---
title: "From 3.8% to 98.2% Filled: What Japan's Skilled-Worker Exams Do and Don't Certify"
description: "Japan's Specified Skilled Worker route holds 438,128 workers across 19 fields, with field fill rates from 3.8% to 98.2%. What a government skills exam certifies, who sets the standard, and why test capacity belongs in the assessment design."
canonical: https://www.assessall.com/blog/from-3-8-to-98-2-filled-what-japans-skilled-worker-exams-do-and-dont-certify
published: 2026-11-07
updated: 2026-11-07
author: "AssessAll Editorial"
source: AssessAll
---

# From 3.8% to 98.2% Filled: What Japan's Skilled-Worker Exams Do and Don't Certify

Japan's Specified Skilled Worker route holds 438,128 workers across 19 fields, with field fill rates from 3.8% to 98.2%. What a government skills exam certifies, who sets the standard, and why test capacity belongs in the assessment design.

A national skills-proficiency exam certifies that a candidate cleared a government-set minimum for one defined occupational field, set by one specific body. It is a licensure-style threshold judgment — pass or fail against a standard — not a rank order, and not a prediction of how anyone will perform in your job. A pass is a floor, not a forecast.

Japan's Specified Skilled Worker (SSW) route is the clearest live demonstration of what that distinction costs employers who get it wrong. It is now a large system: as of the end of August 2026, [Japan's Immigration Services Agency counted 438,128 residents](https://www.moj.go.jp/isa/content/001472175.pdf) on the Specified Skilled Worker (i) status across 19 eligible fields, against an expected-intake ceiling of 805,700 running to March 2029 — a 54.4% fill rate, up 87,422 in twelve months. Two years earlier, [at the end of 2024, the figure was 283,634](https://www.moj.go.jp/isa/content/001462558.pdf), itself a 36.1% jump on the year before.

Every one of those workers passed a skills test and a Japanese-language test to get there. Very few employers using the route can say what either test certifies.

## The headline number nobody reads correctly

Field-level fill rates against the same ceiling range from 3.8% to 98.2%. Food service sits at 98.2% of its 50,000 ceiling. Food and beverage manufacturing is at 77.4% of 133,500. Construction is at 69.1% of 76,000, nursing care at 65.5% of 126,900. Industrial product manufacturing, which carries the largest ceiling of any field at 199,500, is only 32.2% filled. Road transport is at 3.8% of 22,100.

The exam is not what differs. Road transport and several others were only [added to the acceptance list in January 2026](https://www.moj.go.jp/isa/content/001472175.pdf), so newness explains a good deal of the spread. But the wider lesson holds for anyone planning headcount around a credential: the gate at the front of a pipeline is rarely what determines how much talent comes through it. Capacity, demand, wages, employer readiness and test logistics all sit in front of the exam, and none of them appear on a candidate's certificate.

## What a threshold exam is built to decide

The [Standards for Educational and Psychological Testing](https://www.apa.org/science/programs/testing/standards), published jointly by AERA, APA and NCME, make validity a property of a specific interpretation for a specific use — not of a test in the abstract. A credentialing exam is validated for one interpretation: that the holder meets a minimum standard of competence to practise in a defined scope. That is all the evidence supports.

Three consequences follow, and they are routinely ignored:

- **A pass carries no rank information.** Two candidates who both cleared the standard are, as far as the exam is concerned, indistinguishable. If the certificate reports only pass or fail, there is no scale on which to prefer one.
- **The pass mark encodes a policy judgment, not a natural boundary.** Somebody decided what "minimally competent" meant for that field. Who decided, and on what evidence, determines how much the threshold is worth.
- **It was not built to predict your outcomes.** The [EEOC's guidance on employment tests and selection procedures](https://www.eeoc.gov/laws/guidance/employment-tests-and-selection-procedures) puts the burden of job-relatedness on the employer using the test for its own decision. A government exam validated for licensure does not transfer that evidence to your hiring funnel.

## "The SSW exam" does not exist

There is no single SSW skills test. The agency's [examinations page](https://www.moj.go.jp/isa/applications/ssw/nyuukokukanri01_00135.html) routes each field to a different administering body: the Ministry of Health, Labour and Welfare runs the care skills evaluation test and a separate care-sector Japanese test; construction goes to the Japan Association for Construction Human Resources; shipbuilding and road transport both go to ClassNK; food manufacturing and food service share one body, the Organization for Technical Skill Assessment of Foreign Workers in the Food Industry; railways sit with the Ministry of Land, Infrastructure, Transport and Tourism. More than a dozen separate organisations, plus two ministries, set and run these exams.

That matters because standard setting is where a threshold exam earns or loses credibility. Different bodies, different item banks, different panels, different judgments about the minimum. A pass in one field tells you nothing about a pass in another, and treating "SSW pass" as one currency across fields is the same error as reading a 70% on two unrelated internal tests as the same score.

## Watch the language gate move

Japan's language requirement is being adjusted with more deliberate care than most corporate cut scores receive.

The [Japan Foundation Test for Basic Japanese](https://www.jpf.go.jp/jft-basic/e/about/index.html) (JFT-Basic) is a 60-minute, roughly 50-question computer-based test scored from 10 to 250 and mapped to the Common European Framework: 145–174 is judged A1, 175–199 is A2.1, and 200–250 is A2.2. Until July 2026 it reported only A2.2; since August 2026 it also reports A1 and A2.1 — a deliberate widening of the band it can resolve, which is what you do when one threshold is being asked to serve several purposes.

Meanwhile the bar is rising where the consequence of failure is highest. The agency has confirmed that [from 1 April 2027 a "B1-equivalent or above" standard applies](https://www.moj.go.jp/isa/03_00198.html) to Specified Skilled Worker (ii) certificate-of-eligibility applications, and to Specified Skilled Worker (i) in the bus and taxi categories of road transport and the train-crew category of railways — replacing the JLPT N3 requirement for those categories. The cabinet decisions behind those changes are dated 11 March 2025 and 23 January 2026.

Compare that with what the lower levels claim. The JLPT describes [N4 as "the ability to understand basic Japanese"](https://www.jlpt.jp/e/about/levelsummary.html) and N3 as understanding everyday Japanese "to a certain degree". Those are honest, modest descriptions. A regulator that asks for more from a bus driver than from a food-plant operator is doing defensible standard setting: tying the cut to the cost of error in the role. That is the part most organisations skip when they pick a passing mark.

## Capacity is part of the assessment design

In October 2026 the agency published a notice telling candidates to stop waiting for the JLPT. Demand for the [December 2026 sitting was such that all levels, N1 through N5, closed before the application window ended](https://www.moj.go.jp/isa/content/001471906.pdf). The notice points candidates to JFT-Basic instead — an A2.2 or A1 pass satisfies the requirement for Specified Skilled Worker (i) and for the new Employment for Skill Development status due to start in April 2027 — and notes that JFT-Basic runs at 11 locations nationwide, several times a month.

A gate nobody can book is not a gate. It is a queue, and it pushes your hiring timeline out by a full test cycle without changing a single thing about candidate quality. Throughput, sitting frequency and retake windows belong in an assessment design document alongside reliability and cut scores, not in an operations footnote.

## A checklist for hiring against someone else's exam

1. Read the exam's stated purpose, not its name. "Skills evaluation test" tells you nothing about construct or scope.
2. Find out who set the pass mark and on what evidence. An unexplained threshold is an unexamined one.
3. Treat the pass as eligibility. Measure fit, judgment and communication yourself, with an instrument validated for your decision.
4. Never compare passes across fields or administering bodies as if they were one scale.
5. Match the language level to the consequence of error in the specific role, the way the 2027 B1 change does.
6. Check sitting availability and retake rules before you commit to a pipeline timeline.
7. Collect your own post-hire outcome data. It is the only evidence that tells you whether the credential predicted anything in your context.

Points 3 and 7 are where a platform helps more than another certificate. AI-graded scenario assessments with proctoring integrity bands measure the role-specific judgment a licensure threshold was never designed to capture, and AssessAll's pay-as-you-go credits at ₹30 / US$0.50 per assessment let you run that extra layer on a cohort of fifty without a seat-licence negotiation — with 100 free credits on individual signup and 250 on corporate.

**The takeaway:** a national skills exam tells you a candidate was allowed through a door, which is genuinely useful and strictly bounded information. What they will do once inside your organisation is a question you still have to ask yourself.

---

Source: https://www.assessall.com/blog/from-3-8-to-98-2-filled-what-japans-skilled-worker-exams-do-and-dont-certify