Rana Plaza and the Bangladesh Accord

Before the collapse, responsibility disappeared between contracts

Bangladesh's ready-made garment industry connected a large labor force to global brands that could set product specifications, prices, order volumes, and delivery schedules without owning the factories where clothes were made. Factory owners controlled daily work but operated inside those purchasing terms. Human Rights Watch's interviews and review of garment sourcing describe how low prices, short lead times, late changes, and penalties can press suppliers to reduce labor costs or demand more output.¹

That separation of commercial power from workplace authority became lethal at Rana Plaza, a multistory commercial building in Savar that housed five garment factories. The Bangladesh human-rights organization Ain o Salish Kendra (ASK) recorded during the rescue that cracks had appeared on 23 April 2013, that factory authorities had initially declared a shutdown, and that workers reported being forced back the next morning. Shanta, a twenty-year-old sewing operator interviewed in hospital, told ASK that she raised the cracks when her manager called her to work; he assured her he would be there, and the building collapsed about half an hour after she entered.²

At least 1,100 people were killed and more than 2,500 injured. Published totals differ: a 2020 International Labour Review study uses at least 1,132 deaths, while the International Accord's 2026 anniversary statement uses at least 1,138.³ The variation does not change the scale of the disaster, but it is a reason not to turn one later total into false precision.

The immediate organizational failure was not an absence of information. People closest to danger saw it. They lacked protected authority to convert that knowledge into a stop-work decision. Wages, production pressure, building ownership, factory management, brand purchasing, and public inspection combined so that saying no was individually ruinous.

The disaster pulled commercial power into the safety system

On 13 May 2013, global union federations and apparel brands concluded the Accord on Fire and Building Safety in Bangladesh. Its governance gave company and union representatives equal seats and an International Labour Organization (ILO) neutral chair. Disputes could go to final and binding arbitration enforceable in a signatory's home jurisdiction. Suppliers were covered by the system, but were not themselves parties to the agreement.⁴

The design joined mechanisms that are weaker alone. An independent safety inspector directed structural, electrical, and fire inspections; factories received time-bound corrective-action plans; reports and remediation status were made public. Brands had to maintain employment and income during specified closures, negotiate commercial terms that made remediation financially feasible, and end business with factories that refused required repairs. Workers received training, joint safety committees, a confidential complaints channel, and a protected right to refuse unsafe work.⁵

The agreement did not simply require every brand to pay every repair bill. A 2016 study commissioned by the International Finance Corporation and ILO, drawing on 100 corrective-action plans and 32 stakeholder interviews, found that access to remediation finance varied with factory size and buyer relationship: strategic suppliers were more likely to receive buyer support, while weaker or smaller suppliers more often had to self-finance or faced a harder path to credit.⁶ The Accord moved responsibility upstream, but did not erase the bargaining differences among factories.

The current Bangladesh Safety Agreement is implemented by the tripartite RMG Sustainability Council (RSC), with brands, unions, and industry represented in its governance. Its mandatory factory scope centers on cut-make-trim facilities producing for signatories; home textiles, fabrics, and accessories may be listed voluntarily.⁷ The Bangladesh Safety Agreement program remains the authoritative destination for its current factory and program records.

Workers became sensors with a route to remedy

An inspection captures a building at intervals. Workers encounter blocked exits, electrical faults, new cracks, heat, unsafe machinery, and managerial pressure every day. The Accord therefore paired engineering with workplace programs. Joint labor-management safety committees receive training; all-worker meetings explain hazards and rights; workers and unions can use a confidential complaints mechanism; the agreement recognizes a right to refuse unsafe work.⁸ This route matters because it crosses the worker's immediate chain of command. A supervisor who wants output cannot be the final judge of whether the worker may report the supervisor's unsafe instruction.

Worker voice is still a relationship of power, not a training completion. Naila Kabeer, Lopita Huq, and Munshi Sulaiman surveyed 1,500 workers from 240 factories in 2017, interviewing them at home rather than through managers, and followed with focus groups in 2018. Workers in factories affiliated with the Accord, the Alliance, or both were more likely to report structural-safety improvements and workplace participation committees. Yet only a minority of committees were described as highly effective, and 75 percent of the full sample reported some mistreatment in their current factory. The study could not isolate the Accord's effects from the Alliance's in jointly covered factories.⁹

The formal complaints system has boundaries too. The International Accord says its Bangladesh mechanism has handled more than 4,000 occupational-safety complaints. In November 2025 it announced expansion into wages, overtime, contracts, severance, discrimination, and other fundamental labor rights; a February 2026 update said that expansion was on hold pending a phased RSC-board rollout.¹⁰ In a 2025 statement collected by the Clean Clothes Campaign, Bangladeshi union leaders separately argued that employer influence had weakened safety committees, that prices and poverty wages still drove risk, and that an employment-injury scheme needed legal protection.¹¹ Those are labor-advocacy judgments, not neutral program measures, but they identify the power conditions that inspection totals cannot answer.

Physical remediation has nonetheless been extensive. A 2025 peer-reviewed institutional study based on six Bangladesh visits, 143 interviews, and program records reports that, by early 2021, the Accord had recorded 33,253 follow-up inspections across 2,222 factories, identified 145,673 hazards, and marked 93 percent of required remediation complete. The authors note that the quantitative series came from Accord quarterly reports.¹² The International Accord's current page reports a different program frame: as of 30 April 2026, 1,772 covered factories, 49,832 inspections, and 84 percent average remediation progress.¹³ These figures document the institution's reported reach; they are not one continuous independent audit.

Durability remains unsettled. The 2023 Bangladesh Safety Agreement runs through 31 December 2026. In April 2026 the Accord said renewal negotiations were under way and acknowledged that hundreds of factories had not completed all required remediation.¹⁴

Compensation and prevention required different institutions

Safer factories could not repair the injuries, deaths, lost income, and family care created at Rana Plaza. The Rana Plaza Arrangement brought government, brands, employers, workers' organizations, and civil-society groups into a separate claims and compensation process chaired with ILO support. It used a formula informed by ILO Employment Injury Benefits Convention No. 121; by October 2015, more than $30 million had been distributed to 5,109 injured workers and dependents, with a separate medical trust continuing afterward. The first claims submissions nevertheless began about eleven months after the collapse.¹⁵

Aggregate payment does not describe whether a household was repaired. A 2023 worker-centered research brief records survivor Latifa Lipi's account: one of her four sisters who worked at Rana Plaza was killed, she spent most of her BDT 150,000 payment on treatment, and lasting injuries made steady work difficult. An ActionAid Bangladesh-commissioned survey of 200 survivors that year found 54.5 percent unemployed, 29 percent reporting worsening psychosocial condition, and roughly one quarter unable to afford treatment and medicine. Because the report does not provide a population sampling frame for survivors, these numbers describe its 200 respondents, not every survivor.¹⁶

The separation reveals two kinds of accountability. Compensation recognizes a specific harm after it occurs. The Accord reorganizes authority and incentives to prevent recurrence. Neither substitutes for criminal or civil justice, long-term health care, employment-injury insurance, or the survivor's own account of what repair requires. As of 23 April 2026, Bangladesh Sangbad Sangstha reported that the Rana Plaza murder trial had heard 145 of 594 listed witnesses and remained unfinished.¹⁷

Government action also expanded. Bangladesh amended labor law, developed safety committees, strengthened inspection, and assessed export factories through a national initiative alongside the two brand-led programs. The ILO reports that 3,780 export factories were assessed through the government, Accord, and Alliance programs, while Department of Inspection for Factories and Establishments staffing rose from 92 inspectors before Rana Plaza to 393. The same 2023 account identifies unfinished work on union freedom, social dialogue, social protection, and safety beyond garments.¹⁸

Physical safety improved inside a still-extractive system

The Accord is deliberately bounded. It can compel inspection and remediation of covered safety hazards. It does not set a living wage, eliminate excessive production targets, guarantee union freedom, stabilize orders, or cover every subcontractor. The worker survey found the strongest improvements around physical safety while basic wages, workload, mistreatment, and collective voice remained sticking points. Independent labor-law scholarship likewise concludes that the Accord was effective in its mostly first-tier safety scope but could not reach many deeper suppliers or the purchasing practices that constrained investment.¹⁹

That boundary does not diminish the safety achievement; it locates it. In a bounded industry and period, brands and unions created enforceable transnational governance where ordinary contracting had separated control from consequence. Public inspection records made hazards legible. Commercial leverage made remediation harder to ignore. Worker channels gave daily knowledge a route beyond factory management.

The Foxconn production system shows the same division beyond garments: buyer launch deadlines and product authority travel down a contract chain while labor responsibility remains at the supplier. The Accord's distinctive move was to make brand power formally answerable to independent inspection and worker complaint rather than relying on buyer codes and audits alone. This is an analytical comparison of commercial power and supplier labor responsibility, not a claim of direct influence, identical work, or equivalent law and worker experience; the reviewed sources establish none.²⁰

The before-and-after story is consequently larger than better audits. Before Rana Plaza, workers could see danger while every powerful institution retained a way to disclaim it. Afterward, the Accord joined knowledge, enforcement, money, transparency, and union standing in one safety regime. Its unfinished question is whether that model can expand before the next disaster supplies the political will.

Concept fingerprint: commercial power became answerable to safety evidence

Purpose, mission, and institutional legitimacy has supporting weight. Garment production and trade supplied the commercial purpose; the Accord created a narrower collective mission around physical safety. Its legitimacy rests on connecting brand power to worker protection, without turning success in covered hazards into a claim about wages, organizing, or every subcontractor.⁴¹⁹

Authority, legitimacy, and acceptance is defining. Brands and unions became parties with equal governance seats, an ILO chair, and enforceable arbitration; inspectors could require correction; workers received complaint and refusal rights; suppliers remained covered but not parties. The design redistributed safety authority without making each affected worker or factory an equal constitutional participant.⁴⁵

Delegation, decentralization, and responsibility has supporting weight. Brands set commercial terms, factories control daily work, engineers inspect, the RSC implements, committees monitor, and public institutions regulate. The Accord makes responsibility follow selected upstream power, while financing evidence shows that smaller or less strategic factories still carried different remediation burdens.⁶⁷

Coordination, communication, and common understanding has supporting weight. Common standards, factory lists, inspections, corrective-action plans, public reports, training, committees, complaints, and commercial consequences connect firms and workplaces. Coverage and complaints remain bounded, so a common system for enrolled physical-safety risks is not a common understanding of every labor condition.⁵⁸¹⁰

Structure, hierarchy, and scale is defining because the supply chain separates buyer commercial authority from supplier employment and building authority across borders. The Accord overlays that hierarchy with transnational brand-union governance, an independent technical function, and workplace channels. Reported factory and inspection totals measure changing program frames, not one uninterrupted population.⁴¹²¹³

Decision making, judgment, and bounded rationality has supporting weight. Workers judged cracks under wage and managerial pressure; engineers judge hazards and remediation; brands and factories judge financing and sourcing; complaints staff judge remedy. The collapse shows why information without protected refusal is not meaningful choice, while later studies show that committee presence alone does not settle effectiveness.²⁹

Measurement, accounting, and control is defining. Hazard findings, corrective-action deadlines, completion status, factory coverage, inspections, complaints, financing plans, and survivor claims make responsibility visible. Program-reported totals, changing denominators, and aggregate compensation do not independently establish safety in every factory or repair for every household.^12131516

Cooperation, incentives, and organizational equilibrium has limited weight. Enforceable arbitration, continued-sourcing commitments, wage protection during closure, remediation terms, termination for refusal, worker rights, and union standing alter incentives to cooperate around safety. Purchasing pressure, poverty wages, and unequal factory finance remain outside or only partly corrected, so the wider equilibrium is not established.⁵⁶¹

Work design, productivity, and automation has supporting weight, correcting the imported zero. Production targets, deadlines, workload, managerial pressure, safety committees, training, refusal, and factory closure directly affect how garment work is organized. The sources are stronger on safety and purchasing pressure than on task design, pay systems, or automation across covered and uncovered factories.¹²⁹²⁰

Knowledge, expertise, and professional autonomy has supporting weight. Workers carried daily hazard knowledge, engineers held building expertise, unions supplied labor-rights knowledge, and independent inspection prevented factory management from being the final technical judge. Worker knowledge becomes consequential only while complaint and refusal routes remain protected from retaliation.²⁵⁸

Learning, quality, and reliability is defining. Disaster led to a repeatable inspection, remediation, disclosure, training, committee, and complaint system that later moved into successor agreements and the RSC. High reported remediation coexists with incomplete work and renewal uncertainty, so learning is an ongoing institutional capability—not a completed hazard count.¹²¹⁴

Strategy, competition, and adaptation remains at score zero as an independent lens. Brands and factories adapt sourcing, finance, production, and remediation to market and Accord obligations; the institution adapted from the 2013 agreement to later Bangladesh arrangements. The record does not compare competitive outcomes or establish how safety obligations redistribute orders between signatories, non-signatories, and uncovered suppliers.⁶⁷

Innovation, entrepreneurship, and renewal remains at score zero as an independent lens. Binding transnational arbitration joined to independent inspection, transparency, commercial leverage, and union standing was an institutional innovation. Renewal remains provisional because current coverage, rights expansion, unresolved remediation, and agreement duration are still contested.⁴¹⁰¹⁴

Governance, stewardship, and accountability is defining across brand-union rule, ILO chairing, engineering independence, public correction plans, commercial terms, worker channels, state reform, compensation, and criminal proceedings. Prevention, compensation, public regulation, and criminal justice are distinct institutions; none can stand in for all the others.^4151718

Culture, informal organization, trust, and voice has supporting weight. Workers saw danger before collapse but lacked a safe route to stop; later committees and complaints attempted to convert voice into remedy. Survey and union evidence shows that mistreatment, employer influence, retaliation, and weak organizing rights persist, without supporting one uniform culture across all factories.²⁹¹¹

Executive attention, information, and organizational sensing is defining, correcting the imported zero. Public factory lists, hazard reports, remediation status, complaints, sourcing consequences, and governing boards route local safety evidence to brand and institutional authority. The mechanism does not ensure that wages, workload, organizing, or indirect suppliers enter executive attention with equal force.^5131920

Organizational ignorance has supporting weight because contract boundaries allowed brands to know product, price, volume, and schedule while disclaiming building conditions; workers' direct hazard knowledge lacked an enforceable route. The Accord reduced selected ignorance through transparency and complaint, but scope boundaries can recreate it beyond covered factories and risks.¹²⁷²⁰

The declared links to measurement, governance, culture, and executive attention are conceptual lenses grounded in those mechanisms. The benefit-for-all-life link is an ethical audit of how worker safety, family repair, supplier burden, customer visibility, environmental gaps, and future protection are counted. The links do not assert direct influence or classifications adopted by the cited institutions.²⁰

Paths into deeper study

• Reconstruct one worker complaint from the first observation through investigation, remedy, wage protection, and any retaliation.
• Follow the cost of one structural repair among factory owner, brand, financier, worker, and buyer purchasing terms.
• Compare covered and uncovered factories, including indirect subcontractors, to see where responsibility again stops before commercial power does.