An employability rate is the share of candidates in a tested sample who clear a general benchmark test's cut score. It describes a population, not a person, and is not evidence that any individual will perform in a specific role. India's 2026 rate is 56.35%. Graduate unemployment rose too. Both numbers are right.
Two true numbers that describe different things
The India Skills Report 2026 — its 13th edition, produced by ETS with CII, AICTE, AIU and Taggd — put national employability at 56.35%, up from 54.81% in 2025 and 46.2% in 2022. The figure comes from the Global Employability Test administered to more than 100,000 candidates. On its own terms it is a real, decade-long improvement.
Azim Premji University's State of Working India 2026 describes the same young workforce and finds the opposite mood. Two-thirds of unemployed 20–29-year-olds were graduates in 2023, up from 46% in 2017. Fewer than half of all graduates — 48.8% — were employed at all. Only 6.7% held permanent salaried positions and 3.7% worked in white-collar occupations. Between 2004–05 and 2023, roughly 5 million graduates entered the workforce each year and about 2.8 million found employment.
The official Periodic Labour Force Survey for 2025 completes the picture: overall unemployment of 3.1%, youth (15–29) unemployment of 9.9%, and 6.5% among people educated to secondary level and above. Unemployment in India rises with education, not the reverse. PLFS 2023–24 data put graduate unemployment near 13% against 3.2% overall.
So a candidate can be "employable" and unemployed at the same time without either measurement being wrong. That is not a paradox. It is what happens when a population-level benchmark gets read as a hiring signal.
What an employability score actually measures
The Standards for Educational and Psychological Testing, the joint AERA/APA/NCME framework that governs responsible test use, makes the point that resolves the confusion: validity is not a property of a test. It is a property of the interpretation of a test score for a proposed use, and it has to be argued and evidenced separately for each use.
A national employability index has a legitimate, well-supported use: tracking whether the supply side of the labour market is improving, comparing states and disciplines, and telling policymakers where to put money. Nothing in that evidence base supports a second use — deciding whether to interview a particular candidate for a particular job in a particular city.
| | Employability benchmark | Selection assessment | |---|---|---| | Question asked | How does this cohort compare to a national bar? | Will this person do this job well? | | Reference point | Norm-referenced against a large sample | Criterion-referenced against role performance | | Content | General aptitude, language, domain basics | Job-analysed tasks and behaviours | | Validity evidence | Fit to the reported construct | Criterion-related evidence for that role | | Right consumer | Policymakers, universities, workforce planners | Hiring managers, recruiters | | Wrong use | Screening an individual applicant | Claiming national skills trends |
Both instruments can be excellent. Neither substitutes for the other, and a test built to rank a country cannot be repurposed as a filter for a collections role in Coimbatore without new evidence.
Three ways the number gets misused
As a screening proxy. "Only 56% of graduates are employable" becomes an argument for tightening a college-tier or CGPA filter. But the benchmark says nothing about which 56%, and CGPA is not the variable it measured. You end up shrinking the funnel on a variable that carries no criterion evidence for your role — while the actual signal, job-relevant behaviour under realistic conditions, remains unmeasured.
As a target. Institutions coach to whatever test the index uses. This inflates the index without moving the underlying capability — the same failure the State of Working India report documents in vocational training, where Industrial Training Institutes expanded nearly 300% since the 2010s, largely through private providers, while quality fell and ITI locations showed no clear correlation with manufacturing hubs. Volume rose. The training-to-employment link did not.
As an explanation. Reading a rising employability rate alongside rising graduate unemployment as "the skills gap" skips the demand side. When 5 million graduates a year chase roughly 2.8 million jobs, some of that gap is arithmetic, not aptitude. It also skips the supply-side quality problem the same report flags: private colleges average 28 students per teacher and public colleges 47, against an AICTE norm of 15–20.
The role-anchored alternative: a screening checklist
If the national number cannot screen for you, this is what does. The EEOC's guidance on employment tests and selection procedures sets the same expectation US employers work under, and it is sound practice anywhere: a selection procedure should be job-related and consistent with business necessity, with evidence to back it.
- Start from a job analysis, not a competency library. Two or three observable behaviours that separate your strong performers from your weak ones in this role, at this site. Write them down before you look at any test.
- Match content to the behaviour. If the job is de-escalating an irate customer on a voice line, assess a recorded de-escalation, not a vocabulary section. Scenario-based items and work samples generalise poorly across roles and well within one.
- Set the cut score deliberately. Use a documented standard-setting method against the behaviour you defined — not a round number, and not the national percentile.
- Localise the norm. A pass mark calibrated on a metro engineering cohort will misclassify a Tier-2 BPO applicant pool in both directions.
- Protect the measurement. Unsupervised remote screening measures who had help. Integrity signals — the kind AssessAll reports as proctoring integrity bands rather than a pass/fail verdict — let you review flagged sessions instead of silently scoring noise.
- Track outcomes and back-validate. Record scores, then record 90-day performance and retention. After one hiring cycle you have criterion evidence for your own role, which no published index can give you.
- Audit for adverse impact. Compare selection rates across groups before the process becomes precedent, not after a complaint.
- Re-check annually. Roles drift. A cut score set for a 2024 job description quietly stops predicting.
Steps 1, 2 and 6 are where most funnels break. They are also the cheapest to fix: for teams running this the first time, a small, role-specific assessment built to the job analysis — AssessAll's AI-graded scenarios run on pay-as-you-go credits at ₹30 / US$0.50, with free credits on signup — is a lower-commitment way to generate local criterion data than a seat-licence platform bought before you know what predicts.
When the national number is the right tool
If you are planning a campus footprint, deciding which states to recruit in, briefing a university partner on curriculum, or forecasting three-year talent supply, the employability index is genuinely the better instrument — it has a large sample and a consistent decade-long series, which no single employer's internal data can match. Use it for supply-side questions. Just do not let it walk downstream into an individual hiring decision it was never validated for.
The takeaway
India's employability rate going up while graduate unemployment goes up is not a contradiction to be explained away — it is a reminder that a score is only as good as the specific decision it was validated for. Benchmark the country with the index; screen your candidates with something built for your job.