# of Units) N=600+; 2-Sample, 1-Sided Statistical Power Test. Calculated a power based on a 10% Effect and a range fo standard deviations between 40%-70% of Effect size. Produces sample range of n=260 to n=794
C6 Was a power analysis conducted prior to data collection? Yes
C7 Has this research received Insitutional Review Board (IRB) or ethics committee approval? Yes
C8 IRB Number 17.12.2
C9 Date of IRB Approval Dec 15th 2017
C10 Will the intervention be implemented by the researcher or a third party? Researchers
C11 Did any of the research team receive remuneration from the implementing agency for taking part in this research? No
C12 If relevant, is there an advance agreement with the implementation group that all results can be published? No
C13 JEL Classification(s) not provided by authors