Filling in the Unexplained Pay Gap to Build a Better Society—How Mercari Is Tackling the Gender Pay Gap
In order to empower women and promote their participation in the workplace, in July of 2022, the Japanese government introduced a requirement for companies with 301 or more employees to disclose the differences in pay between men and women.
At Mercari, we calculated this overall gender pay gap, which shows the average difference in pay between men and women within an organization. However, in order to ascertain the situation of the gender pay gap at the company more accurately, we also calculated the “unexplained pay gap.” The unexplained pay gap is the portion of the pay gap that remains after accounting for factors such as roles, grades, and job types. Our analysis revealed an overall gender pay gap of 37.5% and an unexplained pay gap of 7%.
In this edition of Mercan, we asked the members involved in taking action to close this gap to reveal how they identified the issues, their analysis process, the kinds of actions they took and the results, and the improvements to be made going forward. The gap uncovered at Mercari is not solely a problem seen at one company, but rather an issue that runs across our entire society. Our hope is that the work we’ve done will inspire others.
Featured in this article
Amy BurkeAfter graduating from the National University of Ireland, Galway with a bachelor’s degree in Irish Studies with Arts, Amy moved to Kagawa Prefecture, Japan to take part in the JET Programme for three years. Her interest in supporting non-Japanese people working in Japan prompted her to shift her career into the HR field. In her current work on Mercari’s I&D Team, she strives to create a fair and inclusive work environment for all, regardless of where they are from.
Hitomi SuwaHitomi majored in social psychology at the University of Tokyo Faculty of Letters. After graduation, she joined a digital marketing support company where she worked as an SEO analyst. After working as the B2B marketer of a data-leveraging support business, she joined Mercari as a data analyst. Having joined the Group as a member of the Product Analytics team, she was in charge of the analysis of the web version of Mercari and the search feature. She now maintains HR data infrastructure and analysis for human capital as a member of the HR System & Data Team.
Are there instances at Mercari of opportunities not being provided equally owing to members’ attributes?
— To get things started, could you talk about the developments that kickstarted the company’s initiative to close the gender pay gap?
@Eimi： So it all started in July 2022, when the government added “gender pay gap” to the list of items related to the empowerment of women that companies are mandated to release publicly. The same year, Mercari received EDGE Assess certification, a globally recognized certification on gender equality. At that time, we also received feedback that it would be better to establish a mechanism for regularly monitoring pay equity. Pay equity is the idea of awarding equal pay for equal work.
To step up to this challenge, we made a point of going beyond the requirements stipulated by the government. Rather than take aim purely at the mean difference that we uncovered in our data, we came up with measures that would enable us to monitor pay equity at Mercari. In the background of these measures is the Mercari Group mission: “Circulate all forms of value to unleash the potential in all people.” Similar to the letter of our mission, we believe it’s important for members from various backgrounds to be able to thrive at work. We felt that there was a need for us to investigate whether there are any inequality of opportunities based on member attributes, to use data to visualize the stages of the employee journey, and to link this information to action.
— Specifically, what were some of the actions you took in launching the initiative?
@Eimi： Sometime around December 2022, we performed the first regression analysis (a method of statistical data analysis). Based on the results of this analysis, we started to think about what sorts of actions we should take. We learned that from the moment a new employee joins the company, there is already a gender pay gap. To look into this point further, we held meetings with stakeholders working in recruitment and related areas to come up with a list of the causes. We also referred to the best practices that EDGE provides.
Our project members came from a diverse range of teams including Analytics, Evaluation & Compensation, Payroll, Recruitment, and I&D. We coordinated with each team to compile data and put together an action plan, and then we brought our findings to executive meetings in March of 2023, receiving final approval for our initiative at the end of April. The process to decide on an action plan took about two months. After that, we once again executed a regression analysis of the latest data ahead of implementing the action plan and disclosure, and then prepared the calculations and other information needed to make adjustments to salaries.
The definition of “compensation” was very broad, and the data was distributed across multiple teams
—So how did you forge ahead with actually analyzing the data?
@suwachan： To start off, we performed regression analysis. Once we had somewhat of an idea of the trends evidenced in the data, we proceeded with two kinds of analysis—analysis for disclosure purposes and analysis for corrective measures.
With regard to the analysis for disclosure, we started by listing the elements that determine a member’s actual salary. In the course of our work, and in the interest of giving due consideration to privacy and other factors, we selected items that would be appropriate to use for analysis. We then acquired data by requesting it from each team and ultimately managed to perform regression analysis.
In our analysis for corrective measures, we used a method called Bayesian hierarchical modeling (a statistical model that configures probability distributions of estimated parameters in a hierarchical manner). In the regression analysis, we were only able to output the overall average correction values. However, with the Bayesian hierarchical model we were able to give consideration to differences for variousgroups and for individuals, so we determined it to be an appropriate model for making individual adjustments. For this project, we also collaborated with a data scientist and expert in Bayesian hierarchical modeling, which allowed us to produce these analysis results.
Hitomi Suwa (@suwachan)
— What would you say were the most difficult points of the analysis?
@Eimi： Looking back, one thing I wish I had done differently involved what happened when I asked @suwachan to take care of the data analysis. I hadn’t clearly defined each item like the scope of analysis and compensation. The definition of compensation was really broad, and we had to judge it multilaterally. For example, regular stock options are included in compensation, but we had to define whether things like referral fees and other such allowances should also be subject to analysis. This was the first time we had worked on this initiative so we didn’t have a grasp of what sorts of definitions and conditions were required; there were also a lot of elements that we hadn’t anticipated. In that sense, this project had a lot of lessons to teach us.
@suwachan：Like @Eimi said, creating solid definitions was tough. The government definitions are not clear on how to handle things like money paid for employee benefits or stock options, so we had to look at what the purpose of this gender pay gap analysis was in the first place and examine this while considering different perspectives like accounting and labor.
There was also a lot of data that was hard for us to obtain, like salary offers and evaluation data from several years ago. Plus, there was a lot of data that was split across teams, so obtaining and aggregating it took a lot of work.
What’s more, regardless of whether or not we could obtain certain data, giving due consideration to people’s privacy was another point that we had to be mindful of. For example, even though we hypothesized that data about things like education and being a primary caregiver would affect the pay gap, one of the tough decisions we faced was deciding whether or not it was necessary to use such data when performing our analysis.
What action did Mercari take to address the unexplained pay gap of 7%?
— So what did the results of your analysis look like?
@suwachan：First of all, we discovered that the “raw pay gap,” which is the name given to the simple average gap in wages of men and women at the company, was 37.5%. What this means is that, on average, for every 100 yen that a man makes, a woman makes 62.5 yen. The likely reason for this is that there are a lot more men who hold job grades or job types with higher salaries. This is the extent of the information that the government has mandated that companies disclose.
Next, we performed regression analysis to analyze pay equity. The result was that there was an unexplained pay gap of 7%.
— What is an unexplained pay gap?
@suwachan：What it means is that even among members with the same attributes and who demonstrate the same level of performance, there exists a pay gap between men and women. By including things like attributes and performance in the variables we analyzed for regression analysis, we were able to control the effect that they had. Even if we did everything possible to eliminate the effects that job types and job grades had on salaries, on average the salaries of women were still 7% lower than men’s salaries. After analyzing where the gap came from, we determined that the cause of the gap stemmed from the annual salaries prospective employees are offered during the hiring process.
— So, what actions did you take based on the analysis results?
@Eimi： In determining what kinds of actions to take, we performed research from a variety of angles. We referred to actions that other companies were taking and what sorts of results their efforts yielded. What we found was that when multinational corporations uncover a gender pay gap, many will make corrections to adjust compensation. One common way of doing this is to not reference the salary from a new employee’s previous company when setting an offer. I think we can call this a best practice.
We also spoke with external consultants, to learn what they said about trends in actions that companies overseas take. For example, in Europe, if the unexplained pay gap exceeds 5%, a labor representative works with the company to investigate the situation and implement corrective measures. Apart from this, we also consulted with EDGE. They provided us with a blueprint for running internal communication, so we referred to that as well.
@suwachan：In terms of the process for determining the individual amounts of salary corrections, we used the Bayesian hierarchical modeling statistical method mentioned earlier to create data for salary recommendations for each individual. Using this data as a reference, we then performed individual adjustments in the same way as our regular process for pay raises.
Based on these calculations, in August of 2023, the company adjusted compensation amounts in order to eliminate the unexplained pay gap that was not caused by such things as differences in roles, grades, or job types. The result of this was that we were able to reduce the unexplained pay gap from 7% to 2.5%.
@Eimi：Going forward, in the short term we are aiming to free the company of any unexplained pay gap that is due only to differences in gender. We’ll introduce periodic monitoring of the wage gap on a six-month cycle that uses regression analysis, and if we find that there is an unexplained pay gap of ±1% or more, the company will consider corrective measures.
We’ll also revise hiring practices so that new hires don’t carry over any wage gap they experienced prior to joining Mercari. These are some of the ways we plan to carry on this initiative and make further improvements.
There are no fix-all measures for the gender pay gap precisely because its social significance is immense
— Because there are still so few precedents of projects like this at companies in Japan, I get the sense that it will be no simple task to get people to readily grasp the social background and causes involved. How did you communicate about your initiative internally at Mercari?
@Eimi：This was the first time for Mercari to tackle the gender pay gap, and, like you mentioned, there are few precedents of companies doing anything like this in Japan. So when it came to disclosing this information, it was obviously crucial for us to find an appropriate communication method.
Now that companies of a certain size are required to disclose information on the gender pay gap, the external disclosure we conducted in September 2023 was laid out as a responsibility we had to comply with. It was challenging to figure out how to convey the information internally and how to communicate it in a way that would get people to understand what we’re doing.
Ultimately, the conclusion that came out of our discussions with company executives was that, based on Mercari’s cornerstone foundation of “Trust and Openness,” openly sharing the background and actions of this initiative with our members was the kind of transparency that makes Mercari who we are.
At an all hands meeting held in July 2023, our company leadership shared information with members belonging to our Japan-based businesses about the analysis results of the gender pay gap and the corrective measures we took. Presenters also took time to respond to questions from our members. In addition, we had foreseen that Managers would receive a flurry of questions from members, and so we also made sure to invest time and care in our communication with managers. Ahead of the company-wide announcement, we gave explanations and took questions about the company’s plans for dealing with the gender pay gap at the regular information sessions we hold for managers regarding compensation. We took the feedback we received from these sessions and incorporated it into the company-wide presentation.
— The analysis and corrections for tackling the gender pay gap were part of a company-wide initiative, so it might not be easy for any and all companies to apply our lessons of this project right out of the box, but do you have any wisdom or advice to share?
@Eimi：Let’s see now. So obviously, the causes and background affecting the gender pay gap will vary depending on the company, so I would probably advise other companies to put together initiatives that suit their own company’s circumstances while referring to the practices and research of others.
If it feels daunting to implement a new policy straight away, I suggest identifying what initial actions you can take immediately using fact-based data, and then taking it step by step. This approach should also make it easier for you to gain the understanding and cooperation of those around you. In all likelihood, there is no perfect solution, and the D&I field evolves in tandem with social change, so I don’t think we need to have all of the answers now.
@suwachan：The thing that I have reaffirmed from being a part of this project is that a gender pay gap can take root regardless of any intention to operate a fair system that sets out to treat all people equally regardless of their gender.
Also, I imagine that calculating the gender pay gap is difficult at any company since it requires the involvement of so many different divisions. However, its social significance is immense. There can be situations where confusion ensues as a result of not knowing which team holds what data, or where there is round after round of discussions trying to come up with clear definitions for things; Gender Pay Gap Analysis is not an area with hard and fast answers, so I think it’s best to be able to consider your organization’s policy based on what your aims are and the characteristics of your company. Consultants offer services to this end, and the government releases formulas that can help as well, so first I would like other businesses to start with what is feasible for them.
@Eimi： This was the first time we had ever worked on an initiative like this, so there were a lot of things that we did not know about. Especially when it came to the analysis process, there were a lot of preparations that we had to take care of involving the analysis conditions and information.
It’s not just about the data results; it’s also important to make resolute decisions together with the project members on how to explain the data in an easily understandable way and how to visualize it so that the message is conveyed clearly at a glance. I’m glad that frank communication allowed us to find the path to solutions.
@suwachan：It was an honor to be involved in a project of such immense social significance. I’m grateful to the stakeholders who have blazed the trail for us thus far. We may have been working to meet the requirement to disclose a gender pay gap, but our goal with this project was not to just disclose this information and stop there. It was important for us to fill in the gaps that occur even where it is unintentional and build a better society going forward. I really think it’s great that Mercari was able to take action as a company, and I hope that we can serve as an example for others to follow.