What is the Human Resources Matching Survey Open Data?
Written by Fractal-kun (Codex) / Edited by Takuya Iwasaki
Japan’s Cabinet Office publishes the results of its Human Resources Matching Survey as open data on e-CSTI, a platform that provides data to support evidence-based policymaking.
e-CSTI provides shared access to science and technology policy data and analytical tools. The Cabinet Office developed it to support evidence-based policymaking, or EBPM. One of its pillars is “visualizing industry needs for human resource development.” The e-CSTI overview explains its purpose and structure. The source pages linked here are in Japanese.
The Human Resources Matching Survey asks working adults about subjects such as the field they studied at university or another educational institution and the specialist knowledge important in their current work. How are learning and work connected? What knowledge do people need in working life? The survey examines these relationships to inform discussions about education and human resource development.
And its response data are publicly available for us to obtain and analyze ourselves.
Analyze the data around your own questions
The first thing we would like you to know about these open data is that you can produce summaries around your own questions.
You may have seen an age distribution, an industry breakdown or counts by educational attainment. But while looking at them, you might wonder: “What would this look like for just this age group?” or “What kinds of work do people who studied this field do?” A published table for that particular combination may not exist.
What is available here is the individual response data from which summary tables are made. The data allow answers about age, education, work and other subjects to be linked to the same respondent. That is why you can combine conditions—for example, people in a particular age group doing a particular kind of work—and calculate a summary for them.
The data are divided into several CSV tables. You can import them into a database or read them with tools such as Python or Tableau, preserving the connections between one person’s answers while analyzing the questions that interest you.
To let you experience what that means, Rokugeisha created “Explore the survey through 11 attributes.”
Start by selecting something that interests you, such as a gender or age category. The other charts change too.
For example, selecting “Male” recalculates the age, industry, occupation and other distributions for male respondents. Select an industry as well, and the results are recalculated for men working in that industry. Each click redraws the other charts for the people matching your chosen conditions.
The count at the top tells you how many people’s responses you are looking at. Select the same category again to clear it.
The 11 charts in this view show the same respondents through different attributes. People in the age chart are also in the industry and occupation charts. That is why a condition selected in one place affects the other distributions.
What appears on screen is a set of summary results. Behind those results are data that can be summarized again and again using different combinations. It is this relationship that we wanted the visualization to make visible.
The tools include all six survey rounds: fiscal years 2014, 2016, 2019, 2021, 2023 and 2025. Select a survey year to explore its respondents. Survey populations and questions differ between rounds, so check the conditions of each survey when comparing years.
Starting with “Is that really true?”
The survey itself began with a question to investigate.
Might there be a mismatch between the people universities educate and the people industry needs? Might the numbers trained in each field, or the knowledge they acquire, fail to align with society’s needs? When such claims are made, where—and in what form—does the mismatch actually occur?
The series now available as open data traces back to a fiscal 2014 survey by Japan’s Ministry of Economy, Trade and Industry (METI) on qualitative and quantitative mismatches in the supply of and demand for human resources between industry and educational institutions. The source page’s list of survey projects documents this history and the subsequent rounds.
In May 2015, METI and the Ministry of Education, Culture, Sports, Science and Technology (MEXT) jointly established an industry–academia–government roundtable on developing science and engineering talent. It brought the ministries responsible for industry and education together with businesses and universities. This history is also described in the summary of METI’s 2017 manufacturing white paper (PDF).
The roundtable’s first meeting introduced a survey intended to clarify what a mismatch between industry needs and educational content would look like, if one existed. The minutes of that meeting (PDF) record a discussion of using a questionnaire for working engineers to examine that relationship.
What Iwasaki sees as important about this effort is that it turned “there may be a mismatch” into something to investigate with data. If differences between fields can be described empirically, discussion of the education and learning opportunities people need can become more concrete.
“Is what people are saying really true? Let’s first find out.” We see that question as one of the ideas at the root of evidence-based policymaking.
Now, the data collected for that purpose are available for wider use. Alongside those who have analyzed them for the original survey objectives, people in many different positions can approach them with questions of their own.
A survey’s population and questions are defined for its purpose. How far it can answer a new question requires comparing that question with what the survey actually asked. Making the data public also makes it possible to examine that potential for yourself.
Data collected to answer particular questions are now available for anyone to consider and test how they might serve their own questions. That is the point we most wanted to convey here.
Bring a question to four different perspectives
The 11-attribute view gives a broad overview of the connections between items. We have also prepared four thematic visualizations to help you explore more focused questions.
Theme: Who does what kind of work?
This view combines age, gender, industry and occupation. How does the distribution of work change when you select an age group? What ages do you find when you select a kind of work?
For industry and occupation, you can start with broad groups and open their detailed categories. Follow the parts that interest you, a little at a time.
Open “Who does what kind of work?”
Theme: How education connects with work
Select educational attainment or university major, then examine the industry and occupation distributions. Among people with the same level of education, does work spread differently depending on their field of study? Looking the other way, what educational backgrounds and majors come together in the same kind of work?
In the Human Resources Matching Survey, the university-major question is not asked of respondents whose highest qualification is high school, a professional training college or certain other institutions. The on-screen explanation also helps you see how the questions asked differ with educational attainment.
Open “How education connects with work”
Theme: Work and income distributions
This view explores annual income alongside gender, age, employment status and the number of employees at the respondent’s employer. A heatmap shows combinations of age and income through colour intensity, with a bar chart of counts by income category below it.
Selecting a cell in the heatmap adds that age–income combination to the search conditions. You can also exclude “Prefer not to answer” and “No response.” The colour scale changes with the conditions, so check the percentages in the legend when you change your selection.
Open “Work and income distributions”
Theme: Where people live and work
The survey asks about residence and workplace, as well as the location of respondents’ high schools and the institutions where they obtained their highest qualification, such as university or graduate school. The 2014 round asked only about residence.
This visualization combines residence and workplace with occupation, age and gender. When you select people living in one prefecture, where do their workplaces extend? Does that relationship differ by work or age group?
The residence and workplace bars use the same count scale. To leave out missing location answers, use the switch that excludes anyone with no response for either residence or workplace.
Open “Where people live and work”
Using the views
Each view has a survey-year selector at the top. When you choose search conditions, the number of respondents matching them appears prominently above the charts. You can immediately see how many people’s responses you are looking at.
Your selected conditions appear in a list. Click the “×” on the right of any condition to remove it individually.
From an unexpected observation to the next question
The four themes are starting points. As you select areas of interest, change conditions and compare results, you may notice a difference or connection you had not expected. “Why is this part different?” “Does the same pattern appear over here?” One question can lead to another.
Different experiences and interests draw people’s attention to different things. Some readers may find insights through approaches we never imagined. Interacting with data can help you discover the question itself. We call tools that let you explore data this way “interactive visualization tools.” Part of the reason for making them is to create opportunities for unexpected observations and new questions.
Taking your own questions further
All five visualizations have a “Save displayed summaries as CSV” button. You can take away the numbers behind a distribution, examine them in a spreadsheet or make another chart. What this saves is the summary results shown in that view.
If you want to combine other items or use your own analytical methods, obtain the analysis data and explanatory materials from the e-CSTI source page. Consult the questionnaires and question descriptions as you work out the summaries needed for your own question.
Follow the source’s conditions of use and credit the source. When publishing processed data or analyses, also identify who carried out that processing or analysis.
Recalculating results yourself may raise questions different from those that occurred to you while looking at the screen. Being able to pursue those questions is another part of the value of having the data available to use.
Towards richer discussion around these data
Iwasaki has worked on the analysis and visualization of this survey and has also been responsible for preparing and creating its open data. This introduction is both a guide for people encountering the data for the first time and a report back to those who have worked on the effort together.
The Human Resources Matching Survey data are now available for anyone to use. Rokugeisha is also creating ways for people to begin exploring them.
We hope more people will learn that these data are public, and more will try using them. People involved in education. People thinking about human resource development. And above all, people thinking about their own learning and work. Each can bring their questions to the same data.
People may interpret a result differently. When that happens, they can continue the discussion with others while checking which survey year, which conditions and how many respondents they are talking about. That is part of what sharing these data makes possible.
We want these data to be used and to give rise to richer discussions. At Rokugeisha, we would like to take on the work of creating those entry points.
We will introduce the questions about specialist knowledge—one of this survey’s most distinctive features—in a separate article.
Please let us know what you notice or wonder about when you try the tools. Your observations can help us decide what visualizations and articles to work on next.
Send your thoughts on this article by email
Writing metadata
| Item | Details |
|---|---|
| Manuscript date | 30 October 2026 |
| Writing and source checking | Fractal-kun (Codex) |
| Concept, questions, structure and editing | Takuya Iwasaki |
| Source material | Conversations between Iwasaki and Fractal-kun about the survey introduction and five visualizations; public Cabinet Office and METI materials linked in the article |
| Published Japanese version | Sixth draft |
| Note | Following Iwasaki’s review, a note was added announcing a separate article about the specialist-knowledge questions. This English translation will be reviewed by Iwasaki after publication. |