Bridging the gap: The effect of gender pay gap reporting on salary
negotiation outcomes
Hazel Ahern-Flynn
11804866
Msc Economics: Behavioural Economics and Game Theory
15 ECTS
Statement of Originality
This document is written by student Hazel Ahern-Flynn who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
1. Introduction and Motivation
Gender inequality is evident in the majority of labour markets. The World Bank estimates that, across 141 countries considered, there is a global loss of 160.2$ trillion in human capital wealth due to gender inequality in lifetime labour earnings (Wodon & Brière, 2018). In a similar vein, estimates of the average gender pay gap across the EU place the figure at 16%, indicating that on average women in Europe earn 84c for every euro made by a man in an hour (Eurostat, 2018).
Figure 1. Unadjusted gender pay gap by EU member, 2016
Note. Reprinted from Gender pay gap statistics, by Eurostat, retrieved from
The issue of reducing gender pay gaps has proven relatively challenging. Recent policies aimed at tackling the issue have included the UK’s new regulation requiring mandatory pay gap recording and reporting for all firms with 250 or more employees (The Equality Act 2010). Likewise, Germany has implemented a similar policy of mandatory wage gap reporting (Buck, 2018). Additionally, Australia, the USA, Belgium, Austria and Iceland all require that firms above a given size report gender wage gaps in some form (Workplace Gender Equality Agency, 2018).
As governments increasingly require gender wage gap reporting, it is therefore timely and relevant to consider whether increasing wage transparency and making wage gap figures publicly available has any effect on the wage gap itself. As such, this thesis seeks to examine the effect of wage gap reporting on salary inequality. More specifically, this thesis will test whether this ‘naming and shaming’ approach to gender pay gaps may create a descriptive norm which encourages the worst performers to decrease their pay gaps.
Gender pay gaps may emerge through several routes: women self-selecting out of the labour force our out of higher paying careers, bias in awarding promotions and raises, and losses in promotions and salary increases due to absence for childcare.
Although gender pay gaps can arise through several mechanisms, the analysis presented here will focus on gender pay gaps as they occur through salary negotiations as this is often the initial point at which gendered pay disparities emerge, which then may persist
throughout and individual’s career.
To this end, the research question considered in this thesis is as follows: Does gender pay gap reporting reduce gender bias in salary negotiations?
Addressing this research question should provide policy makers with further insights into how effective mandatory reporting interventions are in the area of gender inequality and pay gaps in general. If activating descriptive norms proves effective in reducing gender pay gaps, a similar approach may be applied to other inequalities arising from social bias, such as pay gaps between ethnicities.
The results of the analysis presented in this thesis suggest that recording and reporting gender pay gaps may increase average salaries offered to female candidates, which in turn
may reduce real-world gender pay gaps. No significant difference was found in the average salary offered to male and female candidates who attempt to negotiate, but establishing norms around gender pay gaps was effective in improving salary negotiation outcomes for women. It was similarly found that establishing pay gap norms had no significant effect on average salaries offered to male candidates. It may therefore be possible to apply this effect of establishing gender pay gap norms to labour markets where there is indeed a gender pay gap, to increase women’s salaries and so reduce the gap.
2. Literature Review
2.1 Social Norms
As described by Sunstein, (2014, p.4) the use of social norms in a behavioural economics context refers to “emphasizing what most people do”, in an attempt to influence what the individual does. For example, informing people that ‘the majority of people recycle’ may encourage non-recyclers to begin recycling. More specifically, this is a descriptive norm, as it describes peoples’ actual behaviour, rather than an injunctive norm, the behaviour people ‘should’ have. There is a wealth of research supporting the idea that social norms can significantly affect peoples’ attitudes and decisions.
The effectiveness of social norms and their potential for heavily influencing human
behaviour has perhaps most notably been demonstrated by Goldstein et al. (2008) in their study of hotel towel reuse. In their study, Goldstein et al. (2008) provided one group of hotel guests with notes requesting they reuse their towels ‘to help save the environment’ and another with notes stating that 75% of hotel guests reuse their towels. They found 9% higher towel reuse rate in the descriptive norm condition.
Similarly, they repeated the experiment with notes stating the descriptive norm for four different social categories for guests to identify with: gender, guests who had stayed at the hotel, ‘citizens’, and guests who had stayed in the same room as the subject. They found that any of the descriptive norms increased the towel reuse rate compared with the control group but the norm among the group which most closely resembled the guests’ current state, describing reuse rates for people who had stayed in the same room, was the most effective. Goldstein et al. (2008) term these norms, which describe the behaviour of people
in contexts similar to the subjects’ current circumstances, ‘provincial norms’. They hypothesize that provincial norms are more relevant and more effective in changing people’s behaviour because it is usually most helpful to follow the norms of one’s specific environment than more general norms i.e it is more useful to know what other people have done when faced with a similar situation to yours than it is to know what ‘people like you’ do in general.
There have been several recent experiments and policy interventions seeking to use descriptive norms to influence public behaviour. The UK’s Behavioural Insights Team has successfully trialled the use of descriptive norms to increase the rate of taxpayers who pay their taxes on time (Behavioural Insights Team, 2012). 140,000 tax payers were randomly assigned to either a control group or one of several social norm groups. The control groups received a standard letter requesting timely payment with no social norm. Recipients in the treatment groups received a letter stating that the majority of British taxpayers pay their taxes on time and some additionally stated that the majority of people in the recipients’ postcode or town had already paid. Similar to the effect seen by Goldstein et. al (2008), all social norm treatments resulted in a higher rate of tax debt payments than the control group and the rate of payments increased with the specificity of the social norm included in the letter, such that describing payment rates in the recipients town resulted in the highest payment rate.
Descriptive norms have also been effective in reducing alcohol consumption and misuse among student populations. Perkins & Craig (2006) implemented a campaign to reduce misperceptions of alcohol misuse rates among student-athletes. Their analysis found that student athletes tend to over-estimate alcohol consumption rates among their peers and so student-athletes had an inaccurate understanding of social norms around drinking. Perkins and Craig (2006) implemented a descriptive norm intervention based on factual data concerning actual alcohol consumption rates and students’ attitudes toward alcohol consumption. Student-athletes were informed through emails, poster campaigns, campus newspapers, information kiosks and direct training that "The majority (66%) of [this school's] student-athletes drink alcohol once per week or less often or do not drink at all” and "88% of student-athletes at [this school] believe one should never drink to an
2006, p.882). They found that after exposure to more accurate social norms, high quantity alcohol consumption, estimated peak blood alcohol levels, and negative consequences from drinking all declined by 30% or more in student-athletes who were exposed to the
intervention while students with no or lower exposure saw no significant change in alcohol use. Similar results have been found across several studies which sought to reduce student alcohol consumption by correcting misperceptions of the prevalent descriptive norms (Perkins, 2002)(Perkins, Haines, & Rice, 2005)(Haines & Spear, 1996).
The above research demonstrates the strong potential for descriptive norm interventions to further public policy aims (in this case problematic alcohol consumption) and also
illustrates that the effectiveness of descriptive norms is mediated by the degree of exposure individuals have to those norms.
While descriptive norm interventions have shown promising results in influencing public behaviour, it is worth noting that descriptive norms should be credible (and preferably true and verifiable) to be effective. In an analysis of one failed social norm campaign to reduce student drinking through social norm interventions, Thombs et al. (2004) surveyed students targeted by the campaign and found that students did not find the social norms they were presented with credible (i.e the statistic indicating a low use of alcohol was not believable), and that students with a higher personal rate of alcohol consumption were less likely to believe that the social norm was as low as the campaign suggested. Therefore, descriptive norm interventions should ideally be presented in such a way that individuals subjected to the intervention can confirm the norm for themselves. The experimental design
implemented in this thesis, as outlined below, incorporates this insight by providing subjects with information describing the gender pay gaps of several firms of a similar size within a single Industry so that subjects can calculate and confirm the social norms surrounding pay gaps for themselves.
A final example of the potential for descriptive norm interventions to further public policy aims is provided by Nolan et al. (2008). In a study of energy use across 981 households in California, the authors informed households of descriptive norms describing the (usually high) percentage of a recipients’ neighbours which had implemented energy-saving measures such as turning-off lights when not using them. The control messages informed households that taking certain measures to reduce their energy consumption could save
them money. In a comparison of the descriptive norm and cost-saving groups, the authors found that energy use was significantly reduced in the social norm groups, in comparison with the cost saving group despite the fact that cost-saving was in the rational self-interest of homeowners, while conforming to descriptive norms offered no explicit benefit or reward. This study outlines not only the diverse areas of public policy which can benefit from the use of descriptive norm interventions but also that descriptive norms can be even more effective in influencing human behaviour than direct monetary incentives, which a rational actor should theoretically prefer.
To date, the majority of experiments and policy interventions considering social norms have generally focussed on affecting conscious behaviour and decision making. An additional novelty of the research presented in this thesis is the application of descriptive norms to what is most likely unconscious gender bias in salary negotiations. A positive effect resulting from the descriptive norm intervention here may highlight the potential for descriptive norms to influence unconscious biases in other areas.
2.2 Gender Differences in Ultimatum Games
Solnick (2001) considered one-shot ultimatum game outcomes, where 89 pairs of university students were asked to split $10 between proposers and responders, where the proposer proposes an amount to send to the responder. The responder may either accept the amount, or reject it in which case both parties receive nothing.
The ultimatum games were conducted across two treatments; one where players were anonymous and one where the genders of the players are known. Average offers from the proposer did not vary with the gender of the proposer. However, the gender of the
responder did affect the average offer made by proposers. Both male and female proposers on average made higher offers to male responders. Additionally, both female and male responders exhibited higher minimum acceptable offer levels when paired with a female proposer. Solnick (2001) notes that these gender differences in offers and minimum acceptance levels resulted in significantly worse outcomes for women who played the ultimate game. Male responders received offers of an even or greater split (≥$5) 82% of the time, while female responders received offers of an even or greater split only 59% of the time.
From this, Solnick (2001, p.199) concludes that “systemic differences in the expectations and decisions of men and women exist even within this spare negotiating environment, leading to substantial differences in outcomes”. As all proposers offered female responders less on average (despite no difference in responders’ average minimum acceptable offer across genders) and all responders submitted higher minimum acceptable offers when playing against female proposers, Solnick argues that people incorrectly expect that women will be satisfied with less and will be willing to give up more in negotiations. It is further argued that this may explain part of the observed real world gender pay gaps, which can be attributed to differences in bargaining outcomes between men and women.
In direct contrast to the results found by Solnick (2001), Eckel & Grossman (2001) find that offers made by men tend to be rejected more often than offers from women and offers made in women-women pairings are almost never rejected. This is the exact opposite of Solnick’s (2001) finding that women-women dyads are the least likely to reach an
agreement. However, Eckel and Grossman (2001) also found that women’s offers tended to be more generous than men’s. It is therefore possible that the higher observed acceptance rate for offers from women is because female proposer behaviour in the experiment matched expectations that women will give up more in a negotiation, as proposed by Solnick (2001). Supporting Solnick’s (2001) conclusion that women face a different
negotiating environment to men, Eckel and Grossman (2001) did also find that women on average received lower offers from both male and female proposers. Again, this may reflect similar differences in offers women receive in real-world negotiations and may go some way towards explaining the gender pay gap.
As outlined above, no consensus on the effect of gender difference in ultimatum games has been reached. The situation is further complicated by the work of McGee & Constantinides (2013), who find that, while gender differences may exist in one-shot ultimatum games, these differences do no persist past the first round of a multiple-shot ultimatum game. They find that there is some ‘solidarity’ between women-women pairings relative to female-male parings in the first round, where women make higher average proposals to other women and are more likely to propose an even split than female proposers paired with men. However, these gender difference do not persist past the first round.
Several studies have established differences in risk aversion across genders, with women generally being more risk averse than men (Borghans, et al., 2009)(Booth, Cardona-Sosa, & Nolen, 2014). García-Gallego et al. (2012) point out that this difference in risk attitudes is often used to explain difference in ultimatum game outcomes across genders without being explicitly tested for. In their experiment, García-Gallego et al. elicit the risk aversion of male and female participants using lottery panel tests and conduct ultimatum games which are framed as salary negotiations between an employee and employer. They find that women tend to make lower offers and reject offers more than men.
Their experimental results conclude that women are indeed more risk averse than men but their behaviour in ultimatum games is not consistent with risk averse behaviour, where risk averse negotiators should make higher offers to reduce the risk of rejection and maximise their earnings. From this, García-Gallego et al. (2012) argue that gender differences in ultimatum games occur despite rather than because of higher risk aversion among women. While gendered difference in ultimatum game outcomes can shed light on the different negotiating expectations and outcomes women and men face, ultimatum games represent a relatively abstract and artificial environment, where participant actions are limited to a proposal and either complete acceptance or rejection, having no opportunity to negotiate or attempt to bargain for a better outcome after the fact.
The following sections of the literature review will consider other experiments and analysis which more directly examine the effects of gender on decisions enter negotiations and negotiation outcomes.
2.3Gender Differences in Negotiations
Bowles et al. (2007) note previous research indicating that women are less likely to enter negotiations than men. A lower propensity to negotiate is often proposed as an explanation of the gender pay gap, where women get less than men simply because they don’t ask for more. Bowles et al. (2007) contend that women are more reluctant to enter into
negotiations because they are punished for doing so. In two survey experiments, subjects were asked to imagine that they were a manager evaluating a candidate for a position with their firm. Subjects were provided with background information on the candidate. In the first experiment they were provided with a resume and notes about the candidate’s
interview which included the gender of the candidate and whether the candidate had attempted to negotiate for extra compensation or job benefits. Apart from candidate’s gender and decision to negotiate or not, the content of the documents provided were the same across conditions. Subjects were then asked on a scale of 1-7 how likely they thought it was that they or someone else at the company would hire them. They combined the answers to both questions into a composite hireabilty measure.
Using ANOVA to test for differences in the hireability measure between gender of the
subject, gender of the candidate and ask or no ask (whether the candidate had attempted to negotiate) conditions, Bowles et al. (2007, pp.88-89) found a significant negative main effect for candidate negotiation and a significant negative interaction effect between gender of candidate and whether the candidate had attempted to negotiate or not. That is, they found that subjects considered both men and women who attempted to negotiate less hireable than those who did not but the effect was twice as large for women.
Similarly, in the second experiment subjects were asked to review background information, a resume and an evaluation survey for a candidate through an online survey. The
experimental procedure was broadly the same as in experiment one but the interview transcripts provided for the candidate contained either no attempt at negotiating, a ‘moderate’ request for a higher salary or a ‘strong’ request for a higher salary and a bonus. They asked subjects several questions regarding how happy they would be to work with the subject
In the same vein as experiment one, they found a significant main negative effect for candidate negotiation and a significant negative interaction effect between gender of candidate and candidate negotiation, although there was no significant difference between the mean of willingness to work with male candidates who negotiated and male candidates who did not. The authors concluded that women do indeed face a higher social cost than men for attempting to negotiate, and on average are considered less hireable, and people are less willing to work with them as a result.
The experimental design presented in section 3 of this thesis is partially based on the experiments 1 and 2 conducted by Bowles, Babcock, & Lai (2007) as their work
demonstrates how gendered negotiation differences can be successfully studied using remotely administered surveys.
Regarding gender differences in the propensity to negotiate, Leibbrandt & List (2015) conducted a large-scale natural field experiment, where open positions for two real jobs were advertised. After candidates had expressed interest in the jobs those assigned to treatment group one received additional information about the position but no explicit mention of whether salary was negotiable or not. The second treatment group were
explicitly informed that the salary was negotiable. Leibbrandt & List (2015) found that, when wages were not explicitly stated to be negotiable, men were more likely to negotiate than women and women were more likely to offer to work for lower wages than men. However, when candidates were explicitly informed that wage were negotiable, both genders are equally likely to negotiate.
It’s therefore possible that in ambiguous negotiation environments, women self-select out of the negotiation process which in some cases may explain some of the observed
differences in starting salaries between men and women.
Finally, Exley, Niederle, & Vesterlund (2016) conducted a lab experiment to examine whether women negotiate less than men and potential reasons for this. Subjects were assigned to worker-firm pairs. Each firm and worker contributes some amount to the joint revenue, which is determined by the subjects’ performance in a task prior to the
negotiation. Subjects were also assigned either to a forced or optional treatment, where they were either forced to negotiate for a portion of the joint revenue with the firm or had the option to either negotiate or accept the initial computer-generated wage. If
negotiations failed to result in an agreement both the worker and firm received a $5 penalty and the initial computer-generated wage was implemented.
The authors found that, when subjects were forced to negotiate, men and women on average earned equal returns. They also observed that, when not forced to negotiate, women entered negotiations less often than men. However, the authors also found that, in the forced treatment, when women entered negotiations they would not otherwise choose to enter, they earned lower returns than when they have the choice to negotiate or not. When women had the choice to negotiate or not they generally positively selected into
negotiations which were more likely to be successful and so earn them a higher return. The authors conclude that, while there is evidence of women entering negotiations less often than men, this is the result of women being able to correctly identify when it is beneficial to negotiate.
The primary novelty of the research presented in this thesis comes from the combination of both research areas outlined above; the effectiveness of social norms in altering human behaviour and experimental analysis of gender pay gaps and how they arise. The
methodology of this thesis therefore seeks to implement the insights provided by previous descriptive norm experiments to the issue of reducing gender pay gaps.
2. Experimental Design Design
The experiment featured a 2 (descriptive norm: Treatment vs. Control) X2 (Gender of
candidate), between subjects design. Subjects were randomly allocated to either descriptive norm treatment or control groups. Within the treatment and control groups, the genders of both candidates shown to subjects were also randomly assigned, such that is was possible for subjects in either the treatment or control group to have seen one male and one female candidate, two male or two female candidates.
All subjects were asked to imagine that they were a HR manager attempting to fill two open positions at a hypothetical company, ‘TriCorp’. Relevant extracts of both the treatment and control instructions and surveys, along with male and female candidate emails are included in the Appendix.
Control Condition
Subjects in the control condition were then asked to read two emails from candidates who had each been offered one of the open positions with the company. The content of the emails and following questions were identical between male and female candidates other than the name of the candidate and the male or female pronouns used to refer to them in the survey questions.
Each candidate email attempted to negotiate the terms of their contracts. The first candidate requested that their salary be increased from €50,000 to €55,000 to match an
offer they received from another firm. The second candidate requested the ability to work from home two days a week to achieve a better work life balance. The inclusion of two different negotiations requests is to facilitate the comparison of gendered bias in monetary and non-monetary benefit negotiations.
Subjects were then asked to offer the first candidate a salary in the range of €40,000-€60,000, in response to their attempt to negotiate. The range extends above the requested increase and below the initial salary offer to allow the subject to punish or reward the candidate for negotiating, if they so wished. Subjects then answered a series of questions on a scale of 1-5 indicating how reasonable they thought the request was, how happy they would be to work with the candidate and how well a set of words described the tone of the email: Confident, Aggressive, Polite, Assertive, Friendly, Arrogant, Competent, Demanding. Subjects were then asked to review the email from the second candidate, and were asked on a scale of 1-5 how likely they were to agree to the candidate’s request. They were then asked to answer the same questions about reasonableness, willingness to work with the candidate and words describing the tone of the email as before.
Descriptive Norm Condition
Subjects in the treatment group received exactly the same instructions, candidate emails and survey questions as the control group but before being shown the negotiation emails they were shown a table ranking the gender pay gaps of the ‘ten largest firms in the same industry as TriCorp, including TriCorp itself’ from lowest gap to highest. The treatment attempts to activate descriptive norms in the subjects by emphasizing that the firms being compared are of a similar size and from the same industry as the firm they are representing. The actual industry is never specified to avoid unconscious Industry related bias in subjects, for example by associating male candidates more with the construction industry.
There are several measurements of gender pay gap used and published by policy makers, including mean, median, hourly wage, yearly wage etc. To assist subjects in understanding the table and increase ecological validity, they were provided with the following definition of mean gender pay gap, which is similar to the definition used to report UK mean gender pay gap figures:
‘The mean gender pay gap is a measure of the difference between women’s mean hourly
wage and men’s mean hourly wage. For example, with a mean gender pay gap of 10%, when comparing mean hourly rates, women earn 90c for every €1 that men earn.’
The values for the gender pay gaps were generated to be a similar size to those commonly reported in the UK to increase believability. The pay gap table ranks TriCorp as having the 8th out of 10 in terms of pay gap, indicating that 7 of the other 9 firms have a lower gender
pay gap.
As a manipulation check, subjects were then asked a series of simple questions about the pay gap table they were shown, to demonstrate that they understood the information and encourage subjects to focus on the information. Subjects were presented with the following two correct statements which establish a local norm of lower gender pay gaps, relative to TriCorp’s gender pay gap:
1. 70% of the ten largest firms in this industry have a lower gender pay gap than TriCorp.
2. TriCorp has a higher gender pay gap than most of the other firms.
Subjects were then asked to indicate if either question was true or false. The statements establishing a descriptive norm were presented as true or false questions to ensure that the subject was actually engaging with the information via the survey feedback rather than passively reading or skipping it. This also allowed subjects to verify the norm for themselves by checking the ranking table. Additionally, these questions act as a check that subjects were correctly interpreting the ranking table and information they are being shown. Similarly, subjects were asked to selects the firms with the highest and lowest gender pay gaps to determine if they were reading the table correctly.
Gender manipulation check
After the survey, all subjects were asked to identify the gender of each candidate they had been shown, to ensure the gender manipulation was effective.
The number of questions per condition was kept small to reduce the duration of the survey and so encourage a higher number of responses. Longer survey experiments tend to see a
high drop-out rate, where subjects do not complete the experiment. This was effective, as the completion rate among participants was over 90%.
Participants
In total, 135 subjects were recruited using online social network sites and participants were able to refer other participants to the survey through an anonymous URL. To incentivise participation, all subjects were eligible to enter a random lottery for a €50 gift voucher for the online marketplace Amazon.com. The experiment was remotely administered to subjects via the Qualatrics online survey platform. All subjects were asked to provide basic demographic (age, gender, occupation) information at the beginning of the survey. Of the 135 subjects, four were excluded from analysis for failing to identify as true either that ‘70% of the companies listed have lower gender wage gaps than TriCorp’ or that ‘TriCorp has a higher gender wage gap than most other firms’. Answers of false to either of these questions indicate that the treatment of establishing a local norm had not been effective, either because the subjects had misread the information provided or did not find it believable.
A further 13 subjects were excluded because they incorrectly identified the gender of one or both candidates, indicating that the gender manipulation had not been effective or they could not recall the gender by the end of the experiment. The remaining 118 subjects were assigned as described in Table 1.
Table 1.
Subject Assignment
Control Norm Treatment
N=60 N=58
Gender of
Candidate 1 Male Female Male Female
N=26 N=34 N=28 N=30
Gender of
Candidate 2 Male Female Male Female
N=33 N=27 N=26 N=32
The experiment achieved a gender balance in subjects, with 60 female participants and 58 male participants. Tables 2, 3 and 4 further describe the demographics of the sample.
Table 2.
Subjects’ Occupation N Percent
Student 18 15.4%
Employed Full Time 82 70.1%
Part Time 8 6.8%
Retired 3 2.6%
Unemployed 7 6.0%
Table 4.
Subjects' Age N Percent
18-24 41 34.7% 25-34 50 42.4% 34-44 7 5.9% 45-54 8 6.8% 55-65 11 9.3% 65-74 1 0.8%
Since this thesis is concerned with the effect (if any) of pay gap norms in reducing pay gaps, it is possible for the pay gap to be reduced from two ‘directions’ ;Either through increasing female salaries or reducing male salaries (or both at once). Therefore the hypotheses tested in this research are:
Hypothesis 1:
H0: Gender pay gap norms have no effect on average salaries offered to female candidates
Ha: Gender pay gap norms increase average salaries offered to female candidates
Hypothesis 2:
H0: Gender pay gap norms have no effect on average salaries offered to male candidates
Ha: Gender pay gap norms reduce average salaries offered to male candidates 4. Results and Discussion
Salary Offers
The effects of gender of candidate and the norm treatment on salary offers were analysed using two-way ANOVA (analysis of variance).
Results of the 2 (Gender: Male or Female) X 2 (Norm Treatment or no Treatment) ANOVA are presented in table 5 below. There was a statistically significant (P<0.1.) interaction between the effects of gender and pay gap norm treatment on salary, F (1, 114) = 3.52, p = .063. The interaction effect size of 0.114 is considered relatively low, therefore only a small
Table 3.
Subjects' Education N Percent
High School 3 2.5% Some College 13 11.0% Bachelor's Degree 61 51.7% Master's Degree 35 29.7% Doctorate 6 5.1%
proportion of the variation in salaries can be attributed to the interaction of gender and pay gap norm.
No significant main effects of gender or norm treatment were found. Mean salary offers by group are presented in table 6.
Post-hoc Tukey's HSD tests showed that, when subjects were exposed to gender pay gap norms, the female candidate received a higher average salary offer than the male candidate at the 10% level of significance. All other pairwise comparisons were not significant (Full results presented in appendix).
Table.5 ANOVA of Salary by norm treatment and gender of candidate (N=117)
Source df MS F η2 Female 1 82825.59 0.81 0.099 Treatment 1 2898572 2.49 0.066 Female X Treatment 1 8937567 3.52* 0.114 Error 11 4 * P<0.1.
Table 6. Mean Salary by gender of candidate and norm treatment No Treatment Treatment Female Candidate Male Candidate Female Candidate Male Candidate 53038.38 53142.35 54014.63a 52799.11b (2128.51) (1692.70) (2068.70) (1514.16) n n n n 34 26 30 27
Note: Standard deviations reported in parentheses below
means. Different superscripts between means indicate significant mean difference, ab at the p<0.1 level as tested by
TukeyHSD pairwise comparison.
As can be seen from Table 6, activating gender wage gap norms among subjects lead to a statistically significant higher mean salary (€1215 more) being offered to the female
candidate than the male candidate. Additionally, the significant interaction effect indicates that establishing gender pay gap norms does improve negotiation outcomes for women. As such, we reject the 1st null hypothesis outlined above, that pay gap norms have no effect on
salaries offered to female candidates. Figure 2. Interaction plot
The above results provide some tentative evidence that establishing pay gap norms among employers may reduce the gender pay gap.
It seems that the norm treatment was effective in increasing average salary offered to the female candidate relative to the male candidate. As there was no gender pay gap (no statistically significant difference in mean salary offered to male or female candidates) observed in the control condition, this interestingly resulted in a gender pay gap in favour of women. Rather than reducing an observed bias, as expected, the pay-gap norm treatment appears to have created a bias.
However, this result is still useful from a policy analysis perspective. While no gender pay gap was observed in the control condition, gender pay gaps are still observed in real-world labour markets. If gender pay gap norms improve average salary negotiation outcomes for
4 5 0 0 0 5 0 0 0 0 5 5 0 0 0 6 0 0 0 0 Sa la ry 0 1 Norm Treatment Male Female
women, then mandatory reporting and publishing of pay gaps may in fact reduce gender pay gaps.
It is worth noting from table 6 that the pay gap norm treatment did not significantly reduce the average salary offered to the male candidate, indicating that activating gender wage gap norms may improve outcomes for women without making men worse off in absolute terms. Again, this is useful to know from a policy perspective as it supports the idea that gender-pay gap reporting may improve outcomes for women without significantly distorting salaries offered to men. As such we fail to reject the 2nd null hypothesis outlined above: There is no
evidence that gender pay gap norms reduce average salary offers to male candidates.
Assumptions for ANOVA
The TM group (male candidate, received norm treatment) contained a large outlier; one subject offered a salary of €60,000, which is double the salary increase requested and more than three standard deviations above the mean. As two-way ANOVA is sensitive to outliers, the outlier was dropped, leaving 117 observations.
For comparison, 2 way ANOVA results and mean salaries for the sample including the dropped observation (N=118) are presented in the appendix. No significant main effects or interaction effect were found.
Levene’s test indicates that the assumption of equality of variance is met, F(3,112)=0.88, P=0.45. The assumption of independence of observations is met by the between-subjects design of the experiment.
Upon plotting, residuals seem somewhat non-normal (residual plot and histogram of residuals presented in appendix). However, ANOVA, and F tests in general, are robust for validity against non-normality. More specifically, Harwell et al. (1992) have demonstrated that the false positive rate is not affected very much by violation of the normality
assumption.
Since the above results provide some evidence that establishing pay gap norms improves negotiation outcomes for women, it would be useful to know if this is due to a change in
perception i.e if, after being exposed to gender pay gap norms, subjects consider salary negotiations by women to be more reasonable.
To test if there is any difference in the frequency with which subjects consider the salary negotiation to be reasonable a Chi-square test was conducted. Subjects were asked on a 1-5 scale how reasonable they considered candidate 1’s request. This linkert scale data was transformed into three categories. Answers of 1-2 were coded as unreasonable, 3 as neutral and 4-5 as reasonable. The Chi square test was then conducted to test the hypothesis that there is no difference in the distribution of reasonableness ratings across all four
experimental groups; Control female candidate, control male candidate, norm treatment female candidate, norm treatment male candidate. The contingency table presented in appendix.
The Chi-square test found no significant difference at the 5% level in the frequency with which subjects rated the request as reasonable or unreasonable: X2 (6, N = 118) = 2.2582, p=.894486
Tale 7. % of Subjects who rated the negotiation as reasonable, neutral or unreasonable, by gender and norm treatment
No Treatment Treatment Female Candidate Male Candidate Female Candidate Male Candidate Unreasonable 5.88% 3.85% 6.67% 3.57% Neutral 17.65% 23.08% 10.00% 21.43% Reasonable 76.47% 73.08% 83.33% 75.00% total 100.00% 100.00% 100.00% 100.00% As outlined in the literature review, previous studies have found that there is a gendered bias, where people take a more negative view of women entering negotiations than they do men. The above results would seem to contradict this, as we can see that there is no
significant difference in the frequency with which subjects consider male or female candidate’s negotiations to be reasonable in either the control or treatment conditions. Given the norm treatment appears to make no difference in how reasonable subjects consider the negotiations to be, it does not seem as the though the effect of norm
treatment on female salary outcomes is the result of subjects consciously changing their perception of the validity of the negotiation.
Negotiations regarding non-monetary benefits
Recall that candidate 2 attempted to negotiate for the option to work from home two days a week, a non-monetary benefit. Subjects were asked to rate how likely they were to agree to this request on a scale of 1-5. This likert scale data was then transformed into three categories. Answers of 1-2 were coded as do not agree, 3 as neutral and 4-5 as agree to the negotiation request (Frequency table presented in appendix).
Similarly to the analysis presented above, A Chi square test was then conducted to test the hypothesis that there is no difference in the distribution of agreement ratings across all four experimental groups (control male candidate, control female candidate, norm treatment male candidate, norm treatment female candidate). The test was not significant at the 5% level: X2 (6, N = 118) = 5.0232, p=0.540841
From this we can see that there is no significant difference in the frequency with which subjects agree to male or female candidates’ requests for non-monetary benefits. Likewise there is no significant difference in the frequency with which subjects agree to male or female candidates’ requests after being exposed to the pay gap norm treatment. This indicates that, among the subjects in the sample, there was no gender gap in non-monetary negotiation outcomes and the activation of gender gap norms has no effect on
non-monetary negotiation outcomes for either gender.
Table 8.
% Subjects' agreement to non-monetary request by gender and norm treatment
No Treatment Treatment Female Candidate Male Candidate Female Candidate Male Candidate Do not agree 7.41% 27.27% 18.75% 15.38% Neutral 29.63% 24.24% 34.38% 34.62% Agree 62.96% 48.48% 46.88% 50.00% Total 100.00% 100.00% 100.00% 100.00%
While the above results provide only a very loose indication of the (non)effect of gender pay gap norms on outcomes for non-monetary negotiations, they do suggest that, while
mandatory reporting may effect salary negotiation outcomes, another policy instrument may be needed to target any gender gaps in non-salary benefits, such as for health care, additional leave and parental benefits.
5. Limitations and Conclusion
While ANOVA is robust to violations of residuals, this will still likely increase the possibility of a false positive. Given that the effect of gender norms on negotiation outcomes for women was only significant at the 10% level (and this effect was small), these results should only be cautiously accepted and further research should be conducted in an attempt to replicate the effect.
Additionally, there is likely sampling bias among the subjects of the experiment. As the average age of subjects skewed young, it is possible that the sample is not truly reflective of the population that is likely to make hiring and salary decisions. It’s also possible that there are differences in attitudes and gender bias across age groups, such that a different effect (or no effect) of pay gap norms may be seen in a more balanced sample. The risk of sample bias was increased by the nature of the online survey, where subjects were able to refer friends to the survey via anonymous link.
The design of an experiment that can be administered through an online survey was necessary due to limited resources. However, given the somewhat artificial nature of a survey experiment, there may also be limits to the generalisability of the results. Subjects reviewed the pay gap information immediately before making their salary offer so that pay gap, when the norm information is at its most relevant. In real-world settings, hiring managers are unlikely to review the extent of their firm’s pay gap relative to their
competitors immediately before making a salary offer. The results of this analysis indicate that directly informing the subject of pay gap norms may be effective but just how to properly activate those norms in a real-world context remains to be seen.
It is also unusual for salary decisions to be made when reviewing written transcripts. Real world negotiations are more dynamic and the effects of gender on any negotiating bias are more likely to be seen when subjects are face to face and so can read verbal and facial cues.
While the majority of subjects correctly identified the gender of their candidate, a lab
experiment where employers and employees can interact in person is more likely to activate any unconscious gender bias in the subjects. This may explain why no gender pay gap was seen between the male and female candidate in the control condition.
A further limitation is that the research presented here only looks at wage gaps as they occur through initial salary negotiations, assuming the candidates negotiate in the first place. Further research would do well to examine how to reduce gender pay gaps which develop over the course of a career, via promotions and raises, or through candidates self-selecting out of negotiations altogether.
Although there are several limitations to the research presented here, there are still some insights to be gained. As shown in the result of the 2-way analysis of variance, there is a significant (at the 10%) interaction effect between establishing gender pay gap norms and the gender of the candidate for salary offers, such that the female candidate on average was offered €1215 more than the male candidate in the norm treatment condition. This analysis failed to replicate the gender pay gap seen in previous studies, as average salary offers did not significantly differ between the male and female candidates for the control condition and no main effects of gender on salary were observed in the analysis of variance. As a result, establishing gender pay gap norms resulted in creating a bias in favour of the female candidate, increasing their average salary offer relative to the male candidate. However, this still demonstrates that establishing pay gap norms can improve salary negotiation outcomes for women. If this insight were to be applied to a setting where there was indeed a gender pay gap disadvantaging women, it may go some way towards reducing that gap.
Interestingly, it was also shown that the frequency with which subjects rated the salary negotiation as reasonable did not significantly differ across experimental groups. In light of previous research, outlined in the literature review, which indicates that people judge and punish women more than men for attempting to negotiate, one would expect to see different perceptions of reasonableness for the male and female candidate. The analysis presented here failed to replicate the finding that women are seen as too dmanding for entering negotiations.
In conclusion, the above research offers some tentative evidence for the effectiveness of mandatory pay gap reporting in reducing the gender pay gap but further research is needed.
References
Behavioural Insights Team. (2012). Applying behavioural insights to reduce fraud, error and debt. London. Retrieved from
https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/ 60539/BIT_FraudErrorDebt_accessible.pdf
Booth, A., Cardona-Sosa, L., & Nolen, P. (2014). Gender differences in risk aversion: Do single-sex environments affect their development? Journal of Economic Behavior & Organization, 99, 126–154. https://doi.org/10.1016/J.JEBO.2013.12.017
Borghans, L., Golsteyn, B. H. H., Heckman, J., & Meijers, H. (2009). Gender Differences in Risk Aversion and Ambiguity Aversion. Cambridge, MA. https://doi.org/10.3386/w14713
Bowles, H. R., Babcock, L., & Lai, L. (2007). Social incentives for gender differences in the propensity to initiate negotiations: Sometimes it does hurt to ask. Organizational Behavior and Human Decision Processes, 103(1), 84–103.
https://doi.org/10.1016/j.obhdp.2006.09.001
Buck, T. (2018). German employers forced to reveal gender pay gap. Financial Times, p. 1. Retrieved from https://www.ft.com/content/e9f618c0-f210-11e7-ac08-07c3086a2625 Eckel, C., & Grossman, P. (2001). Chivalry and solidarity in ultimatum games. Economic
Inquiry, 39(2), 171–188. https://doi.org/10.1111/j.1465-7295.2001.tb00059.x
Eurostat. (2018). Gender pay gap statistics. Retrieved August 10, 2018, from
http://ec.europa.eu/eurostat/statistics-explained/index.php/Gender_pay_gap_statistics
Exley, C., Niederle, M., & Vesterlund, L. (2016). Knowing when to ask: The cost of leaning in. National Bureau of Economic Research, (No. w22961). Retrieved from
García-Gallego, A., Georgantzís, N., & Jaramillo-Gutiérrez, A. (2012). Gender differences in ultimatum games: Despite rather than due to risk attitudes. Journal of Economic Behavior & Organization, 83(1), 42–49. https://doi.org/10.1016/J.JEBO.2011.06.012
Goldstein, N. J., Cialdini, R. B., & Griskevicius, V. (2008). A room with a viewpoint: Using social norms to motivate environmental conservation in hotels. Journal of Consumer Research, 35(3), 472–482.
Haines, M., & Spear, S. F. (1996). Changing the Perception of the Norm: A Strategy to Decrease Binge Drinking among College Students. Journal of American College Health, 45(3), 134–140. https://doi.org/10.1080/07448481.1996.9936873
Leibbrandt, A., & List, J. A. (2015). Do Women Avoid Salary Negotiations? Evidence from a Large-Scale Natural Field Experiment. Management Science, 61(9), 2016–2024. https://doi.org/10.1287/mnsc.2014.1994
McGee, P., & Constantinides, S. (2013). Repeated play and gender in the ultimatum game. The Journal of Socio-Economics, 42, 121–126.
https://doi.org/10.1016/J.SOCEC.2012.11.012
Nolan, J. M., Schultz, P. W., Cialdini, R. B., Goldstein, N. J., & Griskevicius, V. (2008). Normative Social Influence is Underdetected. Personality and Social Psychology Bulletin, 34(7), 913–923. https://doi.org/10.1177/0146167208316691
Perkins, H. W. (2002). Social norms and the prevention of alcohol misuse in collegiate contexts. Journal of Studies on Alcohol. Supplement, (14), 164–172. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12022722
Perkins, H. W., & Craig, D. W. (2006). A successful social norms campaign to reduce alcohol misuse among college student-athletes. Journal of Studies on Alcohol, 67(6), 880–889. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/17061005
Perkins, H. W., Haines, M. P., & Rice, R. (2005). Misperceiving the college drinking norm and related problems: a nationwide study of exposure to prevention information, perceived norms and student alcohol misuse. Journal of Studies on Alcohol, 66(4), 470–478. Solnick, S. (2001). Gender differences in the ultimatum game. Economic Inquiry, 39(2), 189–
200. https://doi.org/10.1111/j.1465-7295.2001.tb00060.x
Sunstein, C. R. (2014). Nudging: a very short guide. Journal of Consumer Policy, 37(4), 583– 588.
The Equality Act 2010 (Gender Pay Gap Information) Regulations 2017. (n.d.). Retrieved August 10, 2018, from
https://www.legislation.gov.uk/uksi/2017/172/regulation/1/made
Thombs, D. L., Dotterer, S., Olds, R. S., Sharp, K. E., & Raub, C. G. (2004). A Close Look at Why One Social Norms Campaign Did Not Reduce Student Drinking. Journal of American College Health, 53(2), 61–68. https://doi.org/10.3200/JACH.53.2.61-70
Wodon, Q., & Brière, B. de la. (2018). Unrealized Potential: The High Cost of Gender Inequality in Earnings. Retrieved from
https://elibrary.worldbank.org/doi/abs/10.1596/29865
Workplace Gender Equality Agency. (2018). International Gender Reporting Schemes. Retrieved from
https://www.wgea.gov.au/find?keyword=international+gender+reporting
Appendix
ANOVA of Salary by norm treatment and gender of candidate (N=118, includes outlier)
Mean Salary by Gender of Candidate and Norm Treatment for full sample, including outlier (N=118)
No Treatment Treatment
Female Candidate Male Candidate Female Candidate Male Candidate
53038.38 53142.35 54014.63 53056.28
(2128.51) (1692.70) (2068.70) (2014.85)
n n n n
34 26 30 28
TukeyHSD pairwise Comparison of mean salary by experimental group
Tukey Tukey
salary Contrast Std. Err. t P>t [95% Conf. Interval] Group CFvsTF 976.251 474.1334 2.06 0.173 -260.1348 2212.637 CFvsCM 103.9638 493.1285 0.21 0.997 -1181.955 1389.883 CFvsTM -239.271 487.9263 -0.49 0.961 -1511.624 1033.082 CMvsTF -872.287 507.1744 -1.72 0.318 -2194.833 450.2589 TMvsTF -1215.52 502.1177 -2.42 0.079 -2524.882 93.83763 TMvsCM -343.235 520.0918 -0.66 0.912 -1699.465 1012.995
Note: C and T denote control and treatment respectively. F and M denote female and male candidate respectively.
Residual normal probability plot
Total 474791249 117 4058044.86 Residual 454856075 114 3989965.57 treatment#gender 8242024.19 1 8242024.19 2.07 0.1534 gender 5331337.95 1 5331337.95 1.34 0.2501 treatment 5787568.37 1 5787568.37 1.45 0.2309 Model 19935174.3 3 6645058.09 1.67 0.1784 Source Partial SS df MS F Prob > F Root MSE = 1997.49 Adj R-squared = 0.0168 Number of obs = 118 R-squared = 0.0420
Residual frequency histogram
Shapiro-Wilks test of normal distribution by experimental group (N=117, dropped outlier)
0 .0 0 0 .2 5 0 .5 0 0 .7 5 1 .0 0 N o rm a l F [(re s-m )/ s] 0.00 0.25 0.50 0.75 1.00
Empirical P[i] = i/(N+1)
0 1 .0 e -0 4 2 .0 e -0 4 3 .0 e -0 4 4 .0 e -0 4 D e n si ty -5000 0 5000 Residuals
Figures and tables for Perception of Negotiation
Contingency table: Frequency of reasonableness ratings for candidate 1 across experimental groups
Unreasonable Neutral Reasonable Row Totals
CF 2 (1.73) [0.04] 6 (6.05) [0.00] 26 (26.22) [0.00] 34 CM 1 (1.32) [0.08] 6 (4.63) [0.41] 19 (20.05) [0.06] 26 TF 2 (1.53) [0.15] 3 (5.34) [1.02] 25 (23.14) [0.15] 30 TM 1 (1.42) [0.13] 6 (4.98) [0.21] 21 (21.59) [0.02] 28 Column Totals 6 21 91 118 (Grand Total)
Note: Data are presented as: observed value (expected value) [Cell chi-square statistic]. C or T denotes control or treatment group. F or M denotes male or female candidate.
Figures and tables for Negotiations regarding non-monetary benefits
Contingency table: Subjects’ agreement to candidate 2’s request, by experimental group. Results
disagree Neutral agree Row Totals
CF 2 (4.81) [1.64] 8 (8.24) [0.01] 17 (13.96) [0.66] 27 CM 9 (5.87) [1.67] 8 (10.07) [0.42] 16 (17.06) [0.07] 33 TM 4 (4.63) [0.08] 9 (7.93) [0.14] 13 (13.44) [0.01] 26 TF 6 (5.69) [0.02] 11 (9.76) [0.16] 15 (16.54) [0.14] 32 Column Totals 21 36 61 118 (Grand Total)
Note: Data are presented as: observed value (expected value) [Cell chi-square statistic]. C or T denotes control or treatment group. F or M denotes male or female candidate.
treatmentm 27 0.92796 2.118 1.541 0.06160 controlm 26 0.93076 1.980 1.400 0.08079 treatmentf 30 0.92596 2.353 1.770 0.03839 controlf 34 0.94742 1.836 1.266 0.10279 Variable Obs W V z Prob>z Shapiro-Wilk W test for normal data
Survey Extracts
Welcome Instructions
Q3 Thank you for participating in this survey experiment. All responses are anonymous. One
respondent will be randomly selected to receive a €50 Amazon gift card. If you wish to be included
in the draw for the gift card please enter your email at the end of the survey. All submitted email addresses will be deleted after the winner has been selected.
Instructions for control Group
Please imagine you are a HR manager in Tricorp, a large firm. You are filling two open positions in the firm. You have made job offers to one candidate for each position. The candidates are each
applying for a different position and so are not in competition with each other. Both candidates
have responded to their job offer with an email. Please review their emails and answer the questions which follow.
Instructions and Norm Treatment for Treatment Group
Q23 Please imagine you are a HR manager in Tricorp, a large firm. On the next page you will see
some information about average wages at your firm. Please review the information and answer the questions which follow.
Page Break
Below is a list of the ten largest firms in the same industry as TriCorp, including TriCorp itself. They
are ranked by the size of the gender pay gap among their employees, with 1 being the lowest pay
gap and 10 being the highest.
men’s mean hourly wage. For example, with a mean gender pay gap of 10%, when comparing mean hourly rates, women earn 90c for every €1 that men earn
Q30 70% of the ten largest firms in this industry have a lower gender pay gap than Tricorp.
o
True (1)o
False (2)Q31 Tricorp has a higher gender pay gap than most of the other firms.
o
True (1)o
False (2)Female Candidate 1 Email and following Questions Candidate 1: Samantha Mills
Dear Sir/Madam,
However, before I accept the position I would like to discuss the salary and benefits. Although Tricorp is my preferred choice, I have been offered other positions with higher salaries. In fact, the highest offer is 10% more than the offer from here.
I’m very enthusiastic about TriCorp and would be delighted to accept if you could make the offer more competitive by increasing my base salary from €50,000 to €55,000 to match the other company’s salary. As you are aware, I have five years of experience and a proven track record of increasing departmental revenues. I’m certain that I can make a valuable contribution to TriCorp and feel that this is a more acceptable salary for someone with my skills and expertise. I understand that this is more than your initial offer but I’m willing to be flexible to find an acceptable solution. With this in mind, I hope we can come to a mutual agreement.
Looking forward to hearing from you,
Samantha Mills
Q36 What salary would you offer Samantha in response to her email?
Salary in €
40000 55000 60000
Salary ()
Q14 On a scale of 1-5, with 1 being the least and 5 being the most, how reasonable do you think her
request is?
1 (1) 2 (2) 3 (3) 4 (4) 5 (5)
How
Q15 On a scale of 1-5, with 1 being the least and 5 being the most, how happy would you be to
work with her?
1 (1) 2 (2) 3 (3) 4 (4) 5 (5)
How happy
(1)
o
o
o
o
o
Male candidate 1 email
Email Male Candidate 1
Candidate 1: Samuel Mills
Dear Sir/Madam,
Thank you for offering me the position. I’m excited about Tricorp and my future with the company. However, before I accept the position I would like to discuss the salary and benefits. Although Tricorp is my preferred choice, I have been offered other positions with higher salaries. In fact, the highest offer is 10% more than the offer from here.
I’m very enthusiastic about TriCorp and would be delighted to accept if you could make the offer more competitive by increasing my base salary from €50,000 to €55,000 to match the other company’s salary. As you are aware, I have five years of experience and a proven track record of increasing departmental revenues. I’m certain that I can make a valuable contribution to TriCorp and feel that this is a more acceptable salary for someone with my skills and expertise. I understand that this is more than your initial offer but I’m willing to be flexible to find an acceptable solution. With this in mind, I hope we can come to a mutual agreement.
Looking forward to hearing from you,
Samuel Mills
Female Candidate 2 email Candidate 2: Laura Park
Dear Sir/Madam, I was happy to receive your offer for the position with TriCorp. Before we finalise my acceptance I would like to discuss the possibility of changing the terms of my contract. A key benefit of my current position is the flexible working hours. Finding a balance between work and time spent with my family is very important to me and this position would involve longer hours than my current job. To offset this I would like to request the option to work two days a week from home in order to save time on my commute. This would not involve any extra cost for the company and would allow me to devote more time to my job and better focus on my
work. I’m happy to compromise on which days I work from home depending on what is most convenient. Hopefully this arrangement will be acceptable to you.
Thank you, Laura Park
Male Candidate 2 email
Candidate 2: Luke Park
Dear Sir/Madam,
I was happy to receive your offer for the position with TriCorp. Before we finalise my acceptance I would like to discuss the possibility of changing the terms of my contract. A key benefit of my current position is the flexible working hours. Finding a balance between work and time spent with my family is very important to me and this position would involve longer hours than my current job. To offset this I would like to request the option to work two days a week from home in order to save time on my commute. This would not involve any extra cost for the company and would allow me to devote more time to my job and better focus on my work. I’m happy to compromise on which days I work from home depending on what is most convenient. Hopefully this arrangement will be
acceptable to you. Thank you,
Luke Park
Questions following candidate 2 email (pronouns for female candidate)
Q61 On a scale of 1-10, with 1 being the least and 10 being the most, how likely are you to agree to
Laura's request to work from home?
1 (1) 2 (2) 3 (3) 4 (4) 5 (5)
How likely (1)
o
o
o
o
o
Q63 On a scale of 1-5, with 1 being the least and 5 being the most, how reasonable do you think her
request is?
1 (1) 2 (2) 3 (3) 4 (4) 5 (5)
How
Q64 On a scale of 1-5, with 1 being the least and 5 being the most, how happy would you be to
work with her?
1 (1) 2 (2) 3 (3) 4 (4) 5 (5)
How happy
