The Registrar of the National Judicial Data Grid comes in with a tablet under one arm and a single folder under the other, and does not put the tablet down. It shows the Grid’s front page, loaded at 19:21 this evening, and the Registrar has propped it against the District Collector’s box of pins so that you can see the top line from your chair.
“Read the bottom of the folder first,” the Registrar says. “The page has a number on it. It will not be the same number when you finish.”
Pending
- District and taluka courts. 5,16,39,674 cases pending: 1,12,30,161 civil and 4,04,09,513 criminal.
- By age. 40 percent are under a year old, 22 percent are one to three years, 13 percent three to five, 16 percent five to ten, and 9 percent, 47,02,876 cases, are more than ten.
- Last month. 41,39,104 cases instituted; 34,22,324 disposed of.
- Company-law tribunal. About 30,600 cases pending, on 30 benches. DAKSH, a legal research group, found 88 percent of the tribunal’s staff and 84.9 percent of its appellate tribunal’s on contract.
- Insolvency. The statutory limit for a resolution is 330 days. The average is 713.
- Andhra and the national scheme. Smartcards in Andhra were faster and less corrupt with no loss of access; a stricter biometric requirement in the national scheme left 1.5 to 2 million legitimate beneficiaries out at some point.
- 20 August 2024. An advertisement for 45 lateral-entry posts in the civil service is cancelled.
- Economic Survey 2025-26, chapter 16. Proposes “safe spaces” for decisions taken in good faith.
You read the top line twice. Then you do the subtraction the Registrar has already done. Last month 41,39,104 cases came in and 34,22,324 went out, so the stock rose by about 7.2 lakh. Spread over a thirty-day month, that is roughly a thousand more cases an hour. By the time you finish this folder, the page will say a thousand more. This is arithmetic on one month’s figures, not a forecast.
The Registrar is careful about what the number is not. It counts district and taluka courts only. It says nothing about whether any case is a good one, or a delayed one for a bad reason. “It says how many are waiting,” the Registrar says. “Who is waiting and why, you will have to ask someone else.”
The age table is the better clue. Four in ten pending cases are under a year old and six in ten under three, though the 9 percent in the tail has waited longest. The tail is the part with a date on it: 47 lakh cases are more than ten years old.
The Registrar turns the tablet face down and puts a finger on the tribunal line. It is the same pattern in miniature.
Chapter 19 left the insolvency code with its clock: 330 days on paper, 713 on average. The Bankruptcy Law Reforms Committee, which reported in November 2015, had found that creditors recovered about 20 percent in present-value terms and that insolvency took four years. The Economic Survey puts the recovery rate on non-performing loans at 13.2 percent in FY18 and 26.2 percent in FY25. The tribunal that hears the cases has about 30,600 pending, and most of its staff are on contract.
“The law passed,” the Registrar says. “What the bench looks like now is in the folder. Whether it explains the delay, the folder does not say.” It shows a staffing structure, not a measured effect on delay.
The Registrar has left the last two lines of Folder A without comment. They are about a different part of the state.
In Andhra Pradesh, the rollout of biometric smartcards for the rural job guarantee and for pensions was randomised, and the time it saved beneficiaries was alone worth about what the system cost. A follow-on reform to how the job scheme was run raised beneficiary households’ earnings by about 14 percent and cut poverty by about 26 percent, most of it from earnings outside the scheme rather than the scheme’s own payments. The national scheme’s tightening is the other line in the folder, and the next chapter tells it. The same machine, then, with two rollouts. In one the state tested, measured and published; in the other it tightened first.
Folder B is a single sheet in the Registrar’s own handwriting, two columns, headed with two names.
Two maps of the same state
- Kapur (2020). Performance runs from woefully inadequate in core public goods to surprisingly impressive in complex tasks at scale; better where delivery is episodic with a built-in exit than where it is daily and rests on local capacity. This is the design brief: build the thing the state does well, and do not hand it the thing it does badly.
- Pritchett (2009). India’s capable head is no longer reliably connected to the arms and legs; agents of the state routinely do not carry out assigned tasks in health, education and licensing. This is the objection: the brief assumes the arms can be told what to do.
- Pritchett, Woolcock and Andrews (2010). Many countries sit in “capability traps”, where the forms of a functioning state are adopted while the function stays weak and improves very slowly.
“I put Kapur first because he is the more useful for a Prime Minister,” the Registrar says. “He tells you where to start. Pritchett tells you how far the start will go, and he is not wrong either.”
The Registrar does not choose between them. You are asked to.
The Survey’s chapter offers a way through, and you read it twice. It calls for bounded spaces in which officials may try things, and for law that separates good-faith error from corruption. Its own test for any such space is that experimentation be “bounded, reversible, and accountable”. A reader who distrusts the state will hear the first half and not the second. You write both halves on the sheet.
Replay of an earlier scene: Dubey, 2003The Registrar closes the folder on the thing you have been waiting for it to say.
“Two people are in the margin of this folder, and you have met both.” Satyendra Dubey, the highway engineer of chapter 15, shot dead in Gaya in November 2003 after writing a letter alleging corruption on his project; three men were convicted, and the motive is disputed. Jiang Yanyong, the army doctor of chapter 17, who emailed two television stations about the SARS count in April 2003, and in June 2004 was taken into military custody until 19 July. One ended in a trial whose motive is still argued. The other began as an email nobody printed.
“What I ask,” the Registrar says, “is not that you honour either one. It is that the next man or woman who writes the letter is not alone in the room. The Survey asks for protection for the official whose good-faith decision goes wrong. The Rule you write tonight asks for it for the one who raises the alarm as well.”
The choice comes with the folder, not after it, and it is not the sort you would have picked.
You can pass the bill you campaigned on. It is ready, it has a name, and the people who voted for it know the name. Or you can hold it until the institution that must run it is funded in the same budget, which means telling them it waits. A statute without its tribunal, benches and staff is the case on the first page of this folder: 330 days on paper, 713 in the courtroom.
You hold it. That costs you the first season of a promise.
On the Triage Sheet, four stamps go down. R8, the 330-Day Rule: no flagship statute without its institution funded in the same budget. A statutory evaluation unit: nothing scales, and nothing is cancelled, before an evaluation reports. R11, the Warner Rule: protection in law for good-faith decisions and for those who raise the alarm, named here for the first time. R9, the Exceptions-Desk Rule: any digital mandate carries a non-digital fallback and a desk that hears the person it excludes, a rule you will meet again in its own scene.
Triage sheet: 12 of 40 lines filledThe Registrar lifts the tablet and turns it round. The number at the bottom of the page is larger than the one at the top.
The folder under the tablet carries one question. You read it, and leave it where the others lie.
Door D5 · closedCan this state do what you are about to ask of it?