Gender Bias at workplace

Every day in our lives, we come across the word gender bias, which means there is some unfair treatment happening to a group of people who are categorised based on gender. We all know there is always some bias towards men in society. So, to balance that, there are many reservations and new laws in favour of women. In this work, we just wanted to know if there is bias in the workplace or not. We cannot simply conclude this by analysing one situation. So, we have considered multiple datasets and tried to analyse gender bias in the workplace.

We tried to check for gender bias in layoffs and pay gaps using different data sets across the world. In this vlog, we first discuss our analysis using different datasets. Then we conclude if there is any gender bias or not, and then we try to state our solution to remove gender bias.

Analysis using the Glassdoor dataset:

In Kaggle, we found a dataset that was collected from Glassdoor information. The dataset contains job title, gender, age, performance evaluation score, education, department, seniority, base pay, and bonus. This dataset does not contain any names or IDs of the accounts. So, using this data will not raise any issues regarding ethics, and our results do not target any individual, so there will be no issues.

When we categorise them based on gender and draw box plots for their salary and bonus. We found that men are getting a higher salary than women in both base pay and bonuses. We found that there is a higher deviation in the base salary than in the bonus. We feel that this deviation is because the base salary is generally given when they join a company before seeing their work. So base pay is higher for men. whereas a bonus is generally given depending on their work. So, both men and women are getting almost the same bonus, as women can perform equally well as men in most cases.

Using the same dataset, we also tried to analyse how the data's gender bias changes with education qualification, seniority, and department. For that, we first grouped people depending on education (or seniority or department). Then we find the base pay for males and females in each category. The payoff gaps observed for each category are as follows:

If we see the graph, it shows the gender pay gap between males and females in each department. In all the departments, it is observed that men are getting a higher salary than women. It is also found that among all departments here, the engineering department, which contains mainly IT fields, shows the highest pay gaps between males and females.

Here it is observed that with increasing seniority, the gender pay gap between males and females is decreasing. This may be because, with increasing seniority, women also got recognised for their work and got higher salaries.

Here it is observed that with an increase in education level, the gender pay gap between males and females is decreasing. So, women with higher education are getting more value and getting recognised for their work.

In all cases, men are more dominant than women. Men are getting more salaries and bonuses in all situations. So, it is evident that gender bias exists in the workplace.

A t-test is a statistical method used to determine if there is a significant difference between two groups of data. Specifically, it is used to compare the means of two samples to determine if the difference between them is statistically significant or just due to chance. The t-test is a commonly used method in social science research because it is relatively simple to use and interpret. It can be used to analyse data from experiments or observational studies and to test a wide range of hypotheses.

A t-score of 5.392 indicates a large difference between the means of the two groups being compared.

After doing regression analysis, we got an R-value of 0.649. The R-value (also called the correlation coefficient) is a statistical measure that represents the strength and direction of the relationship between the variables we consider. Specifically, an R-value of 0.649 indicates a moderately positive correlation between the variables we consider.

Here, the treatment with ‘1’ represents the experimental group and ‘0’ represents the control group. We have considered only gender as the factor and got the result, which is in negative form, which signifies that there is a lot of difference between pay w.r.t. gender bias.  

Using UK data:

In the UK, there is a rule that every employer who has more than 50 employees should send some information to the government, which contains multiple columns like company name, employer name, address of the office, and other company details. It also contains details such as DiffMeanHourlyPay (the mean % difference between male and female hourly pay), DiffMedianHourlyPay (the median % difference between male and female hourly pay), DiffMeanBonusPay, and DiffMedianBonusPay.

Here, the employer himself reports the important information required for our analysis. So the data here is more authentic. When we analysed this dataset, the results were as follows:

The mean bonus gap is increasing every year. It indicates that the situation is getting worse compared to now. The bonus gap between males and females is increasing every year.

If we observe, the pay gap is constant across the years. It means that there is a fixed rate for males and females. It indicates that even though the years have passed, the ratio of wages for males and females is not changing.

From the above, there is gender bias. We have also analysed the data further for more insights.

If we observe, the number of companies reporting every year is decreasing. It is very bad because companies are not showing interest in showcasing their details. It is a bad indicator for the improvement of the gender gap in the workplace.

We also cluster the companies into groups based on the difference in pay between males and females. Using the elbow method, we found there should be four groups. We divided the companies into four groups so that we could analyse each group more in detail and provide suitable laws or precautions to decrease the gender gap.

Layoffs Data:  

The layoffs dataset was curated using the publicly available data on the "layoffs.fyi" website. This dataset gives information on employees that have been laid off during the 2022–2023 time frame. The dataset contains employee information from Atlassian, Xero, Quora, and Doordash.

From the analysis, we could find that females observed more layoffs as compared to males, and within females, those in leadership roles observed more layoffs as compared to those in non-leadership roles. All these points point to the fact that there was a bias in the layoff process against women in leadership positions. This bias could have arisen due to several factors, such as gender stereotypes or implicit biases held by the decision-makers, a lack of awareness or appreciation of the contributions of women in leadership positions, or a lack of support for women in leadership roles.

Suggested solution:  

We got results in such a way that it clearly shows there is gender bias in the workplace regarding pay. The suggested solution would be to give more work-from-home opportunities so that women can do their office work while at the same time doing their homework.

Limitations:  

• The Glassdoor dataset is not an official dataset published by Glassdoor.
• The datasets used mostly concern the IT industry.
• Layoff data analysis does not consider the employee ratio of companies due to those details being private.
• With only a numeric representation of bias and the impact of multiple features, the study does not put forward any solutions.

Ethics:  

• We have followed these ethical rules:
• We didn’t point out any specific field or company.
• We didn’t use any personal or private data.
• We didn’t use data that was made for some other purpose.