Credential recognition asks whether a qualification earned abroad is equivalent to a domestic one. Skill testing asks whether the person can perform the work to a defined standard, regardless of where they trained. Recognition evaluates documents against a reference curriculum; testing evaluates performance against a job standard. Germany leans on the first. Japan built its fastest-growing work route on the second.
Both are defensible. They fail in different places, and an employer hiring across borders needs to know which question it is actually asking.
Germany: recognition, and what happens when the paper does not match
Germany runs one of the world's most formalised credential-recognition systems. Roughly 71,000 applications for recognition of foreign professional qualifications were filed in 2024, up about 14% on the previous year, according to the federal recognition statistics. Nursing dominates the caseload with 22,425 applications, followed by doctors of medicine at 10,857. India is among the largest countries of origin.
The outcomes are the interesting part. Of 55,332 procedures concluded in 2024:
- Full equivalence: 43.2%
- Compensatory measure required: 46.6%
- Partial equivalence: 8.9%
- Rejected outright: 1.3%
Read that again. The system almost never says no — only 1.3% of procedures end in a negative notice. What it says, about half the time, is not yet. A compensatory measure is an aptitude test or a supervised adaptation period: a skills assessment, arriving after the document review has failed to settle the question.
So the recognition route does not replace testing. In the modal case it produces a referral to testing, with a document-review delay in front of it.
Germany has effectively conceded this in policy. Under the recognition partnership introduced by the reformed Skilled Immigration Act, a worker can enter and take up qualified employment before recognition concludes. The requirements are a home-country qualification of at least two years' training or a degree, German at CEFR A2, and a written agreement in which the employer commits to release the worker for the training needed to close the gap. The permit runs a year and is renewable up to four. The assessment has not been removed — it has been moved from before hiring to during employment.
Japan: define the standard, run the exam
Japan's Specified Skilled Worker route does not ask whether a Vietnamese catering diploma is equivalent to a Japanese one. It sets a standard per sector and administers a common exam.
An SSW (i) applicant must pass a sector skills proficiency test and a Japanese-language test — either the JFT-Basic or JLPT at N4 or above. The skills tests are computer-based and delivered across sending countries as well as inside Japan, covering fields including nursing care, construction, food and beverage manufacturing, food service, agriculture, shipbuilding, aviation and automobile maintenance. Comparability comes from the instrument, not from curriculum mapping.
The scale is no longer marginal. Japan counted 390,296 specified skilled workers at the end of 2025, a record, of whom 7,955 held the higher SSW (ii) status. Growth has been fast enough to hit its own ceilings: the food service sector was given an acceptance cap of 50,000, reached roughly 46,000 by February 2026, and new Type 1 admissions in that sector were suspended in principle from April, expected to hold until the end of March 2029.
That last fact is a caution, not an endorsement. A testing-first system scales admissions quickly, which means the binding constraint moves from assessment capacity to policy quota.
What each model measures — and what the evidence says
Recognition measures inputs: hours, syllabus, awarding institution, scope of practice. Testing measures current performance against a standard.
The selection-science evidence favours the second for predicting job performance. In the revised meta-analytic estimates published by Sackett and colleagues in the Journal of Applied Psychology (2022), the strongest widely used predictors are structured interviews at .42, job knowledge tests at .40 and work sample tests at .33, with general cognitive ability at .31. Every one of those is a measure of what a person does or knows now. None of them is a credential.
That does not make recognition pointless. It makes it the answer to a different question.
When recognition is the right instrument
- Regulated professions where scope of practice is set in law — medicine, nursing, electrical trades. No test result gives a person the legal right to practise; a recognition decision does.
- Where the risk being managed is public safety and the standard is statutory rather than behavioural.
- Where portability matters: a recognition notice follows the worker between employers in a way an employer's internal test result usually does not.
When testing is the right instrument
- Unregulated and semi-skilled roles hired at volume, where no credential exists for most applicants.
- Where the local job standard genuinely differs from any home-country curriculum — customer-facing service norms, local equipment, local safety practice.
- Where the qualification is old. A certificate is dated evidence about a syllabus; it says nothing about what has been used since.
The trap in the middle: a test that does not travel
Testing across borders introduces its own failure mode, and it is not the one most employers worry about. A test translated into another language is not the same test. The ITC Guidelines for Translating and Adapting Tests treat translation as one step inside adaptation, alongside construct equivalence, administration conditions and score interpretation. Skip the adaptation work and you get a test that measures reading fluency in the target language as much as the skill you named.
Two practical consequences. First, keep language load deliberate: if English or Japanese comprehension is part of the job standard, measure it separately and report it separately, rather than letting it contaminate a technical score. Second, remote delivery needs integrity evidence attached to the score, not a pass/fail verdict on the candidate's honesty — AssessAll reports AI-proctoring integrity bands alongside AI-graded scenario responses for exactly this reason, so a reviewer can see how much weight a remote result carries.
A checklist for hiring across borders
- Separate the legal question from the capability question. Recognition answers the first. Only assessment answers the second.
- Write the job standard before choosing the instrument. Sector exams work in Japan because the standard is defined per field, not per school.
- Interrogate the credential's date and scope, not just its existence.
- Adapt, do not translate, and pilot the adapted form before it makes decisions.
- Set the cut score locally. A passing mark is a policy judgment about the local role and applicant pool, not a property imported with the test.
- Attach integrity evidence to every remotely delivered score.
- Check adverse impact by group before launch, in line with the EEOC's guidance on employment tests and the Standards for Educational and Psychological Testing.
- Keep the result portable. A score the candidate can carry to the next employer — a Skill Passport — does for capability what a recognition notice does for credentials, without the twelve-week wait.
The takeaway
Germany's numbers show that document recognition rarely rejects anyone; it defers nearly half of all cases into a test anyway. Japan's show what happens when you put the test first and let the credential go unexamined. If your hiring runs across borders, decide deliberately which question you need answered — and when the answer you need is can this person do the work, buy a measurement, not a translation of someone else's paperwork.