The Last Part
This is an advocacy document. It is the case for the prosecution against Congress governments in power. The selection of material is one-sided on purpose. The evidence in it is not. Every serious claim carries a grade showing how well it is proved, the other side is given its strongest argument, and where the other side wins, this book says so. A second series applying the same test to the government of 2014 onwards is planned.
Eighteen parts ago this book said it would end with a list: proven, alleged, folklore, unknown, and the strongest case the defence can still make. That is what follows.
It is not the list that was planned. Part Eighteen made the planned version impossible, and the reason is worth stating on the first page rather than the fortieth.
Everybody assumes that a consolidation is a total — that the way to end a nineteen-part audit is to gather every finding into one place and present the sum, and that the sum is the verdict.
Part Eighteen took this series’ four tests and applied them to a different government. Three of the four returned the same answers. The instruments predate everybody. The rules are made by nobody. The documents are released by no one.
Which means the findings in this series are two different kinds of thing stacked on top of each other. Some are acts by particular governments on particular dates. Others are facts about a state that outlived every government in the book.
Those cannot be added. Adding them produces a number that looks like a verdict on a party and is actually a description of a country.
So this part has two columns rather than one, and the division between them is not the evidence grade. It is a single question: does this finding survive a change of government? Part Eighteen supplied the test on twenty items and answered yes for fifteen.
The consequence is that the last part of this series is shorter on charges than it was going to be and longer on machinery. That is a worse book for anyone who came to read a prosecution, and a better one for anyone who wanted to know why nothing changes.
There is a second consequence, and it is the reason this part exists in the shape it does rather than as an index.
A reader who has followed eighteen parts has met roughly a hundred findings, spread across seventy-nine years and delivered a few at a time. That is a difficult thing to hold, and the natural response at the end is to want a total — a single figure, a verdict, a sentence that says how bad it was. This part declines to supply one, and the refusal is the substance rather than a caveat attached to it.
What it supplies instead is the material sorted by how well each piece of it is proved, so that a reader can do the weighing. That is a less satisfying object than a verdict. It is also the only one this book has any right to produce, and Chapter One explains why in more detail than most readers will want.
How to read what follows
Chapters Two to Five are the four columns the series promised, in descending order of how well anything in them is established. Chapter Six is the consolidated state ledger. Chapter Seven is what nobody knows.
Chapter Eight is the second column — the findings that survived a change of government — and it is the most important chapter in this part.
Chapter Nine is the strongest case the defence can still make, and it is longer than any equivalent chapter in the series.
Chapter Ten is an honest list of what this book got wrong.
1What This List Is
A list is a claim about completeness, and this one is not complete. This chapter says in what way, so that nothing later depends on a reader assuming otherwise.
1.1 — Four things this list is not
It is not everything Congress governments did wrong. Nineteen parts covering seventy-nine years is a selection, made by one person, from a much larger body of material. Part One conceded this and nothing since has improved it.
It is not a score. There is no denominator. This series has said so in Part One, Part Fourteen, Part Sixteen and Part Eighteen, and repeating it a fifth time is not a formality — it is the single most-ignored sentence in the book.
It is not a comparison. Part Eighteen refused to say one government has been worse than another and this part does not smuggle that verdict back in through arithmetic.
And it is not final. Part Sixteen’s central case is still on appeal after eighteen years. The urea convictions came twenty-three years after the payment. Somebody reading this in 2050 will have evidence this book did not.
The second of those deserves a paragraph rather than a line, because it is the disclaimer that every reader accepts in principle and ignores in practice.
This book contains, on a generous count, something over a hundred findings against Congress governments across seventy-nine years. That sounds like a great many. It is also roughly one and a third findings per year of government, drawn from the entire record of a party that has held office at the Union for fifty-four of those years and in the states for a great deal longer — and there is no comparison figure, for any government, anywhere, because nobody has ever compiled one.
So a reader who finishes this series with the impression that Congress governed unusually badly has drawn a conclusion the book cannot support. What it supports is narrower and stranger: that these particular things happened, that most of them were never resolved, and that the machinery which produced them is still running. Whether another party in the same office for the same decades would have produced a longer list or a shorter one is unknown, and Part Eighteen is the closest this series comes to finding out.
A list assembled from other lists. Its usefulness is entirely a function of whether the things in it were established to the same standard, which is why the grades are printed against every entry rather than summarised at the top.
Its danger is that a list of forty items reads as a stronger claim than forty separate items read one at a time, even when nothing has been added. Length feels like weight.
The defence against that is the grade column. A reader who reads down the grades rather than down the descriptions will get a very different impression of this series than a reader who does the opposite, and the first impression is the correct one.
Which raises the obvious question of whether the scale was applied the same way throughout.
A six-grade scale is only worth something if the same claim would get the same grade in Part Three and in Part Sixteen. There is no way to prove that from inside the book, but there are three checkable signs.
The scale cost the author material in both directions. Part Twelve declined to assert Bofors, which is the most damaging thing available against a Prime Minister in this series. Part Eighteen declined to grade five contested matters against a different government. A scale applied only when convenient does not produce refusals on both sides.
Findings were downgraded as well as upgraded. The Adarsh war-widows claim was Alleged in Part Sixteen and appears in this part’s folklore table. The 1984 death toll is stated at the state’s own lowest official figure rather than the highest available.
And the grades move when the evidence moves. ADM Jabalpur was Court-found in Part Nine and is recorded here as overruled in 2017. The coal convictions are recorded here with the later acquittals attached.
None of this proves consistency. What it establishes is that the scale was doing work rather than decorating conclusions reached some other way, and a reader who wants to test that should pick the claim in this book they most dislike and check what grade it carries.
1.2 — The two columns
Part Eighteen’s Chapter Eight forced a division that this series had not anticipated, and it governs everything after Chapter Five.
| Column one: acts | Column two: machinery | |
|---|---|---|
| What it is | Something a named government did on a date | Something about the state that no government changed |
| Test | Would a different government have done it? | Did it survive a change of government? |
| Examples | The Emergency; the 1969 supersession; the 2008 spectrum allocation; the nine days of December 1992 | Article 356; prosecution sanction; the sitting days of Parliament; the unmade rules |
| Where it appears | Chapters Two to Five | Chapter Eight |
| What it supports | A charge against a party | A finding about a country |
The division is not always clean. Article 356 was built by a Congress government and used by everybody: it is in column two as an instrument and in column one every time a particular dismissal is described. Chapter Eight explains how such items are counted, which is: once each, and the text says which.
Before Part Eighteen this book had one column and thought it was auditing a party.
Applying its own four tests to a different government produced the same answer on three of them, which means for most of what this series measured, the party was not the variable.
That is not a defence of Congress. Column two is longer than column one, and almost every item in it was built by a Congress government during the fifty-four years they held office. The finding is that they built a machine and nobody has dismantled it — which is a heavier charge against the party and a lighter one against any particular ministry.
A prosecutor who discovers in the eighteenth part that his best material is structural has either found something or wasted seventeen parts. Chapter Ten records that this book cannot tell which.
Chapter Ten returns to that question and does not resolve it either.
Before any of that, the objection to assembling a list like this at all.
The objection to this part existing: that a consolidated list is a rhetorical device rather than an analytical one, and that its author knows it.
Eighteen parts delivered their findings a few at a time, each surrounded by the context that made it intelligible and the qualifications that made it honest. A table strips both. An entry that took four thousand words to establish appears as a line, and a reader scanning down Table 3 receives ten Supreme Court findings in the time it takes to read a receipt. That is not consolidation; it is compression, and compression is how a careful argument is converted into an impression.
And the four disclaimers do not fix it. Everybody knows a disclaimer at the front of a document is read once and forgotten by the second table. This book has said so itself about its own denominator problem, five times, while continuing to produce material that invites the comparison it says it cannot support.
The reply, which is partial: the alternative to a consolidated list is not the absence of one. It is somebody else’s, made from the same material without the grades, and this series has spent nineteen parts arguing that the grades are the whole difference between an argument and a slogan.
Where the objection stands, and it stands: this part is the most quotable thing in the series and the most easily severed from its qualifications. Part Seventeen made exactly that complaint about itself and could not solve it either. Two of nineteen parts having the same unfixable defect is worth a reader knowing.
Chapter Ten returns to that question and does not resolve it either.
This list is not complete, not a score, not a comparison and not final. Four disclaimers, all of them load-bearing, and the second one is the most ignored sentence in nineteen parts.
It has two columns, and the division is not the evidence grade. It is whether the finding survived a change of government.
Column two is longer. Almost everything in it was built by Congress governments and nobody has dismantled it — a heavier charge against the party and a lighter one against any particular ministry.
2Proven
This is the top of the scale: things a court convicted somebody of, and things a court found without convicting anybody. It is the shortest substantive chapter in this part and that is the finding.
Four things, each ordinary, which multiply.
Sanction. Certain public servants cannot be charged without permission from the executive. Parts Thirteen, Fifteen, Sixteen and Eighteen all found this operating.
Interlocutory appeals. Almost any ruling during a trial can be appealed separately, and each appeal stops the trial.
Witnesses. A case that takes twenty years outlives the memory, the availability and sometimes the life of the people who saw what happened.
And no clock. There is no statutory outer limit on how long a trial for a public-office offence may take. Parts Fourteen and Fifteen both proposed one; Part Eighteen records that it has never been introduced.
Part Fourteen’s finding, which nothing since has disturbed: delay is not a failure of the anti-corruption process. It is the process.
2.1 — Convicted
| What | Who | When, and how long it took | Part |
|---|---|---|---|
| Telecom contracts | A Union minister and officials | 2002 and 2011. Fifteen years to the second sentence; nine days in custody | Fourteen |
| Urea: ₹133 crore paid in advance, no delivery | Two foreign nationals, a Prime Minister’s nephew, a minister’s son, company officials | 2018. Twenty-three years | Fourteen |
| Coal block allocations | A Coal Secretary, two ministry officials, company directors | 2016 and 2017. Several later acquitted, the last in 2026 | Sixteen |
| November 1984 killings | Individual participants; one life sentence on reversal of an acquittal, a second in 2025 | 2018 and 2025. Thirty-four and forty-one years | Ten, Eleven |
| Secret cremations in Punjab | Six police officials, for the abduction and killing of a human rights investigator | 2005, enhanced 2007. Ten years | Ten |
| Hashimpura, May 1987 | Sixteen, to life, by the Delhi High Court on appeal | 2018. Thirty-one years | Eleven |
| Bhagalpur, October 1989 | Convictions in a case involving around a thousand deaths | 2007. Eighteen years | Eleven |
| The JMM bribery case | A former Prime Minister and a former minister | 2000. Acquitted on appeal in 2002 | Fourteen |
Eight entries. The last one was reversed. Read the third column downwards: the median time from act to conviction across this table is over twenty years.
Everybody assumes that the proven column contains the worst things in a book like this. Convicted is the top of the scale, so the top of the scale must be where the gravity is.
Read Table 2 again. Almost everyone in it is a police official, a company director, a civil servant or an individual participant in a riot. The only minister convicted for anything in seventy-nine years of this series was a telecom minister, over a three lakh rupee bribe, fifteen years late, and he was out on bail in nine days. The only Prime Minister ever convicted was acquitted on appeal.
Now put that beside what is not in this table. The Emergency. Nine days in December 1992. Two lakh crore of presumptive loss across two audits. A hundred and twenty-two licences cancelled. Twenty-one years of coal allocations held illegal. Not one of those produced a conviction of anybody who decided anything.
So the proven column is not where the gravity is. It is where the gravity is not. It contains the cases small enough, individual enough and old enough to have finished.
Part One built a six-grade scale because a book that recognised only convictions would have almost nothing to say about the largest events in Indian public life. Nineteen parts later, Table 2 is the proof that it was right, and it is the proof by absence.
The defence is entitled to make something of that and it is worth putting properly.
Read the third column of that table downwards and the obvious question is whether the delay is political.
Because it does not vary with the politics of the case, and that is the only test that distinguishes a systemic problem from a political one.
The urea convictions took twenty-three years and were obtained under a government of a different party from the one in office when the money was paid. The Hashimpura convictions took thirty-one years and came from a High Court on appeal after an acquittal. The 1984 convictions of 2018 and 2025 came thirty-four and forty-one years on, under two different governments, both hostile to the party in office at the time of the killings.
Every one of those had a government with every incentive to move faster, and none of them moved faster.
Which rules out the explanation that would otherwise be natural — that cases against a ruling party’s people go slowly. They go slowly regardless. The delay is a property of the process, not of who is being prosecuted, and Part Fourteen’s finding stands: it is not a failure of the anti-corruption process, it is the process.
The defence is entitled to make something of that and it is worth putting properly.
The defence reads Table 2 as its best page in this book, and the reading is not unreasonable.
Seventy-nine years, nineteen parts, and eight convictions — of which one was reversed, most are of officials and individual participants rather than of anybody who set policy, and the median time to judgment exceeds twenty years. On any ordinary understanding of what a prosecution is for, that is a prosecution that failed.
The reply, which concedes the arithmetic: the table is short because the Indian criminal process cannot reach decisions taken at the top, not because such decisions were not taken. Table 3 is the same events reaching the same courts by a different route and arriving at findings that stand.
Where the defence keeps something anyway: a book that has to explain why its top rung is empty is a book making a weaker claim than its title suggests. This series does make a weaker claim than its title suggests, and Chapter Ten says so in those words.
2.2 — Court-found
The second rung, and this is where the weight actually sits.
| Finding | Court and date | Still standing? | Part |
|---|---|---|---|
| The privy purse abolition by Presidential order was unconstitutional | Supreme Court, eleven judges, 15 December 1970 | Superseded by amendment in 1971 | Eight |
| The 39th Amendment placing the PM’s election beyond the courts was void | Supreme Court, 7 November 1975, under the Chief Justice who had opposed the doctrine used | Yes | Nine |
| ADM Jabalpur, that the courts were helpless during the Emergency | Supreme Court, 28 April 1976, 4:1 | Overruled in 2017 | Nine |
| The 1959 Kerala dismissal and the 1989 Karnataka dismissal | Karnataka held unconstitutional, Bommai, 11 March 1994 | Yes | Three, Sixteen |
| The 2008 spectrum allocation was arbitrary; 122 licences quashed | Supreme Court, 2 February 2012 | Yes. Never disturbed | Sixteen |
| Coal allocations arbitrary and illegal since 1993; 214 blocks cancelled | Supreme Court, 25 August and 24 September 2014 | Yes | Sixteen |
| The compensation paid to the victims of November 1984 was inadequate, repeatedly | Delhi High Court, 1996, 2005 and 2024 | Yes. In 2024 the administration was called insensitive and callous | Eleven |
| Legislators have no immunity for bribery | Supreme Court, seven judges, 4 March 2024, overruling 1998 | Yes | Fourteen, Fifteen |
| Nine Muslim men were wrongly held for the 2006 Malegaon bombing | Discharged 2016 on a charge sheet by a central agency | Yes | Fifteen |
| The Electoral Bond Scheme was unconstitutional | Supreme Court, five judges, unanimous, 15 February 2024 | Yes. Review dismissed | Eighteen |
Ten entries, eight still standing. Every one is a finding about a decision rather than about a person, which is why none of them produced a conviction and all of them changed the law.
Every entry in that table has a property the convictions do not, and it is the reason this rung carries the weight of the series.
A conviction is a finding about a person, and it binds nobody else. When the telecom minister was sentenced in 2011, the sentence established that one man had taken three lakh rupees; it established nothing whatever about how telecom contracts should be issued, and no subsequent minister was constrained by it in any way.
A court finding against a decision works in the opposite direction. It binds everybody, immediately, and it operates on the office rather than the holder. Bommai did not punish anybody for the Karnataka dismissal; it changed what any future government of any party could do with Article 356, and the effect was visible within two years. Sita Soren convicted nobody and made every legislator in India prosecutable for bribery.
Which is why the two tables in this chapter run in opposite directions. Table 2 is a list of individuals who were reached. Table 3 is a list of rules that were changed. The second is shorter on names and vastly longer on consequence, and a prosecution that measured only the first would have concluded that seventy-nine years produced almost nothing.
Compare the two tables. Table 2 sent people to prison and changed nothing. Table 3 sent nobody to prison and changed the constitutional law of India repeatedly.
Article 356 works differently because of one entry in Table 3. Every legislator in the country is prosecutable for bribery because of another. A hundred and twenty-two licences and two hundred and fourteen coal blocks were cancelled because of two more.
Part One’s decision to build a scale rather than a binary is the single most consequential craft decision in this series, and this is where it pays. A book that recognised only convictions would consist of Table 2.
Which is the shape of the top of the scale, and it is not the shape anybody expects.
Convicted is eight entries, mostly officials and individual participants, at a median of over twenty years. One minister, over three lakh rupees, fifteen years late, nine days in custody. One Prime Minister, acquitted on appeal.
The Emergency, December 1992, the spectrum allocation, the coal allocations — not one produced a conviction of anybody who decided anything.
Court-found is where the weight sits: ten findings, eight still standing, none of which jailed anybody and all of which changed the law.
3Audited and Commission-Found
The middle of the scale, and the rungs that almost nobody uses. An auditor’s finding and a commission’s finding are official, serious, and not legal determinations of anything — which is exactly why they carry the largest numbers in this book and the least consequence.
3.1 — Commission-found
| Finding | Which inquiry | What followed | Part |
|---|---|---|---|
| 110,806 detained without trial; 25,962 officials retired; 1,774 deaths from sterilisation procedures | Shah Commission, three reports to August 1978 | The report was recalled after 1980 and reconstructed from surviving copies | Nine |
| 2,733 killed in Delhi in November 1984; ~3,325 nationally | Ahuja Committee, August 1987; accepted by Nanavati | Four commissions, nine committees, two SITs. 25 convictions | Ten, Eleven |
| Sixty-eight leaders and senior officials held accountable for the 1992 demolition; the executive took no effective measures | Liberhan Commission, 30 June 2009 — seventeen years, forty-eight extensions | Nothing. Its terms of reference excluded the central government | Fourteen |
| Total confusion in processing intelligence alerts; inputs ignored; a proper analysis would have revealed a strong indication | Pradhan Committee, 2009 | Never published | Fifteen |
| 25 of 102 members of a housing society on public land ineligible; 22 proxy holdings; clearances as a quid pro quo | Patil Commission, 2013, 182 witnesses | Sanction to prosecute granted, then quashed for want of fresh evidence | Sixteen |
| 27,000–40,000 killed as a conservative floor in Hyderabad, 1948 | Sunderlal Committee, December 1948 | Never published; surfaced 1988; public from 2013 | Two |
| None of the sampled encounters in Manipur was genuine | Santosh Hegde Committee, 2013 | No prosecutions followed | Thirteen |
Seven inquiries. One report recalled, two never published, one whose terms of reference excluded the body most in need of examination. Not one produced a consequence proportionate to its own finding.
A body appointed by government under the Commissions of Inquiry Act to examine a matter of public importance. It can summon witnesses and documents and it reports to the government that appointed it.
What it cannot do is the whole point. It cannot convict, cannot order anything, and cannot compel its own report to be published. Its terms of reference are written by the government, which decides what it may examine — Part Fourteen found the same device used in 1963 and 1992, twenty-nine years apart, to keep an inquiry away from the body most in need of examining.
And there is no time limit. The Liberhan Commission sat for seventeen years and took forty-eight extensions.
So a commission’s finding is real, official, and attached to no obligation whatever. That is why Commission-found sits below Court-found on this book’s scale and above Alleged.
What those seven inquiries cost, and produced, is worth setting out.
Set the seven inquiries in Table 4 against what they cost.
The Liberhan Commission sat for seventeen years, took forty-eight extensions, held 399 sittings and cost around ₹8 crore. It named sixty-eight people and produced no consequence for any of them, because its terms of reference excluded the government that appointed it.
Four commissions, nine committees and two special investigation teams examined November 1984 across forty-one years. Twenty-five convictions.
Part Eleven’s finding covers this whole table and nothing since has disturbed it: establishing a record has functioned as a substitute for accountability rather than a step towards it.
One feature of Table 4 deserves a note, because it is the closest thing to a pattern in this book that nobody designed.
The seven inquiries were appointed by five different governments across sixty-five years, and the fate of each was decided by whoever held office when it reported rather than by whoever appointed it. The Shah Commission was appointed by a government that lost power before its work could bite, and its report was recalled by the party it examined. The Sunderlal report was made for a government that then held it for a lifetime. The Liberhan Commission outlasted four Prime Ministers.
An inquiry’s usefulness is therefore decided at a moment nobody controls when it is set up — which is an argument for the two rules in Table 12 that would fix it, both of which are ordinary, cheap, and have never been introduced by anybody.
3.2 — Audited
The rung nobody uses, carrying the two largest numbers in the book.
| Finding | Figure, and what it measures | Part |
|---|---|---|
| Spectrum allocated below value, January 2008 | ₹1,76,645 crore presumptive. The CBI’s own charge sheet used ₹30,984 crore | Sixteen |
| Coal blocks allocated without auction, 1993–2010 | ₹1.86 lakh crore final; ₹10.67 lakh crore in the leaked draft | Sixteen |
| Commonwealth Games procurement | Inflated contracts, disqualified competitors, single-tender situations | Sixteen |
| Freight equalisation’s dispersal benefit offset by transport costs | Stated to Parliament by a Union minister citing the committees | Five |
| Adarsh: officials in key posts subverting rules to obtain prime public land | CAG, 2011 | Sixteen |
Part Sixteen’s Chapter Three exists because these two figures were treated as measurements of theft and are measurements of a policy choice against a counterfactual. Both were later partially tested against real auctions and Part Sixteen’s Chapter Eight gives the result.
Those two figures deserve one paragraph of explanation, because they are the most repeated numbers in modern Indian political argument and almost nobody who repeats them knows what they measure.
A presumptive loss is not money that went missing. It is the difference between what the government received and what an auditor calculates it might have received under a different allocation method. It is a subtraction against a hypothetical. Change the assumed method and the figure changes — which is exactly what happened to the coal number, where the leaked draft said ₹10.67 lakh crore and the published report said ₹1.86 lakh crore, using the same underlying blocks.
That does not make the figures worthless. An auditor is entitled to say that an asset was given away below value and to estimate by how much, and both Supreme Court benches found the allocations arbitrary without needing the number at all. What it makes them is an estimate of a policy gap, not an inventory of stolen property — and Part Sixteen’s Chapter Eight tested both against real auctions afterwards, which is the only honest way to check an estimate of this kind.
The distinction cost Part Sixteen most of its force and it is repeated here for the same reason: a prosecution that uses a number it cannot explain is relying on the reader not asking.
There is a test for how much an official finding is worth, and it is not who made it. It is what the state did next.
A Shah Commission report recalled. A Sunderlal report never published and surfacing forty years later. A Pradhan report never published, its stated reason expiring in 2012. A Liberhan report whose terms of reference excluded the central government. A Hegde finding of no genuine encounters, followed by no prosecutions.
Five of the seven inquiries in Table 4 were either suppressed, scoped so they could not reach the answer, or ignored. That is not a coincidence and it is not one party’s habit — the reports span 1948 to 2013 and four different governments held them.
Which is why these grades sit in the middle of the scale and not at the top. A commission finds; a court decides. Only one of those two comes with an obligation attached.
Both rungs invite the same objection and it should be met directly.
The objection to the whole middle of this scale: that Commission-found and Audited are grades this book invented to keep material it could not prove.
A commission cannot convict. An auditor cannot rule. Neither hears the accused with counsel, neither applies a standard of proof, and both report to the government that appointed them. A scale that gives such findings their own rungs — above Alleged — has given official-sounding material a status the law does not recognise.
The reply: the alternative is worse in both directions. Treat these findings as proof and the two largest numbers in this book become theft, which Part Sixteen showed they are not. Treat them as mere allegation and the Shah Commission’s detention figures, the Ahuja death toll and the Hegde finding on Manipur all collapse into the same bin as a suitcase at a press conference.
Where the objection lands, and it lands hard: these two rungs carry the largest numbers and the least consequence, and a reader who does not attend to the grade will read them as proof. This book’s only defence is to print the grade against every entry and to say, as it does here, that the middle of a scale is where a careful reader should be most careful.
Which places these two rungs exactly where Part One put them, and for the reason Part One gave.
Seven official inquiries. One report recalled, two never published, one scoped so it could not examine the central government, one followed by no prosecutions.
The two largest numbers in this series — ₹1.76 lakh crore and ₹1.86 lakh crore — sit on the Audited rung, which means no court ever ruled on them, and Part Sixteen showed what they actually measure.
A commission finds. A court decides. Only one of those comes with an obligation attached.
4Alleged
This chapter is a list of things this book does not assert. It exists because a prosecution document that quietly drops what it cannot prove is a prosecution document that has been edited to look stronger than it is.
4.1 — What is in this column and why it stays there
| The allegation | Why it stayed here | Part |
|---|---|---|
| That the central government directed or connived at the demolition of 6 December 1992 | Thirty-four years, a commission, a criminal trial. Nobody established it | Fourteen |
| That Rajiv Gandhi ordered the Ayodhya locks opened in February 1986 | Contested by a witness from inside his own office. Part Twelve rested its case on 1989 instead | Twelve |
| That Rajiv Gandhi received money from the Bofors contract | No conviction of anybody, ever, under four different governments | Twelve |
| That ₹1 crore in cash was delivered to Prime Minister Rao in 1993 | A suitcase at a press conference by a man facing seventy-two prosecutions | Fourteen |
| That central direction lay behind the police inaction of November 1984 | Forty-one years of inquiry has not established it. Part Ten rested on the fact of inaction instead | Ten |
| Everything sourced to the leaked Henderson Brooks report | The state will neither confirm nor release it. Graded Alleged throughout Part Six | Six |
| Who burned the shrine at Charar-e-Sharief in 1995 | Disputed in both directions, including by a civil servant who was there | Fifteen |
| That any minister of UPA II took money in the four cases of Part Sixteen | No court has found it. The largest case ended in acquittals of everybody | Sixteen |
| The Batla House encounter of September 2008 | Contested in both directions; the challenges have not been sustained in any court | Fifteen |
| CIA or British involvement in the 1959 Kerala dismissal | Part Three deliberately did not rest on it | Three |
| That the Banswara speech of April 2024 was unlawful | Complaints made; no body with power to find made a finding | Eighteen |
Eleven items, several of which would have been the most quotable material in this series. Every one of them was available and every one was left here.
Take the first four rows together and consider what this book would look like if they had been promoted one rung.
It would say that a Congress government arranged the destruction of the Babri Masjid, that a Prime Minister ordered the locks opened and took money from an arms contract, and that another took a crore in a suitcase.
That book would be far more popular than this one and it would be worth nothing, because every one of those four claims can be dismantled in an afternoon by anybody with a search engine, and the dismantling would take the rest of the series down with it.
Part Two’s chapter on this: the true charges are routinely dismissed because they arrive bundled with false ones, and a single thirty-second debunk retroactively poisons every other page.
There is an objection to this discipline and it comes from the same side of the argument as this book.
The objection to this chapter, from the prosecution’s own side: that leaving eleven allegations ungraded is not rigour but timidity. Several of them are believed by very large numbers of people, some are supported by circumstantial material this book itself sets out at length, and a document that assembles all of that and then declines to draw the obvious inference is doing the defendant’s work.
The reply: the inference is available to the reader and this book has laid out the material precisely so that it can be drawn. What it will not do is print the inference as a finding, because a finding carries a grade and none of these has one.
Where the objection lands: on the practical effect. A reader who takes only the graded claims from this series takes a substantially milder view of these governments than the full record would support, and this book has knowingly produced that effect.
Why it stands anyway: Part Two’s finding is that true charges are routinely dismissed because they arrive bundled with false ones. A prosecution that trades accuracy for reach loses both, and always in that order.
Which turns on what the fifth grade is actually for.
Most political writing has two states: things asserted and things not mentioned. This book has six, and the fifth exists to hold a specific class of claim that the other two cannot.
Alleged means: this was said in public, it has not been shown false, and no body with the power to find has found it. All three clauses are load-bearing.
It is not a soft yes. Chapter Four’s eleven entries are not claims this book half-believes; they are claims it declines to make. And it is not a soft no — several of them may well be true, and one may be established after this book is printed.
The difference between this grade and silence is that a reader can see what was considered and rejected, which is the only defence a selective document has against the charge of having quietly dropped what did not suit it.
There is a way to check that this is a discipline and not a convenience.
The test for whether a selective document has buried its weak material is whether the weak material is findable, and in this series it is.
Every entry in Table 6 was set out at length in the part that considered it, with the evidence for it as well as against. Part Twelve gives the Bofors record over a full chapter before declining to assert it. Part Ten sets out the entire case for central direction in November 1984 and then rests on inaction instead. Part Six grades the whole leaked-report material Alleged and then uses it anyway, flagged, because the alternative was to pretend it did not exist.
A claim that has been argued for three thousand words and then not made is a different object from a claim that was never mentioned. The first can be checked by a reader who disagrees. The second cannot be checked by anybody.
Which is the practical difference between a grade and an omission, and the reason this chapter is printed rather than left as an absence.
There is a pattern in that table that is worth naming, because it is not obvious and it recurs.
Nine of the eleven entries are allegations of direction — that somebody at the top ordered a thing that then happened. Not that the thing happened; that is established in every case. That it was ordered.
This is the hardest class of claim to prove about any government anywhere, and for a structural reason. An instruction that would establish it is either never written down or is written down in a file that is not released — and Chapter Seven lists three such files, one of them covering precisely the nine days in which the most consequential of these allegations would have to be established or refuted.
So the eleven entries in Table 6 are not eleven independent failures of evidence. They are one failure, repeated: the Indian state does not release the documents that would settle questions of direction, and in their absence the claim can be neither made nor dismissed. A reader who finds this chapter unsatisfying is having the correct reaction, and the remedy is in Chapter Seven rather than here.
4.2 — The two the defence gets to keep
Two entries in that table deserve to be stated as what they are, which is wins for the other side.
Bofors. Part Twelve set out the whole record — the contract of March 1986, the express bar on agents, the Swedish audit confirming commissions, the Swiss documents of 1997, the chargesheet naming a dead man, the quashing in 2004 before Congress returned to office, the criticism of the investigating agency in 2005, the failure to appeal in time. No conviction has ever been obtained, of anybody, and the case failed under Congress, the National Front, the United Front and the National Democratic Alliance alike. What it established was a mechanism, not a crime.
The spectrum acquittals. Part Sixteen refused to treat a failed prosecution as a proved charge and this part does not reverse that in its final chapter. The appeal admitted in March 2024 is pending. Until a court says otherwise, everybody accused in the largest scandal in modern Indian history has been acquitted, and this book says so in its consolidated list rather than in a footnote.
Eleven allegations this book does not make. Several were the most quotable material available and every one was left on this rung.
Including: that any government arranged the 1992 demolition, that Rajiv Gandhi took Bofors money, and that any UPA II minister took money in the four cases of Part Sixteen.
A book that promoted those four rows would be far more popular and worth nothing, because each can be dismantled in an afternoon — and the dismantling takes the rest of the series with it.
5Folklore
Part One contained a chapter called What Was Made Up and promised that its list would be repeated here. This is that column, and it is the most useful chapter in this book.
5.1 — Why a prosecutor kills his own material
The argument was made in Part One, restated in Part Seventeen and is worth having once more in the last part, because it is the single idea this series would most like a reader to keep.
It is generally assumed that debunking a false claim against somebody helps that person. It is why partisans of every description are reluctant to do it, and why a false claim, once in circulation, tends to stay there unchallenged by the side it favours.
It is backwards, and the arithmetic states in one paragraph. A reader who meets a list of forty charges and finds that one is fabricated does not discard that charge. He discards the source, and with it the thirty-nine he had not checked. The loss is not one item; it is the list, and it is retrospective.
So a fabrication in circulation is not an asset to the side making it. It is an asset to the side it attacks — a permanently available demonstration that the accusers are unreliable, producible at any moment and costing nothing to deploy. Which produces the conclusion this chapter exists for: the fabrications against Congress are Congress’s best defence, and killing them is the most useful thing a prosecutor can do. Nobody else writes this chapter, in either direction, and that is precisely why it works.
There is a second half to the argument that has been implicit for eighteen parts and should be stated plainly at the end.
The fabrications in Table 7 did not appear randomly. Six of the eight attach to exactly the places where the documented record is strongest — the 1984 killings, the Adarsh allocations, the statements of the 2004–2014 governments. A fabrication grows where there is already an argument to be had, because that is where an audience exists for it.
Which produces a perverse result. The stronger the real case against something, the more false material accumulates around it, and the more false material accumulates, the easier the real case is to dismiss. Part Sixteen’s Adarsh chapter is the clearest instance in the series: a CAG finding, a judicial commission with 182 witnesses, 25 ineligible members and 22 proxy holdings — all of it routinely waved away with a single sentence about war widows that nobody has established.
So killing a fabrication is not a concession made in the interest of fairness. It is the removal of the specific object that is protecting the thing this book is trying to prove. That is the whole of the argument, and it is why this chapter sits in the middle of the series’ last part rather than in an appendix.
Before the column itself, the objection to printing it at all.
The objection to this chapter is the sharpest one in the book, and it comes from the author’s own side: that a prosecution which spends a chapter defending the accused has stopped being a prosecution. Nobody appointed this book counsel for Congress. If a claim in circulation is false, the party has lawyers, spokesmen and a national organisation to say so.
And there is a harder version. Killing eight fabrications in a document read by people hostile to Congress does not reduce the fabrications; it launders the rest of the book, by purchasing credibility with material the author never believed anyway. That is a rhetorical technique, not a discipline.
The reply, and it is a partial one: the technique and the discipline are indistinguishable from outside, and the only thing that separates them is whether the corrections cost the author anything. Three of the eight did — including the most quoted political sentence in modern India, dismantled in Part Seventeen’s first substantive chapter, which was the single most useful thing this series had.
Where the objection stands: it is true that this chapter buys credibility. What it says in its own defence is that the credibility was paid for, in material, at a price a reader can check.
5.2 — The column
| The claim in circulation | What the record shows | Part |
|---|---|---|
| A criminal allegation against Rahul Gandhi arising from a 2012 petition | The petition was dismissed. In 2012 a Supreme Court bench found it rested on falsehood and was driven by malice and political vendetta, imposed a ten lakh rupee cost on the petitioner — a former MLA — and directed part of it to the family whose reputation the petition had also damaged. Investigators could not locate that family. Search results conflate a different person of the same common name with the Congress leader, and this conflation is itself a case study | One |
| That Feroze Gandhi was born a Muslim named Feroze Khan | Checkably false. He was a Parsi, son of Faredoon Jehangir Ghandy. Most of the affair rumours about Indira Gandhi descend from this strand | One |
| That a Prime Minister said Muslims have the first claim on the country’s resources | The nine words are real: 9 December 2006, National Development Council. The reading is not supportable on the paragraph, which lists SC/STs, backward classes, minorities and women and children. The Prime Minister’s Office reproduced it the next day | Seventeen |
| That Rahul Gandhi tore up the ordinance in September 2013 | He did not. He said it should be torn up. A photograph of him tearing paper is from a 2012 rally. Major outlets reported the tearing as fact, one has published a correction, and in 2023 the party’s own chief spokesman described the act | Seventeen |
| That Mani Shankar Aiyar called Narendra Modi a chaiwala in January 2014 | The word was not used. He offered space to sell tea. The insult did not require the noun and the substance was worse than the word | Seventeen |
| That Nehru’s remark about Aksai Chin was made in 1962, after the war | 5 December 1961, ten months before. The misdating makes the statement less damaging and is repeated most enthusiastically by the speaker’s critics | Seventeen |
| That the INS Viraat was used for a family holiday in 1987 | Trip real, purpose contested. The ship’s commanding officer, the Southern Naval Commander and the island’s administrator all said it was official. One other retired officer took the opposite view. Verdict: weight low, both sides stated | One |
| That flats meant for Kargil war widows were taken at Adarsh | The society’s stated purpose changed several times across a decade. A fixed allocation for war widows, diverted, is Alleged rather than established — and it is the sentence that lets every documented finding there be waved away | Sixteen |
Eight entries. Three of them were the best material this book had against Congress and two were the best it had against anybody else, which is the point.
Look at what this column cost.
Part Seventeen’s Chapter Two opened by dismantling the most quoted political sentence in modern India — the single most useful thing this book had — and spent its first substantive chapter doing it.
Part Sixteen’s Chapter Six dropped the war widows sentence, which is the most damaging thing attached to the Adarsh case.
Part One graded the Viraat claim as low weight and gave both sides.
Against which: this chapter also kills three claims circulating in the other direction, including two that make a Congress leader look better than the truth does.
A folklore column that only contained things helpful to the author would not be a folklore column. It would be a fifth charge sheet.
Which raises a question this chapter has to answer: if these are all correctable, why are they still here?
Four of the eight entries in Table 7 have been publicly corrected — one by a national outlet issuing a formal correction, one by the speaker himself, one by the Prime Minister’s Office publishing the actual text the following day. All four remain in circulation.
Three things keep them there.
The correction is a different shape from the claim. The claim is a sentence. The correction is a paragraph explaining what the sentence leaves out. Only one of those survives being repeated.
The claim is useful and the correction is not. Nobody needs the true version of the 2006 sentence for anything. Large numbers of people need the false one every week.
And the correction usually comes from the accused. A denial by the person accused is read as a denial, not as evidence — which is exactly why a correction issued by the prosecution, as in this chapter, is worth more than the same correction issued by the defence.
5.3 — What is not in this chapter, and why
One category was excluded from this series entirely, at the outset, and the reason belongs in the last part.
Viral claims about the personal lives of individual leaders — particularly allegations of sexual crime against living named people — do not appear in these nineteen parts in any form, including as a claim with a correction attached. The first row of Table 7 describes a piece of litigation and what a court found about it, and deliberately does not say what was alleged.
The reasoning is not squeamishness and it is testable.
Writing out a false allegation with a debunk attached spreads the allegation. The debunk travels less well than the claim, is dropped when the passage is quoted, and the reader who half-remembers the page remembers the accusation. A correction printed underneath a fabrication is a distribution mechanism with a disclaimer on it.
The inverted handling loses nothing that matters. What makes the first row of Table 7 useful is not what was alleged. It is that a court examined it, found it rested on falsehood and malice, imposed costs, and directed part of them to a family the petition had also damaged — and that investigators could not find that family at all.
That is a complete account of a political fabrication, with a judicial finding attached, and it does not require the fabrication to be restated. Part One established this handling and nineteen parts have kept it.
Which is why the handling has been kept unchanged through nineteen parts and is stated once more in the sources note at the end.
Debunking a false claim does not help the defendant. A reader who finds one fabrication in forty charges discards the source, not the charge — so a fabrication in circulation is an asset to the side it attacks.
Which is why the fabrications against Congress are Congress’s best defence, and killing them is the most useful thing this book does.
Eight entries. Three were the best material this book had. Three more make a Congress leader look better than the truth does, and they are in the same table.
6The State Ledger
Part Five’s Chapter Eight committed every subsequent part to carrying a state ledger and said it would consolidate here. This is it: fourteen parts of entries in one table.
Before the table itself, the objection to keeping one at all.
The objection to any state ledger: that it converts a national government’s record into a set of grievances with postal addresses, and that this is how regional resentment is manufactured rather than measured.
Every large country makes decisions that fall unevenly. A tariff, a rail formula, a language policy or a Plan allocation will always advantage somewhere and disadvantage somewhere else, and listing the disadvantaged places without listing the advantaged ones is not analysis.
The reply, and it is why the last row of Table 8 exists: the beneficiaries are in the ledger. Gujarat and Tamil Nadu appear as the states that received the removal of somebody else’s advantage, which is what freight equalisation actually did.
Where the objection is right: a ledger like this is genuinely dangerous material, it is used by people this book has no wish to supply, and the Hidden Assumption box below is the only defence against that — a state appears here when something was written down about it, not when more happened to it.
One word first about how the entries were carried across.
Nothing in Table 8 is new. Every entry is carried forward from the ledger printed in the part it comes from, and every one of those was sourced there.
That matters because a consolidated table is where errors compound invisibly. A figure loosely stated in Part Five becomes a hard number in Part Nineteen simply by being retyped, and a reader has no way of seeing the softening.
So two rules were applied. No entry is stated more confidently here than it was stated in its own part, and where a part hedged, the hedge is carried. The Punjab row still says the rivers case is weak on the water. The Bengal row still says the sharpest decline runs from 1977 under a government Congress did not control.
A consolidation that quietly hardens its sources is not a consolidation. It is a second draft with the qualifications removed.
6.1 — The consolidated ledger
| State | What was done, and by which instrument | Parts |
|---|---|---|
| Bihar · Jharkhand | Freight equalisation from 1952 confiscated the only advantage the mineral belt had. Jharkhand separated in 2000 by a non-Congress government, after the hole was dug. Coal blocks, screening committee, 1993–2010; 214 cancelled 2014 | Five, Sixteen |
| Odisha · Chhattisgarh · Madhya Pradesh | The same freight rule. The same coal blocks. The overlap of scheduled areas, the mineral belt and the Red Corridor — minerals always, scheduled 1950, extraction from the 1950s, insurgency only from the late 1970s | Five, Thirteen, Sixteen |
| West Bengal | Freight equalisation; ~1,218 factories in 1946, a quarter of India’s. The sharpest decline runs from 1977 under the Left Front, which Congress did not cause | Five |
| Punjab | Canal investment from the 1880s made it the Green Revolution’s target. The rivers: weak case about the water, strong case about the procedure. 1978–1993: Blue Star, President’s Rule for five years, TADA at a 1–2% conviction rate, ~2,097 cremations confirmed in one district. Procurement lock-in never reversed by anybody | Five, Ten |
| Delhi | November 1984: 2,733 dead on the state’s own lowest figure, 587 FIRs, ~240 untraced. The Emergency’s press power cut at 2am on 26 June 1975. Cash on the floor of the Lok Sabha, 2008. The Commonwealth Games | Nine, Ten, Eleven, Fifteen, Sixteen |
| Uttar Pradesh | The affidavit of 27 November 1992 and the nine days. President’s Rule from December 1992. Hashimpura, May 1987 — 31 years to conviction | Eleven, Fourteen |
| Maharashtra | Malegaon 2006 and nine men wrongly held; the July 2006 train bombings, twelve convicted 2015 and acquitted 2025; 26/11 and a report never published; Adarsh; the 1992–93 violence and a commission disbanded then revived | Eleven, Fifteen, Sixteen |
| Jammu and Kashmir · Ladakh | Accession 26 October 1947; the plebiscite offered in January 1948; Article 370 hollowed by Presidential Order for four decades and removed on 5 August 2019 under President’s Rule. Hazratbal ended without a shot; Bijbehara; Charar-e-Sharief. Aksai Chin held since 1962 | Two, Six, Seven, Fifteen, Eighteen |
| Nagaland · Manipur · Arunachal · Assam · Mizoram · Tripura · Meghalaya | AFSPA from 1958, descended from a 1942 colonial ordinance; the 1972 amendment removing the need for state consent; Section 6 unamended for sixty-eight years. Withdrawn Mizoram 1980s, Tripura 2015, Meghalaya 2018; in force elsewhere into March 2026. Mon, December 2021: sanction refused | Thirteen, Eighteen |
| Kerala | The elected government dismissed 31 July 1959 after 27 months — where Article 356 became an instrument. Later a case that breaks the pattern in the freight ledger | Three, Five |
| Andhra Pradesh · Telangana | Hyderabad 1948 and a committee report unpublished for sixty-five years. Mecca Masjid 2007 and compensation ordered recovered. The urea prosecutions | Two, Fourteen, Fifteen |
| Rajasthan · Haryana · Himachal · Karnataka | Three governments dismissed 15 December 1992 and the dismissals upheld. Ajmer 2007; Jaipur 2008 and acquittals in 2023; Samjhauta 2007. Bommai’s own 1989 dismissal held unconstitutional | Fourteen, Fifteen, Sixteen |
| Gujarat · Tamil Nadu · Every state | Freight equalisation’s beneficiaries, who received the removal of somebody else’s advantage. The Right to Information Act 2005; the employment guarantee; food security 2013; GST 2017 | Five, Fifteen, Sixteen, Eighteen |
Thirteen rows covering every state in which this series found something. The credit column is real and is in the last row and in the Jammu and Kashmir and Punjab entries.
Which is the objection. The reason for keeping one anyway is narrower than it looks.
India is a union of states and almost every decision in this book was taken in Delhi and landed somewhere else. A national record that does not say where things landed is describing an abstraction.
Part Five committed the series to this table for a specific reason: it had just found that a national freight formula, uniform on its face, transferred an advantage from one set of states to another for forty-one years — and that this was invisible in any national statistic.
A ledger makes that kind of finding checkable. It also makes it quotable by people who want a grievance rather than an argument, which is the objection above and the reason the box below exists.
With one caution about how a table of this kind is read.
Everybody reading a table like this assumes it is a map of damage — that the states appearing most often suffered most.
It is not, and the difference matters. Look at which states recur. Delhi, five times. Maharashtra, three. The mineral belt, three. Jammu and Kashmir, five.
Delhi recurs because it is where the instruments are operated. Jammu and Kashmir recurs because it is where every constitutional instrument in this book has been tested at some point. Maharashtra recurs because it has the courts, the market and the police force that generate records.
A state appears in this ledger when something was written down about it. The mineral belt appears three times not because more happened there but because Part Five found the pipe, Part Thirteen found the maps and Part Sixteen found the blocks — three separate authors of three separate documents about the same ground.
Which means the ledger is a map of documentation, not of harm, and the states that appear least are the ones about which the least was recorded. Part Thirteen’s finding stands over this whole table: somebody knows every number, and what has never been done is the addition.
Which is a smaller claim than a ledger of this size appears to make, and it is the true one.
One further observation before leaving the ledger.
One row in Table 8 behaves differently from all the others and it is worth naming.
The mineral belt appears three times, in three parts, on three unrelated arguments, and moves the same direction every time. Part Five found a freight formula that removed its only advantage for forty-one years. Part Thirteen found the overlap of scheduled areas, minerals and insurgency — with the minerals there always, the extraction from the 1950s, and the insurgency only from the late 1970s. Part Sixteen found the coal blocks, allocated without auction from 1993 and cancelled in 2014.
Three authors, three decades apart, working on questions that have nothing to do with each other, describing the same ground.
That is what a finding looks like when it is not an artefact of the argument being made. Nothing else in this ledger has that property, and the calculation that would settle it is the first entry in Table 10 and has never been done.
Which is the one entry in the ledger that does not depend on the argument being made about it.
6.2 — What is not in the ledger
Three absences are worth naming, because a ledger’s silences are read as findings whether or not they are meant to be.
The southern states are thin. Kerala appears twice, Tamil Nadu once, Karnataka once. That is not because less was done there; it is because Part Five’s freight argument runs through the mineral belt, Parts Ten and Eleven run through the north, and the constitutional material of Parts Three, Seven and Nine happens to land elsewhere. A series organised around different questions would produce a differently shaped table.
The north-east appears as one row for seven states. Part Thirteen treats them together because AFSPA does, which is a defensible reason for a chapter and a poor one for a ledger. Mizoram, where the statute was withdrawn in the 1980s, and Manipur, where the Hegde Committee found none of the sampled encounters genuine, are not the same entry and are printed as one.
And nothing appears for the states this book did not reach at all. Goa, the smaller union territories, and most of the Himalayan states have no row, which records only that no part of this series had occasion to look.
All three are instances of the same caution: this table maps where the series went, and the series went where the documents were.
Thirteen rows, every state where this series found something, consolidated as Part Five promised.
The mineral belt appears three times across three unrelated arguments — freight equalisation, the three maps, and the coal blocks. It is the only entry that moves the same direction in every part it appears in.
And the ledger is a map of documentation, not of harm. A state appears here when something was written down about it.
7What Nobody Knows
Every part of this series ended with an honest list. This chapter consolidates them, and the result is smaller and stranger than nineteen separate lists suggest.
7.1 — The documents
| Document | Withheld since | Governments that have held it | Part |
|---|---|---|---|
| The Henderson Brooks–Bhagat report on 1962 | 1963 — sixty-three years | Every government since. Part I published abroad in 2014 | Six |
| Home Ministry file notes, 27 November – 6 December 1992 | 1992 — thirty-four years | Congress, United Front, NDA, UPA, NDA | Fourteen |
| The Pradhan Committee report on 26 November 2008 | 2009 — seventeen years. The stated reason expired in 2012 | UPA, then NDA | Fifteen |
| The Sunderlal report on Hyderabad, 1948 | Sixty-five years — eventually released | Every government to 2013 | Two |
| The Shah Commission reports on the Emergency | Recalled after 1980; reconstructed from surviving copies | Congress, from 1980 | Nine |
Three are still withheld, one took sixty-five years, one was recalled after publication. No government of any party has released any of the three that remain, and two of the three describe the conduct of a party now in opposition.
Before the second table, a distinction this book has kept throughout and should state plainly.
Three different states of not-knowing, and this book has been careful to keep them apart.
Withheld: the document exists, the state has it, and release has been refused or simply never made. Henderson Brooks; the 1992 file notes.
Unpublished: the document exists and was made for the government, which has never put it out and has offered a reason that has since expired. The Pradhan Committee report.
Never made: no document exists, nobody is refusing anything, and the components are all public. The six items in Table 10.
The first two have a custodian. The third has nobody, which is why the box below argues it is the harder half of this chapter.
Each of the three documents is asserted here as a live refusal rather than an old one, and that claim can be checked.
Because refusal has left a paper trail in each case, and the trail is public.
Henderson Brooks. Successive Defence Ministers have told Parliament the report remains classified, the most recent refusals citing operational sensitivity — a position maintained about a war that ended sixty-three years ago. Part One was published on a foreign website in 2014; the government neither confirmed its authenticity nor released the official text, which is the position it has held since.
The 1992 file notes. Requests under the transparency law have been refused, and a denial under that Act is an official confirmation that the record exists. The Liberhan Commission’s own report records that it did not obtain the full central-government record — a commission with statutory summoning powers, examining the events those files cover, reporting seventeen years later.
The Pradhan Committee report. The government’s stated reason was that publication would prejudice the trial then pending. That trial concluded in 2012. Fourteen years later the report has still not been published and no fresh reason has been offered.
What none of that establishes is why. A withheld document is consistent with something damaging in it and equally consistent with nobody having taken a decision to release it since the last person who cared retired. This book asserts the withholding and does not assert a motive, and the difference is the same one Chapter Four is built on.
7.2 — The calculations
The stranger half. Six questions this series says would be settled by arithmetic, on data that is already public, which nobody has done.
| The calculation | What it would settle | Part |
|---|---|---|
| Freight equalisation’s cost, state by state, 1952–1993 | Whether the mineral belt’s decline is attributable to the policy or to other causes | Five |
| The number of people displaced by development projects since 1947 | The largest unmeasured fact in this book. Every component exists in project files | Thirteen |
| A national register of terrorist incidents and their judicial outcomes | The conviction rate for mass-casualty attacks in India, which nobody can state | Fifteen |
| Presumptive audit estimates against realised auction outcomes | Whether the two largest numbers in this book were any good | Sixteen |
| A public verbatim archive of political speech | Four-fifths of the arguments in Indian political life. A filing job, not a calculation | Seventeen |
| Declared party income by source, before 2018, during the bond years, and since February 2024 | Whether striking down the bonds made funding transparent or moved it | Eighteen |
Six items. Not one requires a document anybody is refusing to give. Every one could be done by a competent graduate student with the sources already in the public domain.
Everybody assumes that what a country does not know about itself is what somebody is hiding. Nineteen parts of this series have ended by naming a withheld document, which reinforces it.
Put the two tables side by side. Three withheld documents. Six calculations nobody has done.
The withheld documents are a real problem and this book has said so five times. But they are a smaller problem than the second table, because a withheld document has a custodian who can be asked, sued or shamed. An unmade calculation has nobody.
Part Thirteen found an addition where every component existed, nothing was concealed, and no decision existed that anybody had to defend. Part Fifteen found seventeen intelligence alerts forwarded and never assessed because assessment was nobody’s job. Part Eighteen found eleven laws never introduced — not defeated, not lapsed, never brought.
Concealment is a decision and can be reversed. An absence is not a decision at all, and there is nobody to reverse it.
Which is why the second table is the harder half of this chapter and the one this series would most like somebody to take away from it.
That second table is the more vulnerable of the two and the objection to it should be stated here rather than in Chapter Ten.
The objection to Table 10: that six calculations nobody has done is not a finding about the Indian state but about the author’s reading. Somebody may well have done them — in a doctoral thesis, a state planning department, a working paper, a language this book does not read.
That objection is serious and this book cannot fully rebut it. Asserting that nothing exists anywhere is the hardest claim in the series to sustain and Chapter Ten names it as a weakness.
What can be said in reply: the test applied was not whether a calculation exists but whether an authoritative one is available to a citizen who goes looking. On that test all six fail, and several fail conspicuously — no ministry publishes a national displacement figure, and no agency publishes a register of terrorist incidents and outcomes.
And the invitation stands. One counter-example makes a row wrong, and a book that asks for that and then ignores it deserves nothing.
Stated plainly, the residue of nineteen parts comes to this.
Put the two halves of this chapter in the plainest terms available.
Three documents. Sixty-three years, thirty-four years, seventeen years. Every one exists, in a building, with a custodian who could be asked, sued or shamed. Two of the three describe the conduct of a party now in opposition, which means releasing them would cost the present government nothing at all.
Six calculations. Not one requires a document anybody holds. Every component is published. Any of them could be done by a competent graduate student in a summer.
Nineteen parts, seventy-nine years, and this is the residue: three things nobody will hand over and six things nobody has bothered to add up.
The second number is larger than the first, and it is the one nobody argues about.
7.3 — The one question that decides how this book should be read
One item on the unknown list is not a document or a calculation. It is a question, it was raised in Part Sixteen, and it has not been answered.
Why did the prosecutions fail?
Part Sixteen found that in all four of its cases the allocation question went against the government and the criminal question did not stick. Three explanations are available and no more.
One: nothing criminal happened. The allocations were bad policy, badly executed, and the cases failed because there was no crime.
Two: crimes happened and the prosecutions failed. Cases built late, on inference, by agencies without the capacity to prove a bribe.
Three: the cases were brought for reasons that had little to do with proving them.
This book cannot choose between them and has said so twice. The answer determines whether Part Sixteen describes a scandal or the manufacture of one — and by extension whether the largest single body of material in this series is a prosecution or an artefact of one.
Chapter Ten returns to it, and the second series is committed to adjudicating what this one could not.
Each of those three answers implies a different book, and it is worth being explicit about which.
If the first is right — nothing criminal happened — then Part Sixteen is a chapter about policy failure that was mistaken by an entire country for a chapter about theft, and the ₹1.76 lakh crore figure did more damage to Indian public understanding than the allocation it described. That is not an exoneration of the government; arbitrary allocation of public assets is what two Supreme Court benches found, and it stands. But it would mean the scandal was real and the crime was not.
If the second is right — crimes happened and the prosecutions failed — then the finding is about capacity, and it is the bleakest of the three. It would mean the Indian state can establish that a ₹1.76 lakh crore allocation was arbitrary and cannot establish who took anything for it, which is a description of an investigative system that works on documents and not on people.
If the third is right — the cases were brought for reasons unrelated to proving them — then the material in Part Sixteen is evidence about the agencies rather than about the ministers, and this book has spent a part of its length reproducing the output of a process it should have been auditing.
This series cannot choose. What it can say is that the three are not equally convenient: the first favours the accused, the third favours the accused and indicts somebody else, and only the second is what most readers assume without noticing they have assumed it.
Three documents still withheld, for sixty-three, thirty-four and seventeen years. Two of the three describe a party now in opposition, and no government has released them.
Six calculations nobody has done, none of which needs a document anybody is refusing to give.
The second table is the harder one. Concealment has a custodian who can be asked. An absence has nobody.
8The Machine
This is the second column, and the reason this series was worth writing. It contains the findings that survived a change of government — which, on Part Eighteen’s test, is the only evidence that a finding is a finding rather than a coincidence.
8.1 — The instruments
| Instrument | Built | Used since by | Status |
|---|---|---|---|
| Article 356, President’s Rule | In the Constitution; turned into an instrument in Kerala, 1959 | Every government. ~90 times by 1992 | Restricted by Bommai 1994; used in J&K in 2019 |
| Presidential Orders under Article 370 | From 1954 | Congress governments for four decades; the removal itself in 2019 | Spent |
| Amendment of the interpretation clause to reverse a reading | First Amendment, 1951 — its durable product was a procedure with no subject matter | Used on 5 August 2019 | Available |
| Prosecution sanction as a bar | AFSPA s.6, 1958; the Single Directive; state statutes | Every government. Refused at Mon, December 2021 | Unamended |
| Terms of reference that exclude the body most in need of examination | Henderson Brooks, 1963 — Army HQ excluded | Liberhan, 1992 — the central government excluded. Twenty-nine years apart | Never regulated |
| Preventive detention without trial | 1950, in the first year | Every government since, under successive names | In force |
| The ordinance power | Government of India Act 1935 | Every government | In force |
| Disruption as the opposition’s only lever | A property of the rules, not of any party | Whoever is out of power | Unchanged |
Eight instruments. Every one was built by a Congress government or inherited from the colonial state. Not one has been dismantled by anybody, including by governments that campaigned against the use made of them.
Part Nine asked a reader to name one instrument used after 25 June 1975 that did not exist before 12 June. There is not one.
Part Ten found not a single Punjab instrument that was a Punjab instrument.
Part Sixteen found the coal method dating from 1993 and struck down across five governments and three parties.
Part Eighteen found every tool used on 5 August 2019 built by the party this book prosecutes.
Twelve instances of the same finding, across seventy-nine years, four parties and two series of governments. It is the most repeated result in this book and the only one that has been tested against a change of government and held.
One phrase in that answer carries more weight than the numbers around it.
The most important two words in Chapter Eight, and they are easy to read past.
A bill that is defeated has been voted on. There is a division list, a debate, and a set of people who can be asked why they voted as they did.
A bill that lapses has at least been printed, referred, and usually examined by a committee that published a report.
A bill that is never introduced has none of that. No vote, no debate, no committee, no division list, no minister who declined it, and nobody with a position on it to defend. It has not been rejected by Parliament; it has never been put to Parliament, which means there is not a single person in public life who can be held to a view about it.
Eleven of the fifteen rules in Table 12 are in that state. It is the purest form of the finding in Chapter Seven: concealment has a custodian, and an absence has nobody.
There is a mechanism behind Table 11 and it is worth naming, because without it the table reads as eight separate failures of nerve rather than one structural fact.
An instrument of this kind is built by a government that expects to keep using it. It is then inherited by an opposition that spent years denouncing it — and which now holds it. At that moment the incoming government faces a choice between dismantling a tool it has just acquired and keeping a tool it has just criticised, and the cost of dismantling is immediate and certain while the cost of keeping is deferred and hypothetical.
So it is kept, and denounced less. The next election transfers it again. Each transfer adds a government with a reason not to touch it and subtracts one of the constituencies that objected, until the instrument has been used by everybody and is nobody’s responsibility.
Part Nine found this with preventive detention, denounced by every opposition since 1950 and renewed under a different name by almost every government. Part Thirteen found it with Section 6 of AFSPA, unamended for sixty-eight years across governments that campaigned in the affected states on repealing it. Part Eighteen found the twelfth instance and the first in which the user was not the builder, which is the strongest form of the finding because it removes the last available explanation — that the party defends its own handiwork.
One check on that claim is available and it is a simple one.
Because dismantling leaves a trace and nothing in Table 11 has one.
Repealing a statute requires a repealing Act. Removing a constitutional provision requires an amendment passed by two-thirds of each House. Narrowing an emergency power requires either legislation or a judgment. All three are loud, dated, published events that any reader can look up.
For seven of the eight instruments there is no such event, at any point between 1950 and 2026. Preventive detention has been continuously in force under successive statutes. AFSPA’s Section 6 has stood unamended for sixty-eight years. The ordinance power is unchanged since 1935.
The single exception is instructive rather than encouraging. Article 356 was narrowed in 1994 — Bommai — and it was narrowed by a court, over the objection of the government of the day, and not by any Parliament.
Which is the whole of Chapter Eight in one observation: in seventy-nine years, the only instrument in this table that was ever restrained was restrained by the branch that did not build it and could not use it.
8.2 — The unmade rules
Part Eighteen’s Chapter Seven took every rule this series proposed and asked what had happened to it. The consolidated answer is the second half of column two.
Fifteen rules, proposed across seven parts. Eleven have never been introduced in either House.
Not defeated. Not lapsed in committee. Never brought — which means there is no vote, no debate, no committee report and nobody to ask.
The four that moved had a court order, a catastrophe, or an organised constituency behind them. Not one was made because it was a good idea.
And the eleven that did not: compensation for wrongful prosecution, whose constituency is people just released from jail. The communal violence law, unenacted across twenty-two years and three governments. A register of terrorist incidents, which is nobody’s interest. An archive of political speech, which constrains everybody who would have to create it.
Every rule that got made had somebody powerful behind it. Every rule that did not was for people with no way of insisting.
All of which invites the objection that this chapter exists to let everybody off.
The strongest objection to this entire chapter: that it is a sophisticated way of excusing everybody. If the instruments are inherited and the failures are structural, then no government is answerable for anything — and a book that spends eighteen parts building charges and a nineteenth explaining that the charges were really about a machine has argued itself into a position where nobody can be blamed for anything.
The reply, which this book insists on: the machine had builders and they are named. Article 356 did not become an instrument by itself; it was used against an elected government in Kerala in 1959 by a specific administration. The First Amendment’s procedure was legislated. The Presidential Order route into Article 370 was operated for four decades by identifiable governments.
So the finding is not that nobody is responsible. It is that responsibility divides into two: the party that built the inventory, and every government since that has used it and declined to dismantle it. The first is a finding about Congress and it is the heaviest in this book. The second is a finding about everybody, and Part Eighteen tested it.
Where the objection lands: a reader who takes only Chapter Eight and not Chapter Six will conclude that specific governments did nothing in particular, and specific governments did a great deal. Both columns exist for that reason.
The fifteen, and what became of each.
| The rule | What it would have prevented | Status |
|---|---|---|
| A statutory outer limit on trials for public-office offences | The twenty-year median in Table 2 | Never introduced |
| Compensation as of right for wrongful prosecution | Nine men held five years for Malegaon; twelve convicted and acquitted for the 2006 train bombings | Never introduced |
| A communal violence law with command responsibility | The pattern in Parts Ten and Eleven, where nobody senior is ever reached | Introduced 2005, lapsed 2014. Not reintroduced by anybody |
| Terms of reference for commissions to be laid before Parliament before appointment | Henderson Brooks and Liberhan, twenty-nine years apart | Never introduced |
| A statutory deadline for publishing a commission’s report | Sunderlal, sixty-five years; Pradhan, seventeen and counting | Never introduced |
| Removal of executive sanction as a bar to prosecuting security forces | AFSPA s.6, unamended for sixty-eight years; Mon, December 2021 | Never introduced |
| A national register of terrorist incidents and their judicial outcomes | The fact that no one can state a conviction rate | Never introduced |
| A public verbatim archive of ministerial and parliamentary speech | Six of the eight entries in Table 7 | Never introduced |
| A displacement register, project by project | Part Thirteen’s addition that nobody has made | Never introduced |
| Mandatory publication of presumptive-loss methodology alongside the figure | The gap between ₹10.67 lakh crore and ₹1.86 lakh crore | Never introduced |
| A minimum number of sitting days for each House | Legislation passed in minutes across every government in this book | Never introduced |
| Auction as the default for allocating a public natural resource | Spectrum, coal, and twenty-one years of allocations held illegal | Done — after a court ordered it |
| Statutory backing for the biometric identity system | Six hundred million records held for seven years with no statute | Done — after litigation |
| A standing coastal and maritime security command | The approach route of 26 November 2008 | Done — after the catastrophe |
| A statutory right to information | Most of what this book could not establish before 2005 | Done — with an organised constituency behind it |
Fifteen rules. Eleven never introduced in either House by any government. Of the four that were made, one followed a court order, one followed litigation, one followed a catastrophe and one had a movement behind it.
8.3 — What the machine is
Everybody assumes that a structural finding is weaker than a personal one — that “the system failed” is what you say when you cannot name anybody, and that naming somebody is the stronger result.
Chapters Two and Eight, side by side, reverse it.
Chapter Two’s proven column sent people to prison at a median of over twenty years and changed nothing. The one minister convicted served nine days. Every individual named in that table could be removed from public life tomorrow and every instrument in Table 11 would still be there on Monday.
Chapter Eight’s column has no names in it at all and describes the thing that produced most of Chapter Two.
So a personal finding is the weaker result, not the stronger one — it is satisfying, it is legally consequential for one person, and it leaves the machine running. This series spent seventeen parts looking for names and found, in the eighteenth, that the names were never the variable.
Which is the whole argument of this book compressed into one sentence: a party built a machine over fifty-four years, the machine outlived the party’s dominance, and every government since has used it and none has dismantled it.
That charge is heavier against Congress than any individual allegation in Chapter Four, because it is the one thing in this series that is proved — and it is lighter against any particular ministry, because most of them inherited what they used.
Which is where nineteen parts of this series arrive, and it is not where they set out for.
Eight instruments, every one built by a Congress government or inherited from the colonial state, and not one dismantled by anybody.
Fifteen rules proposed; eleven never introduced in either House. The four that moved had a court order, a catastrophe or a constituency behind them.
And the reversal: a structural finding is the stronger result, not the weaker one. Remove every name in Chapter Two and every instrument in Table 11 is still there on Monday.
9The Strongest Case the Defence Can Still Make
Every part of this series has given the other side its strongest argument. This chapter assembles the whole of it, at full length, for the last time — and it is assembled by the prosecutor, which is the only way it gets assembled at all.
9.1 — Seven arguments
One. The proven column is small and the reason is not the courts. Chapter Two: eight convictions, mostly of officials, at a median of over twenty years; one minister, over three lakh rupees, out in nine days; one Prime Minister, acquitted on appeal. Every one of the largest events in this book — the Emergency, December 1992, the spectrum allocation, the coal allocations — produced no conviction of anybody who decided anything. A defence is entitled to say that a seventy-nine year prosecution with that record has not proved what it set out to prove.
Two. Eleven of the most damaging allegations are not made. Chapter Four. Not that the demolition was directed, not that Rajiv Gandhi took Bofors money or ordered the locks opened, not that any UPA II minister took money. Bofors failed under four different governments across two decades.
Three. Eight of the most-circulated claims are false. Chapter Five, and three of them are among the most quoted political claims in India.
Four. The building record is very large. The Constitution and the first elections; the states reorganised on language after a violent demand and not before; the Green Revolution; nationalised banks that reached rural India; the 1991 reforms in thirty-three days; the 73rd and 74th Amendments; the transparency law of 2005; the employment guarantee; forest rights in 2006; education and food as enforceable rights; the 1894 land acquisition Act replaced. Parts Fourteen to Sixteen record the largest body of rights-creating legislation by any Indian government since the 1950s, and it was passed by the party this book prosecutes.
That fourth argument is the longest and it deserves to be shown rather than asserted.
Argument four is the longest and it is the one this book has never set out in one place. Here it is.
| When | What | Why it counts | Part |
|---|---|---|---|
| 1947–52 | A written constitution with judicially enforceable rights, universal adult franchise from the first election, and a boundary and election commission that functioned | Every comparable state that became independent in that decade lost its first constitution within twenty years | Two |
| 1956 | States reorganised on language | Done after a violent demand rather than before it — but done, and it has held for seventy years | Three |
| 1960s–70s | The Green Revolution; agricultural universities; the procurement system | A country importing grain under an American aid programme became a net exporter | Five, Ten |
| 1969, 1980 | Bank nationalisation | Branch networks reached rural districts that commercial banking had never entered. Part Eight is severe about the motive and concedes the result | Eight |
| 1991 | The licence system dismantled in thirty-three days | Under a minority government, three weeks from default, by people who had helped build the system they removed | Four, Fourteen |
| Dec 1992 | The 73rd and 74th Amendments — elected village and municipal councils, a third of seats reserved for women | Passed sixteen days after the demolition. Around 800,000 women in elected office within a decade | Fourteen |
| 2005 | The Right to Information Act | The largest single transfer of power to ordinary people in this series. Most of the documents in Part Sixteen are public because of it | Fifteen |
| 2005 | The employment guarantee — a right, not a scheme | Enforceable by a named person. Part Fifteen’s distinction between the two is the most useful idea in that part | Fifteen |
| 2006 | The Forest Rights Act | The first statutory recognition that people living in forests had claims to them. Part Thirteen sets out how partially it worked | Thirteen |
| 2009, 2013 | Education as a right; food security for around 800 million people; the 1894 land acquisition Act finally replaced | The largest body of rights-creating legislation by any Indian government since the 1950s | Sixteen |
| 2009–14 | The biometric identity system and direct benefit transfers | Six hundred million people in four years, and the architecture on which most subsequent delivery runs | Sixteen, Eighteen |
| 2005–08 | The end of thirty years of nuclear exclusion | A government risked office for a policy, which is rare in nineteen parts | Fifteen |
Twelve entries. Every one is graded and argued in the part it comes from, and several are argued against there as well.
Read Table 13 and Table 11 together, because that is the whole book.
The same party, in the same decades, built both. The government that passed the 73rd and 74th Amendments is the government of Part Fourteen’s Chapter Six. The government that passed the transparency law is the government of Part Fifteen’s Chapter Four. The government that made food and education enforceable rights is the government of Part Sixteen.
Part Fourteen’s Chapter Nine put the problem in one sentence and it has not been improved on: there is no exchange rate between 800,000 women in elected office and 2,000 dead in December 1992.
Nineteen parts have not produced one. This chapter is not an attempt to strike a balance, because no balance can be struck. It is both columns, printed, so that a reader who wants to argue about the weighting has the actual entries in front of them.
Five. The instruments were mostly inherited from somebody else. Preventive detention has colonial parents. AFSPA descends from a 1942 ordinance. The ordinance power comes from the Government of India Act 1935. Part Two’s finding was that what survived least changed into the Constitution was the machinery of control.
The sixth argument is usually waved away, and it should not be.
It reads like faint praise — a party is credited with obeying election results — and in most democracies it would be. Two things make it substantial here.
The first is the counterfactual that was available. In January 1977 the Emergency was still in force, the opposition leadership was in detention, the press was under pre-censorship, and the government held a majority large enough to have amended almost anything. Elections were called anyway, the government lost, and it left. Part Nine is severe about everything preceding that sentence and does not qualify the sentence itself.
The second is the comparison set. Of the states that became independent between 1945 and 1960, a substantial majority had lost their first constitution to a coup, a one-party amendment or a suspension within twenty-five years. India’s founding party suspended civil liberties for nineteen months, under the Constitution’s own provisions, and then held the election that removed it.
That is not an acquittal on anything in Part Nine. It is the difference between an authoritarian episode and an authoritarian state, and this book has been careful throughout to charge the first and not the second.
Six. Where the party lost, it left. 1977, 1989, 1996, 1998, 1999, 2014, 2019. Whatever else is in this book, no Congress government has refused to leave office after losing an election — including the one that had suspended the Constitution eighteen months earlier and then called the election that removed it.
Seven, and it is the strongest. Part Eighteen applied this book’s own four tests to a different government and three of them returned the same answers. For most of what this series measured, the party was not the variable.
Because most of it was produced by this book against its own interest, and the record of that is checkable.
Part Five found the Punjab rivers case weak on the water. Part Six conceded the correction after 1962 worked and the same army won in 1971. Part Ten recorded that the majority of Sikhs who died in Punjab were killed by Sikhs. Part Twelve declined to assert Bofors. Part Fourteen conceded Bommai outright. Part Sixteen conceded the audit figures were soft. Part Seventeen dismantled its own best material in its first substantive chapter.
A concession made against your own case is evidence about your method. A concession made in favour of it is evidence of nothing, because there was no resistance to overcome.
The seven arguments above are what nineteen parts of resistance produced. They are stronger than anything Congress has ever assembled in its own defence, which is the most useful thing this chapter has to say and the most damning thing about the party’s own advocacy.
9.2 — Where the defence fails
On four things. They are narrower than the reputation of this series and every one survives Part Eighteen.
The instruments were built. Not all of them and not the colonial ones — but Article 356 as an instrument in 1959, the Presidential Order route into Article 370, the amendable interpretation clause in 1951, the terms-of-reference device, the sanction bars. A party that governed for fifty-four years and left that inventory behind is answerable for the inventory, and Chapter Eight is the charge that no acquittal touches.
The proven column being small does not mean nothing happened. Ten court findings still stand, five of them against decisions taken at the highest level. The Emergency was constitutional and was still what Part Nine describes.
The delay is the process. Fifteen years, twenty-three years, thirty-one years, thirty-four years, forty-one years. A defence that relies on the absence of convictions is relying on a system whose median time to conviction exceeds the working life of a witness.
And nothing was fixed. Eleven rules never introduced, three documents never released. The party that built the machine had fifty-four years to dismantle it and did not, and Part Eighteen shows nobody else has either.
Those four are worth stating as a single proposition, because taken together they are what this series actually established.
A government is answerable for two different things: what it did, and what it left behind. The defence in section 9.1 is strong on the first and cannot reach the second. It can show that the convictions are few, that the largest allegations were never proved, that the building record is enormous, and that where the party lost it left. None of that touches the inventory.
Article 356 is still available. Prosecution sanction is still a bar and Section 6 of AFSPA is unamended after sixty-eight years. The terms-of-reference device has never been regulated. Preventive detention has been in force continuously since 1950. The interpretation clause remains amendable and was amended in 2019 to do something its drafters did not contemplate. Every one of those was built or converted into an instrument by a Congress government, and every one is on the statute book this afternoon.
That is the charge that survives every acquittal in Chapter Two, every concession in Chapter Four, and the whole of Part Eighteen — because Part Eighteen strengthens it. A tool used by the party that built it can be defended as that party’s judgment. A tool used by its opponents, against it, is a finding about the tool.
Proved: that a party in power for fifty-four years built a set of instruments for governing without consent, used them, and left them in place; that specific governments did specific things on specific dates, of which the Emergency and the events of Part Ten are the gravest; that public assets were allocated by methods two Supreme Court benches held arbitrary; and that at least twenty ordinary laws which would have prevented documented failures have never been made by anybody.
Not proved, and not asserted: that any Prime Minister took money in any case in this book; that any government directed the events of November 1984 or December 1992; that the largest scandals of 2009 to 2014 involved criminality by any minister.
And unresolved: why the prosecutions failed — which is the question in Chapter Seven and the one that decides how the largest body of material in this series should be read.
9.3 — What the defence has never said
One last observation about Chapter Nine, and it is the reason it was worth writing.
The seven arguments in section 9.1 are, this book believes, the strongest available. Several of them are not in general circulation at all. The party they defend has never assembled them — not in an election campaign, not in a manifesto, not in any document this author has been able to find.
What is offered instead, and has been for decades, is denial of the specific charge of the moment, followed by a counter-charge. That is a strategy for a news cycle. It is not a defence of a record, and its cumulative effect is that the strongest case for fifty-four years of government exists only in the pages of a book written to prosecute it.
There is a lesson in that which has nothing to do with Congress. A defence assembled by the accused is discounted before it is read. A defence assembled by the accuser is not. Which means the most valuable thing any partisan document can contain is the part that argues against itself — and the reason almost none of them do is that it costs the author material he would rather keep.
This series has now paid that cost nineteen times. Whether it bought anything is not for the author to say.
That is the verdict of nineteen parts, and it is narrower than the title of the series promises.
Seven arguments for the defence, and most of them were produced by this book against its own interest. They are stronger than anything Congress has assembled in its own defence.
The defence fails on four: the instruments were built; the small proven column does not mean nothing happened; the delay is the process; and nothing was fixed.
What is proved is a machine. What is not proved, and not asserted, is that any Prime Minister in this book took money.
10An Honest List of What This Book Got Wrong
Nineteen parts have ended with a list of what is not known. The last one is turned on the book.
10.1 — The frame
A reader arriving at this chapter expects a list of factual errors — a date wrong here, a figure disputed there — and this book has certainly made some.
Those are not the important ones. The errors in this series are in its frame, and the largest was set before Part One was written.
The founding decision was to treat Congress as one continuous actor across seventy-nine years. Part One conceded that this was a choice made before the analysis, and said it cut both ways.
Part Eighteen showed how. Apply the same four tests to a different government and three of them return the same answers, which means the continuous actor was never the party. It was a state, and the party built most of it and then stopped being the only one operating it.
So the title of this series is wrong, or at least half wrong. A book called The Congress Record spent nineteen parts producing its strongest evidence about something else.
That is not a small error and it is not correctable by editing. It is the shape of the whole thing, and the honest response is not to retitle it but to say plainly, in the last chapter, what a reader should take from it — which is Chapter Eight and not Chapter Two.
Announcing that in the final chapter is itself open to an obvious complaint.
The objection to the whole of section 10.1: that an author who announces his own framing error in the last chapter has performed humility rather than exercised it. If the frame is wrong, rewrite the book. If it is not wrong enough to rewrite, do not claim it is.
The reply: the frame produced the finding. Nineteen parts of looking for a party’s failures is what surfaced the instruments, the unmade rules and the withheld documents — none of which was visible from any other angle, and none of which a book about the Indian state in general would have found, because such a book would have had no reason to trace the same device from 1959 to 2019.
Where the objection is right: a reader is entitled to say that the correct response to discovering your title is half wrong is not to explain it on the last page. This book has no answer to that beyond the one in the author’s note, which is that the machine was built by the party the title names, and pointing away from the builder to point at the building would be its own dishonesty.
Before the rest of them, a word about what this device has been doing for nineteen parts.
Nineteen parts have ended with one and the device has never been defined. The last chapter is the place.
An honest list is not a list of doubts. Doubts are cheap, they can be manufactured indefinitely, and a document that ends by listing them is performing caution rather than exercising it.
It is a list with three properties. Each item must be specific enough that somebody could resolve it — not “the sources are incomplete” but “the Home Ministry file notes for nine days in 1992 have not been released”. Each must be something that would change the argument if resolved, in a direction the author names in advance. And at least some must cut against the author, or the list is a defensive perimeter rather than a concession.
The test for the third is easy to apply from outside. Read the honest list at the end of any part of this series and ask whether resolving every item would make the chapter stronger. If it would, the list is decoration.
Section 10.6 of this part exists to meet that test at the level of the whole series, and it names four things, three of which are outside this author’s control.
10.2 — Six more
The selection. Nineteen parts from seventy-nine years, chosen by one person. A different author applying the same method would produce an overlapping but different book, and there is no procedure that makes the choice objective.
The denominator. Said five times and worth a sixth: this book has no way of comparing what these governments did with what any government would have done, and every implicit comparison in it is therefore unearned.
Part Seventeen. By its own analysis the most quotable and least probative chapter in the series — and it will be the most read. That is a structural failure of the book and naming it does not fix it.
Part Eighteen’s exclusions. Five contested matters left ungraded because they failed this book’s tests, applied to a government whose files have not surfaced. The rule is neutral and its effect at this moment is not. Part Eighteen conceded it could not answer that and this part cannot either.
The recency problem, in reverse. Everything in Parts Fourteen to Eighteen is provisional in a way that Parts Two to Twelve are not. A reader in 2050 will find the last third of this series naive in ways this author cannot predict.
And the method’s own cost. Part Sixteen recorded it best: sixteen parts of careful grading produced, on the largest scandal in modern Indian history, a conclusion that is honest and close to useless as advocacy. That trade was made deliberately in Part One and this book would make it again, but it should not pretend the cost was small.
A reader is entitled to ask how an author would know his own errors at all.
Because they were found by the book’s own procedure rather than by an opponent, and each one is traceable to a specific part that produced it.
The framing error surfaced because Part Eighteen applied this series’ four tests to a government the series was not about. That test was committed to in Part One, before there was any way of knowing what it would return. An author who designs a test in Part One and reports its result in Part Eighteen has at least made the result checkable.
The denominator problem was named in Part One and repeated in Parts Fourteen, Sixteen and Eighteen, each time without being solved. The Part Seventeen problem was stated inside Part Seventeen, in its own Chapter Eight, which found that only two of eleven statements changed anything.
What this does not establish is that these are the only errors, or the largest. A method’s blind spots are not visible to the method. The errors an opponent would find are, by definition, not in this list, and the invitation in the box below is the only thing this book can offer against that.
10.3 — What this book asks of a reader
Three things, and none of them is agreement.
Check one thing. Not the whole book. One claim, chosen by you, preferably one you disliked. Every load-bearing fact here has a date and a source attached for exactly this purpose.
Read down the grades, not the descriptions. Chapter One’s word box explains why. A reader who does the second gets a very different impression from a reader who does the first, and the first one is correct.
And if you find an error, say so. Chapter Seven’s second table is the weakest part of this book, because asserting that six calculations have never been done is asserting a negative about the whole world. One counter-example makes a row wrong, and a book that invites that and then ignores it deserves nothing.
10.4 — What a book like this is for
Nineteen parts is a long time to spend on a question, and it is reasonable to ask at the end what the answer is supposed to be used for.
Not for voting. Nothing in this series tells anybody how to vote, and the material that would — a comparison against what other governments did with the same powers — is precisely what Part Eighteen showed this book does not have. A reader who arrives at the last page holding a voting instruction has extracted something the evidence does not contain.
Not for a verdict on a party either. The four disclaimers in Chapter One rule that out, and the two columns in this part rule it out twice.
What it is for is narrower and, this book thinks, more useful. Eleven ordinary laws would have prevented documented failures and have never been introduced by anybody. Three documents are being withheld and two of the three would embarrass a party currently in opposition, which means releasing them costs the government of the day nothing. Six calculations would settle questions that are argued about constantly and require no document anyone holds.
Every one of those is available to a citizen, a journalist, a legislator or a student, and not one of them requires agreeing with a word of this series’ politics. A reader who thinks this book is unfair to Congress and who then asks a Member of Parliament why a statutory deadline for publishing commission reports has never been introduced has done exactly what the book was written for.
The argument, in the end, is not about which party. It is that a state which knows this much about itself and adds up this little of it is choosing — by not choosing, which is the way most things are chosen — to remain unable to answer questions about its own conduct. That was true under every government in these nineteen parts.
10.5 — What is owed
Part Eighteen’s section 10.5 committed a second series auditing the government of 2014 onwards, in terms specific enough to be checked: the same scale, the same four tests, the same six grades, the credit chapter placed second, the same protection for people against whom nothing has been found, and adjudication of the five questions Part Eighteen declined.
That commitment is repeated here as the last numbered thing in this series, and one clause is added to it.
The second series must contain a part like Part Eighteen — a comparison, applying the same ruler back to the Congress governments of 2004 to 2014 and to whatever has happened since.
If it does not, it is a campaign document, and Part Eighteen will have been the last honest thing in either series.
Across nineteen parts this book made nine forward promises in numbered sections and kept eight in the part that owed them. This is the ninth, and it is the only one a reader cannot yet check.
Chapter Eight of this part is a table of fifteen promises made by governments and never brought. An author who applies that test to others and not to himself has written the wrong book.
10.6 — What would change the verdict
A document that cannot say what would falsify it is not making a claim. So, briefly, the four things that would.
Release of the three withheld files. The 1992 Home Ministry notes are the sharpest: they would either establish central direction, which this book has declined to allege eleven times, or refute it, which would make Chapter Four’s first row a permanent finding rather than an open one. Either outcome improves on what is printed here.
A completed appeal in the spectrum case. Admitted in March 2024 and pending. A conviction would move the largest body of material in this series from Court-found on the allocation to Convicted on the conduct. A dismissal would settle Chapter Seven’s unresolved question in the first of its three directions.
Any one of the six calculations. The freight equalisation figure alone would either establish the mineral-belt finding as the strongest regional claim in this book or dissolve it, and it is the one entry in the ledger that three separate parts arrived at independently.
And a second series that does what section 10.5 requires. If the same ruler applied to a different government over a full nineteen parts returns findings that do not transfer — if the four tests separate rather than converge — then Part Eighteen’s result was an artefact of a single comparison and Chapter Eight of this part is wrong.
Three of those four are outside this author’s control. The fourth is not, which is why it is a promise and not a hope.
One closing observation, and then this series stops.
Every part of it has ended by naming something the state was holding back or had never done. A report from 1963. Nine days of file notes from 1992. Ninety pages from 2009. An appeal pending since 2018. Six calculations nobody has attempted.
The pattern is so consistent across nineteen parts that it stopped being a series of findings somewhere around the thirteenth and became a single finding repeated. The Indian state knows almost everything about itself and has added almost none of it up. Not because anybody decided not to — a decision would have a defender, and a defender can be argued with — but because the addition was never anybody’s job.
That is a smaller and duller conclusion than a prosecution document is supposed to reach. It is also the only one in this book that has been tested against a change of government and survived.
The errors in this series are in its frame, not its facts. The largest is that a book called The Congress Record produced its strongest evidence about something other than Congress.
What a reader is asked for is not agreement. Check one claim. Read down the grades. And if you find an error, say so.
And the finding that survived everything: the Indian state knows almost everything about itself and has added almost none of it up — not because anybody decided not to, but because it was never anybody’s job.
Sources & further reading — Part 19
The Nineteen Parts
One line each, for a reader who wants to know where to go rather than what to conclude. The right-hand column is what the part established, not what it was about — in several cases those are different, and the difference is usually the interesting part.
| Part | Subject | Its central finding |
|---|---|---|
| One | How to judge a government | The six-grade scale, and the argument for killing your own fabrications |
| Two | Partition and the founding | What survived least changed into the Constitution was the machinery of control |
| Three | The first decade | Article 356 became an instrument in Kerala in 1959; the First Amendment’s durable product was a procedure with no subject matter |
| Four | The licence system | A permission regime is worth more to the incumbent than to the regulator |
| Five | Freight equalisation and the states | A formula moves the politics upstream to the design stage and then hides it |
| Six | 1962 | Warnings ignored for years; terms of reference that excluded Army HQ; a report withheld since 1963 |
| Seven | Kashmir | A temporary provision hollowed out by Presidential Order across four decades |
| Eight | The 1969 split and after | A party organisation converted to answer a leader rather than an electorate |
| Nine | The Emergency | Almost nothing about it was unconstitutional, which is worse than the alternative |
| Ten | Punjab and November 1984 | Not one Punjab instrument was a Punjab instrument |
| Eleven | Communal violence and after | Establishing a record has functioned as a substitute for accountability, not a step towards it |
| Twelve | The 1980s | A government that legislated to reverse a court, and an organisation answering upwards |
| Thirteen | Tribal India and AFSPA | An addition nobody has made, where every component exists and nothing is concealed |
| Fourteen | 1991–1996 | The economy reformed in thirty-three days; two commissions twenty-nine years apart with the same device |
| Fifteen | UPA I | Seventeen alerts forwarded and never assessed, because assessment was nobody’s job |
| Sixteen | UPA II | Every allocation question went against the government; every criminal question failed; both are true |
| Seventeen | What was said | Words and acts do not correlate; the most quotable chapter is the least probative |
| Eighteen | The comparison | Three of the four tests were never measuring the party. They were measuring the state |
| Nineteen | The honest list | The machine outlived the party, and nobody has dismantled it |
A Word on Sources
This part contains almost no new material. It is a consolidation, and its sources are the eighteen parts that precede it.
Every entry in Tables 2 to 13 carries the part it came from. A reader who wants the evidence for any line should go to that part, where the date, the court or commission, and the document are set out at length. Nothing has been graded higher here than it was graded there, and where a case has moved since — the coal acquittals of 2026, the spectrum appeal admitted in 2024 — this part uses the later position.
Two entries have been deliberately weakened relative to their original parts. The Adarsh war-widows claim, which Part Sixteen graded Alleged, appears in Chapter Five’s folklore table because a claim that cannot be established and is universally repeated belongs there. And the Bofors material, which Part Twelve set out at length, is stated in Chapter Four as a defence win rather than a charge.
One handling is carried forward unchanged from Part One and is stated again because this is the last chance to state it. Viral allegations of sexual crime against living named people do not appear in these nineteen parts in any form, including as a claim with a correction attached. Chapter Five’s first row describes a piece of litigation and what a court found about it, and does not say what was alleged. A correction printed underneath a fabrication is a distribution mechanism with a disclaimer on it.
How to use nineteen parts
The series was written to be read in order and almost nobody will read it that way, so three shorter routes through it are worth naming.
For the constitutional argument, which is the strongest thing here: Parts Three, Seven, Nine and Eighteen, then Chapter Eight of this part. That is four parts and a chapter, it is about a fifth of the series, and it contains everything that survived Part Eighteen’s test.
For the method rather than the findings — which is what a reader building something similar about any government would want — Part One, Chapter Two of Part Seventeen, Chapter Three of Part Sixteen, and Chapters One and Ten of this part.
And for the material this book would most like acted on: Chapter Seven and Chapter Eight of this part, which between them contain three withheld documents, six unmade calculations and eleven unintroduced laws. None of it requires agreeing with anything else in the series.
A note from the author
I set out to write a prosecution and I have written something else, and I want to be exact about when I noticed.
It was not Part Eighteen. Part Eighteen is where it became undeniable, but the finding was in the book much earlier and I kept writing past it. Part Nine found every Emergency instrument existing before the Emergency. Part Ten found no Punjab instrument that was a Punjab instrument. Part Thirteen found the addition nobody had made. By Part Fifteen I was writing the same sentence in different settings and treating it as a recurring motif rather than as the result.
What Part Eighteen did was take that motif and test it, and it held. Three of the four tests I had built were not measuring what the title of this series says they were measuring.
I have thought about whether the honest response is to retitle the whole thing. I decided against it, for a reason I want on the record so somebody can disagree with it. The machine in Chapter Eight was built by Congress governments over fifty-four years. That is the heaviest charge in this book and it is the one thing in it that is proved. A title that pointed away from the builder in order to point at the building would be its own kind of dishonesty.
But the reader who takes Chapter Two from this part and not Chapter Eight has taken the wrong thing, and I have said so in the last chapter rather than hoping it would be inferred.
Two smaller declarations, because the rule in this series is that they go where the writing could pull. I am Punjabi and Parts Ten and Eleven concern material I have written on before; both parts concede that most Sikhs killed in Punjab were killed by Sikhs, and that is the concession I checked hardest because it is the one I least wanted to make. And I am a critic of the Congress party who has now written eighteen parts against it and one part about somebody else; Part Eighteen’s Chapter One exists because of that and it is the first chapter of that part rather than the last.
The last thing. Chapter Seven contains six calculations that nobody has done, on data that is already public. I cannot do all six. If one person reading this does one of them and publishes it, that will have been worth more than these nineteen parts, and I would rather that sentence were the true one than the flattering one.
Lovepreet Singh
Part Nineteen of The Congress Record — the last part
The Living Archive · misterlove.in
The second series, auditing the government of 2014 onwards on this scale and by this method, is committed to in Part Eighteen, section 10.5, and again in section 10.5 of this part. It must contain a comparison chapter turning the same ruler back on the Congress governments of 2004 to 2014. If it does not, it is a campaign document and this part was the end of the honest one.