Skip to content
Avanti Docs

Pressure-testing the "stuck in between" hypothesis

Internal working note (no access: public frontmatter → withheld from docs.avantifellows.org). 13 September 2026.

The hypothesis, as stated

Section titled “The hypothesis, as stated”

The government has under-invested in degree-level technical education (engineering colleges and graduate-level diplomas), where job rates are high, and left it to the private sector, where fees are high. The private sector cannot fill its seats because too few students can pay; the government sector cannot deliver the quality or scale industry needs. Most government funding goes instead to non-graduate diplomas, polytechnics and ITIs, which are weak sources of employability; the notion that ITIs lead to jobs is false, and over-investment there is a mistake.

Five separable claims: (A) government under-invests in degree-level technical education; (B) private degree seats are high-fee and go unfilled for want of paying students; (C) government degree capacity is small or low-quality; (D) polytechnics and ITIs are weak on employability; (E) public money is concentrated there. This note tests each against the warehouse first (section 1) and the published literature second (section 2). Section 3 is the verdict.

1. What the warehouse says

Section titled “1. What the warehouse says”

All PLFS figures are calendar 2025, first visit, weighted, unless stated. Tables and SQL are in this folder.

1.1 Who reaches each pathway (bears on B and the demand side of A)

Section titled “1.1 Who reaches each pathway (bears on B and the demand side of A)”

Share of each pathway’s holders aged 25–29 who live in the top fifth of households by consumption (data/pathways.json, worth; query in REVERIFICATION.md §6 with the full bucket list):

Pathway From top 20% From bottom 50%
Engineering PG 82% 4%
Engineering degree 74% 7%
Medicine degree 70% 9%
Grad-level diploma, engineering 66% 15%
Other technical degree 55% 15%
Grad-level diploma, other 50% 21%
Polytechnic diploma 42% 22%
Sub-degree diploma, medicine 38% 24%
General graduate 31% 37%
Class 12, stopped 19% 47%

Engineering degrees are a top-quintile phenomenon; polytechnics and nursing-type diplomas are the most equitably reached technical qualifications. Consumption is measured today, not at entry, so these are upper bounds on selection into the pathway, but the ordering is stark and consistent with a price barrier on the degree side. It says nothing about whether the barrier is fees, opportunity cost or preparation.

1.2 Where the empty seats are (bears on B and C)

Section titled “1.2 Where the empty seats are (bears on B and C)”

AICTE 2021-22 approved intake vs enrolment, by institution type:

Level Government + aided Private self-financing
Engineering UG 1.11 L of 1.43 L filled (32% vacant) 6.5 L of 10.0 L filled (35% vacant)
Engineering diploma 2.3 L of 3.4 L filled (32% vacant) 2.6 L of 6.3 L filled (59% vacant)

The private-versus-government gap in vacancy is large for diplomas and absent for degrees. If fees alone emptied private seats, government engineering colleges would be full; a third of them are not. State dispersion is wider than the ownership gap: Tamil Nadu diploma seats 65% empty against Kerala 23%. Vacancy is measured, not explained (see analysis/aicte-aishe-capacity/README.md).

1.3 What government engineering costs (bears on B)

Section titled “1.3 What government engineering costs (bears on B)”

Entry-year fees for an open-category B.Tech seat, per year, from each institute’s own fee page (collegefees_fact_costs, JoSAA set, 2025-26; entry year includes one-time charges, so later years run 20–40% lower):

Group Colleges Median annual fee Range Median hostel + mess
IITs 23 ₹2,27,000 ₹2.0 L – ₹4.1 L ₹52,000
IIITs (PPP model) 27 ₹2,51,000 ₹0.96 L – ₹4.6 L ₹76,000
NITs 30 ₹1,46,000 ₹1.25 L – ₹1.98 L ₹58,000
Other GFTIs 31 ₹1,10,000 ₹9,000 – ₹3.4 L ₹32,000

The “government” engineering system at the top is not cheap: an IIT or NIT year costs ₹2–3 lakh with hostel before waivers (SC/ST tuition is waived; income-linked remission exists). State government engineering colleges are far cheaper but are not in the warehouse at scale (the KCET sample is 25 of 148 colleges). Private fee data is not loaded; section 2 fills it from published fee orders.

1.4 ITI-type training in PLFS (bears on D)

Section titled “1.4 ITI-type training in PLFS (bears on D)”

The narrowest ITI proxy PLFS allows: no technical education, school level Class 10 or 12, formal full-time vocational training of three months or more (voc = '1', voc_typ = '3', voc_dur 3–6; available only from CY2024). Compared with people at the same school level and no training:

School level, age Group In any work Regular salaried Median salary
Class 12, 25–29 ITI-like 68% 34% ₹15,000
Class 12, 25–29 no training 39% 11% ₹14,000
Class 12, 35–40 ITI-like 81% 46% ₹23,000
Class 12, 35–40 no training 54% 16% ₹15,000
Class 10, 35–40 ITI-like 78% 37% ₹22,000
Class 10, 35–40 no training 54% 13% ₹14,000

Sample sizes for the ITI-like cells are 171–592 people. The “no training” group’s low employment is partly women outside the labour force, so the employment gap overstates the training effect; the regular-salaried and pay gaps are the informative ones. On PLFS, formal full-time training after school triples the odds of a regular salaried job and, by 35–40, carries a ₹7,000–8,000 monthly pay premium over the same school level without it. The earlier chat analysis found “no wage premium” using a broader definition (any formal training, including short courses); the narrow definition finds one. Neither is a causal estimate — people who complete a full-time course differ from those who do not.

1.5 Polytechnic outcomes in PLFS (bears on D)

Section titled “1.5 Polytechnic outcomes in PLFS (bears on D)”

At 35–40 polytechnic holders have the highest employment rate of any pathway bar the medical diploma (92%), 59% in regular salaried work, median salary ₹28,000 and expected earnings of ₹23,600 a month — the same as “other technical degree” (B.Ed, law, MBA), above the general graduate (₹14,400) and general PG (₹20,200), below engineering (₹44,900). PLFS does not support “polytechnics are not a good source of employability”; it supports “polytechnics are a good source of employment at modest pay”.

1.6 Size of the government system by pathway (bears on A, C, E)

Section titled “1.6 Size of the government system by pathway (bears on A, C, E)”

From data/ownership.csv (AICTE seats by institution type; AISHE graduating classes apportioned by institution ownership where no seat register exists):

Pathway Government share Measure
Engineering degree 14% (1.8 L of 12.9 L seats) measured
Engineering PG 26% measured
Polytechnic diploma 35% (3.4 L of 9.8 L seats) measured
Medicine degree (health cluster) 22% estimate
Other technical degree 14% estimate
Graduate-level diplomas ~33% weak estimate
General graduate / PG ~40% (probably understated) estimate
Classes 9–12 enrolment 60–65% measured (UDISE+)

Government is a minority provider on every tertiary pathway and a small one in engineering. Whether that reflects under-investment or a deliberate 1990s–2000s policy to let private capacity expand is a question for the literature, not the warehouse.

2. What the published evidence says

Section titled “2. What the published evidence says”

Provenance caveat. This session’s network policy refused every page fetch (a 403 on every domain, official or not), so no document below was read in full. Each citation rests on the search-engine excerpt of the named document at the URL given and is marked [S]; figures from the warehouse are marked [W]. Treat every [S] figure as a lead to be confirmed by opening the URL before it is quoted outside this note.

2.1 Where public money goes (claims A and E)

Section titled “2.1 Where public money goes (claims A and E)”

Read together: the Centre’s largest higher-education lines are degree-level engineering, but at roughly 55 elite institutes with about 37,000 UG seats a year. The tier that holds the other 1.4 lakh public engineering seats and all 3.4 lakh public polytechnic seats gets ₹840 cr a year. PM-SETU alone, at ₹12,000 cr a year for five years, is of the same order as the annual IIT-plus-NIT budget.

2.2 The private engineering sector: fees and empty seats (claim B)

Section titled “2.2 The private engineering sector: fees and empty seats (claim B)”
  • Engineering UG intake is 89% private (11.1 L of 12.5 L seats); vacancy ran 37.6% (2012-13) to 49.5% (2016-17) to 34.1% (2021-22); institutions fell from 3,364 to 2,897. Government engineering colleges’ own vacancy rose from 9.2% to 32.2% over the same decade (AICTE dashboard panel, aicte_fact_intake) [W].
  • AICTE imposed a moratorium on new engineering colleges for 2020-24, its chairman saying no new colleges were needed and hundreds of poor-quality ones were closing each year [S]; 58 colleges were placed in progressive closure for 2025-26 [S]. One snippet reports approved intake back to 14.9 L and vacancy down to 16.4% in 2024-25 — unverified, and if true a sharp cyclical recovery.
  • Fees (section 4 of the page, snippet-level): a state government engineering college costs ₹9,000–90,000 a year; a fee-regulated private college ₹45,000–1.8 L; a private or deemed university ₹2.5–4.5 L; an IIT or NIT ₹1.25–2 L tuition plus ₹50,000–75,000 hostel before waivers. The private-to-government gap at the regulated tier is 2–10×; to deemed universities 5–40×.
  • Affordability: bank education-loan disbursals ₹40,253 cr in FY25, while the sub-₹10 lakh loan book shrank from ₹60,000 cr (2016) to ₹53,000 cr (2019) as big-ticket loans grew (Business Standard, 2020, https://www.business-standard.com/article/current-affairs/decline-in-study-loans-continues-but-big-ticket-loans-on-rise-rbi-data-120012400010_1.html) [S]. No fee-to-income evidence was found.

2.3 Can the government deliver quality and scale (claim C)

Section titled “2.3 Can the government deliver quality and scale (claim C)”
  • A CAG audit found five of eight new IITs placed under 75% of students over 2014-19 [S]; the Parliamentary Standing Committee (March 2025) flagged an “unusual decline” in IIT/NIT placements and over 4,500 vacant IIT faculty posts as “the biggest handicap” [S].
  • Within NIRF’s ranked set, placement does not fall with rank (76% at ranks 1–10, 68% at 151–200), which the repo’s earlier work reads as self-reporting optimism rather than uniform quality (analysis/THREE-REGISTERS.md) [W]. Unranked colleges, four-fifths of graduates, are unobserved.
  • The binding constraint the sources name for the elite is faculty, not funds. Nothing found speaks to the quality of state government engineering colleges.

2.4 ITIs (claim D)

Section titled “2.4 ITIs (claim D)”
  • NITI Aayog, Transforming ITIs (2023): capacity about 25 lakh, roughly 42% filled; a placement figure of 405 of 4,14,247 (0.09%) is relayed in press coverage with the base unclear (https://niti.gov.in/sites/default/files/2023-02/ITI_Report_02022023_0.pdf) [S].
  • DGT tracer study (2018): 63.5% of ITI graduates employed (wage or self), self-reported (https://dgt.gov.in/sites/default/files/pdf/Tracer%20Study%20of%20ITI%20Graduates-Final%20report.pdf) [S]; DGT/STRIVE state tracers (2024): Rajasthan about 50% employed, Karnataka 20% unemployed and women 38% [S].
  • World Bank STRIVE completion report (2025), ₹2,200 cr over 500 ITIs: placement “well below the 50 percent assumed at appraisal” [S]. Commentary on PM-SETU notes that the earlier VTIP and 1,396-ITI PPP upgrades “emphasised buildings over pedagogy” (The Secretariat, https://thesecretariat.in/article/pm-setu-scheme-a-long-overdue-initiative) [S].
  • Azim Premji University, State of Working India 2026: ITIs grew from 3,674 (2005) to 14,582 (2025), 80% private, newer private ITIs scoring lower on quality with a “tenuous link to manufacturing employment” [S]. The MSDE year-end review gives 3,345 government and 11,337 private ITIs, enrolment about 14 lakh [S].
  • Returns research on PLFS: formal vocational training carries an earnings premium of about 11% (IES working paper, 2025, propensity-score matched, down from 28% in a 2016 estimate, https://www.ies.gov.in/pdfs/Vishnu-KVenugopal-march25.pdf) [S]; full-time formal training raises wages significantly while non-formal training does not (IZA DP 15002, 2022, https://docs.iza.org/dp15002.pdf) [S]. Only 4.2% of workers aged 15–59 hold any formal vocational training (PLFS 2025 annual report) [S].
  • Warehouse, narrow ITI-like proxy (section 1.4): regular-salaried share three times that of the same school level without training, and a ₹7,000–8,000 monthly pay premium by 35–40 [W]. The earlier broad definition found no premium; the difference is full-time courses of three months or more versus any formal training.
  • Employer-side “employability” indices (India Skills Report 2025: ITI 46%, polytechnic 33%, engineering 72%) are pass rates on a self-selected online test, not hiring data; they reverse PLFS’s ITI-versus-polytechnic ordering and should not be used [S].

2.5 Polytechnics (claim D)

Section titled “2.5 Polytechnics (claim D)”
  • Warehouse: at 35–40, 92% in work, 59% regular salaried, median salary ₹28,000; formal employment 30.8% at 21–34 against 16.2% for a general degree [W]. Vacancy 49% and rising, concentrated in private polytechnics (59%); roughly half of entrants never pass out; the arithmetic suggests 0.8–1.6 L a year convert to B.Tech through lateral entry, removing the strongest holders from the diploma stock [W].
  • Press on AICTE data reports declining admissions and faculty shortage, with Gujarat at 36% polytechnic seats vacant [S].

2.6 Graduate-level diplomas (claim A)

Section titled “2.6 Graduate-level diplomas (claim A)”
  • No dedicated public programme or budget line was found. Public provision is IGNOU, NIELIT and C-DAC diplomas at ₹5,000–99,000 a course; the private PGDM runs ₹7–9 L a year with no AICTE fee cap [S].
  • PLFS, ages 25–34: 7.7 L holders of the engineering graduate-level diploma, 60% in regular salaried work, unemployment 14%, median salary ₹36,000, and 66% from the richest fifth of households [W]. It behaves like a second, dearer engineering degree bought by those who can.

2.7 Pass 2, web-enabled (13 September 2026) — what was read in full

Section titled “2.7 Pass 2, web-enabled (13 September 2026) — what was read in full”

Full detail in HANDOFF_PASS2.md. The figures below were read on the issuing body’s page unless marked otherwise.

  • PM-SETU (PIB release 2243982, MSDE Lok Sabha written reply, 23 March 2026, read in full): ₹60,000 cr — Centre ₹30,000 cr, states ₹20,000 cr, industry ₹10,000 cr; ADB and World Bank co-finance half the central share; 1,000 government ITIs (200 hub, 800 spoke) plus five NSTIs; six disbursement-linked indicators, the first being employment outcomes of graduates, third-party verified by IIM Indore. Approved investment plans stood at ₹1,237.58 cr in July 2026, about 2% of outlay — committed, not spent.
  • MERITE (PIB release 2154119 / MoE mirror PDF, 8 August 2025, excerpt, three carriers agree): ₹4,200 cr for 2025-26 to 2029-30, including a ₹2,100 cr World Bank loan, for 275 government or aided institutions — 175 engineering colleges and 100 polytechnics — about 7.5 lakh students.
  • Central outlay ratio: 14 to 1 toward ITIs against degree-level engineering and polytechnics combined; about 4 to 1 per institution (₹60 cr against ₹15 cr). Both are Union schemes; state budgets, which carry most government colleges and polytechnics, were not opened in either pass.
  • ITI capacity is bimodal, not uniformly empty (DGT, minutes of the 28th SCAA meeting, 28 April 2026, excerpt of the official PDF; Maharashtra DVET report via Free Press Journal, September 2026, press): ~4.5 lakh seats were de-affiliated in January 2025 for being non-operational 2018–23, and 669 ITIs with 99,196 seats admitted nobody in 2022–24 and face de-affiliation from 2026. Meanwhile Maharashtra filled 85% of seats in 2026, 93% in government ITIs and 71% in private. The oft-quoted 42–48% national utilisation averages full institutions with dead ones.
  • ITI fees, one state pair read (The Tribune, Punjab 2025-26): government ₹3,350 a year, private ₹19,315 — a 5.8× gap, matching the excerpt-level band in the fees table.
  • Institution counts mislead on ownership. Punjab: 145 government ITIs hold 52,000 seats, 183 private hold 24,000. Maharashtra: 418 government hold 95,820, 586 private hold 49,324. Government institutions are about three times larger, so every ownership row apportioned on institution counts overstates the private share of students. This applies to the estimated rows in data/ownership.csv and to the AISHE college-count proxy for general degrees.
  • MBBS seats may be a cycle stale. An NMC notification of 14 July 2026 is reported (aggregator only) to put MBBS at 1,36,939 seats, 63,296 government and 73,643 private, private overtaking government for the first time. Unconfirmed; the warehouse holds the 2024-25 matrix.
  • Nothing on fees for engineering, polytechnics, medicine or general degrees was opened in pass 2; the fees table stays excerpt-level on those rows.

3. Verdict, claim by claim (revised after pass 2)

Section titled “3. Verdict, claim by claim (revised after pass 2)”
Claim Verdict Why
A. Government under-invests in degree-level technical education and graduate-level diplomas Strengthened at the Centre; still a blanket that needs qualifying The only central scheme aimed at state engineering colleges is MERITE, ₹4,200 cr over five years shared with 100 polytechnics, against ₹17,000 cr a year for IITs and NITs and ₹60,000 cr for ITIs. Graduate-level diplomas have no public programme found in two passes. Qualifications: state budgets, which carry most government colleges, remain unopened; both schemes are commitments, not disbursements
B. Private seats are high-fee and go unfilled for want of paying students Fees documented at excerpt level; the mechanism is still a correlation Regulated private fees run 2–10× government, deemed 5–40×; a third of private engineering seats are empty. But government engineering seats are also a third empty, no fee order was opened in either pass, and none of the alternatives — accreditation, location, labour-market signal, loan access — has been examined. The weakest link in the argument
C. Government cannot deliver quality or scale Split it: scale yes, quality contradicted Government holds 14% of engineering seats, so scale is real. On quality, the repo’s NIRF work finds a government engineering seat about 3.5× likelier to sit in a top-100 institute than a private one (26% of government capacity against 7.5% of private). The named constraint at the elite is faculty, not funds
D. Polytechnics and ITIs are weak sources of employability Wrong on polytechnics, too strong on ITIs, and wrong about where the money goes Public money is going to ITIs, not to “polytechnics and ITIs”: polytechnics get 100 slots in the ₹4,200 cr scheme and nothing in the ₹60,000 cr one, making them the least-funded of the three on central evidence. Polytechnic holders have the highest employment rate on the page and twice a general graduate’s formal-job share. For ITIs, PLFS shows real formality and pay gains for completers; the system is bimodal, with government ITIs 93% full in Maharashtra alongside 5.5 lakh dead seats nationally
E. Over-investment in ITIs is a mistake Premise documented; conclusion open 14 to 1 is real. Whether it is a mistake rests on the ITI-versus-polytechnic returns comparison, where PLFS favours the polytechnic, and on execution: PM-SETU is explicitly outcome-linked (employment DLI, third-party verified), which prior schemes were not and which the critique must acknowledge. The honest form: the money is being bet on ITIs rather than polytechnics, with conditionality attached, and the polytechnic has the better outcome profile to start from

The version the evidence now supports. The Centre funds the top of the technical ladder and, at 14 to 1 over everything degree-level and polytechnic, the bottom. The middle — state engineering colleges, polytechnics, post-degree diplomas — is left to state budgets nobody has opened and to fee-charging private providers. That middle is where employment rates are high, where government seats are a third empty for reasons no source explains, and where the fee gap is largest.

What would settle the open points. (1) State technical-education Demands for Grants (Tamil Nadu, Maharashtra, Karnataka, Uttar Pradesh) — until then the 14:1 describes central priorities only. (2) One engineering fee order per state, college-wise, to move claim B from correlation to evidence. (3) The NMC July 2026 notification. (4) ITI counts with dates attached, never averaged. (5) The base of NITI Aayog’s 0.09% placement figure, or its retirement in favour of the DGT minutes. (6) Fee-to-income from CMS-E 2025, which is already in the warehouse and was not run.