Twenty Years After Prakash Singh — The Reform That Was Ordered and Never Delivered

Pritiraj BrahmaFounding Partner, Pritiraj & Partners · 19 February 2026

The Supreme Court ordered seven police reforms in 2006 and held three of them to require no legislation at all. Two decades on no state is in full compliance, around 5.3 lakh sanctioned posts stand vacant, and two-thirds of the shortfall is in the constabulary — the rank that is 86 per cent of the force and offers one promotion in a career. The reform was not refused; it was routed around.

ProvenancePublished 19 February 2026; revised 28 August 2026 to argue from attributed figures and to add the acting-appointment mechanism. Directives are as issued in Prakash Singh v. Union of India, decided 22 September 2006. Compliance assessment is the Commonwealth Human Rights Initiative’s. Vacancy figures are as placed before the Lok Sabha by the Minister of State for Home Affairs, as on 1 January 2020, from Bureau of Police Research and Development compilations; rank-wise distribution is from press reporting of the Bureau’s data rather than a release we can date precisely, and is used to establish distribution across ranks rather than a point-in-time total. Density and constabulary-share figures are from PRS Legislative Research, drawn from 2016 data. Readers should treat the strength figures as several years old; recruitment has continued since.

India has a police reform. It was ordered by the Supreme Court in 2006, three of its seven directives were held to be self-executory, and two decades later no state is in full compliance. Around 5.3 lakh sanctioned posts stood vacant, two-thirds of them constables — the rank that is 86 per cent of the force and offers one promotion in a career. The reform was not refused. It was routed around.

In Prakash Singh v. Union of India, decided on 22 September 2006, the Supreme Court issued seven directives to the states: constitute a State Security Commission to insulate the police from illegitimate political pressure; select the Director General of Police from among the three senior-most officers empanelled by the Union Public Service Commission and give him a minimum two-year tenure irrespective of superannuation; fix tenure for operational field officers; separate investigation from law and order, beginning with towns above ten lakh; establish a Police Establishment Board to decide transfers and postings; create Police Complaints Authorities at state and district level; and set up a National Security Commission at the Union level.

The Court was explicit that three of these — DGP selection and tenure, tenure of other officers, and the Establishment Board — were self-executory, requiring no legislation and admitting no request for further time. Those are the three that have been most comprehensively evaded, and the mechanism of evasion is the subject of this essay.

Six findings anchor this analysis:

1. No state is in full compliance, on the most-cited independent assessment. The Commonwealth Human Rights Initiative’s assessment of state and union territory compliance found not a single case of full compliance, concluding that governments had blatantly rejected, ignored or diluted significant features of the directives. On the DGP selection and tenure parameters specifically, it assessed two states as compliant, two as partially compliant and twenty-four as non-compliant. In 2019 the Court itself observed that states were treating the directions as mere suggestions.
2. The self-executory directive was evaded through a category the judgment did not anticipate. The directive requires a DGP selected from a UPSC-empanelled panel with a guaranteed two-year tenure. States responded by appointing officers to hold charge as “acting” DGP — a post outside the guarantee. Andhra Pradesh, Telangana, Uttar Pradesh and Punjab held such arrangements for around a year or more; Uttarakhand, Odisha, West Bengal and Jammu and Kashmir made acting appointments, prompting the Court to restate that temporary appointments are for exceptional circumstances only. Uttar Pradesh had four acting DGPs across twenty months.
3. The force is chronically understrength, on figures given to Parliament. Answering the Lok Sabha, the Minister of State for Home Affairs placed the sanctioned strength of the state police at 26,23,225 against an actual strength of 20,91,488 as on 1 January 2020 — 5,31,737 vacancies, a fifth of the sanctioned force. Uttar Pradesh alone accounted for 1,11,865 against a sanctioned 4,15,315, followed by West Bengal at 55,294, Bihar at 47,099, Madhya Pradesh at 31,488 and Telangana at 29,492. Nagaland was the only state whose actual strength exceeded its sanctioned strength.
4. The shortfall is overwhelmingly constabulary, and the constabulary has no career. Rank-wise reporting of the Bureau’s data puts roughly 3.61 lakh of the vacancies at constable rank and a little over one lakh at head constable, against 47,271 sub-inspectors, 24,710 assistant sub-inspectors and 5,527 inspectors. PRS Legislative Research records that 86 per cent of the state police is constabulary, that constables are typically promoted once in a service life and normally retire as head constables. An organisation in which nine of ten members have one promotion available has no meaningful performance incentive, and it is precisely there that the missing people are missing.
5. Density remains far below the international benchmark. PRS records sanctioned strength at 181 police per lakh persons in 2016 against an actual 137, with the United Nations recommended standard at 222 per lakh. The Bureau’s own series shows the ratio against sanctioned strength drifting downward — 195.39 per lakh in 2019 against 198.65 the previous year — which is what happens when sanctioned posts rise more slowly than population. Recruitment has since improved absolute numbers; the shortfall is structural rather than cyclical.
6. The investigation function is the one that absorbs the shortage. The 2006 directive to separate investigation from law and order exists because the two compete for the same officer. Where they are not separated, order duties — bandobast, VIP security, election deployment — always win, because they are visible and immediate while investigation quality is neither. PRS records convictions in 47 per cent of cases registered under the Indian Penal Code in 2015, and notes the Law Commission’s observation that poor quality of investigation is among the reasons. The Bureau’s data has also recorded some 66,043 police personnel deployed to protect 19,467 protected persons.

Where the missing police actually are

The vacancy figure is usually quoted as a single number, and the single number invites the wrong remedy. A fifth of the force missing sounds like a recruitment drive is required. The rank distribution says something more specific: the shortfall is concentrated almost entirely in the rank with the least career, the least training investment and the most direct contact with the public. A constable is the state as most Indians encounter it, and the state has left two-thirds of its shortfall precisely there.

This also explains why episodic recruitment drives have not fixed it. A vacancy rate that has hovered around a fifth for more than a decade is not a failure of any particular drive; it is the absence of a standing system that publishes required intake, recruits against it annually, and reports the gap. Where the numbers exist, they exist because a parliamentary question extracted them.

Compliance on paper, and the workaround

The Prakash Singh directives were an unusual instrument: a court prescribing organisational design to executive governments. That is also why implementation has been weak, and the reason is worth naming plainly rather than moralising about.

Each directive removes a specific discretion from the political executive. A fixed DGP tenure removes the power to replace an inconvenient chief. An Establishment Board deciding transfers removes the single most-used lever of control over an individual officer in India — the threat of a posting. An independent complaints authority removes the ability to manage an allegation internally. The reforms were resisted not because they were badly designed but because they were correctly designed.

What the acting-DGP pattern adds is the general lesson about how such resistance actually works in India. Nobody defied the Court. States constituted the commissions, notified the boards, and then appointed officers to a category the judgment had not contemplated, in which the tenure guarantee simply does not attach. A directive that is self-executory in law can still be inoperative in practice if the executive retains the power to define the post it applies to. That is a drafting lesson as much as a political one, and it applies well beyond policing.

The counter-case, honestly stated

Three arguments cut the other way, and the third is the most serious.

First, policing is a State subject under the Seventh Schedule, and there is a genuine federal objection to a uniform national template imposed by judicial direction. States differ enormously in geography, crime profile and political culture, and a Chief Minister’s argument that the best person to head a force may not be among the three senior-most is not self-evidently wrong. The objection has force. It is an argument for legislating a better appointment mechanism, not for holding the post in an acting capacity for twenty months.

Second, several states have made real gains that a compliance scorecard misses: dial-112 integration, forensic capacity, women’s help desks and beat-level digitisation have measurably improved response in some jurisdictions. This is fair, and a Review that judged policing only by its structural directives would miss most of what has actually improved for a citizen calling for help.

Third, and against our own framing: the vacancy figures are dated, and this essay leans on them. The parliamentary answer is as on 1 January 2020; the rank-wise distribution comes from press reporting of the Bureau’s compilations rather than from a release we can date precisely; the density comparison is from 2016. Recruitment has continued throughout. It is entirely possible that the aggregate has improved more than the figures here show, and the honest position is that the Bureau publishes annually, the data is not contemporaneous in public discussion, and the argument would be stronger if the current year’s numbers were as easy to cite as the judgment is.

What we would do

1. Close the acting-appointment route by rule, not by further litigation. A state rule providing that no officer may hold charge as head of police force beyond a fixed short period without a UPSC-empanelled selection would settle in one page what twenty years of contempt proceedings have not. The directive was held self-executory; what it lacked was a definition of the post it attaches to.
2. Recruit against a published rolling plan. Annual intake requirements published five years forward, by rank, with achievement reported against them. A vacancy rate stable at roughly a fifth for a decade is not a recruitment failure but the absence of a recruitment system, and the absence is invisible because nobody publishes the plan it would be measured against.
3. Build a career for the constabulary. Three assured promotion opportunities in a service life, tied to examination and performance rather than seniority alone. This is the single largest untapped source of policing quality in India, it addresses the rank where two-thirds of the shortfall sits, and it costs a fraction of what technology programmes cost.
4. Separate investigation genuinely, with its own cadre and its own head. The operative test is whether an investigating officer can be pulled onto bandobast duty. If she can, the separation is a notification. Publish, per district, the number of officers on the investigation establishment and the number diverted to order duties in the preceding month.
5. Fund the Police Complaints Authorities properly, or admit they do not function. A complaints body without staff, budget or the power to compel is worse than none, because it certifies the absence of redress as the presence of it. Publish complaints received, disposed and acted upon, per authority, annually.

India did not fail to design police reform. It designed it, had it ordered by its highest court, was told three directives required no legislation at all — and then declined to hand over the discretion the reform required, by holding the post in an acting capacity and leaving a third of a million constable posts unfilled. Two decades of that is a finding about how Indian institutions actually change: not by refusing an order, but by preserving the discretion the order was meant to remove.