Ramesh Kumar

In the first budget presented after the Modi government came to power, in July 2014, skill development was identified as a major focus and the “Skill India” programme was announced. The argument was that the Indian labour force, particularly the youth, lack the skills needed by the industry and that is the reason why people are unable to find employment. Subsequently, a number of schemes were announced to improve skills of the job aspirants. In July 2015, government launched a flagship programme, the Pradhan Mantri Kaushal Vikas Yojana (PMKVY), with much fanfare. The objective of the scheme was to give young people industry-relevant training through Short-Term Training (STT) and to certify informally acquired skills through Recognition of Prior Learning (RPL).

Between 2015 and 2022 the Union government released ₹10,194 crore under the Pradhan Mantri Kaushal Vikas Yojana (PMKVY), the flagship scheme of the Ministry of Skill Development and Entrepreneurship. Over its first three phases, the scheme reported that 1.37 crore people had been trained and 1.10 crore certified. On paper it came close to its target of 1.32 crore. From 2016, the scheme was mainly implemented through the National Skill Development Corporation (NSDC), which was provided 75 per cent of the funds, while 25 per cent of the funds were provided to state skill missions.

Recently, the Comptroller and Auditor General did a performance audit of the scheme, tabled as Report No. 20 of 2025, which shows how weak the record behind those figures is. The overall picture is of a programme that got very good at producing certificates. Whether it produced skills or jobs is much less clear. The audit covered the ministry, NSDC and eight states, including inspections of 90 training centres.

Training without a map 

The first key finding of the report is that the scheme had no strategy. The Cabinet was told in 2015 that PMKVY would be a continuing scheme with targets based on skill-gap studies. Yet the ministry prepared no long-term plan, and the National Skill Development Plan promised in 2020 was never produced. Only two of the eight audited states had state-level plans.

As a result, training effort drifted away from where the need lay. The National Policy for Skill Development and Entrepreneurship (2015) estimated that more than 60 per cent of fresh skill demand was in five sectors: construction, logistics, beauty and wellness, furniture, and tourism. These sectors received only 22.7 per cent of PMKVY training. Construction alone was 28 per cent of projected demand but got about three per cent of certifications. Apparel, electronics and retail, meanwhile, absorbed more than 40 per cent of training against less than ten per cent of projected need. Within sectors, effort piled into single job roles. Nine in ten Green Jobs certificates were for safai karmacharis, and two-thirds of Management certificates were for unarmed security guards. In July 2022 the ministry itself told NSDC that choosing job roles without assessing market demand was the main reason for poor record of placement.

Who got trained 

The scheme was supposed to reach unemployed youth and dropouts who met the minimum age and qualification set for each job role. In practice the enrolment system checked neither. CAG found 6.54 lakh certified candidates who were below the minimum age for their role. They included 1,142 persons under the age of 18 years certified for driver and chauffeur roles, which legally require a licence. More than eight lakh candidates lacked the minimum education for their job roles. For roles requiring prior technical qualifications, 85 per cent of those certified did not have them. Nor did the registration system capture whether a candidate was unemployed or a dropout, so there was no way to check that the scheme reached its target group. When auditors looked at data from the current phase, PMKVY 4.0, the problems were still there: another 1.18 lakh under-age candidates had been certified by October 2024.

Data that cannot be trusted 

The scheme was managed entirely through the Skill India Portal: enrolling candidates, assigning trainers and assessors, recording results, issuing certificates and making payments. The portal was therefore the scheme’s main means of enforcing its own rules, and the state of its data shows how well those rules were enforced. The audit found that checks originally built into the portal had been quietly relaxed, and nothing was put in their place.

  • Bank accounts: For 94.5 per cent of the 95.9 lakh candidates in the second and third phases, the bank account field was zero, “Null”, “N/A” or blank. Where numbers were recorded, some were shared: one account was used for more than 2,000 candidates, and entries like “1111111111” were common.
  • Email addresses: Fifty-three email addresses were each used for more than 5,000 candidates.
  • Trainer IDs: The trainer ID was missing or “0” in batches covering 61 lakh certified candidates.
  • No retention policy: There was no data retention policy at all.

CAG’s own online survey of beneficiaries failed: more than a third of the emails bounced. Some 34 lakh certified candidates never received the ₹500 reward promised to them. Uttar Pradesh paid no reward to any of its 74,000 state-component candidates. Throughout, the ministry’s main argument was that payment by direct benefit transfer proved the training was genuine. CAG rejects this logic. “Audit is of the view,” the report says, “that payment to Aadhaar linked accounts does not automatically ensure that payouts were made to bona-fide candidates though Ministry solely relied on this method as evidence of conduct of trainings” [Para 2.4.2.2]. In the Best-in-Class Employer cases, CAG repeats the point: “the payment to candidates does not construe as evidence of conduct of trainings or existence of the agency as BICE” [Para 3.8.1].

The turn to RPL and the “Best-in-Class” employers 

In June 2018 the scheme’s Steering Committee cut short-term training targets and raised RPL targets from 40 lakh to nearly 68 lakh. The minutes give money as the main reason. Meeting the original training targets would have needed another ₹6,600 crore, and certifying existing workers was much cheaper. By the end of the third phase, RPL accounted for roughly half of all certifications.

The most serious findings concern one RPL channel, “RPL with Best-in-Class Employers”. It was meant to let large, reputable employers certify their own workforce without formal training. The eligibility rules were relaxed, and the results were striking:

  • Neelima Moving Pictures was an unregistered proprietorship whose LinkedIn page listed two to ten employees. It certified 33,493 “employees” as actors, animators and editors across eight states. The same photographs appeared as evidence for batches in Bihar and Uttar Pradesh, and the firm no longer existed when auditors looked for it.
  • Radiate Designs had been struck off by the Registrar of Companies and had filed no returns since 2014-15. It certified 15,218 people in media job roles. Its address turned out to be a shop with no signboard, and its actual business was building exhibition stalls.
  • Jaipur Cultural Society certified 56,203 people. The newspaper clippings it submitted had been edited to change venues and dates, one of which was “31 February 2020”.

Together, four such employers certified 1.24 lakh people and earned at least ₹9.96 crore in fees. The checks that should have caught this failed too. NSDC inspection reports reused the same photographs for different batches. Asked to explain, the implementing agencies offered, according to the report, that “persons shown in the photographs were brothers from the same family (3 batches)”, or that “photographs might have been mixed up during uploading process (18 batches)” [Para 3.8.3.1]. One inspector was recorded as physically visiting eight states and completing nineteen inspections on a single day, and 84 per cent of inspection reports had no geo-tag. The component has since been dropped from PMKVY 4.0.

What happened to the jobs 

For short-term training, placement is the outcome that matters. Of 56.14 lakh candidates certified under STT and Special Projects, 23.18 lakh, or 41 per cent, were reported as placed. That figure was 16.7 per cent in the first phase and 13.5 per cent in the third. In the state component of the audited states, placement fell as low as 0.7 per cent in Rajasthan and 6.1 per cent in Bihar. Even reported placements could not always be trusted. In Kerala, two training providers claimed 49 placements with three companies, which between them confirmed five; the appointment letters and salary slips had been forged. Instead of tightening verification, PMKVY 4.0 has removed placement as an outcome of the scheme altogether.

Money and oversight 

The financial findings follow the same pattern. Releases split 84:16 between centre and states rather than the intended 75:25, and a fifth of the funds sent to states lay unspent. In 2014-15, ₹435 crore was drawn from the Consolidated Fund and parked with the National Skill Development Fund so that the grant would not lapse, in breach of the Receipt and Payment Rules. NSDC kept ₹12.16 crore of interest, recovered only after the audit raised it. It charged administrative expenses of more than twice the permitted two per cent. It never passed on ₹32 crore earmarked for district skill committees, the bodies meant to carry out local mobilisation and monitoring. Biometric attendance was used in only 13 per cent of eligible batches, and fewer than one per cent of trainees gave feedback.

Tip of the Iceberg? 

While the Ministry claimed that a number of reforms were introduced in PMKVY 4.0, the CAG notes that the IT validations promised in May 2023 were still not working in October 2024. It also notes that the new phase has replaced cash rewards with kits and stopped tracking placement as an outcome.

A programme judged by the number of certificates issued will find ways to issue certificates. But what the CAG documents goes beyond perverse incentives. Fake employers, forged placement letters, recycled photographs and doctored newspaper clippings are not design flaws. They are the tools of fraud, and they thrived because the bodies meant to check them did not. When auditors asked for proof that a firm certifying 33,000 “employees” existed, the ministry produced a LinkedIn page. Its fallback was always the same: the money had reached Aadhaar-linked bank accounts.

PMKVY is not an isolated case. The same pattern runs through the Modi government’s flagship welfare programmes. Schemes are launched with fanfare and huge targets. Delivery is handed to private intermediaries paid per head, and success is measured by numbers on a dashboard. Rules that might slow the numbers down, such as verification, independent monitoring and records that can be checked, are relaxed or never enforced. CAG’s 2023 audit of Ayushman Bharat–PMJAY told a similar story: about 7.5 lakh beneficiaries registered against a single mobile number, 9999999999, and claims paid for treating patients already recorded as dead. A system built this way invites rent-seeking, and the evidence shows that the invitation is being taken up.

What the CAG found is very likely only the tip of the iceberg. It examined eight states, 90 training centres and a handful of “Best-in-Class” employers, four of which alone accounted for 1.24 lakh dubious certifications. Its reach was limited by the very records that had been allowed to decay, because an audit cannot follow a trail that was never kept. It did not examine how the current phase is being run. Meanwhile, the government’s response has been to stop measuring what embarrassed it. Placement is no longer an outcome, and cash rewards, which at least left a payment trail, have been replaced by kits. Unless skilling and welfare schemes are held to account for verified outcomes (skills, jobs, earnings) and opened to independent scrutiny rather than judged by the government’s own dashboards, the next audit is likely to tell the same story, and probably a worse one.