Findings, and what each one is worth
Every claim this analysis makes, graded. Nothing is asserted here that is not also traceable to a
query in QUERIES.md and an assumption set in ASSUMPTIONS.md.
Confidence is stated in three grades:
- Solid — large cells, statistically separated, survives the sensitivity tests run.
- Directional — the direction is supported, the magnitude is not precise.
- Retracted — asserted in an earlier version, withdrawn on review, recorded here so it is not reintroduced.
Findings are labelled S(olid), DI(rectional) and R(etracted). Assumptions in
ASSUMPTIONS.md use a separate A–G scheme; a bare “D2” in this document always
means Directional 2 unless it says “ASSUMPTIONS”.
S1. A technical diploma delivers far more formal employment than a general degree
Section titled “S1. A technical diploma delivers far more formal employment than a general degree”30.77% against 16.20% — 14.57 points, t = 11.7 on cluster-robust standard errors.
This is the finding the analysis exists for and the only large comparison that is unambiguous. It holds for both sexes (men 33.91% vs 22.81%; women 20.60% vs 9.46%) and survives the strict formal-work definition (ASSUMPTIONS C1 — the sensitivity leaves every ordering unchanged).
The general degree’s entire wage advantage over the diploma is ₹2,000 a month at entry and ₹3,500 at three years — against a 14.6-point lower chance of holding a formal job at all.
Caveat that does not weaken it: not causal. People who take a diploma differ from those who take a degree. And see G2 in ASSUMPTIONS — many diplomas follow class 10, not class 12, so the correct comparison base may be rung 03’s 5.95%, which makes the step larger.
S2. Engineering is far ahead of every other rung
Section titled “S2. Engineering is far ahead of every other rung”52.58% formal employment, t = 6.7 against medicine (the closest rung by SE) and t ≥ 11 against every other. Its informal wage work is 4.29%, the lowest on the ladder against 31.34% at the bottom — the degree changes the kind of work available, not just the odds of having any.
S3. The formal-employment gradient across the ladder is roughly thirty-fold
Section titled “S3. The formal-employment gradient across the ladder is roughly thirty-fold”1.66% at the bottom to 52.58% for engineering — 31.7×, or 40.6× if formality is defined strictly (social security required rather than contract-or-social-security). The headline is therefore a 32–41× range, and the stricter definition widens it.
S4. The gender gap collapses at the diploma
Section titled “S4. The gender gap collapses at the diploma”Across the four school rungs the ratio runs 4.4, 5.1, 5.8, 4.7. Above them it drops to 1.3–2.4 and stays there for every post-school qualification. That step is large, on cells in the tens of thousands, and is among the best-measured things here. At secondary, 68% of women are in domestic duties against 0.36% of men.
What is not supported: that the gap peaks at secondary specifically. Cluster SEs on the ratio are ±0.82 (middle school), ±0.54 (secondary) and ±0.31 (higher secondary), so secondary vs middle school is t = 0.69. An earlier version asserted the peak four times as a named finding. The ordering within the school rungs cannot be established; the drop after them can.
S5. Half of formally employed engineering graduates work in software
Section titled “S5. Half of formally employed engineering graduates work in software”45.8% are software and application developers, at ₹43,700 median. No other rung is close — the largest single occupation elsewhere is medicine’s doctors at 24.2%. Adding electrotechnology and ICT takes the software-adjacent share to 50.7%.
Corrected figure. Published as 50.6% before the denominator bug (V1) was fixed.
Directional
Section titled “Directional”DI1. Below middle school, wages barely grow across a working life
Section titled “DI1. Below middle school, wages barely grow across a working life”Mean wage ₹13,088 → ₹12,704 → ₹13,904 at entry, 30 and 40 — 1.06× over roughly eighteen years.
Downgraded from Solid. An earlier version cited “cells of 4,872 / 8,085 / 18,868” — those are the rung sizes. The formal-wage cells are 52 / 163 / 364. The entry cell of 52 is barely larger than the 37-person cell whose flatness was retracted as an artefact in R3, and this claim is R3’s replacement evidence. What is reasonably firm is the age-30-to-40 leg, ₹12,704 → ₹13,904 on 163 and 364 people: 1.09× over a decade. The entry point should not be leaned on.
DI2. The diploma’s value is mostly the diploma — and its technical premium is entirely engineering
Section titled “DI2. The diploma’s value is mostly the diploma — and its technical premium is entirely engineering”A non-technical diploma delivers 25.80% formal employment against the technical diploma’s 30.77% — both far above higher secondary’s 7.38%. So most of what a diploma buys is not the technical content.
But “the technical field adds a fifth on top” is wrong, and an earlier version of this analysis said it. Split rung 05 by field:
| n | Formal work | |
|---|---|---|
Engineering diploma (tedu_lvl='08') |
3,000 | 33.41% |
| Non-technical diploma (rung 06) | 2,900 | 25.80% |
Agriculture, medicine, crafts (07,09,10) |
921 | 22.51% |
Three of the four technical fields land below the non-technical rung. Rung 05’s 30.77% is an engineering number carried by the 77% of that rung holding engineering diplomas. A technical diploma in anything else does worse than a general one.
Why only directional. The engineering-diploma advantage over non-technical is 7.6 points on cells of 3,000 and 2,900 — real. The agriculture/medicine/crafts deficit rests on 921 people and is not separately tested. And the 05/06 boundary is unresolved (ASSUMPTIONS B2).
DI3. Wage growth rises with education
Section titled “DI3. Wage growth rises with education”Median entry → 3+ years: 1.17× at middle school, 1.15× at secondary, 1.14× at higher secondary, 1.33–1.34× at the diplomas, 1.39–1.40× at general and medical degrees, 1.67–1.68× at other-technical and engineering.
Why only directional. The bottom-rung end of this pattern was an artefact (R3), and the school-rung tenure cells have not been validated for age composition the way the graduate ones have (ASSUMPTIONS D2 — a different numbering; assumptions use A/B/C/D/E/F/G, findings use S/DI/R). The ordering at the top is clear; the exact multiples are not.
DI4. Class 12 is where clerical work becomes available
Section titled “DI4. Class 12 is where clerical work becomes available”Below higher secondary, formal work is factory labour, cleaning, driving and security. At higher secondary, general office clerks are the third-largest occupation and the mix flattens sharply — the top six are 34.9% of formal work against 53.9% at the bottom rung.
DI5. Unemployment rises with education because searching becomes affordable
Section titled “DI5. Unemployment rises with education because searching becomes affordable”0.50% at below-middle-school to 17.87% at other-technical degrees. At the bottom, being out of work means domestic duties or informal labour, not a job queue.
Why only directional. The column is pas='81' as a share of population, not of the labour
force. Engineering’s rate on the labour-force base is 18.1%, not 15.46%. The pattern holds
either way; the levels are not unemployment rates as conventionally reported.
DI6. Medicine is where the gender employment gap disappears
Section titled “DI6. Medicine is where the gender employment gap disappears”32.94% female against 31.75% male. The female lead is not real — t = 0.24 — but the absence of a gap is the finding, and it is unique on the ladder.
Medicine simultaneously carries the widest wage gap of any degree rung (₹25,000 vs ₹35,000), because women in that category are nurses and men are doctors. Doctors are 24.2% of the rung’s formal workers; nursing roles are 31.8%. Medians ₹25,000 vs ₹35,000, a 29% shortfall.
DI7. A general master’s does not out-employ a technical diploma. The other three do.
Section titled “DI7. A general master’s does not out-employ a technical diploma. The other three do.”Restated after the postgraduate rung was split by field. The original claim compared a single merged postgraduate bucket at 29.11% against the diploma’s 30.77% and called them tied. That bucket no longer exists, and the merged figure was a composition artefact: it averaged four rungs that run from 25.57% to 62.12% formal.
Split by field, against the technical diploma at 30.77% formal (46.18% regular salaried):
| Postgraduate rung | Formal | Regular salaried | vs the diploma |
|---|---|---|---|
| General | 25.57% | 34.26% | below it |
| Other technical | 37.26% | 47.30% | above |
| Engineering | 53.13% | 59.33% | well above |
| Medical | 62.12% | 67.50% | highest on the ladder |
So the supportable claim is narrower and sharper than the original: a general master’s — 83% of all postgraduates — does not buy an employment advantage over a technical diploma obtained years earlier. It is the field, not the level, that separates them. The three technical postgraduate rungs all beat the diploma, and none of that was visible while they were averaged together.
The original t-statistics (t = 1.18 against the diploma, t = 1.25 against medicine) were computed on the merged bucket and are NOT carried over: they describe a population that no longer exists here. The per-field comparison above is read off the extract and is not accompanied by a significance test, so it is stated as an ordering rather than a tested gap.
Where the master’s does pay is the wage level: ₹25,000 at entry against the diploma’s ₹16,000.
Retracted
Section titled “Retracted”R1. “A postgraduate degree does not top the list / is fourth by formal employment”
Section titled “R1. “A postgraduate degree does not top the list / is fourth by formal employment””Withdrawn. Asserted five times across the README and rung 11. The diploma–PG gap is t = 1.18 and PG–medicine t = 1.25; those three rungs cannot be ordered. Replaced by DI7.
R2. “Medicine is the only rung where women’s employment exceeds men’s”
Section titled “R2. “Medicine is the only rung where women’s employment exceeds men’s””Withdrawn. 1.19 points on t = 0.24, from 485 women and 400 men, presented as a named finding with a causal story attached. Replaced by DI6.
R3. “Below middle school, time in work buys nothing (1.00×)”
Section titled “R3. “Below middle school, time in work buys nothing (1.00×)””Withdrawn in its tenure form. The rung’s three tenure cells have mean ages 27.4 / 27.1 / 29.6 — they are not entrants versus veterans. Half the 37-person “entry” cell is aged 27–34, job-changers earning a ₹16,000 median. Restricted to ages 21–26 the same rung reads 1.23×, in line with middle school.
Survives in its lifecycle form as S5.
R4. “Nearly half of the other-technical rung is teaching — the B.Ed rung”
Section titled “R4. “Nearly half of the other-technical rung is teaching — the B.Ed rung””Withdrawn. Rested on the occupation denominator bug (V1). Teaching is 23.9%, not 45.9% — a plurality, not a majority. Nearly half that rung’s formal workers are in occupations too small to name individually.
R5. “A third of medical-degree holders are doctors”
Section titled “R5. “A third of medical-degree holders are doctors””Withdrawn. Same bug. It is 24.2%.
R6. “The technical diploma has the steepest lifetime wage growth on the ladder”
Section titled “R6. “The technical diploma has the steepest lifetime wage growth on the ladder””Withdrawn, and so is the same claim made for the non-technical diploma in its own note — the two notes contradicted each other. The lifecycle “Wage ×” column is not comparable across rungs because each rung’s entry window sits at its own completion age, up to four years apart. Re-basing medicine to a common 21–23 entry moves it from 1.71× to 2.83×, from near the bottom to the top.
R7. “Casual wage labour and unpaid family work sit outside these columns”
Section titled “R7. “Casual wage labour and unpaid family work sit outside these columns””Withdrawn. Casual labour is inside the informal-wage column. At rung 01 it is 24.2 of that column’s 31.3 points, so “informal wage work” means daily-wage labour there and something different at engineering. What actually sits outside: unpaid family workers, self-employed with no positive earnings, rentiers, and those unable to work.
Not tested, and therefore not claimed
Section titled “Not tested, and therefore not claimed”- Whether technical diplomas follow class 10 or class 12 — untestable in this release.
- Whether the 05/06 diploma split is real or a coding artefact.
- Any subdivision of the 49-million general-degree rung; PLFS records no field of study for non-technical degrees.
- Anything causal.