Gender and Gender-Role Attitudes in Wage Negotiations: Evidence from an Online Experiment

Basics of the Design

The experiment was implemented via z-Tree unleashed (Duch, Grossmann, and Lauer 2020). At the beginning of the experiment, participants are evenly split between workers and firms and remain in these roles until the end of the experiment. Participants face five negotiation opportunities in their roles as workers or firms. Prior to the first negotiation opportunity, each participant completes a real-effort task that determines the contribution a participant brings to the table for each of the five subsequent negotiation opportunities. A worker’s contribution is $20, $15, or $10, depending on whether the worker’s performance is the highest, second highest, or third highest (compared with two other randomly selected workers). A firm’s contribution is $25 or $20, depending on whether the firm’s performance is the highest or second-highest performance (compared with another randomly selected firm). Any performance ties are randomly broken, which is common knowledge among workers and firms.

Participants are then randomly assigned to worker-firm pairs for each negotiation opportunity. The joint revenue per negotiation is the sum of the worker contribution and the firm contribution. For each negotiation opportunity, a worker’s suggested wage is equal to the worker’s contribution plus a bonus that is randomly and uniformly drawn from the set –4, –2, 0, and 2. A negative bonus (of –4 and –2) represents a favorable negotiation opportunity: if the suggested wage is less than the worker’s contribution, the worker may find it easier to justify the higher wage on the basis of the equity principle. A non-negative bonus (of 0 and 2) represents a less favorable negotiation opportunity: workers may find it difficult to negotiate for a wage equal to or greater than their contribution level.

Workers in the Choice treatment can choose to negotiate with the firm or forgo negotiation by accepting the suggested wage. If a worker accepts a suggested wage, the firm receives the joint revenue minus the suggested wage. Workers in the Always treatment must enter negotiations while observing the suggested wage. In both treatments, workers and firms negotiate over the share of the joint revenue that the worker receives as a wage. Workers and firms have three minutes to reach an agreement. If workers and firms fail to reach an agreement, workers receive the suggested wage along with a $5 impasse penalty imposed on both the worker and the firm. Payment for participation is based on a randomly selected round of negotiation.

The Supplemental Materials online contain further information about the research design, the instructions given to participants, and screenshots of the experiment. Instructions were read online by participants prior to the start of the experiment. ³

Post-experiment Measures

After five rounds of negotiation, participants answer questions about risk and fairness of payoffs similar to those faced by workers deciding whether to negotiate. We use measures implemented by Exley et al. (2020). After eliciting risk preferences and perceptions of fairness, participants complete a short follow-up survey that collects the Bem Sex Role Inventory (Bem 1981), the Gender Role Attitudes Scale (García-Cueto et al. 2015), and demographic information on age, gender, sexual orientation, race-ethnicity, education, household income, marital status, head-of-household status, number of household members, employment status, and U.S. region.

To measure gender, we ask participants to “please choose the description that best fits how you think about yourself: I identify my gender as . . . .” Participants are then shown a five-category response scale, which includes woman, man, transgender, nonbinary, and other (please specify). From these five categories, we construct three dummy variables for male, female, and other.

The Gender Role Attitudes Scale consists of 20 indicators drawn from items used in the Social Role Questionnaire (Baber and Tucker 2006), among others. Items on the scale reflect attitudes that identify gender roles along two dimensions: egalitarianism, or positive attitudes toward equality between men and women (e.g., “People should be treated equally, regardless of their sex”), and sexism, or beliefs that men and women are differently suited for many kinds of roles (e.g., “In many important jobs it is better to contract men than women”). Following García-Cueto et al. (2015), we use these 20 items—each measured on a five-point agree-disagree scale—to construct a single latent factor score, which we call the Traditionalism scale (α = .883). The latent factor score is coded so that low values indicate nontraditional gender-role attitudes and high values indicate traditional gender-role attitudes. Information on the specific items used can be found in Table S1 in the Supplemental Materials online.

We use all other variables in the follow-up survey as control variables. For example, we control for the Bem Sex Role Inventory to disentangle gender-role attitudes from gender traits. This is important because the literature shows that the two operationalizations are correlated (Judge and Livingston 2008).

Data

From June to July 2021, 330 members of the Prolific.co subject pool participated in 33 sessions online via Zoom. ⁴ Prolific is an online platform similar to Amazon.com’s Mechanical Turk, where people can complete tasks, including experiments and surveys, in exchange for payment (Palan and Schitter 2018). Payments on Prolific consist of a fixed hourly wage plus potential incentives earned through participation. Cash earnings ranged from $8.83 to $27.01, with an average of $18.62, based on one randomly selected negotiation round, additional payments from the follow-up activities (i.e., measures of risk aversion and fairness), and an average show-up fee of $9.34 (show-up fees varied by session length).

Following standard Prolific.co procedures, we used their online recruitment system to draw a nonprobability sample of U.S. adults. On average, each session achieved gender balance, with the percentage of female participants ranging from 11 percent to 66 percent across sessions. Given that our focus is on male and female workers, Table A1 in the appendix provides summary statistics of key experimental and demographic variables by (1) all workers, (2) male workers, and (3) female workers. Note that we exclude “other” gender from the analysis, as our analytical goals are to understand male-female differences in wage negotiations. We also observe insufficient observations of other gender to make valid inferences about entry into, and the outcomes of, wage negotiations (see Table 1).

OPEN IN VIEWER

Table 1.

Gender Distribution across Roles and Treatments

	Choice treatment		Always treatment
Gender	Worker	Firm	Worker	Firm
Female	29	24	33	29
Male	45	50	41	48
Other	2	1	4	1

Note: N = 307 (summed across all cells).

Selection into Negotiations

Results from our Choice treatment show that female workers enter at least one negotiation 75.86 percent of the time and that women enter 40 percent of all possible negotiations (or 58 negotiations in 145 negotiation opportunities). Out of the 29 female workers, 7 (or 24.13 percent) never enter a negotiation and 3 (or 10.34 percent) enter all negotiations. By contrast, male workers enter at least one negotiation 88.89 percent of the time, and men enter 56 percent of all possible negotiations (or 126 negotiations in 225 negotiation opportunities). Out of the 45 male workers, 5 (or 11.11 percent) never enter a negotiation and 7 (or 15.55 percent) enter all negotiations. Consistent with prior research, women avoid negotiations more often than men, and women are more likely than men to never enter a negotiation (Kugler et al. 2018).

Next, we examine whether these gender differences in selection differ by the value of the bonus. Figure 1 shows that workers, regardless of their gender, are less likely to enter negotiations as the bonus increases, with male and female workers as likely to enter negotiations under the most favorable (bonus of –4) and least favorable (bonus of 2) negotiation conditions. Table 2 confirms these findings using linear probability random-effects regression models and provides evidence for the moderating effect of gender-role attitudes. ⁵ Substantiating our descriptive analysis, Table 2 shows that (1) female workers are significantly less likely to enter negotiations than male workers (model 1), (2) workers choose to enter negotiations based on the bonus (model 1), and (3) responses to changes in the bonus are similar for men and women (see the statistically nonsignificant Female × Bonus interaction effect in model 3), except for intermediate values of the bonus (see Figure S1 in the Supplemental Materials online). These results are robust to demographic controls (models 2 and 4) and collectively suggest that workers, regardless of gender, understand how the bonus works: increases in the bonus create unfavorable negotiation opportunities that workers avoid.

OPEN IN VIEWER

Figure 1.

Entry Decisions into Negotiations for Male and Female Workers across Values of the Bonus

OPEN IN VIEWER

Table 2.

Entry into Negotiations

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Female worker	–0.200*** (0.074)	–0.369*** (0.090)	–0.204*** (0.079)	–0.389*** (0.091)	–0.195* (0.079)	–0.384*** (0.094)	–0.194* (0.079)	–0.383*** (0.095)
Bonus	–0.071*** (0.011)	–0.064*** (0.013)	–0.069*** (0.015)	–0.055*** (0.017)	–0.079*** (0.014)	–0.067*** (0.018)	–0.078*** (0.015)	–0.066*** (0.019)
Traditionalism scale	–0.064 (0.092)	–0.158 (0.100)	–0.064 (0.093)	–0.153 (0.100)	0.069 (0.121)	–0.046 (0.149)	0.064 (0.124)	–0.050 (0.151)
Female × Bonus			–0.003(0.025)	–0.021(0.027)	0.012(0.025)	–0.002(0.029)	0.012(0.025)	–0.003(0.029)
Female × Traditionalism					–0.178(0.156)	–0.131(0.204)	–0.164(0.168)	–0.120(0.212)
Bonus × Traditionalism					0.060* (0.026)	0.058* (0.029)	0.055(0.030)	0.053(0.035)
Female × Bonus × Traditionalism							0.015(0.056)	0.012(0.061)
Intercept	0.346* (0.177)	1.871(0.997)	0.347* (0.177)	1.850(1.002)	0.309(0.172)	1.771(0.985)	0.311(0.172)	1.778(0.985)
SD (intercept)	0.278	0.313	0.281	0.318	0.284	0.328	0.283	0.333
SD (residuals)	0.387	0.382	0.388	0.382	0.384	0.379	0.384	0.379
Observable design controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Demographic controls	No	Yes	No	Yes	No	Yes	No	Yes
Observations	370	335	370	335	370	335	370	335
Individuals	74	67	74	67	74	67	74	67

Note: Unstandardized slopes and cluster-robust standard errors (in parentheses) computed with linear probability random-effects models. Observable design controls include round fixed effects, worker and firm contributions to the joint revenue, and whether the firm contribution is known as well as fairness perceptions and risk aversion. Demographic controls include Bem Sex Role Inventory, age, U.S. region, race-ethnicity, sexual orientation, education, marital status, cohabitation status, head-of-household status, employment status, and log per capita household income.

*

p < .05. **p < .01. ***p < .001 (two tailed).

Models 5 through 8 in Table 2 provide tests of Hypothesis 1 and Hypothesis 2. Model 5, along with marginal-effects plots, shows that traditional gender-role attitudes moderate the effects of gender and the bonus on entry into negotiations. ⁶ Figure 2, Panel A, shows that as gender-role attitudes increase and become more traditional, the gender gap in negotiations amplifies. Male workers who subscribe to traditional gender-role attitudes (+1 SD and +2 SD on the Traditionalism scale) are more likely to enter negotiations than their female counterparts, supporting Hypothesis 1. Figure 2, Panel B, shows that the negative effect of the bonus is weaker for workers with traditional gender-role attitudes. This suggests that workers are more likely to enter unfavorable negotiation opportunities when their gender-role attitudes are more traditional. While these results provide preliminary support for Hypothesis 2, model 7 shows that a three-way interaction between Female, Bonus, and Traditionalism is statistically nonsignificant. ⁷ In other words, as gender-role attitudes increase, male workers are no more or less likely to enter into (un)favorable negotiation conditions than female workers, thereby statistically rejecting Hypothesis 2. These results are robust to demographic controls (models 7 and 8).

OPEN IN VIEWER

Figure 2.

Marginal Effects of Female (A) and Bonus (B) on Entry into Negotiations

To get an idea of what entry into negotiations looks like across values of the bonus for (non)traditional male and female workers, linear predictions in Table 3 show that nontraditional male and female workers are adept at selecting into negotiations: they enter favorable negotiation opportunities (male = .817; female = .718) and avoid unfavorable negotiation opportunities (male = .040; female = .016). These types of workers are also much more likely than traditional workers to take advantage of favorable negotiation opportunities (male = .531; female = .139). Finally, traditional male workers have the highest level of entry into unfavorable negotiation opportunities (.354) than any other category of workers. This finding is consistent with our expectations about traditional male workers (see Hypothesis 2), despite statistical evidence to the contrary (see model 7).

OPEN IN VIEWER

Table 3.

Select Linear Predictions of Entry into Negotiations by Bonus, Gender-Role Attitudes, and Gender of Workers

Bonus of -4				Bonus of 2
Nontraditional (–2 SD)		Traditional (+2 SD)		Nontraditional (–2 SD)		Traditional (+2 SD)
Male worker	Female worker	Male worker	Female worker	Male worker	Female worker	Male worker	Female worker
0.817	0.718	0.531	0.139	0.040	0.016	0.354	0.036
(0.456, 1.179)	(0.410, 1.026)	(0.250, 0.813)	(–0.263, 0.542)	(–0.320, 0.401)	(–0.284, 0.316)	(0.043, 0.665)	(–0.339, 0.413)

Note: Linear predictions based on model 5 in Table 2 using the delta method (95 percent confidence intervals in parentheses). All continuous covariates constrained to their minimum value and all factor-variable covariates constrained to their base level.

Negotiation Outcomes

Next, we test Hypothesis 3 and Hypothesis 4 by evaluating the outcomes of negotiations. We begin by describing the percentage of agreements reached, how many of these negotiations resulted in gains and losses, and the total profit achieved from entering into negotiations. We compute profits as the worker’s payoff from negotiating minus his or her suggested wage. Profits are set to zero if workers in the Choice treatment choose not to enter negotiations and accept the suggested wage. If no agreement is reached, profits are by design minus $5 (due to the impasse).

In the Choice treatment, entering negotiations is largely beneficial for both genders. For women (men), agreements are reached 75.86 percent (80.16 percent) of the time; 62.07 percent (60.32 percent) of the negotiations result in gains, while 27.59 percent (31.75 percent) result in losses. The average profit from entering a negotiation is $0.86 for women and $0.89 for men. This indicates that negotiating is just as beneficial for women as it is for men, even though women avoid negotiating more often than men.

In Figure 3, we compare the profits earned by men and women in the Choice treatment to the counterfactual of profits earned by men and women in the Always treatment. Because women do not enter 60 percent of all possible negotiation opportunities in the Choice treatment, most female workers earn zero profits in this treatment. But what happens if female workers always negotiate instead of avoiding negotiations 60 percent of the time? Figure 3 shows that forcing women to negotiate in the Always treatment does not help: female workers are unable to avoid negotiations that would have resulted in losses. In fact, the share of losses increases from 11 percent in the Choice treatment to 34 percent in the Always treatment, while the share of gains increases from 25 percent in the Choice treatment to 44 percent in the Always treatment. For male workers, we observe similar patterns albeit to a lesser extent: additional negotiations lead to more losses (from 18 percent to 35 percent) and gains (from 34 percent to 44 percent) in the Always treatment than in the Choice treatment. Overall, the increase in gains and losses is larger for female workers than for male workers in the Always treatment.

OPEN IN VIEWER

Figure 3.

Distribution of Profits among Male and Female Workers in the Choice and Always Treatments

We use random-effects regression models to further explore these results and to systematically test Hypothesis 3 and Hypothesis 4. In support of our descriptive analysis, profits are statistically equivalent for men and women (see models 1 and 2 in Table S2 in the Supplemental Materials online), and forcing workers to negotiate in the Always treatment is no more helpful to women than it is to men (see Female × Always in models 3 and 4 in Table S2). ⁸ Finally, we find no support for Hypothesis 3 and Hypothesis 4, as the Female × Traditionalism interaction and the Female × Always × Traditionalism interaction do not improve model fit (see models 5 through 8 in Table S2). ⁹

Limitations and Directions for Future Work

Some limitations and directions for future research are worth mentioning. Previous research shows that the gender pay gap has myriad sources. Discrimination in job assignments or in refusal to hire (Tomaskovic-Devey and Avent-Holt 2019), the “motherhood penalty” (Correll, Benard, and Paik 2007), and the growing prevalence of “overwork” (Cha and Weeden 2014) undoubtedly contribute to the gender gap in pay. And a number of other factors besides gender-role attitudes influence gender differences in negotiation outcomes, such as perceptions of fairness (Auspurg, Hinz, and Sauer 2017) and women’s fear of potential backlash (Amanatullah and Morris 2010).

From our perspective, negotiation decisions, gender discrimination, and overwork, among others, are the myriad channels through which gender differences in pay can arise. To better understand the factors that influence men’s and women’s negotiation decisions and whether these factors contribute to gender differences in pay, we chose to focus on selection into negotiation and negotiation outcomes. With respect to other factors, it may be useful in the future to explore variables that are known to drive gender differences in negotiation, particularly variables that Exley et al.’s (2020) design purposefully eliminates. For instance, male employers may actually reward male employees more than female employees for their aggressive negotiations (Amanatullah and Morris 2010). If this is the case, aggressive negotiations by male employees, which is more common among workers with traditional gender-role attitudes, may not unduly affect their negotiation outcomes. Female employees, on the other hand, may experience backlash as a result of their aggressive negotiations.

Men’s and women’s bargaining decisions and outcomes depend heavily on the cultural context in which wage negotiations occur. In some countries, like Finland, Denmark, and Norway, collective bargaining takes place at the national and sectoral levels. In other countries, like the United States, France, and Poland, wage bargaining is more decentralized and occurs at the firm level. In this sense, the observed effects should be interpreted as a conservative set of results limited to decentralized wage-bargaining regimes, like those found in the United States and France. Research also indicates that there are cross-cultural differences in gender-role attitudes (Knight and Brinton 2017) and negotiation behavior (Gelfand et al. 2012). Although much of this research focuses on Western and East Asian contexts, it nonetheless reveals starkly different gender-role attitudes and negotiation strategies. In East Asian contexts, the default strategy in negotiations is to avoid offending others, pursue actions that are “socially wise” and “communal,” and avoid behaviors that are overly self-serving (Chiu et al. 2010). By contrast, in Western contexts, where distinct varieties of gender-role egalitarianism vary across countries and over time (Knight and Brinton 2017), the default strategy is to be overly assertive and promote one’s self-interest, even at the expense of others. Because of these cultural differences, our findings may not hold in other cultural contexts, and future research should interrogate the generalizability of our results.

The gender gap in pay has persisted in the United States for decades and shows little sign of changing. Progress toward gender equality in the United States, in other words, has stalled (England et al. 2020). Although more work remains, the present study provides compelling reasons to consider the role of gender-role attitudes in explaining gender differences in wage negotiations. Armed with this evidence, policymakers can implement programs aimed at nudging the behavior of male and female workers toward effective negotiations—negotiations that will benefit the welfare of all workers.