Gender pay transparency

Employees and employers will form a company and affect each other payoffs by making a set of choices. In particular, employers will know the gender and have incomplete information about the potential performance of prospective employees. The incomplete information is based on a screening test that places employees into either Group A or B, depending on whether their performance is better or worse that a defined threshold. Employers will also set the wages and career promotions under a budgetary constraint. Employees will work for their employers, deciding the level of effort they wish to exert, and receive a wage in exchange. Under certain circumstances employees can ask for compensation for unfair treatment.

The devised labour market experiment translates policy options in the following treatments:
- T0: No change of policy (baseline scenario). Employees only know their wage and position and how many men and women work in the company overall. After receiving this information, they are asked if they want to ask for compensation and then they perform the main task.
- T1: Right of employees to information on pay levels. Employees receive information on their wage and position, how many men and women work in the company, and the average wage overall and by gender, but only for their position. This information is provided automatically. After receiving this information, they are asked if they want to ask for compensation and then they perform the main task.
- T2: Reporting on Pay. Employees receive information on their wage and position, how many men and women work in the company, and the average wage overall and by gender in the managerial position and the clerical position. On the basis of this additional information they can request for compensation. They then perform the main task.

In each treatment, employers assign wages and positions twice. The first time they do not receive any feedback on the information available to employees and on their ability to ask for compensation. The second time employers set wages and positions knowing about employees' ability to ask for compensation and knowing employees' level of information about others' wages and positions (treatments).

2020-12-10

2020-12-23

1. Gender pay discrimination under each treatment;
2. Individual level of gender pay discrimination by employer;
3. Employees productivity and inferred effort;
4. Employee’s (costly) action against employer (contentiousness).
5. Employee's gift exchange/intrinsic motivation

1. Gender pay discrimination under each treatment. This is a dummy variable equal to 1 if the average wage for women is lower than the average wage for men, including both clerks wages and manager wages. The average of the variable returns the proportion of employers who are discriminating. We will also construct a mirror variable for discrimination against men.
2. Individual level of gender pay discrimination by employer is a dummy variable equal to 1 if the average women wage in group A is lower than the average men wage in group A (including both clerks wages and manager wages) and/or the average women wage in group B is lower than the average men wage in group B (including both clerks wages and manager wages). In a robustness check this variable may be further refined to take into account ambiguous cases where the intention to discriminate is not clear enough. We will also construct a mirror variable for discrimination against men.
3. Employees productivity in the number of correct strings and the number of attempted strings.
4. Employee costly action against the employer is a dummy equal to 1 if the employee requests compensation
5. Employee's gift exchange/intrinsic motivation is a dummy equal to 1 if the employee continues performing the task after the threshold and the number of strings attempted and correct after the threshold. To disentangle the reasons behind extra effort we include a survey question where we ask why they did not stop working once the threshold was met.

Employers beliefs on employees' performance by gender and group in the screening test and main task.
Employees beliefs on own performance and other employees' performance by gender and group in the screening test and main task.
Employees' second order beliefs on employers.
Risk aversion of employers and employees
Employee's motivation behind asking for compensation
Employers' beliefs on fairness of different wage distribution strategies
Employers' beliefs on potential share of employees asking for compensation
Employees' beliefs on employers' wage strategies

These are all measured in the ex-post questionnaire. They will be transformed in dummy variables or discrete variables.

In each Country the participants are:
-900 employees
-150 employers
One employer will be asked to make decisions referring to 18 employees, leading to a final match by random draw to six employees (for the purpose of calculating the final payment). the 6 employees will perform a task and make some decisions that will affect their employer's payoffs.
There will be 2 experiments:
1. First the one with the employers [EXP1], and once this will be finalized
2. The one with employees (for which the results of EXP1 are necessary) [EXP2]
Employers are asked to set wages for 6 groups of 3 employees (18 profiles). They will also have the possibility to promote 3 of the 18 employees to a managerial position. The wage set for the clerical employees can be high or low, employers have a budget constraint and they cannot assign the high wage to all the employees. The employer knows the gender of the employees and their performance in a screening test, which is similar to the main task the employees will perform. According to the performance in the screening test employees will be assigned to group A (higher or equal than the threshold ) or group B (lower than the threshold). The threshold is the median performance in the pilot.
The task (string task by Isen and Reeve, 2006) has the following key characteristics:
• Not gender stereotypical;
• Real effort task (for the measurement of productivity and inferred effort);
• Allows for “categories” (i.e. managerial/clerical) to which different thresholds for completing the task apply.
Employers get paid based on employees' performance in the task. Employers know that employees can ask for compensation if they think that they were not given a fair wage. A compensation is granted when colleagues who performed better or equal in the screening test receive a higher wage than the one of the employee asking for compensation.
In the treatments, employers are asked to set the wages and positions knowing that:
T0: employees only know their wage and position and are clueless about their wage being high or low
T1: employees know their wage and position and they know the average wage, also by gender, in their own position
T2: employees know their wage and position and they know the average wage, also by gender, in their own position and the number of men and women in the managerial position and their wage.

Under the information set described in T0, T1, T2 employees perform the screening test, the main task and can ask for compensation. In the main task employees need to reach a threshold to get their wage. They can also work beyond the threshold with the purpose to increase the employer's earnings knowing that this extra work will not affect their payoffs.

In each Country the participants are:
-900 employees
-150 employers
One employer will be asked to make decisions referring to 18 employees, leading to a final match by random draw to six employees (for the purpose of calculating the final payment). The 6 employees will perform a task and make some decisions that will affect their employer's payoffs.
There will be 2 experiments:
1. First the one with the employers [EXP1], and once this will be finalized
2. The one with employees (for which the results of EXP1 are necessary) [EXP2]
Employers are asked to set wages for 6 groups of 3 employees (18 profiles). The wage set for the clerical employees can be high or low, employers have a budget constraint and they cannot assign the high wage to all the employees. Employers will also have the possibility to promote 3 of the 18 employees to a managerial position. The employer knows the gender of the employees and their performance in a screening test, which is similar to the main task the employees will perform. According to the performance in the screening test employees will be assigned to group A (higher or equal than the threshold) or group B (lower than the threshold). The threshold is the median performance in a pilot experiment (done with 900 employers and 150 employees).
The task (based on the string task by Isen and Reeve, 2006) consists in identifying strings in alphabetical order and has the following key characteristics:
• Not gender stereotypical;
• Real effort task (for the measurement of productivity and inferred effort);
• Allows for “categories” (i.e. managerial/clerical) to which different thresholds for completing the task apply.
Employers get paid based on employees' performance in the task.
In the treatments, employers are asked to set the wages and positions twice. Before setting wages for the second time employers receive additional information. Employers are told that employees can ask for compensation if they think that they were not given a fair wage. A compensation is granted when colleagues who performed better or equal in the screening test receive a higher wage than the one of the employee asking for compensation.
Employers are also informed that:
T0: employees only know their wage and position and are not aware of their wage being high or low
T1: employees know their wage and position and they know the average wage, also by gender, in their own position. They can then infer if their wage is high or low.
T2: employees know their wage and position and they know the average wage, also by gender, in their own position, the number of men and women in the managerial position and their wage. They can then infer if their wage is high or low.
Employees' profiles shown to each employer are randomly selected out of 12 pre-determined triples. The triples always contain the two genders, but they might contain only one group (A or B). The aggregate distribution of triples is the same across treatments. Gender discrimination can be compared between treatments at aggregate level and within treatment at individual level.
Under the information set described in T0, T1, T2 employees perform the screening test, the main task and can ask for compensation. In the main task employees need to reach a threshold to get their wage. The threshold is lower (120 strings) for clerks compared to managers (170 strings). They can also work beyond the threshold with the purpose to increase the employer's earnings knowing that this extra work will not affect their payoffs. In the employees’ survey it is not straightforward to disentangle the effect of the treatments through a change in discrimination levels (change in employers' behaviour in setting wages and promotions) from the one through employees' reaction to additional information. However, it is possible to compare the effect of the treatments on employees through multivariate analysis by controlling for employees' characteristics (wage and position) and for the observed level of discrimination.

Randomization done by panel provider by a computer using pre-determined quotas.

individual

No

no clusters

2700 employees and 450 employers

150 employers T0 (baseline), 150 employers T1, 150 employers T2.
900 employees T0 (baseline), 900 employees T1, 900 employees T2.

Aggregating the three countries (300 obs) the minimum detectable decrease in discrimination is -0.1558 with respect to a baseline level of discrimination equal to 0.37 with an sd of 0.48. Power=0.8. For each individual country (100 obs) the detectable decrease would be -0.27 Aggregating the three countries (300 obs) the minimum detectable decrease in individual discrimination is -0.1493 with respect to a baseline level of discrimination equal to 0.3 with an sd of 0.46. Power=0.8. For each individual country (100 obs) the minimum detectable decrease in individual discrimination is -0.26 With chi-square test of proportions delta is very similar, respectively -0.1473 and -0.136 when aggregating countries. Aggregating the three countries (1200 obs) the minimum detectable increase in request for compensation is 0.08 with respect to a baseline level equal to 0.37 with an sd of 0.48. Power=0.8. For each individual country (600 obs) the detectable increase would be 0.11 Aggregating the three countries (1200 obs) the minimum detectable decrease in effort is -8.98 with respect to a baseline level equal to 173.44 with an sd of 55.5. Power=0.8 . For each individual country the detectable decrease would be -12.72 Please note that the estimates used for the power analysis are those of a pilot study with a sample size of 900 employees and 150 employers.