The Government Changes, the State Does Not

Politicians fall, slogans change, but Bangladesh's bureaucracy carries much of the same state into every new era.

Bangladesh has changed governments twice since August 2024.

Sheikh Hasina's administration collapsed under a student-led uprising. Muhammad Yunus then led an interim government for 18 months, during which reform commissions were formed, and the International Crimes Tribunal sentenced Hasina and former home minister Asaduzzaman Khan Kamal to death in absentia.

After the February 12 election, Tarique Rahman was sworn in as prime minister on February 17 following the BNP alliance's landslide victory. The changes were consequential. They did not, however, replace the machinery through which Bangladesh is governed.

Ministers changed. Secretaries, deputy secretaries, district administrators, police officers, regulators and clerks largely remained within the service, even when some were transferred, made officers on special duty, or forced into retirement. This is not surprising. A modern state cannot be emptied after every election. Continuity is one purpose of a professional civil service.

The problem is not permanence itself. It is permanence without traceable responsibility. In Bangladesh, politicians receive most of the blame for failed policies and abused power, even though political instructions become state action only after officials examine, record, recommend, authenticate and implement them.

When responsibility is later demanded, the minister can blame the file and the official can blame the order. The system produces a decision, but often no identifiable decision-maker.

The Government Inside the Government

Bangladesh's executive has two layers. The political executive consists of the prime minister, cabinet and ministers. The administrative executive is the permanent service.

Inside a ministry, the chain commonly runs from the minister to the secretary or senior secretary, then through additional secretary, joint secretary, deputy secretary, senior assistant secretary, and assistant secretary.

The field administration runs through divisional commissioners, deputy commissioners, and upazila nirbahi officers. Recruitment currently takes place across 26 Bangladesh Civil Service cadres.

The Rules of Business distribute subjects among ministries and determine when a matter must be referred to the Prime Minister, Cabinet, Finance Division, Public Administration, Cabinet Division, or the law authorities.

The Secretariat Instructions govern the less visible process: Receiving correspondence, opening a file, writing notes, seeking opinions, recording approval and issuing an order. That process is the operating system of government.

A file normally begins below the minister. A desk officer assembles the facts, cites the relevant rule and writes the first recommendation. Senior officers may agree, revise, return or escalate it. Another ministry may add financial or legal concurrence.

Only then does the matter reach the political authority, often in the compressed form of a summary and proposed decision. After approval, officials draft the notification, allotment, licence, contract or sanction that gives the decision legal effect.

This is not clerical work. The first note can determine which facts are visible, whether an exception appears lawful and which alternatives survive long enough to reach the minister. Bureaucratic power lies not only in saying no.

It lies in controlling sequence, language and time: What is placed on the file, what is left out, which rule is treated as binding, and whether a matter moves today or disappears into consultation.

Protection without Public Accountability

The Constitution protects civil servants from arbitrary punishment. Article 135 prevents dismissal, removal, or reduction in rank by an authority junior to the appointing authority and normally requires an opportunity to show cause.

These safeguards are necessary. Without them, every transfer of power could become a purge. But the citizen harmed by administrative failure has no equivalent constitutional machinery for obtaining a prompt explanation, correction or penalty. The imbalance is built into the service framework.

Governments therefore fall back on devices that avoid determining guilt. The most familiar is officer on special duty, or OSD. An official retains salary and status but loses an ordinary posting. OSD can serve legitimate administrative purposes, but it has repeatedly been used as a political holding area.

In March 2025, 517 officials were on OSD, then described as a record; 121 administration-cadre officers had been placed there after the Awami League's fall. Many had not been told of a specific allegation.

The legality of indefinite OSD is already doubtful. In 2021, the High Court held that an official should not remain OSD for more than 150 days: The government should investigate, punish if the allegation is proved, or reinstate the officer. Yet the practice persists because it is easier than a proceeding that ends with a public finding.

Section 45 of the Government Service Act creates another escape route. It allows the government to retire an employee in the public interest after 25 years of service, without a show-cause notice, while preserving retirement benefits.

In July 2026, the government used it to retire 32 former deputy commissioners who had served as returning officers in the disputed 2018 election. They had been OSD since February 2025. The step removed them from service but did not establish, officer by officer, whether they committed misconduct or a crime.

That is the central defect. OSD and compulsory retirement can look like accountability while avoiding adjudication. The public receives no tested account of what happened; the officer receives neither exoneration nor a finding of guilt; and the institution learns little because the conduct is not examined in a reasoned decision.

Criminal accountability encounters a further barrier. Section 197 of the Code of Criminal Procedure prevents a court from taking cognisance of certain offences alleged against qualifying public servants for acts done, or purportedly done, in official duty without prior government sanction.

The rule is meant to protect honest officials from retaliatory litigation. It can also make the executive the gatekeeper of a prosecution against its own personnel. The statutory text contains no clear general deadline for deciding such a request.

The risk is not hypothetical. In 2013, parliament amended the Anti-Corruption Commission Act to require government permission before corruption cases could be brought against public servants, judges and magistrates.

The High Court struck the provision down in January 2014 as discriminatory and inconsistent with equality before law. The attempted shield failed, but Section 197 continues to protect the official-duty boundary in ordinary criminal proceedings.

What the Files Reveal

The disputed elections show how responsibility becomes dispersed. A government inquiry into the 2014, 2018, and 2024 polls concluded that parts of the administration, police, Election Commission and intelligence agencies implemented plans made at the highest level.

It estimated that ballots were stamped overnight in about 80% of polling centres in 2018. The commission described an administration competing to demonstrate loyalty.

The finding matters because a speech from the top cannot engineer an election. It requires returning officers, police deployments, presiding officers, result sheets, transport, communications and records.

Yet the response to the officials associated with that machinery has largely taken the administrative form described above: OSD, transfer or compulsory retirement. Those measures may be defensible as interim management. They are not substitutes for determining individual responsibility.

The S Alam banking scandal shows the same architecture in financial regulation. The Bangladesh Financial Intelligence Unit reported that the group used political influence and institutional control to take over seven banks and a financial institution and allegedly siphoned or laundered about Tk 2.25 trillion through loans and related schemes.

The scale required repeated regulatory failures, not a single exceptional signature. In November 2024, the Anti-Corruption Commission summoned 13 officials from Bangladesh Bank's inspection department over about Tk 3,300 crore in loans channelled through businesses allegedly linked to the group.

It had also questioned 23 Islami Bank officials. A summons is not proof of guilt. It does, however, identify the proper institutional question: What did supervisors and inspectors know, what did their reports say, and why did regulatory action fail?

The enforced disappearance inquiry shows the danger at its most severe. Its final report described disappearances during the previous 15 years as systematic, institutional and politically driven.

The commission received more than 1,800 complaints and earlier reported that the practice was orchestrated through a central command structure. This was not merely a collection of rogue acts. It was an administrative capability involving chains of command, detention sites, transport, records and silence.

These cases differ in gravity, evidence and legal status. They should not be collapsed into one accusation against every public servant. Together, however, they expose the same accountability problem.

State abuse is organized collectively, while punishment is personalised selectively. The political figure becomes the face of the offence. The officials who supplied the procedure, documentation and continuity often disappear back into the institution.

Why Every Government Inherits the Same Bargain

The persistence of this arrangement is easier to explain through incentives than character. Every incoming government needs taxes collected, salaries paid, ports open and order maintained from its first day. It cannot build a new administrative state. It must govern through the one it inherits.

The bureaucracy also needs the government. Promotions, prized postings, deputations, contract extensions and access to powerful offices depend on political confidence. The relationship is therefore reciprocal.

Politicians provide direction, patronage and sometimes protection. Officials provide expertise, legal form, implementation and deniability. Files then fragment authorship.

The minister approved in principle. The secretary forwarded the proposal. Finance commented only on cost. The legal officer vetted form, not policy. The agency acted on ministry instructions.

Everyone participated, but each person's record describes a narrower task. If the decision later becomes indefensible, the file can operate as a chain of mutual alibis.

Successor governments reinforce the lesson when they punish political association rather than proved conduct. Officials learn that the greatest career risk is not administrative failure but proximity to the losing side.

The rational response is caution, upward referral and rapid adaptation to the next centre of power. Neutrality gives way to survival.

Make Responsibility Travel with the File.

Reform should begin with a simple principle: Protection from arbitrary political punishment must be preserved, but protection cannot mean anonymity for an official decision.

First, every major procurement, licence, land allocation, project revision and regulatory exemption should carry a digital responsibility record. It should identify who originated the proposal, supplied facts, gave financial and legal opinions, recorded dissent, recommended approval, made the decision and authenticated implementation.

Liability would still depend on evidence and law. Institutional amnesia would become harder.

The infrastructure partly exists. The government's D-Nothi system is used in more than 15,000 offices and has processed more than 250 million documents, according to a2i. Its audit trail should be made visible to affected citizens, auditors and investigators where secrecy is not legally justified. 'Under process' should identify a desk and a deadline.

Second, OSD and Section 45 retirement should not close an allegation. If wrongdoing is alleged, the government should either open a time-bound disciplinary case, refer suspected crime to an independent investigator, or publish that no case was established.

Transfer may protect an inquiry, but it should not replace one. Third, Section 197 should be narrowed for corruption and abuse-of-power allegations.

A sanction request should receive a written, reasoned decision within a fixed period. If the government does not decide, the sanction should be deemed granted. Refusal should be open to prompt judicial review.

Fourth, promotion into the senior policy ranks should depend less on ministerial confidence. The Public Administration Reform Commission proposed a competitive Superior Executive Service open to officers from different cadres.

Whatever model is adopted, recorded performance, professional competence and integrity should carry more weight than access to the government of the day.

Finally, ministries should publish annual disciplinary statistics: Complaints received, inquiries opened, average disposal time, allegations proved, penalties imposed, losses recovered and matters referred for prosecution. Bangladesh presently discloses too little to distinguish a functioning disciplinary system from selective punishment.

A professional bureaucracy must survive governments. That is one of its virtues. But the official record of a decision should survive with equal force. The uprising changed Bangladesh's rulers; the election changed them again. Neither event could by itself change the incentives inside the file.

The next government will also be tempted to inherit the same bargain: Loyalty in exchange for protection, administrative convenience in exchange for silence. Reform begins when a government chooses to give up that convenience. Until then, ministers will continue to fall while the permanent government learns to serve, and outlast, whoever replaces them.

Md. Abrar Galib is a student of Economics at the University of Chittagong. His writing interests include politics, governance, economic affairs, and human rights.