The Part Where Somebody Has To Be Wrong
Part 6 established that the injury is real and still measurable. That is a necessary condition for a remedy and nowhere near a sufficient one. A disease being real does not make a particular medicine effective.
So this part asks the separate question: does reservation work?
It is the hardest question in the series, because both sides have spent decades asserting answers without evidence and because the evidence, when you actually assemble it, does not tidily support either.
That is the finding, and I have put it at the front rather than the end because I do not want anyone reading three chapters looking for ammunition and stopping.
Three things about method, because in a part made of statistics the method is the argument.
First, I have preferred official data over advocacy data. The most important numbers here come from the Department of Personnel and Training, the Education Ministry's replies to Parliament, and a peer-reviewed paper in a leading economics journal. Where I use a campaigning source, I say so.
Second, every chart gets its strongest criticism. That was the promise in Part 1 and this is the part where it costs something.
Third, I have gone looking for the numbers that embarrass my own reading. Chapter 5 contains the strongest evidence in this series that reservation is not delivering at senior levels. Chapter 8 lists the statistics the pro-reservation side gets wrong, and there are as many as on the other side.
One last thing. Chapter 2 contains a single table that I think is the most informative object in this entire fourteen-part series. It is three rows of government data. If you read nothing else in Part 11, read that.
— Lovepreet Singh
July 2026
1How To Judge A Statistic
Five tests. Most numbers you will meet in this argument fail at least one, including some in this book.
Before any figures, a short method. These five questions will let you assess almost any statistic anyone offers you about reservation.
1. Is it a stock or a flow?
"SC representation in central government is 16.84%" is a stock — everyone currently employed, including people hired forty years ago. "SC share of this year's recruitment" is a flow. A policy changed in 1990 shows up in the flow immediately and in the stock only after decades of retirements.
Both sides quote whichever is more flattering.
2. Is it aggregated across levels?
This is the big one, and Chapter 2 is built on it. A single national figure for "representation in government" can conceal a situation where a group is over-represented at the bottom and badly under-represented at the top. Aggregation hides gradients, and the gradient is usually the story.
3. Correlation or causation?
"SC literacy rose after reservation" tells you almost nothing on its own, because everybody's literacy rose. The question is always: compared with what would have happened otherwise? Very little writing on this subject even attempts an answer. Chapter 6 covers the work that does.
4. Who collected it and what did they gain?
The test from Part 6. A government reporting on its own compliance, an activist organisation documenting a failure, and a party in an election all have reasons. The strongest data is collected for some purpose other than winning this argument.
5. What is the denominator?
"Ten thousand reserved posts are vacant" means nothing until you know out of how many. "OBCs hold 19% of senior posts" is only meaningful against the quota of 27% and against their population share.
Indian caste statistics have a structural problem that no amount of care fixes: there has been no caste census since 1931, so almost every denominator involving OBCs is an estimate. Census 2027 will fix this. Until then, every OBC percentage in this part — including the ones I use — rests on an estimated population base.
SC and ST figures do not have this problem, because those categories have been counted in every census. Where the evidence is strongest in this part, it is usually about SCs and STs for exactly that reason.
2The Gradient
The Government of India publishes its own employment composition. Read down the levels and the argument changes shape.
The Department of Personnel and Training reports the caste composition of central government employment. As of 1 January 2024, across roughly 32.52 lakh employees in 80 ministries and departments, the overall figures were:
Taken alone, those numbers look like a policy that has done its job. SC and ST are above their quotas; OBC is close to it.
Now apply test two from the last chapter, and look at where in the hierarchy those people are.
| Level | SC | ST | OBC |
|---|---|---|---|
| Group A — the senior posts | 14.2% | 6.54% | 19.14% |
| Group C — excluding sanitation staff | 16.75% | 8.94% | 27.29% |
| Safai karamcharis — sanitation posts | Over 66% from SC, ST and OBC combined | ||
| The quota | 15% | 7.5% | 27% |
To the abolition side, it is devastating on their central claim. The argument is that reserved candidates have flooded the system and displaced better ones. Thirty-six years after Mandal and seventy-six after the Constitution, OBCs hold 19.14% of senior posts against a 27% quota — a shortfall of nearly eight percentage points. SCs are below 15% and STs below 7.5%. Whatever is happening at the top of the Indian state, it is not over-representation of reserved categories.
To the retention side, it is equally hard. "Reservation has delivered representation" is not supported at the level where decisions are made. The quota is met in the middle and exceeded in sanitation. It is not met where the power is. A policy justified by Article 16(4)'s test — adequate representation in the services — has not achieved it at the level that matters most, after three generations.
And the sanitation row is the one nobody should be able to read comfortably. Part 6 established that roughly 97% of identified manual scavengers are Scheduled Caste. This table shows the same pattern inside the government's own payroll. The state that runs the reservation policy also concentrates its most degrading work in the same communities.
The report gives SC, ST and OBC representation. It does not report the representation of Economically Weaker Sections, despite a 10% quota running since 2019.
Which means that seven years in, the government does not publicly report whether the caste-blind quota it created is reaching anybody. Anyone proposing to replace caste-based reservation with an economic one should want that number to exist.
3What Merit Measures
The strongest argument against reservation, taken seriously — and the question it has to answer.
The merit argument deserves better than it usually gets. Here it is at full strength.
A competitive examination is the fairest allocation device humans have invented. It is anonymous, it is uniform, it cannot be bribed at the point of marking, and it does not care who your father was. For a country with India's history of patronage and connection, that is not a small thing — it is one of the few genuinely impartial institutions the republic possesses.
Every departure from ranking by score reintroduces a criterion other than performance. And in fields where competence has third-party consequences — medicine, engineering, aviation, the judiciary — the cost of that departure is not borne by the person who benefits from it.
Moreover, an examination is a current measure. It tests what a candidate can do now. Whatever happened to their great-grandparents, the test is administered today, to today's candidates, and the person who scores higher has demonstrated more of the thing being tested.
And the question it has to answer
What does the score measure?
Not rhetorically — literally. An entrance examination score is a measurement of performance on a specific test, at a specific moment, after a specific quantity of preparation.
India has an entrance-coaching industry of enormous size, concentrated in a small number of centres, selling multi-year residential programmes at prices well beyond the reach of most Indian households.
This is not a contested claim. It is a large, visible, advertised market whose entire proposition is that paying money improves the score.
What the coaching point does establish: that a score is not a pure measurement of aptitude, because a substantial and purchasable input raises it. Two candidates of identical ability with different family resources will not score identically. That is arithmetic, not ideology.
What it does not establish: that the score is meaningless, or that caste is the right correction for it. Coaching money tracks income more directly than caste. If the distortion is purchased preparation, the logically matching correction is an income-based one — which is the abolition side's proposal, and Part 10 showed how badly the existing income-based scheme is calibrated.
And the point both sides skip: Part 1 showed that General and OBC candidates closed two marks apart for all-India government MBBS seats in 2025. Whatever the merit argument is about, at the top of the distribution it is not about a large score gap. The gap that is real — the SC gap of roughly 85 marks in that data — is the one that requires the discussion, and it is not the one the argument is usually conducted about.
If the objection is to selection by anything other than demonstrated performance, then it applies with equal force to management quota seats, to donation-funded admissions, to legacy advantages, and to the effect of purchased coaching itself.
The merit argument is coherent and serious. It is only inconsistent when it is applied to one of those and not the others — and in Indian public argument, it usually is.
4The Dropout Claim
More than 13,500 students. Two readings that cannot both be right — and data that does not distinguish between them.
Data reported in 2023 showed that more than 13,500 students from Scheduled Castes, Scheduled Tribes and Other Backward Classes had dropped out of courses at central universities, the Indian Institutes of Technology and the Indian Institutes of Management.
This number is used by both sides, for opposite purposes, and each use is a complete argument.
Admission gained through a lower entry threshold does not confer the preparation to survive the course. Students placed in institutions they were not academically ready for struggle, fall behind, and leave.
On this reading the dropout figure is evidence that the policy harms the people it is meant to help — offering a place that becomes a failure rather than a qualification.
Getting in is not the same as being let in. Students arriving from Hindi-medium or regional-language schooling into English-language technical instruction, without family networks, without money, and in some cases into open hostility, leave for reasons that have nothing to do with capability.
On this reading the figure measures the institution, not the student.
The published figure is a count of departures. It does not record why anybody left, and both explanations produce the same number.
What would settle it: comparative dropout rates for unreserved students at the same institutions; exit interviews; performance trajectories over the course; the language of prior schooling. Some of this exists in fragments. None of it is systematically published.
What can be said honestly: the figure is real, it is large, and it is a problem on either reading. If reading one is correct, admission without support is cruel. If reading two is correct, admission without support is cruel. The two sides disagree about the cause and agree, without noticing, about the remedy.
Nobody in the 2026 argument is campaigning for bridge courses, language support or mentoring. Both sides prefer the dispute.
5The Empty Chairs
Thousands of reserved teaching posts sit unfilled. This is the strongest evidence in this series that the policy is not delivering — and it damages the other side just as badly.
If reservation were the vast redistribution its critics describe, reserved posts would be full. They are not.
As of November 2024, 5,182 teaching posts were vacant across India's central universities, of which more than 2,750 were reserved posts — approximately 740 SC, 464 ST and 1,546 OBC positions.
As of 31 January 2025, roughly 28.56% of sanctioned teaching positions across the IITs, NITs, IIMs, IISERs, central universities and other central higher-education institutions were vacant — some 18,940 posts.
A parliamentary panel reported that 56% of professor positions at the IITs, IIMs, NITs and IISERs were lying vacant.
And the mechanism nobody explains
4,297 reserved teaching positions — professor, associate professor and assistant professor — were left vacant in central educational institutions after the recruitment process had been conducted, through the declaration of a "none found suitable" outcome.
The posts were advertised. Candidates applied. Selection committees sat. And the committees concluded that no candidate from the reserved category met the standard.
There are simply not enough qualified candidates from reserved categories at the doctoral and post-doctoral level to fill senior academic positions. The vacancy is a symptom of a shortage upstream, not of obstruction at the point of hiring.
On this reading, "none found suitable" is an honest finding, and forcing appointments would lower standards.
"None found suitable" is a discretionary judgment made by selection committees whose own composition reflects the imbalance in question. The same institutions do not report comparable difficulty filling unreserved posts.
On this reading, it is the administrative device by which a mandatory quota becomes optional — a legal way of not appointing.
It costs the retention side its comfortable position. Thousands of reserved posts, advertised and unfilled, in the very institutions where representation matters most. Whatever reservation is achieving in Indian higher education, it is not filling the faculty.
But it costs the abolition side something larger. The core grievance is that reserved candidates are taking places that would otherwise go to others. Here are thousands of reserved places going to nobody at all.
If reserved posts sit empty, then those posts were not taken from anyone. The "they got my seat" story requires the seat to have been occupied. In the case of central university faculty, on the government's own numbers, in thousands of cases it was not.
And the honest observation about "none found suitable": I cannot tell you which reading is correct, and neither can anyone else from published data. What is checkable is that the same discretion is not producing the same result for unreserved posts at the same institutions. That does not prove bad faith. It does mean the burden of explanation sits with the institutions, and they have not discharged it.
6Does It Actually Work?
Everything so far has been description. This chapter is the one piece of proper causal evidence, and it gives a clear answer.
Almost all argument about whether reservation works is correlation dressed as causation. Outcomes for Scheduled Castes have improved since 1950 — but outcomes for everyone improved since 1950, so the comparison is worthless on its own.
What is needed is a natural experiment: two groups, similar in relevant respects, one of which received the policy and one of which did not.
It exists.
A study of intergenerational mobility in India using a new measure — the expected education rank of a child born to parents in the bottom half of the education distribution — applied across time, regions and social groups.
The headline result: intergenerational mobility in India has been broadly constant and low since before liberalisation. India is not, in aggregate, a country where children routinely rise past their parents.
But the groups diverge sharply. Mobility has been rising for Scheduled Castes among sons. It has been declining for Muslims. Daughters' mobility is lower than sons' throughout, with less variation between groups.
And the natural experiment
Because the Scheduled Caste list has been amended over time, some communities were added later than others. That creates a comparison: groups that were similar beforehand, one of which subsequently received the policy.
The study finds that groups newly assigned Scheduled Caste status experienced, on average, a 7 to 8 rank point increase in upward mobility over the next two decades.
The authors' conclusion is that affirmative action for Scheduled Castes has substantially improved their mobility.
1. The efficacy question is largely answered. "Reservation does not work" is not a supportable position. The mechanism has a measured causal effect on the outcome it was designed to change, in peer-reviewed work, using a credible identification strategy. Anyone arguing for abolition has to argue that a policy which demonstrably works should nonetheless be ended for other reasons — which is a legitimate argument, and a different one.
2. And the aggregate finding is grim for everybody. Overall mobility in India has been constant and low for decades. Reservation lifted one group's trajectory inside a system that is not generally mobile. That is a real achievement and a small one relative to the problem.
3. Now look at the Muslim result, and read it against Part 4. Mobility has been declining for Indian Muslims over the same period in which it rose for Scheduled Castes. And as Part 4 established, Dalits who convert to Islam or Christianity lose Scheduled Caste status under the 1950 Presidential Order, while those who convert to Sikhism or Buddhism keep it.
The group that receives the policy is rising. A large group excluded from it by a paragraph written in 1950 is falling. That is not proof of causation — many things affect Muslim outcomes in India — but the two findings sit in the same paper and the coincidence is not comfortable for anyone.
7Land, Wealth, Capital
Reservation operates on jobs and college seats. Here is what it has never touched.
Chapter 6 showed the policy works where it operates. This chapter is about where it does not operate — and the scale of what is outside it.
Land
Calculations based on the All-India Debt and Investment Survey give Scheduled Castes roughly 18% of households and 8.5% of land, against upper-caste Hindus at approximately 22% of households and 28% of land.
At state level the pattern repeats. Telangana survey data gives the average General category family roughly 0.42 acres of irrigated land against 0.20 acres for the average Scheduled Caste family.
Business and capital
A survey by the Ministry of Micro, Small and Medium Enterprises found that 12.45% of unregistered non-agricultural MSMEs were owned by Scheduled Castes — and that not one was a medium-sized enterprise.
Ownership exists at the smallest scale and disappears as the scale rises. The gradient from Chapter 2 reappears in the private economy.
Wealth
National Family Health Survey data places roughly 12.3% of Scheduled Castes in the highest wealth quintile.
Analysis of Indian billionaire wealth for 2022 attributed approximately 2.6% to Scheduled Castes and about 88.4% to upper castes.
The All-India Debt and Investment Survey for 2018–19 indicates that upper castes hold close to 55% of national wealth.
Reservation has produced measurable movement in employment and education — Chapter 6 proves it — and essentially none in land, capital or wealth, because it was never designed to and has no mechanism that could.
For the abolition side: this is a serious problem for the claim that reserved categories have been advantaged into dominance. A group holding 2.6% of billionaire wealth and 8.5% of land is not dominant by any measure that involves assets.
For the retention side: it is a serious problem too, and a bigger one. Seventy-six years of a policy aimed at public employment has not moved the ownership of the country. If the objective is equality rather than representation, the instrument is aimed at a fraction of the target.
And for the argument in 2026: both movements are fighting over admissions and appointments. Neither has a position on land, credit, capital or inheritance — which is where the difference between these communities actually sits.
8The Statistics Both Sides Get Wrong
Six each. Same standard, same treatment.
The abolition side
Group A, 1 January 2024: SC 14.2%, ST 6.54%, OBC 19.14% — every one below quota, with OBCs eight points short thirty-six years after Mandal.
VERDICT: Contradicted by the government's own data.General and OBC closed two marks apart for all-India government MBBS seats in 2025. The SC gap of about 85 marks is real and is a different argument from the one usually made.
VERDICT: True for one category, false for the largest.No agreed measure of institutional quality has ever been applied, and Article 335's efficiency clause has never been operationalised. Tamil Nadu has run 69% for three decades without the predicted collapse.
VERDICT: Unmeasured assertion. Nobody has done the study, in either direction.Government employment is a small and falling share of Indian jobs. When lakhs apply for thousands of posts, the constraint is the number of posts.
VERDICT: The pain is real; the address is wrong.India runs one. The EWS threshold was copied from the OBC creamy layer with no study behind it, covers the overwhelming majority of eligible households, and the government does not publish who receives it.
VERDICT: Testable, tested, and not supported by the seven-year record.Groups newly added to the Scheduled Caste list gained 7–8 rank points of upward mobility over twenty years — peer-reviewed, with a credible identification strategy.
VERDICT: The best available causal evidence says the opposite.The retention side
Not where it matters. Group A is below quota for all three categories. Article 16(4)'s test is adequate representation in the services, and at senior level it has not been met.
VERDICT: True in the aggregate, false at the top — and the top is the point.More than 2,750 reserved teaching posts vacant in central universities as of November 2024, and 4,297 reserved teaching positions closed as "none found suitable."
VERDICT: False — and inconvenient for the other side too, since unfilled posts were taken from nobody.Scheduled Caste mobility is measurably rising, Group C representation is at or above quota, and overall central government representation exceeds the SC and ST quotas.
VERDICT: False, and self-defeating. Denying the gains erases the people who won them.It addresses employment and education. Scheduled Castes hold about 8.5% of land, own about 11.4% of enterprises and account for about 2.6% of billionaire wealth. No mechanism in the policy reaches any of those.
VERDICT: True of one dimension, not of the ones that hold wealth.A seven-judge bench permitted sub-classification of Scheduled Castes in 2024 on precisely that finding. Telangana implemented in 2025 and Karnataka in 2026.
VERDICT: Overtaken by events. This is now settled law, not a talking point.As Part 6 conceded at length, registered-case totals measure the reporting system as much as the offence. Better registration raises the same line.
VERDICT: Misuse of data. Chapters 6 and 8 of Part 6 explain why nobody should build on it.Twelve claims. Six from each side. Four marked false on one side and five on the other, with the rest partly true.
That symmetry is not a device. I did not set out to find an equal number; I set out to test the claims each side actually makes most often, and this is where it landed. If it had come out eight to two, I would have printed eight to two.
But notice the difference in kind between the two lists.
The abolition side's errors are mostly about scale and direction — believing reserved categories dominate the senior services when they are below quota, believing the score gap is large when for the biggest category it is two marks, believing the policy does not work when the causal evidence says it does. Those are factual errors about the present, and government data settles them.
The retention side's errors are mostly about sufficiency — claiming representation has been achieved, that seats are being filled, that nothing has improved, that inequality is being addressed. Those are not errors about whether the injury is real. They are errors about how much the remedy has accomplished, and they are made by people defending something that is working less well than they say.
One list is wrong about the facts. The other is wrong about its own success. Those are different failures, and they call for different corrections.
9The Finding Both Sides Will Hate
What the assembled evidence actually shows, stated as plainly as I can manage.
Put every chapter of this part together and a single conclusion emerges that neither camp has an interest in stating.
It works. Groups newly added to the Scheduled Caste list gained 7–8 rank points of upward mobility over twenty years. That is a measured causal effect, in a peer-reviewed journal, and it is the strongest evidence in this argument on any question.
It has not worked at the top. Every reserved category is below quota in Group A, seventy-six years in. Where decisions are made, representation has not been achieved.
It has not touched assets at all. 8.5% of land, 11.4% of enterprises, 2.6% of billionaire wealth. The instrument has no mechanism that reaches ownership.
And it operates on a shrinking share of the economy. Government employment is a small and declining fraction of Indian jobs — and the correspondence study in Part 6 found the sharpest measured discrimination in the private sector, which reservation has never touched.
What follows for each side
The central factual claims do not survive. Reserved categories have not taken over the government; they are below quota at every senior level. The score gap for the largest category is two marks. Reserved posts sit empty in their thousands. And the policy has a demonstrated causal effect.
What survives is the argument that the categories are incoherent — from Part 10 — and that the instrument is badly aimed. Both are true. Neither is an argument for removal rather than redesign.
The efficacy claim is vindicated and the sufficiency claim is not. Seventy-six years of the policy has produced representation in the middle of the government, under-representation at the top, and no movement at all in who owns the country.
Defending the present arrangement as adequate is not supportable on this data. What is supportable is defending it as a working instrument that has been asked to do far more than it can.
10Nine Questions Nobody Asks
The blind spots in this stage of the argument.
1. Why is nobody quoting the Group A numbers?
SC 14.2%, ST 6.54%, OBC 19.14% — published by the government, freely available, and decisive against the central claim of the abolition movement.
The retention side does not use it either, because the same table shows the policy has not delivered where it counts. A number that damages both sides gets cited by neither.
2. Why does the government not publish EWS representation?
Seven years, a 10% quota, and no reported figure. Anyone who believes economic criteria are the answer should be the loudest voice demanding that number.
3. If reserved posts are empty, who exactly lost them?
4,297 reserved teaching positions closed as "none found suitable." Those places went to no one. The grievance requires an occupant.
4. Why has nobody audited "none found suitable"?
It is a discretionary finding, made repeatedly, in one direction. Comparing its frequency for reserved and unreserved posts at the same institutions would settle the question in a month. Nobody has done it.
5. Why does neither movement talk about land?
Scheduled Castes: 18% of households, 8.5% of land. That is a larger and more durable inequality than anything in the admissions argument, and neither side has a position on it.
The plausible answer is that land redistribution is politically impossible and exam cut-offs are not, so the argument goes where the argument can be had.
6. Why is Muslim mobility falling?
The same paper that finds rising mobility for Scheduled Castes finds declining mobility for Muslims. Part 4 established that Dalit converts to Islam lose Scheduled Caste status by a paragraph from 1950.
Nobody has connected these publicly. It is the most important unexplored question in the data.
7. What would falsify "reservation works"?
A fair question to put to my own conclusion. It would be falsified by natural-experiment evidence showing no mobility gain for newly listed groups, or by evidence that gains are entirely captured by an already-advantaged sub-group. The second is partly true — hence sub-classification — and would narrow the claim rather than destroy it.
8. Why does nobody measure "efficiency of administration"?
It is a constitutional condition under Article 335. Seventy-five years on, no metric exists, so the loudest argument against reservation cannot be tested and the loudest defence cannot be either.
9. Why is nobody campaigning for the thing both sides implicitly agree on?
Chapter 4 showed that both readings of the dropout figure imply the same remedy: support after admission. Bridge courses, language help, mentoring, hostels.
It is cheap, it is uncontroversial, and it is the only intervention in this entire book that nobody would have to lose an argument to accept. Neither movement mentions it.
11What The Evidence Can And Cannot Settle
The honest scorecard for Part 11.
| Question | Status | Best current answer |
|---|---|---|
| Does reservation improve mobility? | Yes | 7–8 rank points over 20 years, causal |
| Have reserved groups taken over government? | No | Group A: SC 14.2%, ST 6.54%, OBC 19.14% |
| Has representation been achieved at the top? | No | All three below quota, 76 years in |
| Are reserved posts being filled? | No | 2,750+ faculty vacancies; 4,297 "none found suitable" |
| Is there a large entry-score gap? | Mixed | 2 marks General–OBC; ~85 marks General–SC |
| Does reservation damage institutions? | Unmeasured | No metric exists; Article 335 never operationalised |
| Has it changed asset ownership? | No | 8.5% of land; 2.6% of billionaire wealth |
| Is overall Indian mobility high? | No | Constant and low since before liberalisation |
| Is EWS reaching anyone? | Unknown | Not reported by the government |
| Why do reserved-category students drop out? | Unknown | Count published; reasons not recorded |
The three things nobody can tell you
- Whether "none found suitable" is a shortage or a mechanism. Answerable with a straightforward audit that nobody has commissioned.
- What reservation has cost. There is no measurement of institutional efficiency, so the central cost claim is untested in either direction.
- How the 27% is distributed inside the OBC category. No national sub-classification, and no caste count until 2027.
The best evidence available says reservation works, has not delivered where power sits, has never touched what people own, and operates on a shrinking corner of the economy — which means the honest argument is not about whether to keep it but about what else would have to be done alongside it.
That argument is Part 12, and it is the one both movements have avoided.
12The Five-Line Nichod
Everything in this book, compressed. If you remember nothing else, remember these.
PART 11 — THE DATA WAR
- The gradient is the whole story. Central government, 1 January 2024: at Group A the reserved categories are below quota — SC 14.2%, ST 6.54%, OBC 19.14%. In Group C they are at or above it. Among sanitation staff, over 66% come from those same categories. The higher the post, the fewer; the lower, the more.
- It works, and there is causal evidence. Caste groups newly added to the Scheduled Caste list gained 7–8 rank points of upward mobility over the following twenty years — peer-reviewed, published in 2024. "Reservation does not work" is no longer a supportable position.
- Thousands of reserved posts are empty. More than 2,750 reserved teaching posts vacant in central universities, and 4,297 reserved academic positions closed as "none found suitable" after recruitment. Whoever is right about why, those places were taken from nobody.
- It has never touched what people own. Scheduled Castes: about 18% of households and 8.5% of land; 19.3% of the workforce and 11.4% of enterprises; 2.6% of billionaire wealth against 88.4% for upper castes. Seventy-six years of employment policy has not moved ownership at all.
- The finding both sides will hate: it is an effective instrument aimed at a small and shrinking target. The medicine works. The dose is small, the disease has spread to organs it cannot reach, and both movements are arguing about whether to stop the prescription rather than about what else to prescribe.
Coming next
Case A: abolish it. Written as the best possible advocate would write it — merit, efficiency, caste-hardening, elite capture, the creamy-layer problem, the poor upper-caste problem, seventy-six years, the psychological cost of permanent labelling, and the constitutional case.
Case B: keep it. The same treatment — representation against charity, the vacancy evidence, the persistence evidence, what happened everywhere quotas were removed, the argument that merit is inherited, and the constitutional case.
Then the cross-examination: each side's best ten arguments put directly to the other, with the strongest reply. And the scorecard, claim by claim, marked fact, interpretation or contested — with no summary judgement yet.
A closing thought: the most useful number in this part is the one on the top row. Every reserved category is below its quota in the senior ranks of the Indian government, thirty-six years after Mandal. Both sides have had that figure available for years. Neither has used it, because it is fatal to one and humiliating to the other. — L.S.
Sources & further reading — Part 11
Every factual claim in Part 11, and where it comes from.
Central government composition
Department of Personnel and Training, Annual Report 2024–25, reporting the position as of 1 January 2024: approximately 32.52 lakh employees across 80 ministries and departments; overall representation of SC 16.84%, ST 8.7% and OBC 26.32%. Group A: SC 14.2%, ST 6.54%, OBC 19.14%. Group C excluding safai karamcharis: SC 16.75%, ST 8.94%, OBC 27.29%. Over 66% of safai karamcharis employed by the Union government drawn from SC, ST and OBC categories; some reporting of the same data gives a figure above 76%. The report does not publish representation figures for Economically Weaker Sections.
Faculty vacancies
Education Ministry replies to Parliament: as of November 2024, 5,182 teaching posts vacant across central universities, of which more than 2,750 were reserved — approximately 740 SC, 464 ST and 1,546 OBC. As of 31 January 2025, approximately 28.56% of sanctioned teaching positions (18,940 posts) vacant across IITs, NITs, IIMs, IISERs, central universities and other central higher education institutions. A parliamentary panel reported 56% of professor positions vacant at the IITs, IIMs, NITs and IISERs. 4,297 reserved teaching positions at professor, associate professor and assistant professor level left vacant in central educational institutions following recruitment through a "none found suitable" declaration. Special recruitment drives reported filling 26,751 vacancies by December 2024, including 15,637 faculty positions of which 1,949 were reserved for SCs, 771 for STs and 3,261 for OBCs.
Mobility
Sam Asher, Paul Novosad and Charlie Rafkin, "Intergenerational Mobility in India: New Measures and Estimates across Time and Social Groups," American Economic Journal: Applied Economics, 2024. Findings: intergenerational mobility in India broadly constant and low since before liberalisation; rising mobility for Scheduled Castes among sons; declining mobility among Muslims; daughters' mobility lower than sons' with less cross-group variation. Natural experiment: caste groups newly assigned Scheduled Caste status experienced on average a 7–8 rank point increase in upward mobility over the following twenty years, from which the authors conclude that affirmative action for Scheduled Castes has substantially improved their mobility.
Land, enterprise and wealth
All-India Debt and Investment Survey: Scheduled Castes approximately 18% of households owning approximately 8.5% of land; upper-caste Hindus approximately 22% of households owning approximately 28% of land; AIDIS 2018–19 indicating upper castes hold close to 55% of national wealth. · Telangana survey data: average General category family approximately 0.42 acres of irrigated land against approximately 0.20 acres for the average Scheduled Caste family. · Scheduled Castes approximately 19.3% of the workforce with approximately 11.4% owning an enterprise; Ministry of Micro, Small and Medium Enterprises survey finding 12.45% of unregistered non-agricultural MSMEs owned by Scheduled Castes, with none medium-sized. · National Family Health Survey: approximately 12.3% of Scheduled Castes in the highest wealth quintile. · Analysis of Indian billionaire wealth for 2022 attributing approximately 2.6% to Scheduled Castes and approximately 88.4% to upper castes.
Carried forward from earlier parts
The 13,500 dropout figure from central universities, IITs and IIMs reported in 2023, and the correspondence study result of ten, six and three interview calls, are set out with sources in Part 6. NEET 2025 all-India quota closing scores are set out in Part 1. The Group A shortfall should be read against the 2024 sub-classification judgment and the state implementations described in Part 10.
A note on what is missing — and on corrections
Several figures in this part rest on estimated OBC population denominators, because there has been no caste census since 1931; Census 2027 will change this. Wealth and land figures derive from sample surveys and differ between sources depending on definitions of caste category and of asset. Billionaire-wealth attribution is an analytical estimate rather than an official statistic and is given as such.
Part 11 assesses whether reservation works. It does not present the full case for or against continuing it, which is Part 12, nor a verdict, which is Part 14. If you find an error, I want to know.