A sentiment analysis project
By Stacy Shimanuki
For OIDD 245 under Professor Tambe at the Wharton School
Who are migrant workers in Japan?
Japan has historically had an extremely low foreign population and an even lower immigrant population. In 2018, 2.64 million foreigners resided in Japan according to the Ministry of Justice data, constituting a mere two percent of the Japanese population. Of these migrants, roughly 1.46 million were migrant workers employed at 220,000 Japanese companies, constituting an 11 percent year-on-year increase of both migrant workers and companies employing migrant workers.
However, due to its aging population and need for labor, particularly in the sectors seen as the “3Ks” — kitanai meaning dirty, kitsui meaning demanding, and kiken meaning dangerous — Japan created a migrant worker system under an innocuous name: the Technical Intern Training Program (TITP), which was theoretically meant to place migrants in short-term internships to learn technical skills, but de facto constituted a migrant labor system comparable to the systems in South Korea, Taiwan, Singapore, and Hong Kong.
However, this system received heavy international backlash and media attention for abuse towards its migrant workers; an internal Japanese government investigation found that over 70 percent of employers of migrant workers had violated applicable labor laws. Externally, a United Nations report disparaged the system as “slavery-like.”
To address these issues, as well as the 1.2-million-person labor shortage (due to low birth rates of less than 1.5), Japan overhauled their policies in 2018; and in fact, just this month they have proposed to abolish the TITP entirely.
Why does a sentiment analysis matter, and how is it conducted?
Governments are by and of the people — public opinion both influences public policy, and is influenced by public policy. In Japan particularly, anti-immigration public sentiment is said to be the top barrier to increasing migration and labor migration to this aging population that desperately needs it. Accordingly, I want to examine Japanese public perceptions of migrant workers in Japan, hoping that such analysis can draw people to re-think those perceptions.
In this project, I originally wanted to look at the Japanese public mindset towards migrant workers in Japan through social media to get a pulse on the everyday people of Japan; however, given that the APIs for Twitter recently tightened (and LINE and Instagram have always been more challenging), I have shifted my focus to comparing different media perspectives, since media greatly influences public perception. Specifically, the following sentiment analysis will compare international news, Japanese domestic news, and international organizations like the UN.
I have scraped relevant articles from the Economist to represent international media publications and from Asahi Shimbun to represent Japanese domestic publications (keeping in mind that within Japan, Asahi is known as a leftist publication), which I will use to compare the international vs Japanese perspective. I have also scraped UN Human Rights Council’s National Periodic Review reports (essentially the UN’s grade report on Japan on their current human rights status), including both the UN official report and an NGO stakeholder-created report for the 4 years available (2022, 2017, 2012, and 2008), which I will use to both compare the UN vs NGO perspective, as well as compare those perceptions over time.
I anticipate that the Japanese perspective is harsher on migrant workers than the international one, due to both historical trends and to a fixation on isolated negative incidents that are over-reported in media, such as migrant workers “escaping” (失踪, meaning fleeing their designated workplace and/or overstaying their visa). Even though less than 9,000 migrants do this each year (which represents roughly 2% of overall migrant workers in Japan), I expect that that as well as other “illegal” crimes have an outsized reputation among the people.
So what kind of language and topics pop up in media about migrant workers — and how do they differ by sources?
The following word clouds below (from each media source) can tell us a few main takeaways:
Finding #1: Asahi is more specific and technical (and less ideological) than the Economist.
The Asahi word cloud focuses more on words like “Vietnam”, “trainees” (which is a TITP term), “visa”, and “pandemic”. These words are specific to Japan’s migration policies and revolve around the logistics of migrant workers and its practical implications for Japan. Conversely, the Economist word cloud mentions more words about Japan vs foreignness, focusing on the ideological implications of Japan’s changing stances on immigration at a higher level.
Finding #2: UN National Reports have gotten more specific and issue-oriented over time, and they do not seem to vary greatly from the stakeholder reports.
The more recent national reports in 2022 and 2017 focus on mentioning specific issues, for example words like “indigenous”, “Ainu” (which is an ethnic minority in northern Japan), “Fukushima”, etc. The earlier reports were less detailed, talking more vaguely and broadly about discrimination and larger categories like “violence”, “racism”, “education”, etc.
Additionally, the 4 word clouds for the national reports do not seem to differ significantly from the 4 from the stakeholder reports. This could signify that the NGOs that participate in reviewing human rights for the UN are aligned with the UN, as well as the mere fact that much of the style of writing (and thus language used) is similar.
How do these sources differ in usage of relevant key terms, and what can that suggest about their overall perspective?
The following charts count the frequency of specific keywords in each source selected, with the following main takeaways:
Finding #3: The Economist is more pro-migrant and progressive than Asahi.
First of all, the Economist mention of “abuse” is much higher than Asahi’s, while Asahi’s mention of “illegal” and “foreign” are much higher than the Economist. Essentially, the Economist is more focused on disparaging the Japanese system and society, where Asahi is more focused on disparaging migrant workers themselves for any issues. On an interesting note, neither source use language of “discrimination” at all, which contrasts strongly with the UN.
Second, the Economist refers to those who came to Japan to work as “migrant workers”, which is a term of international migration and labor, whereas the Asahi mentions them more as “interns” or “trainees”. This is in line with Asahi’s language and focus overall as a more specific and Japan-specific publication…but it also could suggest that Asahi acknowledges and validates the TITP as a true “intern training” program rather than the international understanding of the program as a migrant worker system. By using the sugarcoated terminology, Asahi appears to be more pro-government and pro-TITP.
Finding #4: UN reports focus less on “migration” recently than they did previously.
Contrary to my expectations, the UN reports (both in national reports and stakeholder reports) have decreased their mention of “migrants” — that is, the reports have grown to focus on other aspects of human rights and mention migration-related human rights issues less. However, to caveat this, they have started to mention the TITP program specifically to small amount (where they did not mention it all previously) — so it is hard to determine from this term frequency analysis whether the UN is overall focusing less on migration issues in Japan or whether it is just that are growing more specific and reviewing Japan’s actual policies.
Additionally, the UN reports use more distant and formal language compared to both news publications; for example, they mention “discrimination” much more than they mention more harsh terms like “abuse” or “illegal”.
Do these sources overall indicate more positive or negative sentiment surrounding migrant workers?
Looking at the following sentiment scores (generated from the afinn analyzer):
Finding #5: The UN reports experience a general decrease in sentiment over time, with a significant low-point in 2017.
This suggests that the UN has grown harsher and more negative with the status of human rights in Japan over time, including migration human rights issues. The all-time low in 2017 also makes sense, since that is the year that the report disparaged TITP as a “slavery-like” system, which then spurred Japan to reform its systems. However, the continuation of an overall negative sentiment score for 2022 is troubling, as it indicates that perhaps those reforms were not entirely successful or impactful. However, here there is high confounding with other non-migration issues, and it is also likely that the continued dip in sentiment is due not to Japan’s migration reform but instead attention on their poor refugee acceptance record after the Ukrainian refugee crisis.
It is also interesting to note that the national reports are slightly more negative than the stakeholder reports, and wonder if perhaps that is due to a self-censoring from the stakeholders, or additional information that the UN accesses beyond the stakeholder information, or some other reason.
Finding #6: The Economist is more negative in sentiment than Asahi.
The sentiment score for the Economist is significantly lower than for the Asahi, which could suggest that, in line with previous comparisons of the two publications, that the Economist is more concerned with the issues surrounding migrant workers in Japan, whereas Asahi focuses less on a good-bad dynamic and more on the tangible impacts and processes of the systems. This could also be generalized to say that Asahi is more supportive of the status quo of the TITP and migrant work in Japan.
So, what are the main takeaways about media perception of migrant workers in Japan — and what next?
The various types of NLP analysis above suggest (but do not prove per se) a few key takeaways about media perceptions: first, that Japanese domestic publications are much more specific, pragmatic, and positive about the status quo of migrant worker issues. Whether this is due to Japan’s historical tendencies to fact-wash issues (as they did in many history textbooks about Japan’s colonial legacies in Southeast and East Asia in WWII, to great international controversy), or due to an active support for (or at least unwillingness to harshly criticize) Japan’s TITP system and overall migration policies, or general lack of civic engagement, or some other reason, is unclear — but in order for Japanese citizens to have a well-rounded perspective on issues in their own country (and thus influence public policy), the international perspective should be merged with the domestic one. Japanese citizens should know that what the government called “interns” were actually “migrant workers”, and that the number of Japanese employers who abuse migrant workers is much higher than the number of migrant workers who flee or commit crime, and that the world’s organizations and international community disparage this TITP system — and it is the burden and duty of Japanese publications to tell them these things.
Second, the UN reports on Japan have grown more specific, and more negative in sentiment (especially in 2017) — and since that has had a pronounced impact on Japan’s policies and reform in the real world, this should constitute a positive case study for the impact of UN “naming-and-shaming” to influence countries to improve their human rights record.
So what can you do? Unless you become an activist, academic, or policy-maker or policy-influencer on this issue, probably not much.
Except, government is of and by the people. Everyone can be a policy-influencer (or at least micro-influencer) in their own country — so then the burden is on you to be aware of and understand at least the basics about these issues in your own country, to correct any biases and misconceptions and misperceptions, and to advocate at least passively against the wrongdoings in such systems that affect people negatively.
Here, and with future research in my senior thesis in the fall, I hope that Japanese citizens re-think their perceptions (and stance) on migrant workers — for readers in America and elsewhere, I challenge you to draw the connections to similar issues at home, such as discrimination and sentiment against labor migration here. And to expect and challenge media and other information sources to hold themselves to a higher standard when it comes to using language to frame public perceptions in a way that is fair, thoughtful, and productive.
Data sources used:
Datasets from the Japanese Ministry of Justice, Ministry of Health Labor & Welfare, and from JITCO
UN HCR NPR reports on Japan
